Title : NSF 93-133 - Summary of Awards Type : Dir of Awards NSF Org: CISE / IRI Date : December 31, 1993 File : nsf93133 DIVISION OF INFORMATION, ROBOTICS, AND INTELLIGENT SYSTEMS SUMMARY OF AWARDS FISCAL YEAR 1992 NATIONAL SCIENCE FOUNDATION DIRECTORATE FOR COMPUTER AND INFORMATION SCIENCE AND ENGINEERING WASHINGTON, D.C. 20550 The Foundation provides awards for research in the sciences and engineering. The awardee is wholly responsible for the conduct of such research and preparation of the results for publication. The Foundation, therefore, does not assume responsibility for such findings or their interpretation. The Foundation welcomes proposals on behalf of all qualified scientists and engineers, and strongly encourages women and minorities and persons with disabilities to compete fully in any of the research-related programs described in this document. In accordance with Federal statutes and regulations and NSF policies, no person on grounds of race, color, age, sex, national origin, or disability shall be excluded from participation in, denied the benefits of, or be subject to discrimination under any program or activity receiving financial assistance from the National Science Foundation. The National Science Foundation has TDD (Telephonic Device for the Deaf) capability which enables individuals with hearing impairment to communicate with the Division of Human Resource Management for information relating to NSF programs, employment, or general information. This number is (703) 306-0090 (effective October 1, 1993). Facilitation Awards for Scientists and Engineers with Disabilities (FAD) provide funding for special assistance or equipment to enable persons with disabilities (investigators and other staff, including student research assistants) to work on an NSF project. See the program announcement, or contact the program coordinator in the Directorate for Education and Human Resources, Tel. (703) 306-1636. Catalog of Federal Domestic Assistance Number 47,070, Directorate for Computer and Information Science and Engineering. PREFACE The Division of Information, Robotics, and Intelligent Systems (IRIS) is one of the six Organizations which make up the National Science Foundation's Directorate for Computer and Information Science and Engineering (CISE). The others are the Division of Advanced Scientific Computing, the Division of Computer and Computation Research, the Division of Microelectronic Information Processing Systems, the Division of Networking and Communications Research and Infrastructure, and the Cross Disciplinary Activities Office. The IRIS Division is concerned with scientific and engineering research in pursuit of the following goals: (1) to increase scientific understanding of information processes in machines, organisms, organizations, and other systems; (2) to stimulate innovation and enhance U.S. competitiveness in the design of information-based products and the manufacture of information processing machines; and (3) to strengthen the national infrastructure for research, design, and manufacturing in this field. The purpose of this report is to provide the research community and the general public with the award data and the summaries of the research projects supported by this Division in FY 1992. Individually or collectively, these projects have contributed to scientific and engineering advances in various IRIS areas. Some of the research highlights are: o Development of robotics as the science of representing, recognizing, reasoning about, and manipulating physical objects, and performing other physical tasks. o Use of image, voice, and other sensory inputs to study man- machine partnerships and intelligent interfaces between perception and action. o Study and development of reasoning, planning, speech and language understanding, and other cognitive capabilities for computers in the face of imperfect information and a changing environment. o Development of database technologies and knowledge-based systems for the collection, organization, maintenance, distribution, as well as application of machine-interpretable forms of human expertise and know-how. o Development of theories and systems for collaboration in distributed and networked environments, including testbeds for studying and developing electronic environments for the conduct of science, education, and public access of Scientific and Engineering Knowledge. Research in this field is highly interdisciplinary, drawing on mathematics, the computer sciences and engineering, and the biological, behavioral, and cognitive sciences. It reflects the state-of-the-art in microelectronics, large-scale and distributed computation, user input/output, and networking. The organizational chart on page vii shows the current programs and the administrative structure with which we manage these diverse research activities. In addition to the five programs, IRIS also co-manages a Science and Technology Center on Cognitive Science located at the University of Pennsylvania. While this summary of awards represents the accomplishments of the principal investigators, the contributions made by the countless anonymous reviewers who volunteered their time and expertise to help the IRIS staff in the merit evaluation process must be acknowledged. In a sense, the fruits shown in this publication belong equally to them. Y. T. Chien, Director Division of Information, Robotics, and Intelligent Systems TABLE OF CONTENTS PAGE Preface................................................. Table of Contents....................................... Introduction............................................ Analysis of Awards and Expenditures..................... Partnership with Industry under PYI/NYI Programs........ Advisory Committee...................................... NSF's Cross Disciplinary Activities..................... IRIS FY 1992 Special Initiatives......................... Knowledge Models and Cognitive Systems Program.......... Traditional Approaches to Artificial Intelligence..... Highly Parallel Approaches to Artificial Intelligence Natural Language Processing......................... Machine Learning and Knowledge Acquisition......... Cognitive Systems.................................. Special Projects................................... Database and Expert Systems Program..................... Data/Information/Knowledge Modeling................ Information Access................................. Physical and System Aspects....................... System Development and Administration.............. Scientific Databases............................... Special Projects................................... Information Technology and Organizations Program......... Coordination Theory/Collaboration Technology....... Distributed Computing/Shared Environments.......... Impact/Policy...................................... Information/Decision Theory........................ Special Projects................................... Interactive Systems Program............................. Visualization and Interactive Computing (Virtual Reality).............................. Visualization and Manipulation......................... Speech and Natural Language Understanding............. Communication Modalities (Facial Expression Interfaces).................... (Gesture and Stylus Interfaces)................... (Physiological Interfaces)........................ (Acoustic Interfaces)............................. (Tactile and Haptic Interfaces).................... Adaptive Human Interfaces (Intelligent Interface Agents) Information Retrieval Environments Learning, Educational and Decision Environment Special Projects.......................................... Robotics and Machine Intelligence Program............... Computer Vision and Pattern Recognition............. Robotics Perception and Active Sensing.............. Robotics Reasoning, Learning, Planning and Control.. Multi-robot Coordination and Cooperation........... Special Projects.................. Index of Principal Investigators........................ INTRODUCTION This report provides summaries of awards made by the National Science Foundation (NSF) through the Division of Information, Robotics, and Intelligent Systems (IRIS). The summaries are arranged alphabetically (by Principal Investigator) within major categories of research for each of the Division's five Programs. Where awards have received support from other organizations within NSF, those organizations are duly noted. The index at the end of this publication provides an alphabetized list of Principal Investigator's names for easy cross reference by the reader. Request for additional information about specific projects should be addressed to principal investigators. General information on current programs may be obtained from the Cognizant Program Directors or from the Division Office. Information, Robotics, and Intelligent Systems Division [Note to Reader: Organization Chart not included in this electronic document.] ANALYSIS OF AWARDS AND EXPENDITURES The IRIS operating budget for fiscal year 1992 was approximately $26.9 Million. In fiscal year 1986, when the CISE Directorate was created, the budget for IRIS was about $15.09 Million. Over the years, IRIS has experienced continued growth of program activities. This is reflected in increased proposal submissions as well as increases in the number and type of awards made. An analysis of the awards and program expenditures in this fiscal year is presented in the charts below. Chart 1 shows the proposal/award history for the seven-year period, FY 1986 - FY 1992. In it, "new proposals" refers to the proposals received in FY '92 that were for competitive evaluations; "new awards" refers to those new proposals which were selected for FY '92 actions based on the evaluation process. "All proposals" is the total of new proposals and the proposals from the continuing grants submitted for FY '92 actions. Also shown on the following pages are IRIS expenditure profiles by program areas, by award size, and by award type in Charts 2 through 4, respectively. The data and definitions used in these charts were obtained from IRIS's program files and proposal management systems; they may differ from official NSF source documents in details. It is important to note that the majority (about 63%) of the grants made in FY 1992 were for small (less than $50K) to medium-size ($50K-$100K) individual investigator projects (Chart 3). The remainder (about 37%) were funded at higher levels for multi- disciplinary team research or for large-scale experimental work. Also significant is the increasing amount of support for new researchers through the Presidential Young Investigator (PYI) and NSF Young Investigator (NYI) programs and the Research Initiation Awards (RIA) Program, and for other cross-directorate programs designed for development of education and human resources (Chart 4). For listing of all NSF cross-directorate activities, see page xxii. [Note to Reader: Charts 1, 2, 3 & 4 are not included in this electronic document.] PARTNERSHIP WITH INDUSTRY UNDER PYI/NYI PROGRAMS The National Science Foundation holds an annual competition for awards to Young Investigators. Originally through the Presidential Young Investigator (PYI) program, and since FY 1992 through the NSF Young Investigator (NYI) program, recipients receive a five-year award with the base amount of $25,000 per year, and an additional $37,500 granted on a matching basis with contributions from industrial sponsors. The objectives of this program are to attract and retain outstanding young faculty in science and engineering, to provide their research and teaching careers with a strong start and greater freedom to pursue their research interests, to improve the research capabilities of academic institutions and to foster contact and cooperation between academia and industry. A separate competition for a limited number of awards made by the Executive Office of The President is held (The Presidential Faculty Fellow award program). The participation of industrial sponsors is greatly appreciated. We believe their contributions to this program help make the scientific establishment of the United States stronger and also contributes to the strength of our economy and competitiveness. The IRIS Division thanks the contributors and recognize them by publishing the names of awardees and their industrial sponsors in this section of the Summary Awards. [Note to Reader: Chart 5 is not included in this electronic document.] IRIS ADVISORY COMMITTEE DECEMBER, 1992 Peter Allen Charles Dyer Dept. of Computer Science (Committee Chair) Columbia University Dept. of Computer Science 500 W. 120th Street University of Wisconsin New York, NY 10027 Madison, WI 53706 Phone: (212) 854-8186 Phone: (608) 262-1965 E-Mail: allen@cs.columbia. E-Mail: dyer@cs.wisc.edu edu George A. Bekey Paula Hawthorn Computer Science Dept. HP Labs Univ. of Southern California Building 1U, MS 14 941 W. 37th Place 1501 Page Mill Road Los Angeles, CA 90089-0782 Palo Alto, CA 94304 Phone: (213) 740-4501 Phone: (415) 857-6833 E-Mail: bekey@pollux.usc.edu E-Mail: hawthorn@hplabs. hp.com Alan Biermann John Hopcroft Dept. of Computer Science Dept. of Computer Science Duke University Cornell University Durham, NC 27706 4130 Upson Hall Phone: (919) 660-6539 Ithaca, NY 14853 E-Mail: awb@cs.duke.edu Phone: (607) 255-7416 E-Mail: Hopcroft@cs. cornell.edu William J. Campbell Elaine Kant Code 934 Schlumberger Laboratory NASA GSFC Computer Science Greenbelt, MD 20771 P.O. Box 20015 Phone: (301) 286-8785 Austin, TX 78720-0015 E-Mail: Campbell@nssdcb. Phone: (512) 331-3737 gsfc.nasa.gov E-Mail: kant@slcs.slb.com Brian Carlisle Nils Nilsson ADEPT Technology, Inc. Robotics Laboratory 150 Rose Orchard Way Cedar Hall San Jose, CA 95134 Stanford University Phone: (408) 434-5011 Stanford, CA 94305 E-Mail: None Phone: (415) 723-3886 Fax Number: (408) 434-5077 E-Mail: Nilsson@cs. Stanford.edu Gary M. Olson Stanley Reiter Cognitive Science and Center for Mathematical Studies Machine Intelligence Lab. Northwestern University University of Michigan 3-014 Levdrone Hall 701 Tappan Street Evanston, IL 60201 Ann Arbor, MI 48109-1234 Phone: (708) 491-2531 Phone: (313) 747-4948 E-Mail: sreiter@casbah.acns. E-Mail: gmo@csmil.umich.edu nwu.edu Mari Ostendorf Avi Silberschatz College of Engineering Department of Computer Sciences Boston University University of Texas at Austin 44 Cummington Street Austin, TX 78712 Boston, MA 02215 Phone: (512) 471-9706 Phone: (617) 353-5430 E-Mail: Avi@cs.utexas.edu E-Mail: mo@buenga.bu.edu Joseph Pasquale Donald E. Walker Dept. of Computer Science Bellcore, MRE 2A379 and Engineering 445 South Street, Box 1910 Mail-C-014 Morristown, NJ 07960-1910 University of California Phone: (201) 829-4312 /San Diego E-Mail: walker@flash. La Jolla, CA 92093 bellcore.com Phone: (619) 534-2673 E-Mail: pasquale@cs.ucsd.edu Leonard Wesley Artificial Intelligence Center SRI International 333 Ravenswood Avenue Menlo Park, CA 94025 Phone: (415) 859-3368/6470 E-Mail: Wesley@ai.sri.com NSF CROSS DISCIPLINARY ACTIVITIES Computational Science and Engineering (CSE) Educational Supplements (ES) Experimental Program to Stimulate Competitive Research (EPSCOR) Facilitation Awards for Scientists and Engineers with Disabilities (FAD) Faculty Awards for Women Scientists and Engineers (FAW) Minority Research Initiation (MRI) - Research Initiation Minority Research Initiation (MRI) - Planning Grant Minority Research Initiation (MRI) - Extension Award Presidential Young Investigators (PYI) Research Assistantships for Minority High School Students (MHS) Research Experiences for Undergraduates (REU) - Sites Research Experiences for Undergraduates (REU) - Supplements Research Initiation Awards (RIA) Research in Undergraduate Institutions (RUI) - Equipment Research in Undergraduate Institutions (RUI) - Research Research Opportunity Awards (ROA) - Supplements for Small College Faculty Research Opportunities for Women (ROW) - Research Initiation Research Opportunities for Women (ROW) - Planning Grants Research Opportunities for Women (ROW) - Career Advancement Small Business Innovation Research Awards - (SBIR) Software Capitalization (SC) Visiting Professorships for Women (VPW) NSF Young Investigator Awards (NYI) IRIS FY 1992 SPECIAL INITIATIVES Intelligent Material Handling Systems - (IMHS) Intelligent Control - (IC) (Will be Awarded in FY-93) Grand Challenges: High Performance Computing and Communications - (HPCC) Research on Scientific Databases - (SDB) MESSAGE FROM THE PROGRAM DIRECTOR The budget of FY 1992 for the Knowledge Models and Cognitive Systems (KMCS) Program reached $6.65 million. Some notable highlights are the increase in RIA (Research Initiation Award) funding, the increase in joint funding with other programs in the Foundation, the increase in REU (Research Experiences for Undergraduates) funding, and the participation in special NSF initiatives. The number of new RIA awards were 13 (one is jointly funded with the Robotics and Machine Intelligence Program). With two new NYI awards, the KMCS Program supported 15 new young researchers in FY 1992. The number of joint funding reached more than 29 with various programs on multi-disciplinary research activities. The number of REU awards was 10. In FY 1992, the KMCS Program participated in two NSF Initiatives - HPCC Grand Challenge Applications and Intelligent Control. A number of grants under these two initiatives were made with contributions from the KMCS Program. KNOWLEDGE MODELS AND COGNITIVE SYSTEMS PROGRAM The research focus of the Knowledge Models and Cognitive Systems (KMCS) Program is Artificial Intelligence. Major program areas are: (1) traditional approaches (2) highly parallel approaches, (3) natural language processing, (4) machine learning and knowledge acquisition, and (5) computational emulation of human cognition. Traditional approaches are characterized by symbolic representations and a high degree of structures imposed by the programmer, in an attempt to represent complex entities and carry out complex tasks involving search and reasoning. Highly parallel approaches - such as parallel distributed processing, neural computing, connectionism, genetic algorithm, and emergent computation - may have similar long-term goals traditional approaches, but take a different direction which is based on a high degree of parallelism among relatively simple processing units connected according to various patterns. Natural language processing* includes: (i) computational aspects of syntax, semantics, and lexicon, (ii) discourse, dialog, and generation, and (iii) system issues. The distinction between the first two often involves such intersentential concerns as topic, plan, and situation. System issues include the interaction and unified treatment of various kinds of modules. Machine learning and Knowledge Acquisition emphasize problems of automatic acquisition of knowledge and assimilation of information. Emphasis is on the design of suitable knowledge structures, static (rule-based) and dynamic (neural-network-based) control paradigms as well as the _______________ *Starting FY 1993, natural language processing will be moved into the Interactive Systems Program and will be merged with speech understanding. development of various algorithms which embody the dynamic change in learning processes. In these four areas, research may be enhanced by the scientific understanding of the nature of high- level human cognition. Moreover, computational emulation of human cognitive capabilities enables the development of intelligent computer hardware and software systems. New research opportunities include, but are not limited to: New approaches for knowledge representation and processing Automated knowledge acquisition Hybrid traditional and highly parallel paradigms Human cognitive modeling Time-critical intelligent systems Machine learning for perception and language High performance computing and communications for AI, language and cognition Intelligent systems in dynamic and uncertain environment Integrated multimedia (visual and language) systems Large-scale knowledge bases KNOWLEDGE MODELS AND COGNITIVE SYSTEMS FISCAL YEAR 1992 RESEARCH PROJECTS TRADITIONAL APPROACHES TO ARTIFICIAL INTELLIGENCE IRI-9208920 Adrion, Richard; Stankovic, John; Riseman, Edward; Ramamritham, Krithi; and Grupen, Roderic University of Massachusetts, Amherst $50,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $435,000) Intelligent, Real-Time Complex Computing Systems This is the first year of a three-year continuing award. The effort focuses on coordinated research in robotics, vision, real- time AI, and real-time software systems, in the context of a robotic assembly testbed. The emphasis is on flexible manufacturing automation involving active perception, planning, and cooperative activities among agents in real time. The project investigates integration of dexterous manipulation and vision, with cooperation among static and mobile robots and humans, and addresses current limitations of robotic systems in dealing with uncertainties and rapidly changing environments and tasks (such as occur in short-run production). Activities include development of multiple resolution representations which permit arbitration between local reflexive and global combinatoric strategies; studying tradeoffs between solution quality and computational speed in real-time systems; modeling cooperation and communication to achieve goal-oriented behavior; integrating architectures and algorithms for extracting relevant environmental information for control of distributed robotic manipulators; implementation of active perception to support model-based, goal oriented sensing for manufacturing assembly operations; constructing high level symbolic approaches to reasoning about geometry; and implementing learning mechanisms which model the environment based on experience over a general class of tasks to guide perception, planning, and multi- agent cooperation. IRI-9249058 Ashley, Kevin University of Pittsburgh $65,000 - 12 mos. PYI: Symbolic Reasoning About Relevance This is the third year base and second year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-9058441. Intelligently assessing relevance is a central issue in AI. Whether in diagnosis, problem solving, or explanation, AI systems must determine the information in a database that is relevant to the task. Assessing relevance is difficult because it depends on context. This work focuses on reasoning symbolically about relevance of cases or features in context in a computationally tractable way. Integrating expanded domain theories, where such theories are often ill-defined or partial, into the assessment of relevance is a primary goal of the work. The ultimate goal of this approach is to design an intelligent tutoring system that will train students to argue with cases. IRI-9211662 Baral, Chitta University of Texas, El Paso $89,956 - 36 mos RIA: Research in Knowledge Representation and Common Sense Reasoning The aim of the proposed research is to answer several questions regarding adequately formalizing common sense reasoning. There will be three steps. First, to evaluate and classify some of the knowledge representation languages, in particular, to compare various semantics of logic programs and develop benchmark programs that distinguish them. Second, to consider features that are not handled by the traditional nonmonotonic formalism and suggest methods to handle them. The particular features considered are: defining the consistency of a set of default and strict rules, and being able to infer new defaults from a set of default and strict rules. Third, to develop a prototype system to answer queries with respect to more general knowledge representation language than can be handled by the systems developed so far, in particular, to follow the lead of systems based on the well-founded semantics of normal logic programs to develop a prototype system to answer queries with respect to the well-founded semantics of a default theory. The significance of the proposed research is that if successful it makes theoretical advances in formalizing common sense reasoning and provides a prototype implementation of query answering system for a more general formalism. IRI-9122401 Coombs, Michael New Mexico State University $80,000 - 24 mos. Dynamic Control for Problem Solving Tasks This project is concerned with dynamic strategies for the so called Model Generative Reasoning (MGR) architecture that will enable MGR application programs to cope with both data and knowledge limitations imposed by unstructured task environments with greater effectiveness, i.e., task environments in which problems are typically ill-posed, and data are typically sparse, noisy, and uncertain relevance. IRI-9245527 Cooper, Gregory University of Pittsburgh $55,374 - 12 mos. Learning Probabilistic Networks From Databases This is the second year funding of a two-year continuing award IRI- 9111590. This research investigates Bayesian methods for constructing probabilistic networks from databases. Its main focus is on constructing Bayesian-belief networks. Primary goals are to (1) develop methods for calculating the probability of a Bayesian belief-network structure given a database of cases, (2) identify the most probable belief-network structure given a database of cases, and (3) perform probabilistic inference by taking a weighted average over the inferences of multiple belief-networks. Potential applications include computer-assisted hypothesis testing, automated scientific discovery, and automated construction of probabilistic expert systems. Methods will be explored for integrating prior knowledge with data, handling missing data, and discovering hidden (latent) variables. Of particular concern is the development of computationally efficient algorithms. The methods developed will be empirically evaluated using databases from several domains. IRI-9244420 D'Ambrosio, Bruce Oregon State University $83,471 - 12 mos. Dynamic Uncertainty Management for Large Scale Problem Solving This is the second year funding of a three-year continuing award IRI-9100530. Current approaches to uncertainty management in AI and expert systems largely ignore the dynamics of large-scale problem solving. For example, much research has focused on the problem of computing the answer to a single query or set of queries, given a static problem model. However, advanced work in large scale problem solving, such as story and image understanding, reveals that such queries are interspersed in a larger stream of model specifications/reformulations and evidence assertions. This research has as its goal the understanding of the functional requirements for uncertainty management at this higher, task-oriented level, and the development of uncertainty management systems which provide efficient support for these requirements. IRI-9249856 Dechter, Rina University of California, Irvine $62,500 - 12 mos. PYI: Characterization of Tractable Sub-Problem in Automated Reasoning This is the second year base and first year matching of a five-year Presidential Young Investigator continuing award IRI-9157636. This work will expand on the understanding and exploitation of tractable models of reasoning. To explain how people perform so well on tasks that are theoretically intractable, we must assume that approximation methods, based on tractable models, cover a significant part of intelligent activities and, hence, should serve as the cornerstone of automated reasoning systems. This principle provides the motivation for this research. This work will: (1) establish a formal relational basis for connectionist models and neural networks and, using the language of constraint networks, quantify the powers and limitations of these models in terms of expressiveness and computational efficiency, (2) establish a formal basis of causal theories using the language of relational algebra and (directed) constraint networks, (3) identify tractable classes of logic programs, default knowledge bases and temporal reasoning sub-languages, recognizable via the topological features of their specifications, and (4) apply constraint network techniques to real life problems such as scheduling, planning and diagnosis. IRI-9210906 Elgot-Drapkin, Jennifer Arizona State University $30,000 - 12 mos. RIA: Tractable Non-Monotonic Inference: Focusing One's Reasoning Using Step-Logic This is the first year of a two-year continuing award. This research is directed at the investigation of formal mechanisms to focus the reasoning and handle contradictions in formal common sense reasoning framework. An important contribution of the research is the framework it will provide for truly real time common sense reasoning. The proposed research is directed at enhancing the step logic formalism to provide a formal model to real time common sense reasoning. Step logic is a formalism designed to model the ongoing process of deduction. What makes the step logic formalism so unique is that the logic explicitly allows unsound inferences (i.e. contradictions can arise). This is done in the interest of efficiency (that is, to avoid the consistency check problem). The objectives of the research are to investigate ways of focusing one's reasoning, to develop a means of focusing the reasoning of the step logic formalism, to investigate methods of handling contradictions efficiently, and to develop a more general method of handling contradictions within the steplogic formalism. The attainment of these objectives will result in a non-monotonic reasoning formalism that is not only tractable, but real time. The research issues will be addressed both in terms of theory as well as implementation. IRI-9207633 Freuder, Eugene C. University of New Hampshire, Durham $91,563 - 36 mos. (Jointly funded with the Information Technology and Organizations Program and the Cross Disciplinary Activities Office - Total Award $186,250) Constraint-Based Reasoning: Computation and Representation Constraint-based reasoning has been used in many areas of artificial intelligence: vision, language, planning, diagnosis, scheduling, configuration, design, temporal reasoning, defeasible reasoning, truth maintenance, qualitative physics, logic programming, and expert systems. This research focusses on the constraint satisfaction problem paradigm, which underlies many of these applications. The research objectives are: (1) characterizing tractable problem classes, (2) developing new algorithms, (3) addressing knowledge representation issues, (4) addressing knowledge acquisition issues, and (5) studying extensions of the basic constraints satisfaction problem paradigm. IRI-9240163 Gelernter, Herbert SUNY, Stonybrook $103,995 - 12 mos. Implementation of a Very Large Knowledge-Based Domain Specific Heuristic Problem Solving System This is the second year funding of a two-year continuing award IRI- 9107599. The Synchem system is a large knowledge-based domain specific heuristic problem solving program that is able to find valid synthesis routes for organic molecules of substantial interest and complexity without on-line guidance on the part of its user. Synchem requires an extensive knowledge base to make it routinely useful. However, it is very difficult to engage domain experts to the long-term dedication and intensity of commitment necessary to create a genuinely productive knowledge base. In order to deal with the debilitating knowledge-base bottleneck, machine learning programs will be devised that can use large computer-readable databases of specific reaction instances to provide training examples for algorithms designed to extract the underlying reaction schemata through inductive and deductive generalization. This approach will be augmented by the methodology that is usually described as explanation-based learning. Since the individual reaction entries in most databases are often haphazardly sorted and classified, another machine learning program will partition the input databases into coherent reaction classes using the methodology of conceptual clustering. To make it possible for Synchem to deal with the explosion of reasonable pathways that would be the result of a successful effort to greatly expand the knowledge base, the inference engine will be reformulated for distributed parallel search of the problem space. IRI-9244428 Gelfond, Michael University of Texas, El Paso $42,310 - 12 mos. Representing Properties of Actions In Extensions of Logic Programming This is the second year funding of a two-year continuing award IRI- 9103112. This is joint work with Dr. Vladimir Lifschitz, at the University of Texas, Austin. The goal of this research is to develop, study and implement formalism for representing properties of actions in declarative extensions of logic programming. The theoretical part of the project is based on the mathematical theory of nonmonotonic reasoning. The experimental part uses deductive database systems. The specific aims are (1) to develop methods for representing properties of actions using both negation as failure and classical negation, and to study their relation to the representations based on circumscription and other nonmonotonic formalism, (2) to implement these representations using deductive database systems, (3) to extend this work to the models of action that include continuous time and concurrence, and (4) to investigate the possibility of using abductive logic programming for the automation of reasoning about the past. IRI-9210925 Goel, Ashok Georgia Institute of Technology $30,000 - 12 mos. RIA: An Adaptive Approach to Qualitative Modeling in Design This is the first year of a two-year continuing award. An adaptive approach to qualitative modeling of physical devices will be investigated. The behaviors of a physical device, in this approach, are acquired by modifying and combining structure- behavior models of known devices and behavioral models of generic domain mechanisms. The structure-behavior model of a specific device explicitly specifies its structure, its functions, and the internal causal behaviors that show how its structure achieves its functions. The behavioral model of a generic domain mechanism specifies its function and model of a known device is revised by model-revision plans, where each plan accommodates a specific type of structural or functional difference between the new and the known devices. The process of model revision is focused by knowledge of the internal causal behaviors of the known device. The research will evaluate the adaptive approach to qualitative modeling for solving the task of design verification. The design- verification task is to predict whether a proposed design will deliver the functions desired of it. This research will contribute to qualitative modeling by providing a unified view of qualitative modeling, model memory, and model learning. IRI-9207262 Haddawy, Peter University of Wisconsin $40,594 - 12 mos. Decision-Theoretic and Symbolic Planning This is the first year funding of a three-year continuing award. As AI planning systems are applied to more realistic problems, the issue of choosing among alternative plans arises. The planner typically cannot prove that any alternative will or will not be successful, and so it must choose among plans on the basis of how likely they are to achieve their aims and how efficiently they do so. Decision theory supplies a normative model for making these choices, but is not directly applicable to the planning problem as AI has posed it, and more importantly does not provide a computational framework within which to make the choices. In particular, decision theory offers no advice on how to generate or modify plans. Decision theory and symbolic AI planning algorithms therefore offer complementary strengths: the former provides a rich representation for expressing choice among alternatives and a theory of how to make those choices rationally; the latter provides a computational method for generating and improving plans, but under restrictive assumptions. The two approaches will be integrated to produce a practical decision-theoretic planner. The research contains theoretical and computational aspects, the former oriented toward reconciling the decision-theoretic and AI problem representations, the latter oriented toward developing an algorithm that efficiently builds, evaluates, and chooses among plan alternatives. IRI-9206733 Hanks, Steven University of Washington $74,264 - 12 mos. Decision-Theoretic and Symbolic Planning This is the first year funding of a three-year continuing award. As AI planning systems are applied to more realistic problems, the issue of choosing among alternative plans arises. The planner typically cannot prove that any alternative will or will not be successful, and so it must choose among plans on the basis of how likely they are to achieve their aims and how efficiently they do so. Decision theory supplies a normative model for making these choices, but is not directly applicable to the planning problem as AI has posed it, and more importantly does not provide a computational framework within which to make the choices. In particular, decision theory offers no advice on how to generate or modify plans. Decision theory and symbolic AI planning algorithms therefore offer complementary strengths: the former provides a rich representation for expressing choice among alternatives and a theory of how to make those choices rationally; the latter provides a computational method for generating and improving plans, but under restrictive assumptions. The two approaches will be integrated to produce a practical decision-theoretic planner. The research contains theoretical and computational aspects, the former oriented toward reconciling the decision-theoretic and AI problem representations, the latter oriented toward developing an algorithm that efficiently builds, evaluates, and chooses among plan alternatives. IRI-9247040 Hanks, Steven University of Washington $4,000 - 12 mos. REU: Modeling a Dynamic and Uncertain World This is a REU supplemental award to IRI-9008670. This will support undergraduate participation in the ongoing research described. AI planners have traditionally made certainty and simplicity assumptions which have obscured two important distinctions: that the time at which a plan is executed differs from the time at which the plan is built, and that the planning agent's model of the world will differ from the real world. Relaxing these assumptions introduces the problem of maintaining a planning agent's world model, which will be both incomplete and dynamic. A framework for representing and maintaining such a model, has been developed which takes into account reports from external sources (e.g. sensors), the agent's proposed actions or plans, and external forces that may aid or confound those plans. Extensions to this framework will allow the agent to reason about (1) execution-time reports from sensors, thus allowing it to detect planning failures and learn from bad predictions, and (2) complex casual structures, thus allowing it to represent more realistic physical systems (like the ones studied in the qualitative-physics literature). IRI-9244377 Henschen, Lawrence J. Northwestern University $72,514 - 12 mos. Theorem Proving in Paraconsistent Logics This is the second year funding of a two-year continuing award IRI- 9015251. There is currently no way for knowledge based reasoning systems to handle inconsistent information, although inconsistent information arises in many important application such as medical diagnosis and other expert systems. Most often during the construction of a knowledge based system, if the knowledge engineer is presented with information, which may come from a number of sources, that is inconsistent, the knowledge engineer simply chooses, sometimes incorrectly, to discard certain pieces of information that have been identified as the cause of the inconsistency. The consequence of such a practice may be disastrous since first of all the engineer may be throwing away important information and secondly, in many instances, the existence of inconsistent knowledge may itself convey important information. However, since most reasoning systems are based on classical two valued logic, they have no way to correctly handle inconsistent information. A class of logics called annotated logics, has been proposed by several researchers as an alternative model or reasoning that allows for inconsistencies. These logics have been shown to provide nice, theoretical foundations to a variety of formalism in computer science. This work will study the use of annotated logics as a practical computational model. It will implement annotated logic systems that can serve as an environment in which knowledge base containing inconsistencies can be built and in which inferences can be computed. This work should provide a nice bridge across the gap that currently exists between the theory and the practical utility of annotated logics, and provide a reasoning basis for many important practical applications. IRI-9244233 Indurkhya, Bipin Boston University $57,361 - 12 mos. Modeling the "Redescription" Process in the Context Of Proportional Analogies This is the third year funding of a three-year continuing award IRI-9105806. It has been known for quite some time that metaphors and analogies can redescribe an object, thereby creating new perspectives on it. However, this phenomenon has not yet been properly addressed in cognitive science and AI research. Previous work has identified and characterized this process of redescription that underlies creative instances of metaphor and analogy. This work will implement a computational model of redescription in the proportional analogy domain. In this framework, an object is seen as an element of a finitely generated algebra, and a "description" of the object is defined to be any possible way in which that object can be generated from other objects by applying appropriate operators. Typically, an object can be generated in many ways and, consequently, has many descriptions. In designing the model, the problem then, is to find efficient algorithms and heuristics which select appropriate descriptions of the objects so that the analogy relation is comprehensible. IRI-9247017 Indurkhya, Bipin Boston University $4,000 - 12 mos. REU: Modeling the "Redescription" Process in the Context of Proportional Analogies This is an REU supplemental award to IRI-9105806. This will support undergraduate participation in the ongoing research described. See the above award for a description of the project. IRI-9210997 Kambhampati, Subbarao Arizona State University $90,000 - 36 mos. RIA: Exploring Fundamental Utility Tradeoffs in Plan Reuse Despite the attractive generality of domain independent planning techniques, their inefficiency has severely hampered the attempts to scale them to complex real world planning domains. One very promising solution to this involves enabling the planner to improve performance from experience by reusing previously generated plans to solve new planning problems. Developing effective plan reuse frameworks is thus currently a very active area of research in automated planning, machine learning and case based reasoning communities. Successful design of such planning systems requires a thorough understanding of the fundamental tradeoffs between storage, retrieval and modification in plan reuse. In the previous work, a domain independent planning framework called PRAIR which facilitates flexible modification of existing plans to solve new planning problems in the context of hierarchical least commitment planning has been developed. Experiments with PRAIR have demonstrated its potential to bring about order of magnitude improvements in planning performance by incrementally modifying existing plans in a variety of domains. PRAIR framework thus promises to be an ideal test bed for investigating the issues of utility of reuse. This research, a unified reuse framework, is proposed based on PRAIR modification framework, and is used to study the utility tradeoffs involved in plan reuse. This framework will also be used to study ways of integrating reuse with other speedup learning techniques such as abstraction and search control rules. The effectiveness of the resulting reuse framework will be demonstrated by experimenting with the classical planning domains and more complex real world applications such as process planning. Results from this research will help to design plan reuse frameworks with demonstrably favorable utility tradeoffs thereby facilitating scaling up of domain independent planning techniques of more complex domains. This research will thus have a significant impact on automated planning, machine learning, and case based reasoning communities. IRI-9242106 Kolodner, Janet L. Georgia Institute of Technology $76,161 - 12 mos. A Case-Based Approach to Creative Design This is the third year of a three-year continuing award IRI- 8921256. Design is a knowledge-intensive and experience-intensive task. For the most part designers create new designs by adapting a combined previous designs. While much research in the past few years has gone into investigating the processes involved in routine design, almost none has addressed issues related to creative or non-routine design. This work will investigate the processes involved in creative design and building experimental systems that do creative design in the domains of meal planning, architecture (office design), and mechanical design. This investigation will capitalize on previous work in design, and will look at the ways in which routine problem solving processes can be used or extended to create novel solutions to problems. It will also extend previous work on case-based reasoning, looking at how previous cases can be used in novel ways. The work will concentrate in four areas: design specification restructuring, exploration of design alternatives, merging of several design alternatives, and non- routine adaptation. The first two are necessary for what engineers call conceptual design, a design stage in which the design specification itself is defined. The last two look at how fixes or patches to partial designs can be done in novel ways. An understanding of design processes will enable the creation of the right kinds of design tools to help designers with their tasks and will help to build the kinds of tools and to develop the kinds of curricula that can best be used to train designers of the future. IRI-9119825 Korf, Richard E. University of California, Los Angeles $68,100 - 12 mos. Best-First Minimax Search This is the first year funding of a three-year continuing award. This research is concerned with a fundamental problem solving method - heuristic search, and an important heuristic search algorithm - the best-first search. A new best-first search algorithm is developed, whose memory requirement is only linear in the search depth, at the cost of expanding some nodes more than once. The algorithm runs faster than classical best-first search due to its simple structure and reduced overhead. It removes the memory limitation of best-first search, and opens up a host of new applications: combinatorial optimization problems, optimal decisions under real-time constraints, selective search algorithms for two-player games, and difficult constraints satisfaction problems. IRI-9244429 Lifschitz, Vladimir University of Texas, Austin $72,558 - 12 mos. Representing Properties of Actions in Extensions of Logic Programming This is the second year funding of a three-year continuing award IRI-9101078. This is joint work with Dr. Michael Gelfond at the University of Texas, El Paso. The goal of this research is to develop, study and implement formalism for representing properties of actions in declarative extensions of logic programming. The theoretical part of the project is based on the mathematical theory of nonmonotonic reasoning. The experimental part uses deductive database systems. The specific aims are (1) to develop methods for representing properties of actions using both negation as failure and classical negation, and to study their relation to the representations based on circumscription and other nonmonotonic formalism, (2) to implement these representations using deductive database systems, (3) to extend this work to the models of action that include continuous time and concurrence, and (4) to investigate the possibility of using abductive logic programming for the automation of reasoning about the past. IRI-9246630 Loui, Ronald P. Washington University $4,000 - 12 mos. REU: Applications and Investigation of Formalism for Resource- Bounded Argument This is a REU supplemental award to IRI-9008012. This will support undergraduate participation in the ongoing research described. This research is aimed at studying various aspects of resource- bounded defeasible reasoning and the applicability of the formalism. The scope of this investigation includes: 1. defeasible reasoning about resource-bounded decisions; 2. control of resource-bounded argument; 3. case-based defeasible reasoning in law and projection from small samples; 4. applications of existing statistical defeasible reasoning, specifically, to existing HyperText and Diabetes projects, and 5. mathematics and philosophy of defeasible reasoning. The research is also expected to have an impact on such related work as nonmonotonic reasoning and logic programming. IRI-9240459 McCartney, Robert University of Connecticut $25,817 - 12 mos. Case-Based Planning and Execution in a Time-Critical Environment This is the second year funding of a two-year continuing award IRI- 9110961. This project will design, implement, and evaluate a system that integrates planning, execution, and learning in real time. The four main objectives to this research are: (1) develop a usable execution monitor that will control plan execution in real time; (2) develop the methods to plan directly from episodes in memory, (3) automate the acquisition of cases from execution; and (4) develop techniques for learning using a weak domain theory. The system's domain is meal planning and preparation. The system will plan and assist in the preparation of meals, then incorporate the results of these into its case memory. A central concern here is to design a system that is efficient enough to operate in real time. This will be attained in part by placing computational restrictions on time-critical processes, and in part by using a case-based approach to plan generation and repair that use episodic representations of previous executions. The detail in the resultant plans and repairs simplifies both execution monitoring and case acquisition. Learning is done at noncritical times, and is done without a strong domain theory by attributing expectation failures to failures in the assumptions used to generate the expectations. IRI-9247018 McCartney, Robert University of Connecticut $4,000 - 12 mos. REU: Case-Based Planning and Execution In A Time-Critical Environment This is an REU supplemental award to IRI-9110961. This will support undergraduate participation in the ongoing research described. See award above, for a description of the project. IRI-9245325 Mostow, David J. and Ellman, Thomas P. Rutgers University, New Brunswick $68,652 - 12 mos. Idealization-Based Discovery of Search Heuristics This is the second year funding of a two-year continuing award IRI- 9017121. Many computational problems can be solved by searching through a well-defined but huge space of candidates provided good heuristics are available to guide the search. Thus automated discovery of search heuristics would have both practical and scientific importance. In this approach, the generation of heuristics is modelled as the application of successive transformations to an initial problem specification. This research addresses a number of issues raised by the long-term goal of scaling up to more realistic application domains. To refine the model, studies of perturbation experiments aimed at understanding when and why it works will be carried out. To enrich the space of generated heuristics, extensions and integration of transformations so as to rederive known scheduling and routing heuristics and the discovery of new ones will be undertaken. To explore this space more efficiently, novel methods for evaluating potential heuristics will be employed. Expected outcomes of this work include engineering advances in automated methods for discovering heuristics, as well as scientific advances in understanding their underlying principles. IRI-9200918 Pearl, Judea University of California, Los Angeles $100,000 - 12 mos. Probabilistic Networks for Automated Reasoning This is the first year funding of a three-year continuing award. This research is concerned with the extending the capabilities of probabilistic networks as the basic mechanism of representing and managing uncertainty in automated reasoning applications. The project consists of three parallel investigations: 1. formalization of relevance and causation using extensions to the theory of graphoids, 2. mechanization of qualitative non-monotonic reasoning based on sound probabilistic semantics, and 3. The automatic recovery of causal structures from data. IRI-9120310 Porter, Bruce W. and Maguire, Bassett J. University of Texas, Austin $10,048 - 12 mos. (Jointly funded with the Application of Advanced Technologies Program - Total Award $160,048) A Flexible Explanation Facility for Advisory Systems This is the first year funding of a three-year continuing award. This proposal is concerned with the development of a computer system for tutoring college-level biology. While tutoring systems have been built for biology before, new capabilities: flexible instruction, model simulation, and visualization, are proposed. Flexible instruction will permit dynamic generation of instructional material that is appropriate for each teaching situation. Simulations of models of biological systems will explain and predict such systems' behavior. Visualization will dynamically illustrate these systems. Synergistically combining these capabilities, this tutoring system will allow students to ask questions, request clarification or elaboration, and inquisitively experiment. IRI-9244217 Prieditis, Armand E. University of California, Davis $29,180 - 12 mos. Discovering Effective Admissible Heuristics by Abstraction: Developing a Quantitative Theory Relating Abstractness to Effectiveness This is the second year funding of a two-year continuing award IRI- 9109796. Admissible heuristics are an important class of heuristics worth discovering. They guarantee shortest path solutions in search algorithms such as A* and they guarantee less expensively produced solutions with a bounded increase in solution path length in search algorithms such as dynamic weighing. Several researchers have described how admissible heuristics can be generated from abstracted versions of a given problem, ones from which certain details have been removed. This work aims to develop a quantitative theory that relates abstractness to the effectiveness of the resulting heuristics and then empirically validate that theory. Such a theory will enable us to predict how much complexity reduction can be expected from using abstraction-derived heuristics. Ultimately, this theory will result in a better understanding of how effective admissible heuristics can be automatically discovered. IRI-9210030 Provan, Gregory University of Pennsylvania $100,000 - 36 mos. (Jointly funded with the Cross Disciplinary Activities - Total Award $104,998) MRI: Diagnostic Reasoning Using Sequential Decision Models Formalizing diagnostic reasoning is a necessary aspect of creating useful diagnostic tools. Even though much progress has been made, there are several deficiencies in the existing formalization. These include the definition of diagnostic reasoning as a single-stage process which is independent of the use of the diagnosis, of tests which can be done to clarify the hypothesized diagnoses, or of the utility associated with the treatment of the abnormalities. This research proposes a sequential decision framework for diagnostic reasoning which overcomes many of these deficiencies. A key new intuition for this approach is modeling diagnostic reasoning as a process which consists not solely as a single step of computing a diagnosis (as there is considerable controversy over what a diagnosis actually is, even given the same input data); instead, it is a multi-stage decision process in which decisions about testing and treatment, as well as scarce resource and utility considerations, are crucial parts of the model. This framework extends most existing formalization of diagnostic reasoning to incorporate important aspects of the diagnostic process. Two applications of the framework to diagnostic systems are investigated in this research. The first application (which has been implemented) is a Markov decision network model for the diagnosis of acute abdominal pain. The second application is a sequential process of consistency-based diagnosis interleaved with tests, for diagnosing Boolean circuits. IRI-9208429 Pu, Pearl University of Connecticut $29,433 - 24 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $59,433) RIA: An Efficient Case-based Assembly Sequence Generation System Automatic assembly sequence generation (ASG) is important for efficient manufacturing and concurrent engineering. There are two main difficulties with developing tools for ASG: (i) the combinatorics makes solutions by blind search intractable and (ii) criteria for optimal assembly sequences are difficult to formalize. Earlier work of the PI investigated the use of case-based reasoning techniques to simultaneously address both problems with encouraging results especially concerning the extensibility to covering large sets of problems. The practical objective of this project is to verify these results on a larger prototype, and thus demonstrate the usefulness of case-based assembly sequence generation for practical applications. The theoretical objectives of this project are to investigate new methods for adaptive learning and case matching which will form part of the new prototype. The result of this work is a fully implemented case-based reasoning system for a large class of assembly devices. The conceptual results involved in building such a system should prove useful to research in intelligent design and manufacturing systems and in case-based reasoning field. IRI-9120851 Rounds, William C. University of Michigan, Ann Arbor $28,820 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $57,639) Natural Language Techniques for Information Systems This is the first year funding of a three-year continuing award. This project is concerned with mathematical techniques from natural language processing systems, specifically from unification-based grammar formalism, to help in the design of data and knowledge bases. These techniques include the use of feature theory, feature logic, hyperset theory, and type systems from polymorphic functional programming languages. The goal of the project is to use these techniques to move knowledge and data models closer together, while providing for at least some aspects of heterogeneity in knowledge bases. IRI-9246627 Russell, Stuart J. University of California, Berkeley $4,000 - 12 mos. REU: Research on Real-Time Decision Making: The RALPH Project This is a REU supplemental award to IRI-8903146. This will support undergraduate participation in the ongoing research described. The objective of the proposed research is to understand the effect of limited computational resources on the design of optimal decision making systems, so as to be able to build agents that are robust in the face of complexity, and capable of acting in real time. The theoretical phase involves the completion of a normative theory for reasoning about computations. The approach is to treat computations as actions, with outcomes of varying probabilities and utilities using decision theory to choose between them. The theory will be incorporated in a multilevel decision theoretic architecture, which extends the idea of explicit metalevel reasoning to cover recursively all aspects of the decision making process. The approach has been applied successfully to game playing systems, and will be modified to apply to single agent problems such as robot motion planning. However, metalevel reasoning is often expensive, and the cost can only be amortized over future problem solving if a compilation mechanism is available. This work will therefore study methods for compilation of reasoning processes occurring anywhere in the hierarchy. It will extend explanation based learning techniques to cover the case of choosing best actions, where probabilistic time accuracy tradeoffs play a significant role. IRI-9241766 Schubert, Lenhart University of Rochester $68,031 - 12 mos. The Representation of Unreliable General Knowledge for Narrative Understanding This is the second year funding of a two-year continuing award IRI- 9013160. Narrative understanding requires the use of large amounts of generic knowledge about word meanings, the world, and narrative structure. Such knowledge is typically unreliable ("defeasible"), and raises difficult representational problems. The aim of this research is to develop a direct, natural, formally interpretable representation for this kind of knowledge, taking English formulations of the knowledge as an initial representation, and translating systematically (ultimately, automatically) into the formal representation. Certain classes of English generic sentences can already be mapped directly into the logic, but the research will greatly expand these classes, allowing virtually any generalization expressible in English to be directly formalized. The result will be not only theoretical insights into the nature of unreliable general knowledge, but also further development of an extensible narrative understanding system, whose general knowledge base can be expanded through English input. IRI-9244459 Sedelow, Sally Y.; Sedelow Walter A.; Talburt, John R.; and Kent, Robert E. University of Arkansas, Little Rock $101,145 - 12 mos. A Foundational Knowledge Representation for Knowledge-Based Systems This is the second year funding of a two-year continuing award IRI- 9114068. This research concentrates on the modelling, extension, and use of an associative knowledge base ranging over the English Language as a whole. Our existing topological model of English semantic space has been employed experimentally for information analysis, including the exclusively rule-based resolution of semantic ambiguity. Specific thrusts of this work include further research on the global structure of the knowledge base, especially as to implications of relaxing, and also increasing, the strength of the connectivity definition -- for on this definition depends the elaboration of the semantic distance metrics. Such semantic distance metrics, and other structural components, are implicit in the proposed efforts to 1. automatically add specialized terms (for narrow domain knowledge applications) to appropriate semantic sub-spaces; 2. explore the relationship between non-monotonic reasoning and semantic spaces in the knowledge base (e.g., how shifts among associative sub- nets relate to the rejection of "old" information and the addition of new information?); 3. explore the possibility that the associative memory structure of the knowledge base implicitly already contains information of the type currently explicitly and manually pre-encoded for knowledge bases (implicit availability would enable a system either to access such information automatically or to bypass it); 4. explore the relationship between the associative knowledge base and the generic human machine learning models incorporated in current natural computation, neural computing, and genetic computing; and 5. precisely establish uses for Rough sets as a multi-valued knowledge representation structure for English semantic space when there is a degree of uncertainty in the knowledge encoded. IRI-9241829 Shortliffe, Edward H. Stanford University $89,086 - 12 mos. Modeling Time in Belief Networks This is the second year funding of a two-year continuing award IRI- 9108385. This project addresses the problem of modeling time in dynamic domains given incomplete and uncertain information about the domain. The first objective is to construct a dynamic model within a belief-network paradigm and to demonstrate how well known time series concepts-such as backward smoothing forward filtering and forecasting - are implemented in this model. The dynamic model will be generated semiautomatically given a belief-network that models the time-invariant relations of the domain. This will provide a semiautomatic method for extending existing belief network models to dynamic belief-network models that can be used in applications where consideration of the time evolution of system variables is crucial to making valid inferences about the domain. The second objective is to design an efficient randomized approximation scheme (RAS) for probabilistic inference in belief- networks to be employed by the dynamic model. Certain features unique to a RAS, compared to other stochastic simulation algorithms for probabilistic inference, make the RAS desirable as an inference algorithm for a dynamic model. For example, in dynamic domains, the time required to make a decision enters the utility of the decision when this time becomes comparable to the expected time in which the system changes sufficiently to outdate a decision. A RAS provides an a priori bound on the running time required to achieve a predefined level of accuracy in the output. This information can be used to reduce the loss of utility due to delayed decisions. Existing RASs for probabilistic inference in belief-networks are known to have a poor worst-case behavior, although it is conjectured that they have efficient average-case complexity. This research will characterize the class of belief-networks or which existing RASs run efficiently, then will extend these algorithms to handle cases that fall outside of this class. The new RASs developed will subsequently optimize specifically for computing inferences in the dynamic models developed. IRI-9241766 Shortliffe, Edward H. Stanford University $85,277 - 12 mos. Dynamic Model Selection Under Time Constraints This is the second year funding of a three-year continuing award IRI-9241766. This research addresses the problems of modeling at the appropriate level of detail for the purpose of controlling physical or biological systems under time or computation resources constraints. The research is to develop, implement and test a method of dynamic selection of models for use in real-time applications. First the PIs develop and organize multiple simplifications of a detailed (base) model of the domain by applying combinations of domain-specific simplifying assumptions. They create an explicit representation of these assumptions and develop methods to reason dynamically with the assumptions to find the simplest model with valid assumptions. Then the PIs test the hypothesis that the simplest model with valid assumptions is optimal according to a decision-theoretic definition of optimality. The system is implemented in the domain of human cardiopulmonary physiology, with the control task of finding settings for a ventilator (mechanical breathing device). The research will provide improved methods of applying complex models in time-critical applications, enable the use of previously intractable models in real-time applications, and merge numeric simulation and control methodologies with formal decision theory. IRI-9244215 Subrahmanian, Venkatraman University of Maryland, College Park $29,271 - 12 mos. RIA: Reasoning About Inconsistency This is the second year funding of a two-year continuing award IRI- 9109755. First order theories (and hence databases, knowledge bases, and logic programs) could be inconsistent in many ways. For example, during the construction of an expert system, we may consult many different experts. Each expert may provide us with a group of rules and facts which are self-consistent. However, when we coalesce the facts and rules provided by these different experts, inconsistencies may arise. Thus, a framework for reasoning about databases that contain inconsistent information is necessary. This work will study at least three alternative theoretical frameworks for carrying out such forms of reasoning. Such a study includes rigorously characterizing the meaning of inconsistent theories from a non-classical model-theoretic stand-point. Fixed-point characterizations of the different logical consequence relations, together with procedures for computing answers to such queries also come under the scope of the study. A prototype implementation which serves as an experimental test-bed for evaluating computational aspects of these alternative semantical frameworks will be developed. IRI-9245554 Tong, Christopher and Ellman, Thomas P. Rutgers University $68,652 - 12 mos. Knowledge Compilation: A Divide-and-Conquer Approach This is the second year funding of a two-year continuing award IRI- 9245554. Knowledge compilation (KC) converts explicitly represented domain knowledge into an efficient algorithm for performing a task. KC technology could have a major impact in fields depending on efficient problem-solving algorithms. For example, productivity in circuit and mechanical design has grown increasingly dependent on efficient computer-aided design (CAD) tools. KC techniques can accelerate CAD tool development and improve CAD tool reliability. This work will develop knowledge compilation techniques for a variety of CAD applications. Initially the work will focus on spatial configuration problems. Such problems arise in many domains, including VLSI design, mechanical design, and architectural design. It will focus on synthesizing search algorithms composed of such components as generators, testers, and hill climbing patchers organized into one or more levels of abstraction. The research will develop a divide- and-conquer approach to KC: A problem specification is partitioned into parts; each part is assigned to be implemented as a particular type of algorithm component. This approach should avoid the control problems that arise in unrestricted transformational models of knowledge compilation. Expected benefits of the research include articulated principles for the synthesis of design algorithms, knowledge compilers implementing such principles, and new CAD algorithms resulting from their application. IRI-9241671 Truszczynski, Miroslaw and Wiktor, Marek University of Kentucky $77,000 - 12 mos. Nonmonotonic Logic Commonsense Reasoning and their Algorithmic Aspects This is the third year funding of a three-year continuing award IRI-9012902. This work studies non-monotonic logics and their application to common sense reasoning about knowledge and belief. It addresses both theoretical and algorithmic issues of non- monotonic modal logic. The results will increase understanding of non-monotonicity and will be applicable to non-modal formalism of non-monotonic reasoning such as default logic, logic programming, and truth maintenance systems. HIGHLY PARALLEL APPROACHES TO ARTIFICIAL INTELLIGENCE IRI-9212191 Anderson, Charles W. Colorado State University $29,461 - 12 mos. RIA: The Generality and Practicality of Reinforcement Learning for Automatic Control This is the first year funding of a two-year continuing award IRI- 9212191. Recent discoveries of the theoretical relationships between reinforcement learning and dynamic programming suggest exciting possibilities for developing automatic controllers that learn with experience to follow optimal control strategies. Combining reinforcement learning algorithms with the adaptive structure that neural networks provide results in theoretically optimal controllers that have more flexibility, and thus are more general, than current adaptive control techniques. However, for reinforcement learning networks to be practical, the efficiency with which they learn must be improved. In previous work, the PI identified one cause of slow learning to be difficulty of discovering useful features by the hidden units of the network. This difficulty has also been recognized within the supervised learning paradigm and a number of alternatives to the common error back propagation algorithm have been shown to significantly reduce learning time. These ideas will be extended to the reinforcement learning paradigm and their potential for reducing the learning time of reinforcement based networks will be explored. The objective is to alleviate the problem of training hidden units and to identify any remaining limitations of reinforcement learning networks that restrict their generality and practicality as real time control techniques. The methods will include both simulation studies and implementations as controllers of physical systems. ECS-9214866 Barto, Andrew G. and Ydstie, B. Erik University of Massachusetts, Amherst $0 - 12 mos. (Jointly funded with the Division of Electrical and Communications Systems and the Robotics and Machine Intelligence Program - Total Award $51,155) Reinforcement Learning Algorithms Based on Dynamic Programming This is the first year of a five-year continuing award ECS-9214866. This project will investigate aspects of a class of reinforcement learning algorithms based on dynamic programming (DP). Although these algorithms have been widely studied and have been experimented with in many applications, their theory is not developed enough to permit a clear understanding of the classes of problems for which they may be the methods of choice, or to guide their application. Research at the University of Massachusetts has made considerable recent progress in relating these methods to the most closely related conventional methods and in understanding the factors that influence their performance, both successful and unsuccessful. These methods may provide the only computationally feasible approaches to very large and analytically intractable sequential decision problems. The objective of this project are 1) to continue development of DP-based reinforcement learning methods and their theory, 2) to investigate their computational complexity, and 3) to define the characteristics of problems for which they are best suited. IRI-9210327 Boser, Bernhard University of California, Berkeley $30,000 - 12 mos. RIA: A Pattern Classification Method with Optimized Generalized Performance This is the first year funding of a two-year continuing award IRI- 9210327. The proposed research is directed towards finding classification algorithms that are trained from examples and achieve good generalization performance. A new training procedure is investigated that adjusts the complexity of the classifier to the problem at hand in order to optimize the tradeoff between perfect memorization and good generalization. The new algorithm applies to large variety of classification functions, including polynomial classifiers, two layer neural networks, or radial basis functions. Initial experiments on optical character recognition problems demonstrate both its performance relative to other learning algorithms, and its efficiency regarding time and memory requirements. Topics for investigation include the estimation of the VC dimension and generalization performance on a wide variety of problems in order to obtain a robust algorithm that performs well even when no expert knowledge in the training procedure to further improve performance will be considered. The ability of the algorithm to find atypical, misclassified, or meaningless pattern will be investigated to automatically clean databases. IRI-9247335 Carpenter, Gail A. Boston University $40,101 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $80,203) Analysis of Neural Networks for Adaptive Pattern Recognition This is the second year funding of a two-year Creativity Extension to award IRI-9000530. The PI has made creative accomplishments in the field of artificial neural systems - relating to the development of the Adaptive Resonance Theory (ART) and a resulting series of neural networks for computationally-efficient and/or fast real- time learning and pattern recognition. These networks are derived from behavioral and neural analysis of human and mammalian learning, and a math/computationally based approach to rigorous implementation of analogous learning methods in the artificial networks. These results have gained wide recognition and have quickly found their way into interesting applications, including a very larger CAD-based group technology program at the Boeing Company that is expected to streamline the design process by reducing parts inventory. Other application include pattern recognition in seismic, medical, military (target recognition), and robot navigation applications at Lincoln Labs; and a number of machine learning/expert system applications, including application to development of the Medicare Uniform Clinical Data Set at the University of Nevada School of Medicine. A new approach to supervised learning, involving combination of self-organizing modules, has recently been developed, and shows promise of performance exceeding that of the backpropogation and genetic algorithms. Other work is exploring coding and recognition of temporal sequences of events, with applications to speech perception, motion planning, and object recognition, and a new exploration of noise-resistant invariant filtering for rapid image recognition. In the extended period, research will focus on enhancing the ability of ART networks to learn new patterns rapidly while preserving the recognition capabilities for previously-learned patterns. This will include methods and structures to combine information from multiple sources, multi-scale (coarse-fine) recognition, hierarchical architectures that reduce overall connectivity requirements, and robustness to noisy data. The project will also explore local-logic implementation of fuzzy ART algorithms, and study implementations of medium-term memory, as well as continuing pursuit of temporal processing and rapid invariant filtering. IRI-9203532 Cottrell, Garrison and White, Halbert University of California, San Diego $73,235 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $79,848) Active Selection of Training Examples for Network Learning This is the first year funding of a three-year continuing award. This project is concerned with techniques for active selection of training examples for neural network learning, while simultaneously growing the network to fit the data. The approach uses a statistical sampling criterion, Integrated Mean Squared Error, to derive a "greedy" selection criterion which picks the next training example that maximizes the decrement in this measure. This selection criterion is usable for a wide class estimators. A practical realization of this schemes for multi- layer neural networks is demonstrated. DBS-9209432 Elman, Jeffrey L. University of California, San Diego $13,493 - 12 mos. (Jointly funded with the Division of Behavioral and Cognitive Sciences - Total Award $53,973) Implicit Learning of Sequential Inputs: Developing a Computational Model This is the first year of a three-year continuing award. Many of the most important things people learn are learned without explicit instruction; often, people are not even aware that they are learning something. One of the most dramatic example of this phenomenon is language. Virtually all speakers of a language learn to speak and understand before formal instruction even begins, and the learning process seems to be largely unaffected by attempts of adults to guide it. This style of learning contrasts with the more explicit learning which also occurs when people consciously attempt to master a domain, which typically involves explicit formulation and testing of hypotheses. Although considerable research has been carried out on explicit learning, implicit learning has been investigated only recently. Furthermore, most of this research has been experimental, with relatively little work focussed on developing theories of computational models. The purpose of this research is to develop both a theory and a testable computational model of implicit learning of sequential behaviors. The research will use experimental techniques to ask the following questions: Under what conditions is implicit learning triggered? Are there domains in which implicit learning is more effective than explicit learning? What are the constraints on the types of thing which can be learned with implicit learning? A second component of the research will focus on trying to understand the possible mechanism for implicit learning by developing a computational model. This model, based on an artificial neural network architecture proposed by Elman, has been shown to have interesting properties which resemble the behaviors of humans engaged in implicit learning tasks. Elaboration of this model should allow for better understanding of the circumstances which facilitate learning of domains such as language and might make it possible to structure training environments in order to maximize learning. This work should also provide a foundation for the construction of machine- based systems for learning domains currently only well-mastered by humans. IRI-9248391 Forrest, Stephanie University of New Mexico $62,500 - 12 mos. PYI: Computational Systems Based Upon Aspects of the Immune System This is the second year base and first year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-9157644. On-line intelligent systems (natural or artificial) must respond to real-time constraints, synchronize themselves with their environments, and be capable of adapting their behavior continuously to dynamic environments. Furthermore, adaptive mechanisms must play a central role at many levels if we are to utilize large-scale and highly complex computing environments effectively. These statements provide the basis for this research. Self-adjusting data structures are a simple example of how adaptive principles can be exploited within a traditional computing framework. A more radical approach uses insights from nonlinear dynamics to design systems in which many low-level interacting adaptive units produce a global dynamics that computes some useful function. This latter approach is refer to in this effort as "emergent computation." Classifier systems and other computational ecologies provide evidence that such systems are feasible. The immune system provides an excellent model of adaption operating at the local level and of useful behavior (e.g. recognizing antigens) emerging at the global level and will be used as the domain of analysis for this work. IRI-9204655 Geller, James New Jersey Institute of Technology $29,521 - 12 mos. RUI: Efficient Reasoning With Massive Parallelism and Hybrid Techniques This project is concerned with building a fast and theoretically well founded reasoner - a massively parallel transitivity-tree reasoner - for general AI. AI reasoning algorithms are often intractable, that is they are too slow for any practical problem sizes to be of real value. A widely used approach to overcome these problems has been to create special purpose reasoners. Such reasoners solve only a very limited set of problems, and are too restricted in architecture. This research is aimed at designing a reasoner that is more general than current special purpose reasoners and faster than existing general reasoners. This reasoner will have a well defined interface to a general purpose reasoner. IRI-9123692 Haussler, David and Warmuth, Manfred University of California, Santa Cruz $99,973 - 12 mos. Quantitative Analysis of Learning Algorithms This is the first year funding of a three-year continuing award. This project is concerned with developing a theoretical foundation for learning in neural networks and artificial intelligence in the direction of statistical decision theory and analysis of algorithms. The proposed work includes several aspects. First, further extensions and applications of the statistical techniques of Vapnik and Chervonenkis are pursued. Second, a coherent Bayesian methodology is developed for learning that includes the Valnik-Chervonenkis theory and the recent statistical physics approached to learning. Third, weighted majority and other on-line learning strategies are developed, including extensions to non- stationary learning environments. Finally, analysis and application of unsupervised learning and feature discovery techniques are performed. IRI-9244479 Levinson, Robert University of California, Santa Cruz $47,461 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $94,922) Adaptive Pattern-Oriented Chess This is the second year funding of a two-year continuing award IRI- 9112862. Although chess computers now are competitive at/ master and grandmaster levels, that is where their resemblance to human players ends. Psychological evidence indicates that human chess players search very few positions, and base their positional assessments on structural/perceptual patterns learned through experience. Morph is a computer chess program that has been developed to be more consistent with the cognitive models. The learning mechanism combines weight-updating, genetic algorithms, and explanation-based and temporal-difference learning to create, delete, generalize and evaluate graph patterns. An associative pattern retrieval system organizes the database for efficient processing. The main objectives of the project are to demonstrate capacity of the system to learn, to deepen our understanding of the interaction of knowledge and search, and to build bridges in this area between AI and cognitive science. To strengthen the connections with the cognitive literature the system's knowledge is to come from its own playing experience, no sets of pre-classified examples are given and beyond its chess pattern representation scheme little chess knowledge such as the fact that having pieces is valuable (let alone their values) has been provided to the system. Further, the system is limited to using only one ply of search. CCR-9115603 McAloon, Ken and Tretkoff, Carol CUNY Brooklyn College $25,000 - 24 mos. (Jointly funded with the Division of Computer and Computation Research - Total Award $174,991) Highly Parallel Constraint Logic Programming for AI and MIP Applications This research has a three-fold goal: (1) to develop tools for building intelligent systems that require tightly coupled logical and numerical computation; (2) to apply mathematical programming and constrained optimization techniques to do logical inference and, (3) to test the efficacy of highly parallel constraint logic programming systems in this arena. Constraint logic programming (CLP) provides a natural paradigm to link logical computation and numerical computation. This research is based on compact CLP languages which are supersets of linear and integer programming and support the data structures necessary for mathematical programming rather than on constraint based extensions to Prolog. The fit between the applications and constraint logic programming opens up several avenues for parallelism, since logic programming systems have direct hooks into both or-parallelism and and-parallelism. IRI-9244551 Mozer, Michael C. University of Colorado, Boulder $37,500 - 12 mos. PYI: Study of Attention in Perception and Cognition within the Connectionist Paradigm This is the second year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-9058450. The connectionist paradigm provides the initial assumptions which form the basis of this work on characterizing mechanisms of the mind in terms of large networks of autonomous processing elements. Specific aspects of this work include models of human cognition, connectionist AI, and connectionist learning algorithms in an interleaved fashion. This work focuses specifically on attention which will be studied in perception and also in higher levels of cognition. IRI-9249516 Mozer, Michael C. University of Colorado, Boulder $25,000 - 12 mos. PYI: Study of Attention in Perception and Cognition within the Connectionist Paradigm This is the third year base funding of a five-year Presidential Young Investigator continuing award IRI-9058450. See award above for a description of the project. IRI-9242193 Touretzky, David and Thomason, Richmond Carnegie Mellon University $90,588 - 12 mos. Inheritance Theory and Knowledge Bases This is the third year funding of a three-year continuing award IRI-9003165. Theories of defeasible reasoning previously developed and supported will be further investigated. Expansion will include networks with expressive enhancements: nets with limited cyclicity, Boolean combinations of nodes in nets, study of arguments which interferes with a given argument by disabling inferences leading to a particular conclusion, hydra inheritance, higher order links, semantic foundations for defeasible inheritance, and additional work on roles and relations. This is a basic research on reasoning focusing on inheritance and having significance in non-monotonic methods. IRI-9211419 Wang, DeLiang Ohio State University Research Foundation $30,000 - 12 mos. RIA: Segmentation and Recognition of Complex Temporal Patterns This is the first year funding of a two-year continuing award IRI- 9211419. Temporal information processing underlies various kinds of intelligent behaviors, including hearing and vision. A neural network framework for segmenting and recognizing complex temporal patterns is proposed. Processing of temporal segmentation is based on the idea that segmentation is expressed by synchronization within each segment and desychronization among different segments. Each segment becomes an input to the recognition network that explicitly encodes neighborhood or topological relations of local features of the input, and recognition is based on the graph matching method. To cope with problems embedded in time, the network to be constructed be codes time explicitly. Multiple temporal patterns are segregated into different segments that are activated alternately in the time domain. The network is able to recognize complex temporal patterns, and recognition is invariant to distortions of time intervals (time warping) and to changes in the rate of presentation. The network will be tested for both neural plausibility and computational effectiveness. Results of this project will provide new computational principles that might be used by the brain to process temporal segmentation and recognition. Also, they will provide effective methods for solving technical problems indispensable in real-time continuous auditor pattern recognition. NATURAL LANGUAGE PROCESSING IRI-9244338 Ahlswede, Thomas Central Michigan University $22,086 - 12 mos. Evaluating Automatic Resolution of Ambiguity in Text This is the second year funding of a two-year continuing award IRI- 9110364. A necessary part of language understanding by people or computers is to determine the meaning of ambiguous words. Efforts at automatic disambiguation of text have so far been evaluated by ad hoc, manual checking of their results. As more disambiguation systems are developed and put to use, a rigorous method of evaluation becomes necessary. This work will study text disambiguation by human informants, to aid in developing both a formal model of the process and a metric or correctness for automatic disambiguation. The study will include the selection of representative texts and development from these of a disambiguation test to be administered to a large sample (over 100) of informants, from university students to professional lexicographers. Analysis of the test results will provide insight into the human disambiguation process that will contribute to the evaluation and refinement of current models of disambiguation, both human and computation, and eventually to the development of new models. The test results themselves will form a statistical profile of the disambiguation of English text, as a benchmark for efforts at automatic disambiguation. IRI-9246629 Allen, James F. and Carlson, Gregory University of Rochester $4,000 - 12 mos. REU: Plan-Based Models of Discourse This is a REU supplemental award to IRI-9003841. This will support undergraduate support participation in the ongoing research described. This project aims to develop a formal and implementable theory of the communicative interactions that can occur within a man-machine dialogue. It will develop a plan- based model of the interaction - explicitly representing the differing beliefs of the system and the user, as well as their mutually shared beliefs arising from the dialogue, and representing the sentences in the dialogue explicitly as actions both as speech acts reflecting immediate intentions, and as discourse acts, signalling the structure of the dialogue. The goal is to develop a domain independent model of dialogue interaction, focusing on those aspects that allow the system and user to monitor how well the dialogue is understood. In particular, the intent is to account for the prevalent acknowledgement behavior in dialogue, and to provide a general account of clarifications and correction subdialogs. The work also will examine how the planning model affects the actual interpretation of the sentences in the dialogue, concentrating on the effects the model has on speech act interpretation, and the interpretation of tense and aspect, and the generic/specific distinction. As a result, the plan-based approach to language will be brought into much closer correspondence to the actual linguistic behavior apparent in natural dialogues. IRI-9210608 Batali, John University of California, San Diego $28,850 - 12 mos. RIA: Analysis of Discourse Events This is the first year funding of a two-year continuing award IRI- 9210608. The development of a formal model relating the syntactic, semantic, and discourse-level analyses of texts will be investigated. This model will be organized around the notion of "discourse events" which involve the creation, modification, or manipulation of the propositional structure contributed by sentence constituents. Along with the development of the formal models, computational support for the analysis to texts in terms of the model will be implemented. With such support, a large collection of analyzed texts and utterances will be accumulated. The model will also serve as the specification for the design and implementation of integrated natural language understanding programs. The primary methods to be employed will include unification-based grammar formalisms and logic programming techniques. The theoretical background for the approach will be based on the analysis of "speech-acts" by Searle, the "projection- semantics" of Halvorsen and Kaplan, and the lexical functional" approach to grammar developed by Kaplan and Bresnan, as well as the techniques for implementing all of these, developed by the PI. It is important to understand the contributions that all linguistic levels make to the workings of texts. The discourse events model is meant to serve as a way to organize the analysis of those contributions. IRI-9246628 Boggess, Lois and Hodges, Julia Mississippi State University $4,000 - 12 mos. REU: Automatic Extension of a Knowledge Base Through Natural Language Text Analysis This is a REU supplemental award to IRI-9002135. This will support undergraduate participation in the ongoing research described. This research will develop a flexible knowledge representation medium and sufficient natural language processing tools to allow the automated extension of an expert system's knowledge base through access to machine-readable natural language texts such as dictionaries, thesaurus, and technical references. The work on an early prototype has used the Merck manual as a source [Merck 1977]. Initial work will include the building of a parsing system consisting of a core grammar and the means to expand the core grammar semi-automatically. Using this parsing system, the effort will develop the incremental augmentation of the knowledge system directly from the analysis of the text. The ability of an expert system to extend its own knowledge base by extracting information from machine readable text is valuable for a wide variety of systems, including expert systems for large technical applications in a domain in which the information is continually being updated. Other projects rely on human intervention during the process that extracts new knowledge [Lenat et al. 1986; Humphrey 1989; Antonacci et al. 1989]. This makes dealing with problems such as contradictory information much easier. In contrast, here the feasibility of automating the actual extraction of information, in a disciplined way, from a structured technical reference will be examined. IRI-9122026 Carberry, Mary Sandra University of Delaware $177,113 - 36 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $179,916) An Incremental Tripartite Model of Consultation Dialogues and Negotiation Subdialogues This project is concerned with the modeling of cooperative expert consultation dialogues within a plan-based framework, with emphasis on the recognition of problem solving and communicative intentions and the handling of negotiation subdialogues. The research will develop a rich tripartite dialogue model that differentiates among domain, problem solving, and communicative intentions yet captures the relationships among them; beliefs and intentions will be explicitly ascribed to the participants and each kind of intention will be recognized in increments as the dialogue progresses. IRI-9120788 Dorr, Bonnie University of Maryland, College Park $51,830 - 24 mos. Interlingual Machine Translation and the Lexicon This research is concerned with the applicability of a lexical- based framework to the problem of interlingual machine translation. There are two tasks relevant to this goal. The first task is the augmentation of an existing lexical-semantic representation to include temporal, aspectual, and spatial information, all of which are necessary for adequate machine translation. The second task is the construction of routines for automatic acquisition of lexical entries based on this representation. In general, the project aims to test hypotheses that support the view that a lexical-based parametric framework can be built to accommodate interlingual machine translation, and provide an adequate basis for capturing temporal, aspectual, and spatial information. IRI-9241998 Grosz, Barbara J. and Gordon, Peter C. Harvard University $39,581 - 12 mos. (Jointly funded with the Language, Cognition and the Social Behavior Program - Total Award $79,162) Centering of Attention, Pronominal Reference, and Discourse Coherence This is the third year funding of a three-year continuing award IRI-9009018. This work addresses fundamental problems in three facets of discourse processing, (1) the focus of attention of discourse participants, (2) the use and understanding of pronominal reference, and (3) the coherence of discourse. The research focuses on the interaction among these facets within a discourse segment. It will refine and expand current computational theories of discourse coherence and reference at this local level. The approach is interdisciplinary: it includes the development of algorithms that embody the theory and psychological investigations to determine constraints on parameters of the theory (e.g. the ordering of candidate referents for pronouns). It will empirically test the claims of the theory as a model of human language use, and will compare the performance of algorithms for handling pronouns employing the constraints of this theory with alternatives based on other theories. The results will contribute to computer science and the cognitive sciences in three major arenas: the construction of more fluent and robust natural-language processing systems; the development of more computational sufficient psychological theories of discourse processing; and, potentially in the longer term, the design of more sophisticated computer languages (e.g. database-access languages, system-interface languages). IRI-9244235 Hamburger, Henry J. George Mason University $89,685 - 12 mos. Foreign Language Tutoring and Learning Environment This is the second year funding of a three-year continuing award IRI-9020711. In this collaboration, investigators from GMU and MIT will pursue a computational tutoring and learning environment for foreign language learning. GMU will enhance its naturalistic, immersion- style foreign language experience. Based on the premise that language is most thoroughly and memorably learned in the course of meaningful use, this system integrates dialogue and rich graphical interaction in the context of everyday scenes. The MIT group has over four years built extensive theory-based natural language processing software with language independent semantics and multi-level error handling especially designed for tutorial purposes. The collaboration will have important research benefits in both directions. The MIT language software will greatly enhance the GMU immersion system and permit a proper study of its efficacy. The error diagnosis capabilities will permit an investigation of student modeling, with potential long-term consequences for the study of both intelligent tutoring systems and psycholinguistics. IRI-9108363 Ide, Nancy M. and Veronis, Jean Vassar College $47,639 - 12 mos. RUI: English/French Computational Lexicography This is the second year funding of a three-year continuing award IRI-9108363. This award supports a project in computational lexicography which is an international collaborative effort of Vassar's Computer Science Department and the Groupe Representation et Traitement des Connaissances (GRTC) of the Centre National de la Recherche Scientifique (CNRS) in Marseille, France. The project involves the construction of a large-scale lexical and textual database from machine readable materials in English and French, with the goal of extracting information that can be used for automatic language processing tasks. A number of machine-readable materials in English and French, including dictionaries in typesetter's format but also corpora and other linguistic data, have been acquired by the project and installed at each site. Software tools are being developed to structure, access, and analyze these data. The project involves the structuring of machine readable dictionaries and other linguistic data into a consistent lexical database, and the extraction of lexical knowledge from these data for use in natural language understanding and translation. IRI-9247041 Ide, Nancy and Veronis, Jean Vassar College $4,000 - 12 mos. REU: English/French Computational Lexicography This is an REU supplemental award to IRI-9108363. This will support undergraduate participation in the ongoing research described. See the above award for project description. IRI-9244213 Joshi, Aravind University of Pennsylvania $71,515 - 12 mos. Research in Natural Language Processing: Mathematical and Computational Investigations in Constrained Grammatical Formalism This is the second year funding of a two-year continuing award IRI- 9016592. This is a joint award with Dr. Vijayshanker at the University of Delaware. This team has proposed several major research tasks in natural language processing with special emphasis on several mathematical and computational aspects. The work clearly has a formal and mathematical character. The various computational structures and strategies the team has developed and will develop, need to be investigated mathematically because such investigations shed light on the descriptive and processing powers of these formalism. These two aspects, i.e., the development of the structures and strategies, and their mathematical investigations, are very much interrelated. It is believed that natural language processing backed up by a formal framework, and mathematical investigations grounded in empirical studies are two very productive areas of research. The researcher will focus on mathematical investigations only to the extent to which the results have important implications for natural language processing. IRI-9201987 Kwasny, Stan C. Washington University $63,905 - 12 mos. Natural Language Processing and Connectionism This is the first year funding of a three-year continuing Zaward. This project is concerned with the following questions: (1) investigate hybrid symbolic/subsymbolic architectures for systems that process natural language, (2) develop and study representational schemes that support both traditional structures as well as ambiguity within those structures, (3) directly challenge the determinism hypothesis of M. Marcus in creating recurrent networks that require information to be extracted from the sentence while processing left-to-right, and (4) determine the generalization capabilities of various deterministic, connectionist processing systems. Incorporating these ideas, a family of demonstration systems will be developed. IRI-9240501 Martin, James H. University of Colorado, Boulder $29,438 - 12 mos. The MetaBank: A Knowledge-Base of Metaphoric Language Conventions This is the second year funding of a two-year continuing award 9109859. The frequent and conventional use of non-literal language has been a major stumbling block for natural languages processing systems since the early machine translation efforts. Metaphor, metonymy and indirect speech acts are among the most troublesome phenomena. Recent computational efforts addressing these problems have taken an approach that emphasizes the use of systematic knowledge about non-literal language convention. This work will construct the MetaBank,: an empirically derived and theoretically motivated knowledge-base of English metaphorical conventions. More generally, this effort can be seen as an attempt to develop methodologies for empirically capturing language conventions in usable knowledge-base forms. IRI-9244221 Moore, Johanna University of Pittsburgh $29,208 - 12 mos. Contextual Issues In Tutorial Dialogues: A Study of the Effects on Previous Explanations This is the second year funding of a two-year continuing award IRI- 9113041. This work investigates the effects of discourse context on explanations in tutorial settings. When human tutors engage in dialogue, they freely exploit all aspects of the mutually known context, including the previous discourse. Utterances that do not draw on previous discourse seem awkward, unnatural, or even incoherent. Previous discourse must be taken into account in order to relate new information effectively to recently conveyed material and to avoid repeating old material that would distract the student from what is new. Strategies for using the discourse history in generating utterances are of great importance to research in natural language generation for tutorial applications. The goal of this work is to produce a computational model of how tutors exploit the discourse history in instructional settings, and to implement this model in an intelligent tutoring system that maintains a dialogue history and uses it in planning its explanations. IRI-9113064 Passonneau, Rebecca and McKeown, Kathleen R. Columbia University $84,512 - 24 mos. Constraints on Reference and Segmentation in Conversational Discourse This project is concerned with two interdependent problems in discourse processing. The first is to develop an explicit declarative representation of referring relations to be used in the generation and understanding of discourse anaphors. The goal is to account for the mutually constraining effects of linguistic form and the dynamically updated discourse model. The second is to determine how to identify the distinct discourse actions that comprise a discourse, and how to relate these actions in a dynamically constructed hierarchical model of intentional structure. IRI-9249061 Shieber, Stuart N. Harvard University $100,000 - 12 mos. PYI: Use of Constraints in Natural Language Processing Systems This is the second year base and first and second year matching of a five-year Presidential Young Investigator continuing award IRI- 9157996. The focus of this work is to increase the efficacy of communication with computer through natural language and other media. In particular, three central problems confront researchers: robustness, fluency, and modularity. These issues can be addressed through the use of systems of constraints as the underlying method for encoding the structures of natural language. The results of investigations of these issues will be tested through the implementation of prototype systems. IRI-9249517 Steedman, Mark J. University of Pennsylvania $37,110 - 12 mos. (Jointly funded with the Interactive Systems Program - Total Award $74,220) Computer Synthesis of Contextually Appropriate Intonation for Spoken Language This is the second year funding of a two-year continuing award IRI- 9018513. The structure imposed upon spoken sentences by intonation seems frequently to be orthogonal to their traditional surface- syntactic structure. The involvement of two apparently uncoupled levels of structure in Natural Language grammar appears to complicate the path from speech to interpretation unreasonably, and to thereby threaten a number of computational applications in speech recognition and speech synthesis. This work will complete the specification of a generative theory of intonation in relation to syntax and discourse function, and instantiate the theory by constructing a "discourse-model driven" utterance generator and speech synthesizer. Such a device is to be contrasted with "text- to-speech" generation. The input to the system will be a representation of discourse context characteristic of a simple natural language query system. On the basis of such representations, the system will be capable of constructing a variety of contextually appropriate intonation contours for any given sentence covered by its grammar. Such a device is expected to considerably improve comprehensibility and naturalness of the synthesized speech. IRI-9242204 Swartout, William; Bateman, John and Paris, Cecile University of Southern California $76,388 - 12 mos. Phrasing This is the third year of funding of a three-year continuing award IRI-9003087. When humans use language, they show an essential responsiveness to their hearers. When language is automatically generated, it is similarly necessary to ensure that language is appropriate for its intended audience. Much previous research on computational linguistics has focused on building user models and selecting information from a knowledge base to present to the user. It is important, however, that the phrasing of a text be also tailored to the hearer - otherwise it may be just as ineffective as texts which rely on knowledge the hearer does not have. The realization that situations systematically affect language can greatly help the tailoring process: given appropriate classifications of situation-types, the consequences of these situations on language can be specified, and suitable language can be produced. In a pilot study that involved generating explanations for an expert system tailored to three user types, previous work has shown that it is possible to specify the type of language required in a given situation and to build a text planning system that uses that specification to control phrasing of its text. This research will extend that work. The goal is to create a framework within which it is possible to gain systematic control over phrasing by designing methodologies and mechanisms by which the systematic relationship between situation and language can be found, represented, and used to tailor generated text. IRI-9247415 Tomita, Masaru Carnegie Mellon University $37,500 - 12 mos. PYI: Multilingual Natural Language Processing This is the fourth year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-8858085. This research centers on computer systems for natural language processing, especially multi-lingual translation systems and real time natural language processing systems. The goals include the development of knowledge-based translation systems for English, Japanese, French, and German. The significance of this research is that automatic processing of natural language is a vital component in computer systems interface for the future. Multilingual translation systems have fundamental theoretical and economic importance. IRI-9249227 Tomita, Masaru Carnegie Mellon University $25,000 - 12 mos. PYI: Multilingual Natural Language Processing This is the fifth year base funding of a five-year Presidential Young Investigator continuing award IRI-8858085. See above award for description of the project. IRI-9249057 Vijayashanker, K. University of Delaware $47,579 - 12 mos. Research in Natural Language Processing: Mathematical and Computational Investigations in Constrained Formalism This is the second funding of a two-year continuing award IRI- 9016591. This is a joint award with Dr. Joshi at the University of Pennsylvania. This team has proposed several major research tasks in natural language processing with special emphasis on several mathematical and computational aspects. The work clearly has a formal and mathematical character. The various computational structures and strategies the team has developed and will develop, need to be investigated mathematically because such investigations shed light on the descriptive and processing powers of these formalism. These two aspects, i.e., the development of the structures and strategies, and their mathematical investigations, are very much interrelated. It is believed that natural language processing backed up by a formal framework, and mathematical investigations grounded in empirical studies are two very productive areas of research. The researcher will focus on mathematical investigations only to the extent to which the results have important implications for natural language processing. IRI-9123336 Wilensky, Robert University of California, Berkeley $50,000 - 12 mos. A Discourse-base Theory of Inference for Natural Language Processing This is the first year funding of a three-year continuing award. This project is concerned with developing a model of natural language understanding in which the inference machinery is based on discourse principles. In particular, a theory of text interpretation is developed. This theory computes an interpretation that conforms to plausible goals of a speaker and is most likely to be intended given the utterances than any other interpretation. The system is implemented by postulating an access/selection integration paradigm. It is allowed for probabilistic considerations to be compiled into how elements are indexed, rather than exist as part of a run-time mechanism. IRI-9245496 Wilks, Yorick and Guthrie, Louise New Mexico State University $139,600 - 12 mos. A Lexical Knowledge Base From A Machine Readable Dictionary This is the second year of a three-year continuing award IRI- 9101232. Machine-readable dictionaries (MRD's) contain knowledge about language and the world essential for large-scale tasks in natural language processing (NLP). This knowledge, however, collected and recorded by lexicographers for human readers, is not presented in a structured manner for MRD's to be used directly as tools for NLP tasks. What the NLP research community needs is machine tractable dictionaries (MTD's); that is MRD's transformed into a format for NLP tasks. Previous work has developed a range of tools for extracting information from MRD's for NLP tasks in general. This work will combine these methods based on their respective strengths. It will define a methodology which combines various elements of these methods into a single coherent procedure. Different forms of knowledge are derived from each individual method and each is incomplete in certain respects; yet the combination of these methods yields better results than could any individual method. In particular, the goal is to construct an MTD as a data-base of lexical facts, together with a network of statistically-related word sense. The MTD will support the automatic construction of lexicons for many types of NLP systems, from those using traditional methods based on syntax and semantics to those based completely on statistical methods. MACHINE LEARNING AND KNOWLEDGE ACQUISITION IRI-9244476 Clancey, William J. Institute for Research on Learning $76,925 - 12 mos. Modeling the Limitations of Knowledge Representations in a Knowledge Acquisition Tool This is the second year funding of a two-year continuing award IRI- 9022293. Knowledge engineering is concerned with the development of qualitative computer models of processes, and procedures for applying these models to practical problems of diagnosis, design, control, etc. of complex systems. Knowledge acquisition research focuses on the development of computer tool for facilitating this modeling process, especially the accumulation of general representation and reasoning procedures that can be used for many problems. This work studies the foundational issues of formalization and human cognition that lie behind reuse of representation in new settings for new purposes. Situated cognition research suggests that people have capabilities for new representations that today's computers lack. To deal with this, the work will represent the assumptions that justify changes to a model, so a knowledge acquisition program, but especially knowledge engineers themselves, will be able to intelligently reconstruct the origin and meaning of past representations, and hence more readily adapt them to new contexts. The main idea is to study how people use a layered history of cases, representations, and annotations to reinterpret and re-represent their past work as they discover and argue for changes to an evolving model. The result is improved failure-analysis techniques, coupled with a better understanding of what aspects of the knowledge engineering process cannot be automatized using current methods, but which appropriate tools can nevertheless facilitate. IRI-9249054 Dietterich, Thomas Oregon State University $37,046 - 12 mos. PYI: Formulations of Machine Learning Programs This is the fifth year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-8657316. The research continues to seek ways to improve the understanding of both the foundation and the practical aspects of building machine learning programs. The focus is in three directions: (1) finding ways by which learning programs can use the knowledge they already have gained additional knowledge; (2) understanding the role of active experimentation in aiding the learning process, and (3) analyzing the constraints the different representation languages place on the learning process. To make progress in each of these areas, controlled experiments are conducted to compare different learning methods, experimentation strategies and representation languages. IRI-9149815 Dietterich, Thomas Oregon State University $25,000 - 12 mos. PYI: Formulations of Machine Learning Programs This is the fifth year base funding of a five-year Presidential Young Investigator continuing award IRI-8657316. See above award for a description of the project. IRI-9247015 Epstein, Susan L. CUNY Hunter College $4,000 - 12 mos. REU: Learning Search Control Strategy This is a REU supplemental award to IRI-9001936. This will support undergraduate participation in the ongoing research described. The best current programs for two-person perfect information games play only a single game; they search exhaustively and they do not learn. The goals of this project are to demonstrate further the viability of a weak theory as a search paradigm and to facilitate the construction of computers that learn entire categories of tasks, rather than requiring individualized instruction. This research develops a program, HOYLE, that can play any such game correctly and, with experience, can learn to play it extremely well. Instead of extensive search, HOYLE learns a search control strategy for each new game under the guidance of its weak domain theory: a combination of procedural knowledge about game playing, declarative knowledge about specific games, a language and framework for strategic elements, and a set of narrow but expert perspectives called Advisors. As it plays a new game, HOYLE selectively constructs, organizes, and reformulates a knowledge base for each game from its playing experience. Under a novel architecture that employs a variety of control strategies for move selection, HOYLE combines that knowledge base with its Advisors to produce a coherent, incisive, steadily improving strategy for the game. A prototype of this multifaceted learning has proved itself remarkably effective in a varied but limited domain. HOYLE addresses, both in its design and in its implementation, important questions in theory formation, collaboration of experts, conflict resolution, experimental design, and operationalization. IRI-9211045 Etzioni, Oren University of Washington $30,000 - 12 mos. RIA: Explanation-Based Learning: Finding Better Explanations Via Partial Evaluation This is the first year of a two-year continuing award. Controlling search is a central concern for AI. Overcoming a combinative search in realistic planning, design, and reasoning problems requires large doses of domain-specific search-control knowledge. Explanation-Based Learning (EBL) has emerged as a standard technique for acquiring search-control knowledge. Previous EBL work has produced impressive demonstrations but has also uncovered a fundamental problem- EBL frequently constructs overly-complex explanations that yield ineffective control knowledge. This research describes a solution: integrating EBL with partial evaluation to improve EBL's explanations. In standard EBL systems, the problem solver's behavior on a training example determines what EBL explains and how. Partial evaluation, in contrast, performs a global analysis that often yields simpler and more general explanations. In previous work, STATIC (a partial evaluator written by the PI) was pitted against PRODIGY/EBL, a state-of-the-art EBL system. When tested in PRODIGY/EBL's benchmark problem spaces, STATIC generated search-control knowledge that was up to three times as effective as PRODIGY/EBL's, and did so twenty-six to seventy-seven times faster. Since STATIC's analysis in not focused by training examples, however, it may flounder when confronted with large and complex problem spaces. The PI intends to design and build a hybrid system , called DYNAMIC, that will overcome the weaknesses of both approaches. DYNAMIC will identify learning opportunities a-la-PRODIGY/EBL, but generate explanations a-la-STATIC. The detailed studies of the two systems suggest that DYNAMIC will significantly out perform both, and yield insights in two fundamental questions: how to improve machine-generated explanations, and what is the appropriate role of training examples in explanation-based learning? IRI-9209795 Hirsh, Haym Rutgers University, Busch Campus $28,716 - 12 mos. RIA: Scaling Up Version Spaces This is the first year of a two-year continuing award. This grant addresses a problem central to building intelligent systems, namely how to extract knowledge from data. This problem, known as "inductive concept learning," has received much attention in the machine learning community, and provides an alternative approach to the labor intensive and time-consuming knowledge-acquisition bottleneck by foregoing interaction with an expert altogether, and instead acquiring knowledge from case libraries. Although version spaces are one of the best-known conceptual tools for concept learning, they suffer from three limitations that restrict their use as a practical tool for learning: computational intractability, noise-intolerance, and representational inadequacy. This work proposes a three-layered approach to overcome these limitations so that version spaces can be applied to practical problems while maintaining the attractive properties that make them a useful conceptual and analytical tool for concept learning. IRI-9257592 Kaelbling, Leslie P. Brown University $24,154 - 12 mos. NYI: NSF Young Investigator This is the first year base funding of a five-year NSF Young Investigator continuing award. The focus of the proposed research is the design and implementation of autonomous agents-computer programs that have a sustained interaction with a dynamic, incompletely predictable environment. Programming such agents is very difficult, largely because the programmer rarely actually knows the correct behavior for the agent. For example, typically, the programmer ends up learning, during the debugging process, how the agent's sensors and effectors interact with the environment. Thus, the agent, rather than the programmer, should learn this from experience. This learning, called reinforcement learning, will be investigated. In particular, two special problems will be studied: that they require the input and output spaces to be completely enumerated (and have polynomial time complexity) and that they cannot use prior information about the world, information which ought to be able to speed and deepen the learning process. IRI-9242109 Kibler, Dennis University of California, Irvine $61,989 - 12 mos. Learning Stepping Stones for Problem Solving This is the second year of a two-year continuing award IRI-9001756. Approaches to acquiring problem-solving knowledge have focused on achieving general results with simple problem domains. These domains have had small search spaces, few subgoal interaction, and only Boolean constraints. Real-world application domains have many subgoal interactions, large search spaces, and real-valued performance constraints as well as Boolean constraints. Stepping Stone is a new general learning problem solver for solving difficult real-world problems. Stepping Stone solves these problems by decomposing them into simpler subproblems, and then learning to deal with the interactions that arise between them. An initial version of our approach demonstrates this in the classical tile-sliding domain. This work will extend the system and demonstrate its capability in real-world problem-solving situations such as VLSI design and job-shop scheduling. By experimenting with both classical problem-solving domains and real-world applications it will demonstrate that the problem solver is both powerful and general. IRI-9209577 Kulkarni, Sanjeev R. Princeton University $30,000 - 12 mos. RIA: Extensions of Learning Models and Applications to Signal Processing and Geometric Reconstruction This is the first year funding of a two-year continuing award. Recently there has been a great deal of work on formal models of machine learning such as the probably approximately correct (or PAC learning model). This model is a precise framework attempting to capture the notion of learning from examples. Recent progress in machine learning and statistical inference on paradigms such as the PAC model has provided fundamental results on the amount of data needed for function approximation in a completely non-parametric setting. The applicability of these paradigms is limited by the assumptions on the data gathering mechanisms and the performance criteria. Some of these assumptions will be relaxed to allow the extended learning paradigm to be applied to areas such as signal/image processing and geometric reconstruction. The approach is to place mild assumptions on the function classes while allowing more flexibility in the sampling and error criteria. Specifically, the extensions proposed are to allow deterministic sampling strategies, sampling over noncompact domains, and learning with respect to general performance criterion. The extended model will be applied to a variety of problems in signal processing and geometric reconstruction to provide information complexity results for some classical and new reconstruction/estimation problems. In the area of signal processing, the framework will be applied to problems dealing with tomographic image reconstruction, multiresolution signal processing, and classical sampling theorems. In the area of geometric reconstruction, applications to stochastic geometry and shape form probing problems. The approach will provide results on the fundamental capabilities and limitations of reconstruction as well as sample size bounds for these applications. IRI-9245567 Michalski, Ryszard and Tecuci, Gheorghe George Mason University $75,875 - 12 mos. A Theory and Methodology of Multistrategy Learning This is the second year of a two-year continuing award IRI-9020266. The last several years have seen a great expansion and diversification of research directions and approaches in machine learning, and a simultaneous interest in developing systems that integrate various learning strategies. Such a situation creates a need for analyzing and clarifying the relationships among different strategies and approaches, and building a theoretical basis for the implementation of multistrategy learning systems. This research attempts to develop a theoretical framework for describing diverse learning processes, based on a general notion of inference (hence, inference-based theory). According to this theory, a system learns from input information by trying to understand it i.e., to relate it to its background knowledge (BK). Based on this theory the researcher plans to develop a multistrategy task-adaptive learning (MTL) system, that synergistically integrates different learning strategies. Given an input and a goal of learning, an MTL learner applies the strategies that are most appropriate according to the relationship between the input and learner's BK in the context of the learner's goal. The MTL methodology is intended to integrate ultimately such learning strategies as empirical learning, constructive induction, explanation-based learning, abduction, learning by analogy and abstraction. IRI-9244458 Mooney, Raymond J. University of Texas, Austin $54,331 - 12 mos. Refining Concepts and Domain Theories by Combining Explanation- Based and Empirical Learning This is the second year funding of a two-year continuing award IRI- 9102926. Recently, two general approaches have emerged in machine learning research. Empirical learning involves inductively acquiring concept descriptions by examining the similarities and differences among a large number of examples. Explanation-based learning (EBL) involves using existing knowledge to explain and generalize single examples and thereby acquire operational concept descriptions. However, both of these approaches have important weaknesses. Standard empirical methods cannot take adequate advantage of existing knowledge. Standard explanation- based methods cannot deal with domain theories which are incomplete or incorrect. Our goal is to develop robust and efficient methods that combine explanation-based and empirical techniques to refine imperfect concepts and domain theories to account for a set of empirical data. IRI-9241425 Pirolli, Peter University of California, Berkeley $61,160 - 12 mos. Strategies and Mechanisms for the Construction and Refinement of Programming Knowledge: A Unified Computational Model of Learning This is the third year funding of a three-year continuing award IRI-9001233. This project will develop a cognitive model addressing human learning in the domain of computer programming. Building on prior research and additional psychological studies to be carried out, the comprehension of instructional text is formulated as a problem solving process in which the goal is to construct mental models of domain entities, process principles, and problem solving heuristic. New concepts introduced in these texts may be given meaning by explaining their relation to examples presented in instruction. One major goal of this project will be to model differences among learners in the self-explanation of examples. Specifically, preliminary studies show that learners differ in the way that they explain examples to themselves, and these differences have significant correlations with subsequent programming performance. Another major goal will be to study and develop a model whereby learners reflect on their problem solving and generate new strategies. Both of these areas have received little study in cognitive science. Our results should be important to the psychology of learning in complex formal domains, the acquisition and modification of knowledge in machine learning, and instructional research. IRI-9247737 Porter, Bruce W. University of Texas Austin $37,500 - 12 mos. PYI: Machine Learning and Evolutionary Problem Solving This is the fourth year matching amount funding of a five-year Presidential Young Investigator continuing award IRI-8858802. The research focuses on the construction and use of computer programs which are capable of learning and solving problems with considerable autonomy. Such programs are capable of acquiring significant knowledge and evolving into effective "expert systems." The significance of this research is that effective computer systems in a wide variety of applications must utilize extensive and specific knowledge about the application - not just general algorithmic or problem solving techniques. IRI-9249055 Porter, Bruce W. University of Texas, Austin $25,000 - 12 mos. PYI: Machine Learning and Evolutionary Problem Solving This is the fifth year base funding of a five-year continuing Presidential Young Investigator continuing award IRI-8858802. The research focuses on the construction and use of computer programs which are capable of learning and solving problems with considerable autonomy. Such programs are capable of acquiring significant knowledge and evolving into effective "expert systems." The significance of this research is that effective computer systems in a wide variety of applications must utilize extensive and specific knowledge about the application - not just general algorithmic or problem solving techniques. IRI-9204473 Rendell, Larry A. University of Illinois, Urbana $162,549 - 36 mos. Techniques for Learning Hard Concepts through Constructive Inducition Although mechanized induction is relevant for science and technology, standard methods, such as similarity-based learning, are seriously limited for learning hard concepts from poorly understood domains. Several important domains, such as protein folding, would benefit from more advanced treatment of data and partial knowledge to construct new features that help similarity- based learning. The goal of this project is to improve methods for feature construction, using increased interaction between computer and user to take advantage of the strengths of each. Two thrusts of this project are more dynamic construction and more interactive refinement. For dynamic construction, the problem of constructing representations is decomposed into several subproblems for forming and selecting components for new features. Components are dynamically determined by information contained in automated analysis of data and knowledge. For interactive refinement, the approach allows partial specification of "hunches", which can represent domain-dependent or independent heuristics. These fragments of knowledge can be updated by the learning systems and re-examined by the user. IRI-9212190 Schlimmer, Jeffrey C. Washington State University $56,644 - 24 mos. RIA: Learning Models for Database Checking As part of a series of research projects aimed at marrying machine learning technology to economically important problems, this project in concerned with the task of finding errors in databases. Because many database situations do not afford the luxury of a surplus programmer, machine learning methods are used to construct prescriptive data models. These models predict appropriate values for data attributes, and if actual values differ, an alarm is raised. Two basic research issues in database consistency checking are studied. First, fundamental limitation of a common class of learning methods is identified, then a new learning formulation is proposed. Second, this research reiterates the need to incorporate expert knowledge into the learning process, and a space of possible approaches. Understanding of the resulting new methods will lead to advances in the quality and applicability of database consistency systems, and as a consequence, and to improve performance in the wide variety of systems that rely on high quality data. IRI-9242108 Shavlik, Jude W. University of Wisconsin, Madison $61,156 - 12 mos. Integrating Explanation-Based and Neural Approaches to Machine Learning This is the third year funding of a three-year continuing award IRI-9002413. To be considered intelligent, machines must be capable of learning. Symbolic and neural approaches to machine learning both have been heavily investigated. However, there has been little research into the synergies achievable by combining these two learning paradigms. This work will develop a hybrid system that combines the symbolically-oriented explanation-based learning paradigm with the neural back-propagation algorithm. In this system, the initial neural network configuration is determined by the generalized explanation of the solution to a specific classification of planning task. This research addresses the problem of choosing a good initial neural network configuration and overcomes problems that arise when using imperfect theories to build explanations. Most real-world problems can never be formalized exactly. However, there is much to be gained by utilizing the capability to reason approximately correctly. Explanation-based learning provides a way to profitably use casual models of the world, while neural networks provide a way to refine roughly-correct concepts. This research on combining these two learning paradigms promises to broaden the applicability of machine learning techniques, producing learning algorithms that are not brittle and which can produce concepts whose accuracy improves through experience. IRI-9248595 Weld, Daniel S. University of Washington $62,500 - 12 mos. PYI: Model Reformulation and Machine Learning This is the fourth year base and third year matching amount of a five-year Presidential Young Investigator continuing award IRI- 8957302. This research focuses on fields of machine learning and automated reasoning. Because computers require an explicit model of a system to perform reasoning and the character of the reasoning depends critically on the nature of the model, great leverage can be obtained by enabling programs to reformulate their models dynamically. The investigator has pursued reformulation in his work on aggregation and the comparative analysis techniques of differential qualitative analysis and exaggeration. He will now incorporate ideas from machine learning to aid in the reformulation task. One specific project is to generalize current schemes for model reformulation. The limits of applicability for each model will be represented explicitly in a language based on the simplifications implicitly assumed by the model. A second project is a restricted form of theory formation: the automatic construction of world models. Here the approach is to generalize previous work on measurement interpretation and view application by characterizing he program's views with the modeling language described above. COGNITIVE SYSTEMS IRI-9100149 Arkin, Ronald Georgia Tech Research Corporation $15,000 - 24 mos. (Jointly funded with the Robotics and Machine Intelligent Program - Total Award $119,901) Cooperation and Communication in Multi-Agent Reactive Robotic Systems This research is to study communications and control requirements for multiple autonomous robots to cooperate in accomplishing a task, where the robots are organized heterarchically instead of using the more common hierarchical (master/slave) control. Each behaves reactively, in accordance with sets of relatively primitive motor schemes. The resulting organization does not require an explicit world model, and should be more robust and hence more suitable for operations in remote and/or hazardous environments. A key research issue is to evaluate the efficacy of such systems operating under constrained inter-agent communications. Such agents could be more economical to operate. The goal of the research is to understand and develop control and communication mechanisms that produce robust yet efficient cooperative robot behavior. IRI-9247016 Baird, Bridget Connecticut College $4,000 - 12 mos. REU: Real-Time Tracking of Simultaneous Musical Inputs Using a Multiprocessor This is a REU supplemental award to IRI-9010793. This will support undergraduate participation in the ongoing research described. An expert musician, when playing with other musicians, is able to listen to simultaneous inputs and synthesize that information in order to decide on the correct position in the musical score. The research objective of this project is to investigate algorithms that would allow an artificial system to simulate this real-time tracking ability. A working computer performer, capable of playing interactively with live musicians, and implemented on a multiprocessor, will be the vehicle for testing and developing the tracking algorithm. Since the computer performer will need to examine complex patterns of notes from several sources and also to inspect simultaneously several possible locations in the musical score, parallel processing and pattern matching considerations will be important. This research will provide insight into methods of parallel processing and pattern matching; will give musicians a new, practical interactive tool for performance; and will be an innovative application of multiprocessors. IRI-9248672 Ballard, Dana H. University of Rochester $65,000 - 24 mos. (Jointly funded with the Information Technology and Organizations Program and the Robotics and Machine Intelligence Program - Total Award $225,000) Animate Robotics Vision This award is the fourth year of a four-year creativity extension IRI-8903582. Based on creative accomplishments in active vision and reinforcement learning. The PI has proposed extensions to his work in combining deictic primitive behaviors (primitives that dynamically reference the world rather than depending on models) to produce the complex behaviors needed for intelligent robotic agents to perform practical tasks in active vision. Topics will include economical search of three-dimensional space for small objects, speedup of the learning algorithms, and development of a set of qualitative grasping primitives, integrated with visual feedback, as a natural extension of the vision primitives already developed by the PI. This last is a promising alternative to the fine, open loop control used in conventional robotic systems. IRI-9244433 Barnden, John New Mexico State University $89,400 - 12 mos. Reasoning Usefully About Mental States During Discourse Understanding: A Metaphor Based Approach This is the second year funding of a three year continuing award IRI-9101354. This Artificial Intelligence project investigates part of the problem of discerning the coherence relationships within natural language discourse. The relationships of interest are between discourse components that describe or presuppose mental states (propositional attitudes). The mental states can be those of the speaker, listener, or any person mentioned by the discourse, and can be states of belief, intention hope, desire, etc. The focus of this work is on descriptions that rely implicitly on common sense models of the mind that people often deploy, such as the prevalent model of a mind and its beliefs as respectively a physical container and some physical objects within it. Common sense models are frequently relied upon in mental-state descriptions and are important factors in discourse coherence relationships, yet previous detailed work on representing and reasoning about mental states has largely skirted the issues. This research relates strongly, however, to work by psychologists and linguists on metaphor, and metonymy, because the common sense models are metaphorical and metonymy is often involved as well. The project includes the development of a prototype AI discourse- interpretation system, as well as major data collection activities: collection of information about mental-state language through experiments with human subjects, and through semi-automated examination of diverse natural language texts, including definitions and examples in machine-readable dictionaries. Annotated compendia of examples from these texts will be produced. The project is also concerned with the relationship between mental- state descriptions in English and those in other language, notably Spanish, with a view to contributing to machine-translation research. IRI-9209365 Kalita, Jugal University of Colorado $89,896 - 36 mos. RIA: Semantics of Action Verbs for Animation Control of Task Performance Language has evolved rich vocabularies, especially action verbs and their modifiers, that can felicitously describe simple as well as complex movements performed by humans and others. The overall long term goal is to model the structure of human-like agents, instruct them in natural language to perform simple physical tasks, and display their actions being performed in terms of computer graphics animation. The research addresses selected problems, issues and methods of solutions in the overall endeavor. This research will investigate how natural language relates to concepts of space, motion and force, and use these relations in the specification of lexical semantics of action verbs and their modifiers. This research will be further strengthened by the claims of obtaining useful lexical entries for an extended set of action verbs and their modifiers. It will investigate salient functional, structural and other properties of manipulated objects and the physical environment in general, required for motion based on the structure, function and other relevant constraints. This is crucial to actually executing tasks specified by natural language instructions in varying physical situations. Meaning entries (given in natural language) for action verbs in human dictionaries determine the nature of processing necessary to use such entries for possible acquisition of new computational lexical entries. DMS-9121266 Mumford, David Harvard University $10,000 - 12 mos. (Jointly funded with the Computational Mathematics Program and the Robotics and Machine Intelligence Program - Total Award $149,000) Mathematical, Computational and Biological Aspects of Vision This is the first year funding of a three-year continuing award. A visual signal, as recorded on the retina of an animal or by a TV camera, differs from many signals analyzed by engineers in that it is produced by a world with many overlapping objects and shadows, and, to "decode" the signal, these must be teased apart so as to reconstruct the world geometry. Most classical techniques smooth over the discontinuities in the signal produced by this multiplicity of effects, making it harder to separate them and reconstruct the world geometry. The PI plans to use new techniques involving variational problems for discontinuous maps to attack some of these problems. He also seeks to compare this mathematical appoach with the neural techniques by which animals solve problems. To do this, he will study "neural net" implementations and formulate experiments to study how close these nets are to the true neural activities in animals. The goal is to understand mathematically one of the most remarkable cognitive abilities of living organisms. Vision is a focal point of research activity in computational science, as well as HPCC. IRI-9116399 Nilsson, Nils J. Stanford University $32,905 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Information Technology and Organizations Program - Total Award $98,715) Research On Autonomous Agents This is the first year funding of a three-year continuing award. This research is concerned with developing autonomous, cooperating, adaptive computational agents. Expected applications of this research are in the control of mobile robots capable of performing delivery, maintenance, and/or construction tasks and of agents that gather and manipulate symbolic information over computer networks. The research will concentrate on bounded-time action computation and learning mechanisms and on combining these components in integrated agents able to function cooperatively in dynamic, uncertain environments. IRI-9258517 Ristad, Eric S. Princeton University $25,000 - 12 mos. NYI: NSF Young Investigator This is the first year base funding of a five-year NSF Young Investigator continuing award IRI-9258517. The proposed research is concerned with language modeling and handwriting. The goal of the language modeling research is to represent the computations performed in the comprehension, production, and acquisition of human languages. Computations may be described at various levels of abstraction: as a sequence of state transitions (least abstract), as an algorithm (more abstract), and as a computational problem (most abstract). The first step in the proposed research is to characterize the three central computations of human language (comprehension, production, and acquisition) at the highest level of abstraction, as computational problems. Each such problem statement is a computational theory of a portion of language. The total sum of all such problem statements is a comprehensive computational theory of human language. The goal of the handwriting recognition research is to accurately model the human handwriting process. Such a model is, crucial to improving current handwriting recognition technology. To accomplish this, the statistical constraint inherent in handwriting, including physical constraints arising from the biomechanics of the pen/hand system, motor-perceptual constraint arising from the shared desire to communicate efficiently and reliably, and constraint due to the ad- hoc convention of handwriting, such as the chosen alphabet and language, are explicitly modeled. IRI-9123720 Sklansky, Jack University of California, Irvine $45,000 - 36 mos. (Jointly funded with the Information Technology and Organizations Program and the Robotics and Machine Intelligence Programs - Total Award $330,000) Biologically Inspired Intelligent Classifiers This project will advance knowledge about design of intelligent agents. There are two objectives. The first objective is the development of mathematical insights and procedures for automating feature discovery and learning by evolution -- two biologically inspired forms of long-term learning -- in the construction of automatic pattern classifiers. The second objective is the incorporation of automated feature discovery and evolution in automatic analyzers of large-scale image data banks and time- varying visual signals. This project will work to achieve these objectives by building on the Principal Investigator's and his students' techniques of adaptive hyperplane placement, window training (an extended form of stochastic approximation), relaxed branch-and-bound search, and genetic feature selection. The project will implement and test the resulting mechanisms of long- term learning in three-stage neural classifiers of handwritten numerical characters, tree classifiers of medical images in a large data bank, and adaptive segmenter of moving objects in digitized video. This project will advance the technology of long-term learning in intelligent machines, leading to machines that can detect and explain coherence buried deeply in enormous amounts of noisy multidimensional data. This project will also advance the technology of multiple-sensor robotic vision by automating the discovery of highly discriminating combinations of features from diverse sensors. SPECIAL PROJECTS ASC-9217041 Berwick, Robert C.; Bizzi, Emilio; Bulthoff, Heinrich H.; Jordan, Michael; Wexler, Kenneth; Poggio, Tomaso; Rivest, Ronald L.; Winston, Patrick H.; and Yang, Woodward Massachusetts Institute of Technology $125,000 - 12 mos. (Jointly funded with the New Technologies Program, the Division of Behavioral and Cognitive Sciences, the Division of Advanced Scientific Computing, the Robotics and Machine Intelligence Program and the Experimental Systems Program - Total Award $600,000) HPCC: High Performance Computing for Learning This is the first year funding of a five-year continuing award in the High Performance Computing and Communications (HPCC) Initiative's Grand Challenge Application Groups competition. In this award to Berwick, Bizzi, Bulthoff, Jordan, Wexler, Poggio, Rivest, Winston, and Yang at MIT, the research project has been designed explicitly to push the High Performance Computing algorithmic and architectural envelope via a CM-5 and VLSI testbed and to address many of the HPCC goals. It will advance new algorithms and software for a broad class of optimization and learning problems, tested on and directly driving operations system and architectural changes on the CM-5 (working with one of the CM- 5's key architects). The learning problems addressed are essentially an entire class of modeling/optimization problems that intersect with nearly all HPCC Grand Challenge Problems. IRI-9245074 Carbonell, Jaime G.; Tomita, Masaru; and Nirenburg, Sergi Carnegie Mellon University $94,312 - 12 mos. Machine Translation Summit The Machine Translation Summit is the third international meeting for exchange of scientific, technological and user-oriented information on machine translation. It is being held for the first time in the U.S. where there is increasing research interest in the area. Further stimulation to the research area will be achieved by including partial support for student participation through scholarships. This meeting is jointly funded by the DARPA Speech and Language effort. IRI-9208831 Cole, Ronald; Zahorian, Stephen; and Hirschman, Lynette Oregon Graduate Institute $4,000 - 12 mos. (Jointly funded with the Interactive Systems Program - Total Award $35,762) NSF Workshop on Spoken Language Understanding This is a critical planning workshop to determine research problems in the area of spoken language. The main goals of the workshop are: (a) it identifies the most important areas of research in speech and natural language understanding, with particular attention to those not currently addressed (b) to determine how NSF can benefit from other programs, such as the DARPA speech and natural language program, and yet distinguish itself from these programs; and (c) to produce a set of recommendations to NSF to help guide opportunities. The workshop participants represent a balance of researchers in the areas of speech recoginition, natural language understanding and human-computer interaction. Participants includes researchers with NSF grants and researchers in government and industrial laboratories. Discussion took place in plenary sessions and in breakout groups, which focused on three researh areas: robust systems; in the interaction of speech and natural language; and human-computer interaction. IBN-9202477 Cruickshank, Alexander and Kagan, Herbert Gordon Research Conferences $4,000 - 12 mos. (Jointly funded with the Integrated Animal Systems Program - Total Award $144,000) Summer Gordon Research Conference 1992 The Gordon Research Conferences are held every summer to discuss the recent findings and current thinking in various areas of science. Although participation is limited, the attendees usually consists of established scientists, young investigators who are making important contributions in these research areas and postdoctoral fellows. The meetings are held at schools in small isolated New England towns away from other distractions, in an atmosphere conducive to the exchange of ideas. These conferences have a recognized value for presenting the latest advances in research and also indicating the future trends which are likely to be most productive. IRI-9216094 Hendler, James and Agrawala, Ashok K. University of Maryland, College Park $5,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Robotics and Machine Intelligence Program - Total Award $15,000) Workshop on Artificial Intelligence in Real-Time This workshop on "Artificial Intelligence in Real-Time" is aimed at examining how AI systems can both be supported and can help to support real-time operating systems. This area is of critical importance as AI systems need to function in the support of such critical applications as nuclear power plant control, aircraft operation, hospital life support systems, and military command and control, among others. The workshop, to be held at the University of Maryland in the Spring of 1993, brings together members of both AI and real-time communities to explore issues of mutual interest. IRI-9123156 Hunter, Lawrence; Shavlik, Jude W. National Library of Medicine $9,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Database Software Development Program - Total Award $19,000) Creating an Infrastructure for Intelligent Systems in Molecular Biology Research applying intelligent systems technology to problems in molecular biology is expanding rapidly. Techniques such as artificial intelligence, neural networks, object-oriented databases, large-scale computer modeling, and robotics are being successfully applied to problems in genetics, protein structure, development, evolution, and many other aspects of biology. The rapid growth in this interdisciplinary research has created a need for infrastructure development and program planning. The proposed meeting will bring together about two-dozen active researchers in the field with representatives of NLM and NSF to discuss issues such as coordination of conferences and publication venues, distribution of research support, interdisciplinary training programs, and the sharing of data and other resources. In addition to facilitating communication, the goal of the meeting is to produce a report describing community needs and a set of standing committees to help address those needs. Intelligent systems technology is relevant to several key problems in the HPCC Program. Directly, advanced software technology and algorithms require intelligent systems technology; and indirectly high performance computing systems will be developed based on novel intelligent software algorithms. This proposal addresses some of the key issues in the problem domain of molecular biology, which have implications in many other problem domains of the HPCC Program. DBS-9247119 Joshi, Aravind K. and Gleitman, Lila R. University of Pennsylvania $580,000 - 12 mos. (Jointly funded with the Language, Cognition and Social Behavior Program - Total Award $1,160,000) Center for Research in Cognitive Science This is the second year funding of a five-year continuing award DBS-8920230. This award to the University of Pennsylvania establishes a Science Technology Center for Research in Cognitive Science. The Center for Research in Cognitive Science unites a diverse and richly interconnected group from many traditional disciplines of computer science, linguistics, mathematics, philosophy, and psychology. The goal of the research is to understand the processes and mechanisms by which human beings acquire knowledge about their environment, store and retrieve that knowledge, communicate it to others, and apply it to carry out actions and manipulate their environment. The research is organized into three separate but highly interrelated themes: perception and action, language learning, and language processing. Research in the area of perception and action spans from the processes involved in the first stages of visual and auditory representation of spatial and spectral information, to higher order representation of more complex attributes and to the storage and retrieval of such representations by the organism as they are used in goal-oriented actions. The study of language learning focuses on how children develop the abstract representations of language on the basis of their visual and auditory perceptions. The research in language processing combines investigation of formal systems with investigation of computational models all in the context of empirical study of a wide range of natural languages. Significant features of the perception and action research are its increasing fidelity to actual neural computation and its sophisticated computational modeling and related potential for contributing to artificial intelligence technology. The languages learning research has significant potential for technological spin-off in machine learning and automatic acquisition of lexical and grammatical information for language systems, crucial to the development of grammars sufficient for the robust analysis of unconstrained text. The language processing research will have significant impact of the technological base for human-computer interaction, in particular the design of natural language interfaces for database and expert systems and knowledge-rich systems in general. This center will stimulate enhanced activity in precollege education and in the development of human resources. IRI-9221943 Malkani, Hogan J. Tennessee State University $5,000 - 12 mos. (Jointly funded with The Engineering Systems Program - Total award $20,000) Workshop on Fuzzy-Neuro Systems A five day (August 9-13, 1992) workshop on Fuzzy-Neuro Systems was organized by the College of Engineering and Technology at Tennessee State University to introduce and provide hands-on experience to the faculty and graduate students from Historically Black Colleges and Universities (HBCU) who have either engineering and or computer science degree programs. IRI-9201138 Motro, Amihai George Mason University $4,000 - 18 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $39,750) Uncertainty Management in Information Systems: From Needs to Solutions Uncertainty pervades real world scenarios, and must therefore be incorporated into every information system that attempts to provide a complete and accurate model of the real world. Yet, present generation information systems have very limited capabilities in this regard. On the other hand, many theoretical models have been proposed for the management of uncertainty, but often without thorough understanding of the actual problems faced by researchers, developers, designers and users of information systems. This project concentrates on a planning workshop that will bring together leading researchers in the scientific communities of information systems (e.g., database systems, information retrieval systems, expert systems, office information systems) and uncertainty modeling (often working within frameworks such as mathematical logic, probability theory, fuzzy set theory, possibility theory, and evidential models) to study the needs of the information systems community and to tap the expertise of the uncertainty modeling community, for solutions that respond to these needs. This workshop will promote true dialogue between these two scientific communities; it will establish the state of the art in this field, and will set the course for future research. The project, which is co-sponsored by the European Esprit program, will foster collaborations between researchers in the United States and Western Europe. IRI-9218448 Nilsson, Nils J. and Rumelhart, David E. Santa Fe Institute $0 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Robotics and Machine Intelligence Program - Total Award $33,990) HPCC: Approaches to Artificial Intelligence The field of artificial intelligence (AI) has as its goal the development of machines that can perceive, reason, communicate, and act in complex environments much like humans can, or possibly even better than humans can. Even though the field has produced some practically useful machines with rudiments of these abilities, it is generally conceded that the ultimate goal is still distant. That being so, there is much discussion and argument about what are the best approaches for AI in the sense of laying the core foundations for achieving ultimate goals as well as best in the sense of producing practically useful shorter-term results. Thus, a number of different paradigms have emerged over the past thirty-five years or so. Each has its ardent advocates, and some have produced sufficiently many interesting results so as not to be dismissable out of hand. Perhaps combinations of these approaches will be required. In any case, the advocates of these approaches often feel that theirs is the "breakthrough" methodology that deserves special support. In order to acquaint researchers and others with these paradigms and their principal results, the Santa Fe Institute organizes a workshop to which advocates actively working in the different approaches will be invited. Such a workshop allows leading workers to understand each others' points of view and the relationships among them. Specifically the workshop is not organized around what classically have been the main sub-disciplines of AI, namely machine vision, natural language processing, knowledge representation, planning, expert systems, and so on. Instead, different perspectives are presented on how the whole, or major parts, of AI ought to be attacked. ASC-9217091 Taylor, D. Lansing and Fahlman, Scott E. Carnegie Mellon University $100,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program, the New Technologies Program, the Long-Term Projects in Environmental Biology Program, the Division of Advanced Scientific Computing and the Division of Environmental Biology - Total Award $599,213) HPCC: High Performance Imaging in Biological Research This is the first year of a five-year continuing award in the high Performance Computing and Communications Initiative's (HPCC) Grand Challenge Application Groups competition. This award is to research and develop an Automated Interactive Microscope (AIM). The AIM will combine the latest technologies in light microscopy and reagent chemistry with advanced techniques for computerized image processing, image analysis, and display, implemented on high- performance parallel computers. This combination will produce an automated, high-speed, interactive tool that will make possible new kinds of basic biological research on living cells and tissues. While one milestone of the research will be to show the proof-of- concept of AIM, the ongoing thrust will be continued development as new technologies arise and the involvement of the biological community. INT-9123796 Tripathi, Satish K. University of Maryland, College Park $2,500 - 12 mos. (Jointly funded with the Division of International Program - Total Award $45,250) Cooperative Research in Computer Science, Indo-U.S. Workshop August 4-6, 1992, Bangalore, India This project supports participation of a U.S. team of computer scientists and engineers in a Indo-U.S. workshop on "Cooperative Research in Computer Science" to be held in Bangalore, India, August 4 to 6, 1992. The areas of expertise to be represented include: robotics and computer vision, computer systems and networking, software engineering, and artificial intelligence. The objective of the workshop are to: present papers by U.S. and Indian participants about the recent advances in the various topical areas, identify research needs in each of these area, identify suitable areas for Indo-U.S. collaborative research, and establish firm collaborations between U.S. and Indian scientists. This is the second Indo-U.S. workshop in the area of computer science. The first was held in 1989 in the city of Hyderbad, India with participation from India's main governmental and private research organizations with interest in computer research and applications. Since that time a number of collaborative activities have resulted. The present is designed to continue and intensify this mutually beneficial interaction and collaboration. The project meets the objectives of the U.S.-India Cooperative Science Program. IRI-9212592 Wah, Benjamin W. University of Illinois, Urbana $0 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Interactive Systems Program - Total Award $35,395) HPCC: Workshop on HPCC: Vision, Natural Language and Speech Processing and Artificial Intelligence This proposal is concerned with a workshop on High Performance Computing and Communications (HPCC) for Grand Challenge Applications - Vision, Natural Language and Speech Processing, and Artificial Intelligence. This workshop will bring together 22 invited experts from academia and industry, with the goal of identifying near-term and long-term problems in supporting these grand challenge application problems. Many traditional results in these grand challenge applications involving vision, natural language and speech processing, and artificial intelligence were developed without the availability of HPCC systems. On the other hand, computer architects design HPCC systems without considering requirements in vision, natural language and speech processing, and artificial intelligence applications. In this workshop, key issues and potential approaches/research directions for the next five to ten years will be identified, with the goal of answering the following problems: (1) What are grand challenge applications in vision, natural language and speech processing, and artificial intelligence that can be benefited by the availability of HPCC systems? (2) How should HPCC systems be designed so they can support grand challenge applications in these areas? The interdisciplinary HPCC initiative has created a need for infrastructure development and program planning. The meeting brings together 22 experts in the fields with representatives of NSF to discuss current research issues and formulate interdisciplinary research frameworks. IRI-9244738 Wilks, Yorick New Mexico State University $1,154 - 12 mos. (Total Award $12,000) US-Japan Workshop on Integrated Comprehensive and Generation Systems in Multi-Media Environments The history of artificial intelligence has been marked by a continuing specialization and diversification of its areas of research. At times, some sort of reaction to developments is needed, one in which theorists can step back and consider the implications of theories in subareas of artificial intelligence that are potentially important to the development of integrated systems. All areas of artificial intelligence, such as natural language understanding and computer vision, have made significant progress, but the field lacks a high level of integration. The goal of this workshop is to bring together top-level researchers from both the US and Japan to explore potential development of an integrated multi-media environments, as well as to promote better understanding and cooperation on this problem between the two countries. IRI-9207927 Wilks, Yorick New Mexico State University $15,000 - 12 mos. Cooperation on Research and Development of Reusable Multi-Lingual Lexicons for Natural Language Processing This is an ESPRIT/NSF Cooperative Activity project supported jointly by NSF to Yorick Wilks of NMSU and by CEC/DGXIII/ESPRIT to Antonio Zampolli of University of Pisa, Italy. The building of reusable, multilingual lexicons for natural language processing is essential for the Grand Challenge Problem - Speech/Language Understanding. There are extensive machine translation and other linguistic research efforts in EC, especially those under the auspices of Espirit. This joint project is expected to provide a start of means for U.S. researchers to form a collaborative relationship with their European counterparts. The results of this joint work will be shared with the general research community on both sides of the Atlantic. Funding of the European colleagues is provided by EC/Esprit. MESSAGE FROM THE PROGRAM DIRECTOR In FY92, the Database and Expert Systems (DBES) Program continued its special emphasis on scientific and engineering databases. The highly interdisciplinary initiative þResearch on Scientific Databasesþ fosters coupling between database technology and scientific or engineering research for the advancement of both. The first announcement released in FY91 resulted in 21 awards jointly funded by relevant NSF scientific and engineering programs. The second announcement brought in 100 responses, with the awards planned for FY93. Regular proposals in the area of scientific and engineering databases are welcome at any time. Proposals for development of courses for bridging the gap between domain scientists or engineers and database scientist in an effort to build effective scientific databases are of interest. The DBES research community has participated in the High Performance Computing and Communications Initiative Grand Challenge Application Group. Awards in these initiatives are denoted by SDB and HPCC, respectively, in the DBES Summary of Awards. A number of research planning workshops were funded in FY92, and these are listed in the General Research Section. In addition, two workshops that were funded in FY91 were held in FY92: (1) Workshop on Future Directions in Text Analysis, Retrieval and Understanding, Chicago, Illinois, October 11-12, 1991. NSF Grant IRI-9114210, Jitender Deogun, Univ. of Nebraska, Lincoln and Vijay Raghavan, University of Southwest Louisiana; (2) NSF Workshop On Visual Information Management Systems, Redwood, CA, Feb. 24-25, 1992. NSF Grant IRI-9112753, Ramesh C. Jain, University of Michigan. Information on workshop reports can be obtained from Maria Zemankova (mzemanko@nsf.gov). Proposals for planning workshops intended to stimulate innovative research are also encouraged. New emphasis is being given to research related to developing electronic libraries where this term is used to capture the various concepts of a future digital knowledge network, based on the National Research and Education Network (NREN). Research addressing issues in representation, storage, distribution, interoperability, access and delivery of different information forms -- text, image, sound, videos, large databases, software tools, educational aids, scientific or engineering instruments, etc., in order to satisfy a wide variety of users with specific needs is encouraged. The DBES Program will continue supporting new research ideas and will play an active role in fostering collaboration with other NSF programs, other agencies and industry in order to achieve the goal of building widely accessible intelligent information systems for the enhancement of human endeavors. DATABASE AND EXPERT SYSTEMS PROGRAM SCIENTIFIC SCOPE The Database and Expert Systems Program supports research fundamental to the design, implementation, development, management and use of databases, information retrieval and knowledge-based systems. The aim is to build a "new generation" of distributed, interoperable, multimedia, intelligent, active information systems capable of sophisticated and efficient information processing. Research pertinent to this goal involves investigation of novel concepts, or combination and extension of conventional systems. It hinges on basic and applied research in databases, knowledge-based systems and information retrieval systems and includes a wide scope of related areas, ranging from artificial intelligence methodologies to techniques for effective utilization of hardware technology. Projects supported by the Database and Expert System Program can be divided into interrelated areas: (1) data/information/knowledge modeling; (2) information access, (3) physical and system aspects; and (4) system development and administration. Data/Information/Knowledge Modeling This area provides foundations for new, more expressive data / information / knowledge modeling. Topics include object-oriented systems; temporal, spatial, pictorial, multi-media databases; special-purpose scientific databases, including metadata representation; full-text systems; deductive databases; constraint- based systems; active systems; and knowledge-based systems. Issues considered include type systems; declarative extensions; database and knowledge-base evolution, integrity and validation; inheritance and exceptions; and management of uncertainty arising from imprecisions in data or knowledge. (Basic research in formal models of knowledge and information is supported in the Knowledge Models and Cognitive Systems Program, IRIS. Related research in languages is considered in the Programming Languages and Compilers Program in the Division of Computer and Computation Research). Information Access The aim of research in this area is to design more intelligent and efficient access methods. Research topics include query language design; data analysis tools, including automation of knowledge acquisition from databases; enhanced query processing (e.g., statistical sampling, approximate queries, cooperative answering, employment of feedback, thesauri or semantic nets in information retrieval); browsing; knowledge-based query optimization; and resolutions of incompleteness and inconsistencies in heterogeneous systems. (Related research placing emphasis on human interfaces, e.g. user modeling, data visualization or natural-language interfaces, is supported in Interactive Systems Program, IRIS. Research into principles for building multi-user collaborative systems is covered in the Information Technology and Organization Program, IRIS). Physical and System Aspects The objective of this area is building high performance systems through addressing issues in reliable storage, access and manipulation of actual data. Research topics include indexing and hashing algorithms; persistent object storage; main-memory systems; associative memory; cache-memory; utilization of optical storage; tertiary storage management; distributed and heterogeneous systems; extensible systems; real-time or constrained-time/space query processing; parallel processing; concurrency control; long duration transaction processing; fault-tolerant systems; backup and recovery. (Related research is also supported in Computer Systems Architecture Program and Software Systems Program in the Division of Computer and Computation Research, and in the Division of Networking and Communications Research and Infrastructure.) System Development and Administration This category involves development of methodologies for specification, implementation, verification, maintenance and management of information systems. Research includes work in specification languages, including areas of CAD/CAM and scientific databases; extension and development of design tools and environments; reconfigurable systems; fast prototyping; modularization of large heterogeneous systems; security issues; audit trails; and providing integrated interfaces for related manipulations or tasks. (Related issues are supported in the Information Technology and Organizations Program, IRIS and in the Software Engineering Program in the Division of Computer and Computation Research.) DATABASE AND EXPERT SYSTEMS PROGRAM FISCAL YEAR 1992 RESEARCH PROJECTS DATA/INFORMATION/KNOWLEDGE MODELING IRI-9119310 Borgida, Alex Rutgers University $73,849 - 12 mos. Research on Families of Description Logic KBMS This is the first year funding of a three-year continuing award. Knowledge Base Management Systems (KBMS) based on Description Logics perform deductions with intensional descriptions, thus providing novel services, and now finding commercial applications. This research considers facilitating the specification and implementation of a family of such KBMS, through the extension of a core system, ProtoDL, by the addition of new description constructors. Specifically, the objectives include (1) a formal notation for specifying the semantics of Description Logics; (2) a system, with a score of procedures that need to be completed for every new constructor being added to the KBMS; (3) heuristics linking the implementation and specification. Thereafter, given a particular application (in the domain of Software Information Systems -- the chosen area for experiments), the prospective KB developer determines suitable extensions to the description language and its inferences, with the help of a "knowledge language engineer"; the latter then augments the ProtoDL implementation by filling in and modifying the appropriate procedures/modules. The advantages of such extensible KBMS include: (1) meeting the need of many practical applications for domain-dependent constructors (e.g., dealing with time, plans, etc.); (2) a novel approach to the expressiveness vs. computational intractability stand-off: starting with a limited language and extending it in an application- dependent way. IRI-9212074 Chen, Weidong Southern Methodist University $28,520 - 12 mos. RIA: Reasoning about Database Updates This is the first year funding of a three-year continuing award. Updates are essential in database applications that require reasoning under dynamic situations. The objective of the project is to enhance current database systems with strong capabilities of modeling and processing changes. An update calculus and an update algebra have been developed for specifying dynamic queries, and have been shown to be equivalent to each other. This project investigates algorithms for optimization and efficient evaluation of dynamic queries. Hypothetical queries are a special case in which new databases after updates are not needed for answers. Transformations of these queries into static ones are explored in order to take advantage of existing querying technology. This project also pursues simplification and inference techniques for consistency checking so that data integrity can be preserved after updates. Results of the project will extend database technology to a wider class of applications that exhibit complex dynamic behavior. IRI-9242024 Chomicki, Jan Kansas State University $32,282 - 12 mos. RIA: Dynamic Integrity Constraints in Databases This is the second year funding of a two-year continuing award IRI- 9110581. Dynamic integrity constraints arise in many traditional and non-traditional database applications; however, such constraints are not supported by current database management systems. This project focuses on efficient methods for checking dynamic integrity constraints in databases. Past Temporal Logic and Past Metric Temporal Logic are proposed as constraint languages, making it possible to formulate both qualitative and quantitative temporal conditions. The project includes the development and the implementation of a method to reduce dynamic integrity checking to static integrity checking. If this method is used, past database states do not have to be stored for the purpose of dynamic integrity checking. Instead, compressed historical information is stored in every database state as auxiliary relations. The definitions of the needed auxiliary relations are automatically derived from the constraints. Techniques that optimize constraint checking and issues that arise from the integration of this new method with existing database management systems will also be addressed. The proposed research will significantly enlarge the application scope of database management systems and will make it easier to integrate these systems with real-time processes. IRI-9200898 Grant, John Towson State University and Minker, Jack University of Maryland, College Park $39,970 - 12 mos. RUI: Research in Database Foundations and Logic Programming at an Undergraduate Institution This is the first year funding of a two-year continuing award. This research addresses three important topics dealing with the foundations of knowledge bases: view updates, combining knowledge bases, and inconsistent knowledge bases. Algorithms are developed to provide solutions to several problems in this subject. The view update problem is concerned with the appropriate translation of the user's request to update a particular view into an update of the underlying relations. Algorithms that have previously been developed for stratified disjunctive databases are refined by using specialized data structures and extended to more general types of knowledge bases. Many different but related knowledge bases have been constructed in recent years. Previous research on combining knowledge bases considered primarily the case of definite deductive databases. Those algorithms are extended to more general types of knowledge bases using the stable and well-founded semantics. Sometimes a knowledge base may become inconsistent. Several different semantics have previously been proposed to define the meaning of such knowledge bases in order to localize inconsistencies. Algorithms are produced to compute answers to queries under these semantics using theorem-proving techniques. The research will be useful for the development of advanced knowledge based systems; it will also enhance the research environment at Towson State University, an undergraduate institution. IRI-9120330 Henrion, Max; Heckerman, David and Horvitz, Eric Institute of Decision Systems Research and Shortliffe, Edward H. Stanford University $193,439 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $200,000) Practical Decision Analytic Methods for Large Knowledge-Based Systems This is the first year funding of a three-year continuing award. Expert systems that perform diagnosis are being used increasingly in medicine, aerospace, and other high-stakes domains. Research has recently shown that heuristic schemes widely used in such systems to encode uncertain knowledge -- used in diagnosis and decision making -- can produce faulty reasoning. Schemes based on probability and decision analysis have been proposed as sounder; however, they have often then been dismissed as intractable. The goal of this research is to develop and evaluate new probabilistic and decision-analytic schemes, that are both sound and tractable. The planned products comprise methods for encoding uncertain knowledge in probabilistic form; for encoding judgments on values, costs, and preferences; efficient algorithms for probabilistic inference decision making; and methods to generate comprehensible explanations of systems conclusions. These techniques are demonstrated and evaluated by application to two large knowledge- based systems, including a reformulation of QMR (Quick Medical Reference), a system for diagnosis in internal medicine; and a system for fault diagnosis and repair of gas turbines for power generation. These techniques are expected to lead to diagnostic expert systems that are easier to build and modify, that are more reliable, and that generate diagnostic test strategies that minimize unnecessary and expensive testing. IRI-9245136 Hull, Richard B. University of Southern California $81,224 - 12 mos. Theoretical Investigation of Information Sharing for Semantics and Object-Oriented Databases This is the second year funding of a two-year continuing award IRI- 9107055. This research is developing languages and mechanisms for restructuring and merging information from disparate databases, and is studying the theoretical foundations of such languages. The research uses the PI's formally defined ILOG languages, which permit the declarative specification of database restructuring. Certain technical properties of these languages provide a strong connection to logic programming. The research is also developing extensions to ILOG for data merging; a central problem here involves determining when object identifiers (OIDs) in two or more distinct databases refer to the same object "in the world", and developing appropriate mechanisms for "merging" the OIDs in this case. Specific theoretical questions under study include: attempting to characterize when two (restructuring or merging) programs are equivalent; exploring algebraic properties of programs (e.g., associativity, distributivity); developing algorithms for the translation of queries against restructured, merged databases; and for incrementally updating such databases if they are materialized. Experimental work will be performed to test the algorithms and view update methodologies under realistic conditions. The results obtained by the research will provide a crucial component for all database applications which involve data sharing. CCR-9243590 Kifer, Michael; Ramakrishnan, I.V. and Warren, David S. SUNY, Stonybrook $22,000 - 12 mos. (Jointly funded with the Software Systems Program - Total Award $154,732) Design and Implementation of a Higher-Order Logic Programming Language This is the second year funding of two-year continuing award CCR- 9102159. HiLog is an untyped logic with a higher-order syntax and a first-order semantics. This project extends the theory of Hilog, design a Horn-clause logic programming language based on Hilog, and develop an efficient implementation of this language. The goal is to develop a more powerful system than is currently available for declarative programming. HiLog permits variables over predicates and thus supports a more powerful programming and metaprogramming style. It also lets the programmer construct new predicate names using function symbols, which leads to a simple and clean methodology for modular programming in Hilog. However, the semantics of the new language is first-order so that an implementation of efficiency comparable to Prolog is achievable. The planned implementation uses newly developed indexing techniques to increase its efficiency. It supports a low-level (WAM-based) implementation of the extension-table evaluation technique to provide a complete logical evaluation of queries. The extension- table techniques extended to support a more complete implementation of negation in logic programs based on the well-founded semantics. The system will be a powerful prototyping tool that can help in the development of robust software more quickly. IRI-9243990 Klir, George SUNY, Binghampton $50,000 - 12 mos. Principles of Uncertainty and Information in Expert Systems This is the second year funding of a two-year continuing award IRI- 9015675. The principal objective of this research is to develop adequate principles for dealing with uncertainty in expert systems. The focus is on four relatively new mathematical theories of uncertainty: possibility theory, fuzzy set theory, random set theory, and the Dempster-Shafer theory of evidence. In each of these theories, as in probability theory, well-justified functions that measure uncertainty are now well established. These functions are crucial for the following three principles of uncertainty to be investigated: principles of maximum uncertainty, minimum uncertainty, and uncertainty invariance. According to the principle of maximum uncertainty, conclusions in any form of ampliative reasoning are determined by maximizing relevant uncertainty function subject to constraints that represent information available. The principle of minimum uncertainty is fundamental to a variety of problems involving unavoidable increase of relevant uncertainty, such as simplification or conflict-resolution problems. According to this principle, solutions are determined by minimizing relevant uncertainty function within constraints representing the requirements of each problem. The principle of uncertainty invariance is concerned with information-preserving transformations between distinct theories of uncertainty. This project is developing mathematical and computational foundation of the three principles as well as investigating their practical role in expert systems. IRI-9244411 Kumar, Akhil Cornell University $15,895 - 12 mos. RIA: A Model and Techniques for Integrating Rules and Data in a Common Framework in a Knowledge-Based System This is the second year funding of a two-year continuing award IRI- 9110880. The major strength of relational databases lies in their ability to process data in a set-at-a-time manner. An important drawback, however, is their inability to handle rules. An information system must allow users to define, represent and manipulate rules just as easily as data and treat them in a common framework. Such a system is called a knowledge-based system. The objective of this research is to investigate the feasibility of developing such a framework for integrating rules and data within a relational database system. This research develops and evaluates techniques for integration in this manner, and highlights the strengths and weaknesses of such an approach. Consequently, it focuses on the following issues: extending a relational language like SQL to express rules, storing and manipulating rules in an efficient manner, developing query processing algorithms (this will also involve designing new data structures), implementing the algorithms and evaluating their performance. This work draws to a large extent on existing literature and theory in the area of relational database systems, deductive databases, and also on existing work in the areas of artificial intelligence and expert systems. The findings will be useful for the designers of next- generation database systems. IRI-9210220 Lobo, Jorge University of Illinois, Chicago $30,000 - 12 mos. RIA: Updates in Deductive Databases This is the first year funding of a three-year continuing award. The process of updating a view in a relational database usually requires the translation of an update into updates over the underlying base relations. This process can be complex since the translation may not be unique. A generalization of the view update problem in relational databases is equivalent to the update of an intensional relation in a deductive database in which only the extensional database can be modified. The aim of this research is to design and implement update algorithms for stratified normal deductive databases by incorporating disjunctions in the extensional part of the database. The algorithms consider single insertions, single deletions and transactions. A fundamental structure for the implementation of the algorithms are interpretation trees that are used to compute answers to queries in disjunctive deductive databases. The research also addresses the issue of integrity constraints in disjunctive databases. So far, constraint satisfaction in disjunctive databases has been relatively unexplored; nevertheless, it has been shown that there exists a strong connection between disjunctive databases with constraints and the stable model semantics for normal logic programs. Hence, the development of algorithms to check integrity constraints will indicate directions to define procedures to answer queries in normal logic programs under the stable model semantics. IRI-9244474 McLeod, Dennis University of Southern California $81,945 - 12 mos. Time in Object Databases This is the second year funding of a two-year continuing award IRI- 9021028. The research project involves the design, development and experimental demonstration of a model that integrates time with object databases. Preliminary evidence suggests that existing notions of temporal data must be modified because of the transition from the relational to the object model. As such, this model is the first step towards the synthesis of an integrated object model that supports the temporal aspects of data modeling in addition to the structural and dynamic ones. Investigations to be undertaken include: (1) the study of the semantics of time and the development of a set of temporal principles to be used as an underlying basis for the design of our temporal object model; (2) the study of the temporal aspects of objects and the development of a model for integrating time with objects and the formalization of the model; (3) the augmentation of the data modeling power of object and semantic modeling concepts, particularly versions, with temporal modeling concepts; (4) the extension of an archetype extensible object database model to incorporate notions of time; and (5) the development of suitable user interfaces. Anticipated applications of these investigations are in engineering, business, medical information systems, scheduling, computer-aided design and computer-integrated manufacturing. IRI-9247324 Minker, Jack University of Maryland, College Park $104,993 - 12 mos. Artificial Intelligence Logic Programming and Deductive Databases This is the third year funding of a three-year continuing award IRI-8916059. The research covers theoretical investigations, experimental work, and practical techniques in logic programming and deductive databases. Research is conducted on extending well- founded semantics of logic programs to general well-founded logic programs and then to disjunctive well-founded logic programs. Their model theoretic, proof theoretic and fixpoint semantics are investigated. Proof procedures are planned for implementation. In deductive databases, alternative definitions of null values and to develop their model theoretic, proof theoretic and fixpoint semantic are investigated. Proof procedures that are developed are considered for implementation. In the area of disjunctive deductive databases, bottom-up and top-down computation methods are to be investigated, algorithms that are free of implications are studied, and finally the general case with implications is considered. Computing answers to negated queries and conditions under which one might reasonably compute answers in disjunctive databases are also investigated. In separate efforts, investigations into approximate answers to disjunctive logic programs and research in parallel inference systems are conducted. The research will be useful for the development of knowledge base systems and deductive databases in which the application requires disjuncts as part of the theory. IRI-9245820 Parker, Stotts D. University of California, Los Angeles $4,000 - 12 mos REU: Stream Databases This is a REU supplement to the award IRI-8917907. Ms. Claudia Amador, an undergraduate student, is developing software for the Stream Databases project. Specifically, she is implementing the SVP stream-set parallel data processing model and stream processing tools for the Bop system. This draws on her background in operating systems and artificial intelligence software, and gives her experience in both research and systems development. Today many applications deal with large quantities of data that take the form of a stream -- an ordered sequence of data items. Streams arise in many forms, including text files, genetic sequences, sound tracks, geological strata from drill holes, videotape, histories of sensor values, and so on. There is a serious need for a new approach in stream data processing. Today's database systems are not able to manage this data. The approach used in this project develops database systems from the established paradigm of stream processing, in which transducers (functional transformations) operate on streams. Under this approach databases are treated as streams, and queries or data analyzers are compositions of transducers that transform input data streams to output streams as needed. The objective of this research is to clarify the foundations of the resulting stream database paradigm. Important problems include adapting database concepts to the stream processing environment, characterizing useful stream pattern match queries, formalizing the optimization of stream database transducers, and finding the limits of the stream database approach. Beyond general uses in stream data processing, this research will have applications in DBMS/KBMS integration, database systems supporting data exploration, and a variety of special purpose database systems -- including systems that manage genetic sequences, event histories, and on-line data. IRI-9210332 Passino, Kevin M. Ohio State University $30,000 - 24 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $60,000) RIA: Modeling, Analysis, and Design of Expert Control Systems This award provides funding for the first two years of a three-year continuing award. Recently, there has been a significant increase in the use of techniques from applied Artificial Intelligence (AI) such as expert systems to implement the control functions for complex industrial processes. This research is showing that the design of such expert controllers (controllers implemented via expert systems) can be accomplished using the same design philosophy as that used for fuzzy controllers. Furthermore, to ensure that such expert controllers can be trusted in critical environments (e.g., aircraft control, spacecraft control, process control) this research is developing a discrete event system (DES) theoretic framework for the modeling and analysis of expert control systems. In particular, a mathematical model is introduced that can represent "rule-based" expert systems and a wide class of processes. Techniques from DES theory are being developed for the analysis of reachability (to study inference chains), cyclic behavior (to verify that the expert system will not get stuck in circular reasoning), and stability properties (to verify critical properties related to the safe operation of expert control systems). The application of the approach to the solution of a load balancing problem in flexible manufacturing systems and process control problems is being investigated. Furthermore, implementation issues for expert controllers are being investigated via the implementation of an expert controller for a flexible link manipulator. IRI-9249747 Ramakrishnan, Raghunath University of Wisconsin, Madison $62,500 - 12 mos. PYI: Declarative Database Programming Languages This is the third year base and second-year matching funding of a five-year continuing PYI award IRI-9057562. This work aims to develop a declarative paradigm of programming, based on flexible constructs that can deal efficiently with large volumes of data. Commercial query languages and logic programs both extend relational algebra, which is a simple fragment of first-order logic. These languages are declarative, in that their semantics can be specified non-operationally. Programs can thus be viewed as high-level specifications and are easier to develop, understand and maintain, and there is much promise for parallel execution. Current implementation techniques that support logic programming cannot handle large volumes of data, and techniques for relational query languages cannot implement recursion, to name two limitations. This work explores an approach based on program transformation followed by fixpoint evaluation that combines the focusing properties of logic program oriented techniques with the efficient set-oriented processing of relational techniques. In addition to the challenges of efficient implementation, many aspects of the declarative programming paradigm, such as the right combination of features, the role of data abstraction and object identity, the semantics of negation and updates, and facilities for program development and debugging are also being studied, and a prototype language called Coral is being developed. This research will yield techniques for integrated query processing and database design. IRI-9209029 Ross, Kenneth A. Columbia University $42,439 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $46,439) RIA: Declarative Features for Deductive Databases This is the first year funding of a three-year continuing award. Deductive databases provide a declarative framework based on first- order logic for writing queries. Entries in the database define the answer to a query without indicating how to answer the query. By allowing recursion, deductive databases yield higher expressive power than classical relational databases. The two main problems addressed in this project are semantics and optimization. This research studies semantics for deductive databases extended with aggregate operators such as "minimum" and "sum." Extensions of such semantics to programs with second-order constructs are also considered. The project is assessing how effectively second-order features can be incorporated into present-day relational databases. Optimization techniques for deductive databases with aggregates, both new techniques and extensions of existing strategies are also be investigated. This project will contribute fundamental knowledge about how to provide declarative features, such as aggregation and second-order constructs, within deductive databases. IRI-9120851 Rounds, William C. University of Michigan, Ann Arbor $28,819 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $57,639) Natural Language Techniques for Information Systems This is the first year funding of a three-year continuing award. This project is concerned with mathematical techniques from natural language processing systems - specifically, from unification-based grammar formalism - to help in the design of data and knowledge bases. These techniques include the use of feature theory, features logic, hyperset theory, and type systems from polymorphic functional programming languages. The goal of the project is to use these techniques to move knowledge and data models closer together, while providing for at least some aspects of heterogeneity in knowledge bases. IRI-9249350 Samet, Hanan University of Maryland, College Park $25,000 - 12 mos. (Jointly funded with Robotics and Machine Intelligence Program - Total Award $95,000) Spatial Data Acquisition and Processing This is the second year funding of a three-year continuing award IRI-9017393. The efficient processing of spatial data plays an important role in solving problems in computer vision, robotics, and computer graphics. The acquisition of spatial data as well as its processing is investigated. The acquisition is concerned with data useful for cartographic applications (e.g., terrains) and maps in general. In the case of maps, the interest is in the acquisition of relative spatial information (e.g., based on a map's legend) rather than precise information (e.g., locations of cities, roads, etc.). An investigation into the concept of a hypermap and the related representation issues is conducted. Other problems include the investigation of hierarchical surface representations, large spatial databases (especially those including line segments). Attempts are made to parallelize any algorithms that are developed and they will be tested on the Connection Machine. IRI-9244401 Su, Jianwen University of California, Santa Barbara $30,968 - 12 mos. RIA: Dynamic Aspects of Semantic and Object-Oriented Databases This is the second year funding of a two-year continuing award IRI- 9109520. In recent years, the need for developing techniques and methodologies for the design, representation and manipulation of transactions or behaviors has become clear. This research focuses on the study and development of theoretical foundations for object-based database manipulation languages. Several fundamental problems concerning dynamic constraints and update transactions in the context of semantic and object-based models are investigated. One aspect is the formal analysis of object behaviors under a behavior modeling construct used to compose and structure transactions in projects TAXIS, INSYDE, DAIDA, etc. The initial study is based on a simple object-oriented data model including class hierarchies, attributes ranging over printable values, and methods. Another topic is the study of behavior modeling primitives as new kinds of dynamic constraints and also compare them with other kinds of dynamic constraints. More general database models (e.g., to allow attributes ranging over classes) are also considered. The investigations into these problems yield a better understanding of these modeling constructs and insights into the impact of Object Identifiers on database manipulation languages. IRI-9242971 Ullman, Jeffrey Stanford University $94,927 - 24 mos. Research into the Design and Implementation of Logical Database Languages This is the second year funding of a two-year continuing award IRI- 9016358. The new generation of database applications, such as design databases and scientific databases, and integrated, heterogeneous databases, requires more powerful query languages, able to support declarative programming, ad-hoc queries, complex objects, and other features not found in conventional DB languages. The GLUE/NAIL language provides, for the first time, all these capabilities in one language. It consists of a "semideclarative" language, GLUE, which is essentially logical rules connected by conventional sequencing, procedures, and modules, and NAIL, a completely declarative "view facility" for GLUE. To make the language efficient, certain optimization techniques, such as "magic sets" and "regular recursion optimization" must be used, and they must be extended to handle certain other features. These include negation in subgoals, following the well-founded semantics, aggregation (e.g., sum, avg.), and second-order logic of support sets and other object-oriented capabilities. IRI-9102513 Van Gelder, Allen University of California, Santa Cruz $15,423 - 12 mos. (Jointly funded with the Programming Languages and Compilers Program - Total Award $30,423) The Well-Founded Semantics for Knowledge Bases This is the first year funding of two-year continuing award. Properties of the well-founded semantics for (possibly) unstratified knowledge bases are investigated for the purpose of implementing this semantics on a wide class of practical logic programs and expert systems. This project continues development of algorithms for automatically recognizing sets of rules that can be guaranteed to terminate with the appropriate execution strategy. The project also studies methods by which the compiler can optimize execution by discovery of determinacy in the rules. In the presence of determinacy, the correct choice among alternatives can always be made immediately, making backtracking unnecessary. The study focuses on the use of mode information and inductive type definitions. This project continues implementation of prototype analyzers, compilers, and interpreters, building upon and extending algorithms and software developed during prior research. A major difficulty today in building large-scale rule-based systems is choosing effective control strategies for the various subsets of rules, and current systems place virtually all of this burden on the programmer. An inferior choice can easily produce a nonterminating or extremely inefficient system. Progress in this research will facilitate the development of tools that automatically choose effective control strategies in most cases, and advise the programmer of unclear cases. IRI-9249260 Winslett, Marianne University of Illinois, Urbana $62,500 - 12 mos. PYI: Research on Reasoning about Change in Knowledge-Based Systems This is the fourth year base and third year matching funding of a five-year continuing PYI award IRI-8958582. The focus of this project is on formal and experimental approaches to the problem of reasoning about change in knowledge-based systems. Recently much formal work has addressed the problem of formal semantics for updating propositional theories, but little effort has yet gone into the problems encountered when trying to use these semantics in real-life applications. This project addresses those problems, concentrating on architectures for implementing formal update semantics for knowledge bases, and on extending semantics defined for propositional theories to the full first-order case. The extension to first-order logic requires means of updating beliefs held about equality, and also raises questions about the possibility of maintaining consistency of the knowledge base under the update semantics currently in vogue. To develop a satisfactory implementation architecture for the inherently exponential problem of propositional theory revision, one must find search pruning heuristics that lead to acceptable performance in typical update requests; this project is using examples from the domain of qualitative physics to ensure salability of the architecture. The potential applications of this work include semantics and implementations for use in planning, reasoning about the effects of actions, counterfactual reasoning, diagnosis, and belief revision. IRI-9103925 Yager, Ronald R. Iona College $30,069 - 12 mos. Possibility Theory in Expert Systems This is the first year funding of a three-year continuing grant. This research focuses on the development of a formal system, based upon possibility theory for the representation and manipulation of uncertainty in expert systems. The use of possibility theory allows a natural representation of linguistic information as well being an appropriate setting for handling partial matching. One emphasis of the research is the development of tools for the general representation of imprecise functional information. Here use is made of the possibility theory ability to interact with other knowledge representation formalisms such as default reasoning and probability theory. A second emphasis of this research is the development of a tool to help in the validation of knowledge-based systems containing uncertain information. Particular attention is paid to the discovery of potential inconsistencies in a knowledge base. The approach used is based upon the idea of reflecting back on the input. A very fundamental uncertainty principle is the specificity-correctness tradeoff. What this principle says is that in providing information we generally must make a tradeoff between being very specific and running the risk of being incorrect, or being unspecific and in turn assure ourselves of being correct. In most expert systems one desires both correctness and specificity. In this research the measure of specificity is extended to environments in which there exists an underlying similarity relation. The significance of this research is that it provides tools for builders of expert systems which can be used to formally and systemically include uncertainty in their systems. INFORMATION ACCESS ECD-9209623 Adrion, Richards W.; Lehnert, Wendy; Croft, W.B.; Lesser, Victor R. and Rissland, Edwina University of Massachusetts, Amherst $50,000 - 12 mos. (Jointly funded with the Industry/University Cooperative Research Centers Program - Total Award $300,000) State/University Cooperative Research Center for Intelligent Information Retrieval This is the first year funding of a four-year continuing award. In an increasingly serviced-oriented and information based economy, organizational decision makers must have immediate access to information. However, the technology needed to facilitate rapid, flexible, and effective access to this information has not kept pace with the dramatic advances in advances in hardware and storage technology. Furthermore, the significant advances in intelligent information retrieval have not been widely transferred to commercial applications. The State/ Industry /University Cooperative Research Center on Intelligent Information Retrieval at the University of Massachusetts at Amherst is addressing the scientific issues that are key to efficient access, management and control of information. The initial research projects are related to international commerce and medical information systems. The core research program includes four thrust areas: (1) information retrieval; (2) natural language processing, (3) case-based reasoning; and (4) control architectures. Eight companies and three medical institutions comprise the industrial partnership of the Center. The industrial members are involved actively in targeting new technology, identifying appropriate stages for moving research results out to the companies, and devising technology transfer implementation plans. The Center is collaborating with the Massachusetts Computer Science Education Consortium of five Massachusetts based educational institutions to implement a visiting faculty program, and in curricula development. IRI-9209025 Badal, Dushan Z. University of Colorado, Colorado Springs $50,000 - 12 mos. SGER: Image Signatures This project investigates image signatures which are compact representation of images. The research focuses on the investigation of different neural network architectures, such as backpropagation or cascade-correlation, for their capability to generate image signatures. The project is based on two ideas. One idea is to use the notion of atomic images (images containing one object) and their corresponding atomic signatures, as well as composite images (images containing more than one object) and their composite signatures. Another idea is to train neural networks on atomic images and/or their signatures and then to have such networks to recognize arbitrary compositions of atomic images and/or their signatures. The results of this research will form the foundations for an efficient access of images in databases and retrieval of images by their content. ISI-9160088 Bjorson, Robert D. Scientific Computer Associates $0 - 12 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $49,349 SBIR, Phase I: Developing Parallel Database Search Technology Management and search of large databases is a computational bottleneck in many domains. One attractive approach to eliminating it is to exploit recent advances in parallel computer hardware and processing. However, the development of software that permits this represents one of the major challenges facing computer scientists. This project proposes to develop a generic approach to parallel database management that should be especially suitable for large databases of complex objects in which even single object comparisons may be computationally costly. (Examples include databases of image data, in which database search may require numerous comparisons of complicated images.) This approach is based on the use of Linda, a commercial product of Scientific Computing Associates, Inc. Linda systems provide convenient access to parallelism from within familiar high-level languages such as C or Fortran. In Phase I, a fast text retrieval system is developed for a synchronous multiprocessors. This will serve as proof of concept for more general software. Due to the use of Linda, the technology developed will be not only efficient, but also portable across vastly different parallel architectures including both explicitly parallel computers and local-area networks of workstations functioning as parallel "hypercomputers". IRI-9249259 Faloutsos, Christos University of Maryland, College Park $62,500 - 12 mos. PYI: Access Methods for Multimedia Databases This is the fourth year base and third year matching funding of a five-year continuing PYI award IRI-8958546. This project examines storage and retrieval methods for non-traditional data types. (1) Geometric data: Methods for managing large collections of geometric objects (points, lines, rectangles etc) are designed and analyzed. Fractal-based transformations of such objects into points are considered, as they achieve clustering of geometric objects, thus reducing the number of (expensive) disk accesses. The most promising method has been implemented and it is being analyzed mathematically and subsequently compared to older competitors. Results in geometric objects management find applications in numerous areas such as in Geographic Information Systems (GIS), robotics, Engineering Information Systems (EIS). (2) Text and multimedia filing systems: The goal is to design and implement a document retrieval system. Several prototypes have already been implemented, using versions of signature files, as well as B-tree inverted indices. A separate module that handles animated images is currently being integrated with the text retrieval module. Signature methods are suitable for magnetic and optical disks, as well as for parallel machines (e.g., the Connection Machine); the animation module is useful for CD-I (Compact-Disk Interactive) applications, for CAI (computer-aided instruction), for electronic encyclopedias, multimedia electronic mail etc. IRI-9119446 Graefe, Goetz University of Colorado, Boulder $75,000 - 24 mos. and IRI-9118360 Maier, David Oregon Graduate Institute of Science and Technology $99,542 - 24 mos. The REVELATION Project: Query Processing in Behavioral Object- Oriented Databases This is an interinstitutional collaborative research of Goetz Graefe and David Maier. This award provides funding for the first two years of a three year continuing award. Many new data management systems are being developed to support applications with advanced data management requirements, such as extended relational models, persistent programming languages and object-oriented databases. Such systems enhance semantic expressiveness, but they must offer good performance to be viable commercial technologies. An area particularly needful of improved performance is query processing, where new modeling features such as encapsulation, type hierarchies, complex values and object identity make conventional set-processing methods difficult or ineffective to apply. The overall goal of the Revelation project is to expand query processing to address and exploit these new modeling extensions at all levels of query processing, from schema management, through optimization and physical planning, to runtime support. For encapsulation, a Revealer is incorporated, a trusted system component that can access type implementations. Heterogeneity in collections arising from subtyping and polymorphism is handled by an Annotater that reasons across schema definitions. Complex values and identity are addressed both at the logical algebra level, with new operations to deal with ordered structures such as sequences and arrays, and at the physical level with new operators such as one to assemble complex objects. This project will construct a prototype query processor, exploiting existing software technology such as the Volcano optimizer generator and the Volcano query evaluation system. IRI-9248907 Ioannidis, Yannis University of Wisconsin, Madison $62,500 - 12 mos. PYI: Database Support for Scientific Data This is the second year base and first year matching funding of a five-year continuing PYI award. Optimization of complex queries in future database systems is a major focus of this work. The applicability of randomized algorithms for query optimization is investigated for several types of database systems, e.g., relational, object-oriented, and parallel systems. Emphasis is placed on identifying the abstract characteristics of cost functions and data models that determine the effectiveness of such algorithms. Of special concern is also the propagation of errors in parameters that affect the query optimizer's decisions. This research leads to developing techniques for query optimization in future systems where current technology is inadequate. Another focus of this work is the study of specialized database systems for managing data from scientific experiments. The emphasis is on graphical user interfaces and semantic heterogeneity. Collaborating scientists from many disciplines provide guidelines for the goals of this effort based on the needs of their laboratories. This research develops tools that enable scientists to use database technology effectively for their experiments. Although this work is driven by the specific needs of scientific databases, its results can be applied to solve many similar problems faced by next generation database systems in general. IRI-9113736 Ioannidis, Yannis and Naughton, Jeffrey F. University of Wisconsin, Madison $79,716 - 12 mos. Optimization of Complex Queries This is the first year funding of a three-year continuing award. Significant progress in the optimization of complex database queries requires fundamental advances in three distinct but interrelated areas: (1) evaluation plan search algorithms, (2) query cost estimation, and (3) cost estimate error propagation. This project represents an effort to improve the state of the art in each of these three areas. The project itself is organized in three phases. The first phase consists of the analysis and performance modeling of search algorithms and sampling cost estimation strategies. The second phase generalizes the results obtained in the first phase to a wider variety of environments, including object-oriented queries and/or relational queries in parallel environments. Finally, the third phase explores the interactions between the results and techniques developed in the first two phases. These interactions are complex, and currently very little is known about them, yet they are critical to the performance of the optimizer. The anticipated results from the projects will handle the capabilities of conventional relational systems and will also provide the foundations for future developments in query optimization for deductive, object-oriented, and parallel database systems. IRI-9245627 Jain, Ramesh C. and Weymouth, Terry E. University of Michigan, Ann Arbor $127,826 - 12 mos. A Visual Information Management System: Image Databases, Knowledge and Visualization This is the second year funding of a two-year continuing award IRI- 9110683. This research extends query language and user-interface technology to include context-based queries for image data. A context is a domain dependent body of knowledge which encodes the translation between, on one hand, user-level semantics and object classes and, on the other hand, database syntax and entities. A prototype image information system provides a testbed for the investigation of issues in image indexing, similarity measures, and knowledge-guided incremental querying. A set of image processing and feature extraction routines supplies the base for image indexing; this base is augmented with a data model that accounts for objects, their images, the image features, and events in the world involving the objects. A knowledgebase and the associated user-interface support navigation within the data model to guide query formation and refinement. This research develops new concepts in data modeling, image indexing, and user-interfaces, which will enable users to access the image data in such applications as global monitoring and scientific imaging, based on concepts, context, similarity, and image features. IRI-9246263 Jain, Ramesh C. and Weymouth, Terry University of Michigan, Ann Arbor $4,000 - 12 mos. REU: A Visual Information Management System: Image Databases, Knowledge and Visualization This is a REU supplement to the award IRI-9110683. This REU supplement supports Ms. Jill Kliger's work on the conceptual and graphical design of a query mechanism for InfoScope. The demands of the InfoScope system include data in multiple media formats, tracking of objects over time, and example-based querying. Relevant issues include clustering, navigation vs. broad query specification, user-definable keywords, and pictorial query specification. In addition to query refinement, the undergraduate student is investigating visualization of results, including the integration of alphanumeric, sketch, image, and motion data. IRI-9241981 Salton, Gerard Cornell University $81,746 - 12 mos. Multi-level Processing of Natural Language Texts for Information Retrieval This is the third year funding of a three-year continuing award IRI-8915847. The existing methods for handling natural-language texts for information retrieval are applicable only to relatively homogeneous texts of approximately equal length and style covering related subject matter. The available methodologies for text analysis and indexing, text classification, text linking of related text excerpts, and text retrieval need to be extended to cover large heterogeneous texts, such as collections of messages of the kind now being circulated on electronic networks, and collections of encyclopedia articles. In these cases, the articles differ in length and subject matter, and often also in style and structure. Such articles may also include extraneous materials, such as quotations from other texts, and bibliographic citations and cross-references. Two main tasks are undertaken, consisting of an automatic text analysis and text linking system capable of recognizing text excerpts covering related subject matter. When similar text portions are appropriately linked, a network structure is produced for each text collection which then guides the user during the text retrieval operations by producing reading prescriptions in response to particular user queries consisting of chains of related texts or messages. The results obtained in this research are expected to lead to the implementation of new text handling systems in practical environments that are of interest to a wide variety of users. IRI-9244421 Salzberg, Betty J. Northeastern University $68,672 - 12 mos. Non-Traditional Access Methods This is the second year funding of a two-year continuing award IRI- 9102821. Database management systems efficiently organize, access and manipulate enormous quantities of data for traditional business applications. This efficiency is based on indexing records by one attribute, such as social security number. However, new applications in both science and business require indexing on several attributes. Astronomers and geographers want data to be organized spatially; businesses want to study trends over time. This project continues work on spatial and temporal indexes which have already been demonstrated to be efficient. A new node consolidation and concurrency algorithm will be applied to the Holey Brick tree, a spatial index with guarantees for space utilization and query speed. This will enable scientists to make sophisticated queries on spatial data while records are being inserted and deleted. The Time Split B-tree (a time-and-key based indexing system) will be enhanced by limiting the time-interval size and by managing the timestamping of records. Last, ways are being found to create indexes and do other reorganizations of data while the DBMS is online. The results of this project will aid in the management of anticipated non-traditional applications on massive collections of data in both science and in business. IRI-9245473 Shenoi, Sujeet University of Tulsa $30,000 - 12 mos. RIA: Designing Multilevel Relational Databases for Security Control This is the second year funding of a two-year continuing award IRI- 9110709. A well-designed database must balance security-control and operational efficiency. Consequently, the principal objective is to develop a design theory for multilevel databases unifying security (access-control and inference-control) and efficiency aspects in a dynamic environment. This project identifies the key constraints imposed by these aspects and formulates practical guidelines for database design. Emphasis is placed on data dependencies fundamental to inference-control and anomaly- elimination. The approach employed deals with a lattice-theoretic relational model enforcing database security by abstracting or clouding sensitive information in user views. The model uses "contexts" constructed from natural equivalences. Contexts induce multilevel relations capable of holding abstraction, and support a rich and highly secure data language. Contexts also incorporate semantics in abstraction; this helps balance user convenience and database security. The lattice-theoretic framework simplifies the investigation; classical design issues are also readily mapped to their multilevel formulations. The investigation also focuses on an Oracle implementation of the security-control model. Special consideration is given to implementing the query-handling facility. The results of this research will have a significant impact on the design of databases handling abstract information, and particularly those supporting AI and expert systems applications, as robust and efficient security-control is a critical issue facing database and knowledge-base designers. IRI-9244550 Shenoi, Sujeet University of Tulsa $4,000 - 12 mos. REU: Designing Multilevel Relational Databases for Security Control This is a REU supplement to the award IRI-9110709. This REU supplement is helping to support the Honor's Thesis research of an undergraduate student, Mr. James P. Oly. His work involves the design and implementation of abstraction-based query languages for security-control. Mr. Oly has a strong background in language design, having already completed courses in compiler construction and denotational semantics. The project is providing the student with valuable research experience and will better prepare him for his graduate studies and his ultimate research career. IRI-9211060 Sieg, John C., Jr. University of Massachusetts, Amherst $29,647 - 12 mos. RIA: Collaboration between Query Optimizers and Query-Plan Evaluators This is the first year funding of a two-year continuing award. New applications -- such as automatic knowledge discovery -- are making new demands on database systems. The constraints on collaborations between query optimizers and query-plan evaluators in conventional database systems severely limit the use of database systems for these new applications. The award supports research into unconventional kinds of collaborations between query optimizers and query-plan evaluators, including the migration of selected optimization decisions to run-time, the scheduling of multiple- implementation queries, and the migration of the use of prefetch hints from the query-plan evaluator to the query optimizer. The performance of the techniques is being extensively analyzed using an integrated rule-based optimizer and database system simulator. This research is expected to result in more efficient designs for a new generation of more flexible, high-performance database systems. IRI-9244405 Siegel, Michael and Madnick, Stuart E. Massachusetts Institute of Technology $93,427 - 12 mos. Identification and Reconciliation of Semantic Conflicts Using Metadata This is the second year funding of a two-year continuing award IRI- 9012189. With the development of complex information systems, the integration of heterogeneous systems, and the availability of numerous online computer data sources, it has become increasingly important that methods be developed that consider the meaning of the data used in these systems. For example, it is important that an application requiring financial data in US dollars does not receive data from a source that reports in another currency. This problem is further complicated by the fact that the meaning of data may change at any time. To deal with this problem, information systems must have the ability to represent data semantics and detect and automatically resolve semantic conflicts. This research focuses on the representation and manipulation of data semantics (i.e., metadata). This research develops: (1) a rule-based approach to metadata specification; (2) query processing techniques that identify semantic conflicts; and (3) methods that use conversion routines to automatically resolve these conflicts. When applied to source-receiver systems, these techniques create a dynamic system environment that permits changes in data semantics. Similarly, in heterogeneous information systems these techniques define a dynamic schema integration environment where the integrated system can continue to operate as the component systems independently modify data semantics. IRI-9249669 Silberschatz, Abraham University of Texas, Austin $90,236 - 12 mos. Parallel Evaluation of Deductive Database Queries This is the second year funding of a two-year continuing award IRI- 9106450. This project concerns parallel evaluation of deductive database queries by processors interconnected by a communication network. Deductive database systems are key to the future design of complex applications like knowledge bases, expert systems, and scientific applications. Yet the potential of such systems is compromised by poor performance in environments where the data cannot fit in main memory. Parallelization is investigated as the means to improve performance, with attention on Datalog, a logic- based deductive language that naturally expresses queries that arise in advanced database applications. Based on three parameters of discriminating functions, discriminating variables and hash functions, a framework for the parallelization of Datalog programs is designed. Discriminating functions and variables which establish only required interconnections between parallel processors, thereby minimizing the overhead of unnecessary communication, are investigated. Hash functions are designed which capture the division of labor between processors and provide maximum performance by effectively balancing the workload. Although parallel systems promise enormous computing resources, their performance rests on the design of sophisticated parallelizing compilers that can make effective use of such resources. This research makes a pragmatic contribution to parallel system performance by providing an approach to the automatic parallel evaluation of an important subset of deductive database programs, Datalog. IRI-9244413 Srivastava, Jaideep University of Minnesota $19,638 - 12 mos. RIA: Query Optimization of Parallel Relational Databases This is the second-year funding of a two-year continuing award IRI- 9110584. This research addresses the problem of relational query optimization for general purpose parallel machines. The principle objective of this project is to develop a query optimizer that will generate efficient query plans for parallel machines. The three main components of a database query optimizer are its query plan representation, its cost model, and its search algorithm. This project addresses the issue of developing a plan representation that incorporates intra- and inter-operator parallelism, as well as pipelining. Furthermore, cost models will be developed which distinguish between the total work done, i.e., the (sequential) time it would take on a uniprocessor and total (parallel) time taken on a multi-processor. The cost model depends on the query and database parameters, as well as architecture parameters. A cost- based search algorithm selects efficient (low cost) query plans. Since the search space is much larger than that found in uniprocessor query optimization, this research will develop new heuristics to prune the search space. Furthermore, since a multiprocessor is available, applicability of parallel processing to the optimization process itself is also investigated. The results of this research will have an impact on efficient query processing in parallel relational databases. IRI-9241259 Stonebraker, Michael; Rowe, Lawrence A. and Katz, Randy H. University of California, Berkeley $142,000 - 12 mos. Database Support for a Three Level Store This is the second year funding of a three-year continuing award IRI-9107455. Traditionally, DBMSs have assumed that all accessible data resides on magnetic disk with small amounts of data occupying space in a main memory cache. Future object managers will be called on to manage very large object bases in which time critical objects reside permanently in main memory, other objects are disk resident, and the remainder occupy tertiary memory. Moreover, there may be more than three levels of storage present, and some of them may exist at remote locations. This project investigates the architecture of a DBMS that can support a multi-level store in a single software system. The approach is to generalize and apply ideas from file systems, query optimization, and abstract data types to multiple levels of storage. In addition, since storage may exist on multiple computer systems, ideas from distributed data bases must also be applied. The significance of this research is that the issue of massive size is addressed. Specifically, a new generation of DBMSs is unfolding that will support more general applications than the business data processing ones addressed by relational DBMSs. The next frontier is to overcome the dramatic size increase that these new applications entail. IRI-9246432 Stonebraker, Michael R.; Rowe, Lawrence A. and Katz, Randy H. University of California, Berkeley $4,000 - 12 mos. REU: Database Support for a Three Level Store This is a REU supplement to the award IRI-9107455. Two undergraduate students are gaining valuable experience in participating in a large, complex multi-person development team. POSTGRES is one of the few such projects, and previous undergraduate participants have obtained valuable skills for performing "software development in the large". They are also expected to participate in the POSTGRES research group, and get a good look at how research in database systems is carried out. IRI-9244412 Tsotras, Vassilis Polytechnic University $29,350 - 12 mos. RIA: Support of Historical References in Database This is the second year funding of a two-year continuing award IRI- 9111271. Supporting historical references is an important problem in several areas of computer science and engineering. The significance of the database versioning problem (i.e., keeping and accessing old versions) has been particularly mentioned in recent advances in the database community. In the rapidly growing area of object oriented programming, historical references enable programmers to create new objects based on previous object versions. Traditional approaches to supporting historical references require either large space or long reconstruction times, making the extensive use of such references prohibitive. This project provides efficient support of historical references, i.e., fast reconstruction to any past state without sacrificing a large amount of storage space. New, optimal ways are investigated to compress temporal data, while still providing practically random access to any past reference on these data. Moreover, the problems of distributing the history, keeping the history of a system that evolves like a graph, keeping histories in limited space and other general historical queries (not indexed by time but instead using historical correlation) are analyzed. The results of this project will impact the use of the recorded past history for different programmers or database users. Other applications of these results to be considered, include network management tools, animation storing and archiving the UNIX directory system. IRI-9111988 Yu, Clement T. University of Illinois, Chicago $65,999 - 12 mos. Query Optimization in Heterogeneous Environments This is the first year funding of a two-year continuing award. Currently, users have difficulties in retrieving information which is stored in different types of database systems. Most users are not familiar with different query languages and may not be aware of the locations of data. The objective of this project is to facilitate query formulation and to speed up query processing in heterogeneous database environments, i.e. where different types of database systems, specifically, relational, object-oriented, deductive, and textual systems co-exist in an organization or across different organizations. To enable a user who is familiar with only a single query language to access information stored in different types of database systems, translators of query languages are being built. An intelligent data dictionary is constructed to facilitate schema integration and query translation. Query processing performance is optimized by application of logic-based methods, using knowledge about the types of queries previously processed. Parallel processing techniques across different systems are employed to provide further speedup. The resulting heterogeneous query processing system will find wide applicability as the networking of different database systems becomes widespread. PHYSICAL AND SYSTEM ASPECTS CCR-9242928 Bernstein, Arthur J. SUNY, Stonybrook $25,000 - 12 mos. (Jointly funded with the Software Systems Program - Total Award $86,329) High-Throughput Distributed Database Systems This is the second year funding of two-year continuing award CCR- 9101524. The objective of this research is to develop concurrency control algorithms for transactions in distributed database systems which achieve high throughput and low response time. Data replication is a mechanism that has been proposed for this purpose. Unfortunately, the requirement that transactions behave serializably forces considerable synchronization among replica sites and hence adversely impacts throughput and response time. This work will consider the tradeoff between serializability and throughput/response time by allowing bounded violations of the database's integrity constraints. Such violations occur when conflicting transactions are not synchronized. Under these circumstances a transaction is ignorant of the effects of some prior transactions, and so the results it returns and the modifications it makes to the database are based on incomplete information. Both the deviation from serializable behavior and the throughput/response time are directly related to the extent of ignorance. Concurrency control algorithms which bound the extent to which a transaction can be ignorant of conflicting transaction will bound the deviation from serializable behavior. Since violations of integrity constraints are not permissible in all applications, the research involves study of the characteristics of applications in which violations can be tolerated. IRI-9245687 Carey, Michael University of Wisconsin, Madison $62,500 - 12 mos. PYI: Extensible Database Systems with Improved Performance This is the fifth year base and fifth year matching funding of a five-year continuing PYI award IRI-8657323. This research is focused on improving database system performance, particularly for distributed databases, distributed transaction management and real- time database applications. The performance of concurrency control algorithms, distributed join algorithms and load balancing techniques for distributed database systems are all investigated, in addition to techniques such as main memory databases for achieving high performance. Another major aspect of the research is the development of extensible database systems, including the necessary storage management facilities and programming tools. The EXODUS system, under development with this research, is intended to support the rapid implementation of database management facilities for database applications that are poorly served by current relational database systems. The EXODUS group is building "database system generator" software, including a set of versatile kernel database facilities and software tools to simplify the construction of customized database management systems. Research on distributed database systems and their performance is vital for guiding the designers of future database systems. While relational database systems are now a commercial reality, and their distributed counterparts are starting to become available, significant challenges remain for understanding and improving their performance. Extensible database systems such as those investigated here are essential to go beyond relational databases and support a wide range of emerging database applications: computer-aided software engineering, scientific data management, image/voice data management, and data-intensive AI applications. The success of large-scale efforts in these areas will depend on flexible, high-performance data management facilities such as those developed in EXODUS project. IRI-9201643 Eich, Margaret H. Southern Methodist University $60,000 - 12 mos. and IRI-9201596 Gruenwald, Le University of Oklahoma $60,000 - 12 mos. Main Memory Database Recovery Issues This is an interinstitutional collaborative research award of Margaret Eich and Le Gruenwald. This is the first year funding of a three-year continuing award. The increasing size and decreasing cost of semiconductor memory has prompted research into databases which are memory resident. These Main Memory DataBase (MMDB) systems are aimed at high throughput applications such as airline reservation systems, phone switching databases, and other real time systems where the availability of the memory resident data is crucial. To achieve this high availability with volatile RAM requires a backup archive database on disk as well as efficient algorithms to checkpoint the database to the archive and to recover it from the archive to the main memory after a system failure. It has been shown that MMDB systems often perform better with deferred update techniques where data to be updated is first placed in a special nonvolatile shadow area and only at transaction commit time is placed in main memory. This project investigates both checkpointing and reloading of MMDB databases with the use of more conventional immediate update (IU) techniques as well as deferred update (DU) strategies. The best checkpointing approaches are to be determined as are partial reloading strategies. A partial reload allows the database to be brought online after a system failure faster because not all of the MMDB is reloaded prior to bringing the system up. This research represents the first examination of techniques for partial reloading of MMDB systems. Due to the volatility of RAM, the high throughput needs of MMDB applications, and the potential for partial reloading to dramatically increase the uptime of database systems, the results of this research will be applicable in a variety of real-time databases. IRI-9244402 Elkan, Charles University of California, San Diego $30,904 - 12 mos. RIA: Flexible Distributed Concurrency Control by Reasoning about Transactions This is the second year funding of a two-year continuing award IRI- 9110813. Increased intelligence and flexibility in a distributed database system are achieved in this project using new automated reasoning algorithms. In particular, concurrency control is done by logical analysis of updates and queries to determine whether they conflict. The algorithms for reasoning about query and update conflict are based on a comprehensive theory of query and update independence developed recently by the principal investigator. The specific focus of the project is the design and implementation of a new database system that is fully distributed in the sense that neither concurrency control nor any other function are necessarily performed at a central site. Fully distributed concurrency control is achieved by explicit agreement between sites as to which transactions may proceed, which should be delayed, and which should be restarted. Explicit common knowledge of active transactions allows the system to adapt heuristically to changing patterns of access to the shared database. This explicit common knowledge is achieved by syntactic and semantic reasoning about queries and updates at individual sites which then execute a distributed consensus algorithm to share their conclusions. IRI-9248814 Elmagarmid, Ahmed K. Purdue University $62,000 - 12 mos. PYI: Heterogeneous Distributed Database Management Systems This is the fifth year base and fourth year matching funding of a five-year continuing PYI award IRI-887952. The InterBase project addresses problems inherent in an environment consisting of distributed, autonomous and heterogeneous database systems. Such an environment is often the natural result of the shifting priorities and needs of an organization as it acquires new DBMSs that are designed independently and often run over differing operating systems and hardware platforms. Although each new system may facilitate a short term requirement, it increases the complexity of those global applications that access data and services from several different systems. The solution proposed by the InterBase project encompasses the design and implementation of a global heterogeneous database interface and the study of fundamental issues in consistency, correctness criteria, transaction models and algorithms for the proper execution of transactions. In the last year, a prototype of InterBase that integrates Sybase, Ingres, Guru, Dbase IV, and various other non-database systems has been built. In addition, a new more flexible correctness criterion for concurrency control that can be used in this environment without violating local autonomy of the underlying systems has been designed. The effects of autonomy and its meaning on such systems was studied. A new transaction model that is more suited for this environments has been developed. This year, the Principal Investigator and research assistants focus their attention on consistency, replication and the generalization of quasi serializability. These concepts are expected to be incorporated into the current prototype in order to demonstrate their feasibility. IRI-9258362 Ghandeharizadeh, Shahram University of Southern California $25,000 - 12 mos. NYI: Physical Database Design in Parallel Next Generation Database Management Systems This is the first year base of a five-year continuing NYI award. A fundamental concept in database management systems (DBMSs) is that of physical data independence, i.e., the ability to modify the physical database scheme without requiring a rewrite of the software that implements the DBMS or the application programs that use the data. This concept enables a DBMS to support multiple applications and at the same time optimize the physical design of each database to provide each application with the highest performance available. This research focuses on alternative implementations of database management systems and investigates: (1) physical data independence; (2) tools and techniques that can optimize the physical design of data for a given application; and (3) techniques for dynamic reorganization of data in the presence of changing patterns of data access. The main objective is to maximize the performance of a system, and whenever appropriate, to utilize parallelism. Some of this research has been conducted in the context of the relational data model. This research project concentrates on the object-oriented data model and other current and emerging data models (i.e., a family of logical constructs for representation and manipulation of data). This research is expected to significantly enhance the performance of the current and next generation DBMSs. IRI-9244410 Ghandeharizadeh, Shahram University of Southern California $30,141 - 12 mos. RIA: Physical Data Design in Multiprocessor Database Systems This is the second year funding of a two-year continuing award IRI- 9110522. Machines during the past decade, parallel database systems have gained increased popularity due to their high performance, scalability, and availability characteristics. With the predicted future database sizes and complexity of queries, it is essential that these systems effectively utilize hundreds or thousands of processors. Several studies have repeatedly demonstrated that both the performance and scalability of a parallel database system is contingent on the physical layout of the data across the processors of the system. Recently, the design of a new declustering strategy, termed Multi-Attribute GrId declustering (MAGIC), was introduced. MAGIC is superior to the current declustering strategies because it can partition a relation using two or more of its attributes. In addition, it minimizes the overhead of using parallelism to allow the system to scale to a large number of processors. This project focuses on: (1) analysis and extension of MAGIC for executing complex queries; and (2) incorporation of algorithms for dynamic on-line reorganization of a relation in the presence of update queries and changing workload characteristics. The approach involves development of analytical and simulation models first, followed by a prototype implementation. This research enhances the scalability of a parallel database machine and improves its performance for executing complex queries. IRI-9248906 Naughton, Jeffrey University of Wisconsin, Madison $62,500 - 12 mos. PYI: Theory and Implementation of Database Systems This is the second year base and first year matching funding of a five-year continuing PYI award IRI-9157357. The long-term goal of this project is to increase the performance and functionality of database systems, with a special emphasis on parallel database systems. Three specific topics currently being addressed are clustering/declustering of data on secondary storage, probabilistic methods in query cost estimation and query evaluation, and transparent checkpointing techniques in parallel persistent programming environments. The research consists of theoretical studies to gain an understanding of the key problems involved, followed by analytic studies to compare and contrast alternative solutions to these key problems, followed by prototype implementations of the most promising solutions. IRI-9247505 Ouksel, Aris University of Illinois $4,000 - 3 mos. REU: Transaction Management in Interpolation-Based Files This is a REU supplement to the award IRI-9010365. The undergraduate student is participating in development of concurrent algorithms for spatial structures and is investigating the portability and the impact of the new system on currently existing commercial systems. The student is gaining a valuable research experience in current concurrency as well as recovery issues in databases systems, while enhancing the scope of the research project. This project implements mechanisms for cooperation and communication within and between complex database transactions with the aim of increasing the degree of parallelism as well as the level of recovery control during failures. A transaction is defined as a partially ordered set of primitive database operations. The approach exploits the spatial properties of a multidimensional data structure, the Interpolation-Based Grid File (IBGF), which represents objects -- points, boxes, or otherwise. It provides a capability to learn important access path information from live processes, prior to accessing the data, and a technique to allow sharing of this information among processes. Throughput will be greatly improved by using this capability to anticipate potential conflicts and to reduce access redundancy. This research develops servers, responsible for any transaction on the IBGF structure, which incorporate this access path information passing capability. A transaction manager is built on top of IBGF servers. Methods developed by other researchers are modified to take advantage of IBGF spatial properties to improve performance. The portability of the results to a commercially available system such as INGRES is considered. The results provide tools for the design of advanced transaction systems. IRI-9209406 Ramakrishna, M. V. Michigan State University $30,582 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $33,082) RIA: Analytical and Practical Performance of Hashing Functions This is the first year funding of a two-year continuing award. Hashing techniques are used extensively in databases and file structures, text retrieval systems, and in hardware. This research is addressing the fundamental question of the relationship between the results of probabilistic analysis of hashing techniques and their practical performance. The goals include investigating: (1) if and how the analytical performance can be achieved in practice; (2) the effect of different distributions of the key sets; and (3) application of hashing and perfect hashing to hardware implementation. Preliminary results obtained thus far are promising and indicate that the analytical performance can indeed be achieved in practice. These fundamental results are useful in other areas such as in signature files for text searching and parallel hashing. By addressing the long open question in the area of hashing, this research enables the users of hashing techniques to have a formal basis for file design and confidence in the expected performance of the resulting file structures. IRI-9244478 Ramamritham, Krithi University of Massachusetts, Amherst $65,000 - 12 mos. A Framework for Studying Extended Transaction Models This is the second year funding of a two-year continuing award IRI- 9109210. Although powerful, the transaction model adopted in traditional database systems is found lacking in functionality and performance when used for applications that involve reactive, long- lived and collaborative activities. The following are some of the issues that arise in the context of the extended transaction models that have been proposed to address these shortcomings: (1) correctness properties of a model vis a vis visibility, consistency, recovery, and permanence (e.g., traditional transactions guarantee failure atomicity, serializability, and permanence.); (2) similarities and differences between two transaction models; (3) constraints on using two models in conjunction; and (4) mechanisms needed for managing extended transactions. These issues motivated the development of ACTA, a comprehensive formalism which characterizes the effects of transactions on each other and on objects in the database. ACTA's ability to deal with existing extended transaction models and correctness notions is indicative of its generality. The need for extended transactions emerges from the demands of applications, such as CAD/CAM, software development environments, object-oriented databases, and distributed operating systems. By providing a formal framework in which to capture the properties of extended transactions, and thereby identifying the needed transaction management mechanisms, this work is producing results that will impact these information-intensive application areas. IRI-9212087 Sechrest, Stuart University of Michigan, Ann Arbor $29,014 - 12 mos. RIA: User-Level Physical Memory Management This is the first year funding of a three-year continuing award. A machine's physical memory is a resource for which many subsystems, including databases, persistent object storage managers, and multimedia file systems, are potential competitors. As memory size grows, optimizing the diverse uses of physical memory will require diverse management policies. These policies must, however, coexist, and the claims on memory for different uses must be balanced. This research investigates a system architecture for balancing the requirements of memory competitors, while allowing a greater variety of management policies to be implemented. The system architecture permits pageout daemons situated outside the kernel to control pools of physical memory. This allows subsystems such as databases much greater control over memory management policy. A balance manager within the kernel controls the size of these pools. A simulator is used to explore the interactions of management policies, while a prototype system, supported by an extended version of Mach 3.0, allows the experimental validation of the simulation results. Successful completion of this project will provide a framework for managing large physical memories in future machines. Using the balancing strategies identified in this research, subsystems for memory- intensive application will be more readily developed and installed on new machines without ad hoc performance tuning. IRI-9247758 Sellis, Timoleon University of Maryland, College Park $62,500 - 12 mos. PYI: Research in Query Optimization, Active Databases, and Learning Mechanisms for Relational Database Management Systems This is the third year base and second-year matching funding of a five-year continuing PYI award IRI-9057573. This project focuses on research in three areas: query optimization, active databases, and learning mechanisms for relational database management systems. Query optimization is necessary to provide fast access to large databases; in this project, techniques that allow the optimization of multiple queries, thus providing higher throughput for the system are developed and implemented in a prototype system. In contrast to the traditional single command-at-a-time optimization, a multiple-query processing strategy avoids redundant page accesses by accessing data common to more than one query, only once. The research on active databases investigates methods that allow the specification, efficient monitoring and processing of triggers in relational systems. Such methods are essential in contemporary systems and allow the database management system to "react" to changes performed. The use of machine learning is explored at the buffer management and query optimization levels in order to improve the performance of database systems. In particular, schemes that will adapt the above sub-systems based on the query and user environment, are investigated. The contributions of the research in the above three areas lie in improving performance of relational database systems in contemporary applications, such as deductive and engineering databases. IRI-9114197 Stankovic, John A.; Towsley, Don F. and Ramamritham, Krithi University of Massachusetts, Amherst $99,998 - 12 mos. Time Constrained Databases This is the first year funding of a three-year continuing award that is a renewal of a two-year award IRI-8908693. Real-time database systems are needed in many important applications such as program trading in the stock market, telecommunications, computer integrated manufacturing, advanced process control, intelligent highway systems, and various multimedia applications. Most current real-time database technology consists of fairly simple extensions to basic database protocols and is only applicable to soft real- time applications. This work extends real-time database technology to more sophisticated levels by utilizing an abstract data type model, nested transactions, distributed systems, and both soft and hard real-time constraints. The work compares the use of an active database paradigm with the more traditional approach. Integrated solutions across concurrency control, commit protocols, CPU scheduling, I/O, deadlock resolution, and communication are stressed. The research approach is largely experimental and uses testbeds built with prior NSF awards. Results from this project can create more efficient, profitable, and safer real-time database applications. IRI-9247187 Wehrmeirster, Robert M. Data Parallel Systems, Inc. $0 - 12 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $127,102) SBIR, Phase II: Single Instruction Stream Multiple Data Stream (SIMD) Parallel Algorithms for Nested Relational Database Implementation A class of massively parallel computers has been introduced over the last several years in research projects and as commercial products. These computers, designated as Single Instruction Stream - Multiple Data Stream (SIMD), offer an increase in performance and a significant improvement in cost/performance. Applications that can exploit this high degree of parallelism can enjoy the economic benefits of this cost performance breakthrough. The majority of the initial applications on these machines is in the hard sciences and engineering fields. However, the Phase I results showed that these machines can be used for fast and efficient implementations of relational and post-relational database technology. The company is working with a leading SIMD computer manufacturer, MasPar Computer Corporation to integrate our Phase I results into a complete database solution. Phase II tasks focus on implementation of a complete and demonstrable prototype. The same cost/performance metrics that make this product a strong competitor in the commercial database market, lead to the next generation of database technology that allow for storage of complex data objects. This post-relational database model allows for object-oriented database implementation in diverse fields such as CAD/CAM systems, genetic mapping, and image analysis applications. SYSTEM DEVELOPMENT AND ADMINISTRATION IRI-9117095 Barnes, Julie Bowling Green State University $29,998 - 12 mos. and IRI-9117089 Mamrak, Sandra A. The Ohio State University $30,000 - 12 mos. The Role of the Intermediate Form in Data Translation This is an interinstitutional collaborative project of Julie Barnes and Sandra Mamrak. In the data translation arena, there is considerable speculation about the relative merits of the pairwise model of data translation versus the intermediate-form model. If translation is desired between some number of specific encoding schemes, the pairwise model of translation requires that translators be written between every combination of the encoding schemes, taken two at a time. The intermediate-form model requires that a new, generic encoding scheme be used as an intermediate vehicle. Then, for each encoding scheme, two translators need to be written, one to and one from the intermediate encoding scheme. These can then be composed to form a translation bridge between any two specific encoding schemes. This project proposes two experiments to evaluate these models of data translation. The first experiment is designed to compare the quality of the translators in each case. The second experiment is designed to compare the effort required to generate the translators in each case. The ultimate goal of the work is to provide general, definitive guidelines for choosing a model of data translation. These guidelines will allow the myriad practitioners who do data translation to optimize their efforts by choosing the model that best supports their individual translation environment. IRI-9210588 Chrysanthis, Panos K. University of Pittsburgh $28,700 - 12 mos. RIA: Understanding Autonomy and Consistency in Multidatabase Systems This is the first-year funding of a three-year continuing award. Multidatabase systems respond to the need of organizations to support applications that require information maintained by several autonomous and possibly heterogeneous database systems. A fundamental requirement underlying multidatabase systems is the preservation of the autonomy of the individual database systems. However, autonomy requirements cause problems in maintaining traditional database consistency in these systems, the reason being that autonomy and consistency are conflicting requirements. Attempts to strike a balance between these two requirements have focused on new correctness criteria. However, very little analysis has been done with respect to (1) the semantics of autonomy requirements and (2) their implications for both consistency requirements and implementation considerations. The goal of this research is to systematically analyze autonomy and consistency by applying a uniform specification technique to express autonomy requirements and correctness criteria, thereby explicitly showing their relationships and identifying the tradeoffs between them. By understanding these tradeoffs and their effects on different parts of a database system, this work will help in the development of multidatabase systems and will have an impact on the integration of other systems in which autonomy and consistency are requirements. IRI-9210200 Davis, Karen C. University of Cincinnati $35,000 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $40,000) RIA: Specification and Update Semantics of Views for CAD Databases This is the first year funding of a three-year continuing award. This research addresses computer-aided design (CAD) data modeling and view construction, as well as updating both materialized views and the underlying database in support of design activities. Database support provides a means to share data between CAD tools and multiple designers; however, only a small portion of the contents of a database may be germane to a particular design application. A copy of the relevant data (a materialized view) may be stored and modified locally during a design step. Maintaining consistency between a database and views of it entails ensuring that changes to the database are reflected in the view data, and vice versa. Database view technology to efficiently and correctly support extraction and manipulation of relevant views of design objects will be developed, including a formal CAD data model and an extended object-oriented query algebra for view specification. Algorithms for view maintenance will be derived from the formal semantics. Database support for CAD offers opportunities for concurrent design by simplifying the maintenance and sharing of technology-related data. View processing provides a conceptual framework for unifying a CAD software environment; both the input and output of a tool are views of a design. This research seeks to facilitate interoperability of tools and cooperation between designers via CAD database view technology. IRI-9257578 Musen, Mark A. Stanford University $25,000 - 12 mos. NYI: Design of Knowledge-Based Systems from Reusable Components This is the first year base of a five-year PYI continuing award. The construction of knowledge-based systems is extremely labor-intensive. Unfortunately, current tools for building knowledge-based systems do not allow developers to take significant advantage of previous knowledge bases when building new systems. The research supported by this award involves design of a novel architecture for the construction of knowledge-based systems. The architecture allows system builders either to select an existing problem-solving method from a library, or to construct a new problem-solving method from a set of predefined building blocks. The developer then maps the data requirements of the problem-solving method to elements of reusable domain ontologies that define models of particular application areas. Both the domain ontologies and the problem-solving methods that are used to build knowledge-based systems are stored in libraries that future developers can access when implementing new applications. The research thus addresses strategies to index and retrieve problem-solving methods and domain otologies, and to integrate these components within new problem-solving frameworks. The methodology developed in this project allows construction of new classes of tools for building knowledge-based systems and defines an infrastructure that supports the reuse of knowledge-base components at increasingly high levels of abstraction. IRI-9209252 Storey, Veda C. University of Rochester $28,727 - 12 mos. RIA: Commonsense Reasoning in Database Design Systems This is the first year funding of a three-year continuing award. Human database designers apply their accumulated experience and commonsense reasoning abilities to each new design task. Recently, a number of expert systems have been developed that attempt to automate the database design process. Although these systems may have a high degree of expertise in database design, they typically know very little about anything else. As a result, they often have to ask questions that appear unnecessary or trivial, thus detracting from their credibility as experts and increasing the effort required of the user. The objective of this research is to add commonsense reasoning capabilities to a proven database design expert system. The commonsense reasoner will employ a simple, generic model of the business environment that it can extend semi- automatically with facts "learned" during each design session. The project involves conceptual work on the nature and representation of commonsense knowledge as the basis for a later implementation. The implemented system will be tested to assess the value of the commonsense facility. This research is expected to aid in designing computer systems that combine a high degree of expertise in a particular subject with general knowledge about the world at large and have the ability to apply both to solving problems. IRI-9257293 Yen, John Texas A&M University $25,000 - 12 mos. NYI: Using Fuzzy Logic to Deal with Qualitative Requirements and Uncertainty in the Environment This is the first year funding of a five-year continuing NYI award. Intelligent systems often need to deal with two kinds of uncertainty: (1) system requirements that are qualitative in nature; and (2) uncertainty about the state of the external environment. The primary objective of this research is to develop sound and practical techniques for dealing with these issues. To address the first issue, fuzzy-logic-based methodologies for specifying and validating qualitative requirements are being developed. Explicitly capturing the elasticity of the system's requirements facilitates the exploration of various trade-offs during the design stage and enables a more realistic validation of the implemented system. To address the second issue, systematic modeling techniques for designing hybrid autonomous intelligent systems are being developed. These techniques use fuzzy logic to integrate AI symbolic problem solving with the numeric processing exhibited by neural networks and model-based control. Potential industrial applications of such hybrid systems range from the petrochemical process control to autonomous vehicle systems and automated manufacturing systems. Methodologies and techniques developed in this research project will not only enhance the quality and the adaptability of the next generation of intelligent systems, but will also reduce the cost for designing and maintaining them. SCIENTIFIC DATABASES IRI-9117030 Baclawski, Kenneth; Salzberg, Betty J.; Futrelle J., Robert P. and Pescitelli, Maurice J. Jr. Northeastern University $102,383 - 12 mos. (Jointly funded with the Database Activities and Computational Biology Program - Total Award $204,383) SDB: Data/Knowledge Bases for Biological Papers and Techniques This is the first year funding of a three-year continuing award. The Biological sciences produce a complex research literature equivalent to thousands of megabytes every year. Research papers have extensive and detailed information about procedures, information not captured in keywords or tables. New techniques are developed for storing and querying this data and knowledge. This project combines the ongoing work in the Biological Knowledge Laboratory, which is amassing and codifying biology papers, with new technology for managing knowledge and databases. The focus is on the logical text structure (title, abstract, sections, paragraphs) and the Materials and Methods sections (chemicals used, genetic strains, equipment, procedures). The text structure can be treated as a schema in an object-oriented data model. The Materials and Methods database is a complex knowledge base involving time- ordered events, computed rather than stored attributes, references to methods in other papers, etc. The techniques and concepts of object-oriented databases are used for both types of information. Innovative indexing schemes are developed to aid in search. This new object-oriented database technology advances the understanding of the structure of scientific knowledge and data and provides new tools for research scientists to access the complex knowledge contained in the research literature. IRI-9116798 Bourne, Philip; Pu, Calton; Koetzle, Thomas; and Abola, Enrique Columbia University $142,200 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award - $373,339) SDB: An Object-Oriented Toolbox for Use with the Protein Data Bank (PDB) This is the first year funding of a three-year continuing award. The Protein Data Bank (PDB) contains the atomic structure of macromolecules. As of October 1991 there were 790 structural entries (196 Mbytes), if current growth rates persist, this number could grow to 10,000 by the end of the decade. The data provide opportunities for understanding biological function through, for example, comparative structural research. This work addresses several challenges in first making the PDB more accessible to molecular biologists and crystallographers in particular, and second assisting in the management of increasing amounts of data. Several software developments are being undertaken in parallel, but share the same class libraries. First, a new object-based PDB storage format provides suitable access to the levels of substructure found in macromolecules. Second, object-based software tools that interrogate and manipulate structural data, and assist in structure verification are being derived from existing structured programs. Finally, a high-level query language provides intuitive and direct interaction with the PDB. Each aspect of software development proceeds by prototyping followed by iterative cycles of testing in the laboratory and code modification. This work integrates the state-of-art database research results such as object-oriented databases and knowledge bases, software engineering results such as component and glue collaborative work such as extended transaction models to support cooperative scientific research. These tools could potentially precipitate the discovery of new structure-function relationships by permitting data query in a more intuitive fashion. IRI-9247836 Bourne, Philip; Pu, Calton; Koetzle, Thomas and Abola, Enrique Columbia University $3,500 - 3 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award - $7,500) REU: An Object-Oriented Toolbox for Use with the Protein Data Bank (PDB) This is an REU supplement to the award IRI-9116798. Two undergraduate students are involved in the project. The computer science student is assisting in the development of software "glues" for the Toolbox, to test, evaluate, and demonstrate the "glue" development methodology. The biochemistry student is working with a research group to determine the contents of a software suite to verify the accuracy of a protein structure determination. Their combined knowledge of protein structure and software engineering is applied to the development of the appropriate software suite. IRI-9245629 Chu, Wesley W.; Cardenas, Alfonso and Taira, Ricky University of California, Los Angeles $96,781 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Biological Instrumentation and Resources Division - Total Award - $206,781) SDB: A Knowledge-Based Multi-Media Distributed Database System This is the second year funding of a two-year continuing award IRI-9116849. This research deals with the use of domain and application knowledge to merge and manage scientific multi-media data from multiple sources. The primary goal of this research is to develop methodologies to integrate knowledge with bio-medical image databases and to provide approximate, summary and conceptual query answering via generalization, specialization, and association operations. Further, logical indexing techniques for retrieving images by semantic contents and the requirements of the query language constructs for supporting such cooperative query answering will be developed. Such features are important in aiding scientists to extract new scientific knowledge from the raw data. These concepts will be validated in a testbed linked with the bio- medical image databases. The joint research between the investigators from the Computer Science Department and Medical School will assure the prototype system and experiments used are of direct interest to biological and medical applications. The new methodology enables access to the vast storehouse of images by content and features rather than by artificial keys, such as a person's ID. Such capabilities enhance the image databases by characterizing the objects functions or dysfunctions and behavior which will lead to improved image analysis and diagnosis techniques. The outcomes of this research should be extensible to other image and multi-media database applications and result in an experimental knowledge-based image database system that will be made available to the research community. IRI-9244835 Chu, Wesley; Cardenas, Alfonso F. and Taira, Ricky University of California, Los Angeles $24,165 - 24 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $39,165) A Knowledge-Based Multimedia Distributed Database System This supplement is composed of a REU and an instrumentation supplement to the award IRI-9116849. The REU supplement involves two undergraduate assistants in reading pertinent documents and attending the project meetings to understand the various research and proof of the concept issues. The students are also working on coding and documentation of various building blocks to implement the system, such as the data model, the query language, or image storage/retrieval routines. The instrumentation supplement provides the necessary computing environment for the project. ASC-9217384 Crutcher, Richard M. University of Illinois, Urbana $11,998 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program, Division of Advanced Scientific Computing, Computational Mathematics Program and DARPA - Total Award $450,000) HPCC: Radio Synthesis Imaging: An HPCC Application This is the first year of a five-year continuing award in the High Performance Computing and Communications Initiative's (HPCC) Grand Challenge Application Groups competition. This award is for the direct implementation of three computing recommendations of the Astronomy and Astrophysics Survey Committee. The objective of this project is to implement a prototype of the next generation of astronomical telescope systems - remotely located telescopes connected by high speed networks to very high performance, scalable architecture computers and on-line data archives. The very computationally intensive algorithms for calibration and imaging of radio synthesis array observations will be optimized and new algorithms which utilize the massively parallel CM-5 are developed. IRI-9116999 Ferrin, Thomas E.; Langridge, Robert and Cohen, Fred E. University of California, San Francisco $50,000 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award $201,891) SDB: An Object File System for the Macromolecular Workbench This is the first year funding of a three-year continuing award. The objective of this project is the development of an object oriented-database system for supporting application programs used by scientists involved in the rational design of therapeutic agents and studies of specific interactions in peptides and proteins. This work is focused specifically on the data storage and access requirements this class of applications, and is the outgrowth of one component of a project much larger in scope entitled "Macromolecular Workbench: Computer-Assisted Analysis of Protein Design and Function" supported by DARPA. Specific work includes refinement of the data communications protocol used between the Macromolecular Workbench (MMWB) data manager module and object file access library, as well as reimplementation of the data manager in the C++ programming language in order to take advantage of the natural correspondence between scientific data and actions performed on this data as specified in the original MMWB data manager design. The design provides for a unified user interface and machine independent object oriented data access. The result of this project will have significant impact on other scientific applications software, and therefore on the scientific community as a whole, as this software will now be able to take advantage of a unified database system, thereby making possible a "toolkit" of scientific software with each tool capable of simultaneously directly manipulating common shared data. IRI-9247892 Ferrin, Thomas E.; Langridge, Robert; and Cohen, Fred E. University of California, San Francisco $2,864 - 12 mos. SDB: An Object File System for the Macromolecular Workbench This is a supplement to the award IRI-9116999 to correct for a difference in the overhead rate. IRI-9245481 Fox, Edward A.; Hix, Deborah and Heath, Lenwood S. Virginia Polytechnic Institute $156,090 - 12 mos. SDB: A User-Centered Database from the Computer Science Literature This is the second year funding of a three-year continuing award IRI-9116991. This project involves developing and testing a hypermedia database based on the computer science literature. Human-computer interaction and information storage and retrieval research are being applied to develop a new paradigm of user- centered database development and access. Local testing is being done by faculty and students at VPI&SU, with additional testing over the Internet. Computer science literature provided by the Association for Computing Machinery is being transformed to better serve the needs of a wide spectrum of computing practitioners. Based on interviews with computer professionals and analysis of existing publications, "objects" used in computing are being identified so that a tailored object-oriented database including algorithms, algorithm descriptions, bibliographic records, full texts of articles, reviews, data, page images, and related multimedia can be stored and appropriately linked for searching and browsing. New minimal perfect hashing and graph partitioning algorithms provide efficient access to the large, distributed graph of objects. As a result of this project, it will be easier to locate and manipulate information in the resulting electronic archive, which will facilitate reading, writing, annotating, and sharing. Software and methodology used in this project will aid other fields as they develop sophisticated electronic archives with multimedia capabilities. IRI-9244677 Fox, Edward A.; Hix, Deborah and Heath, Lenwood S. Virginia Polytechnic Institute $14,971 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $29,941) A User-Centered Database from the Computer Science Literature This is an instrumentation supplement to the award IRI-9116991 that provides the necessary computing environment for the project. IRI-9246567 Fox, Edward A; Hix, Deborah and Heath, Lenwood S. Virginia Polytechnic Institute $8,000 - 12 mos. REU: A User-Centered Database from the Computer Science Literature This is a REU supplement to the award IRI-9116991. The two undergraduate students employed by the Envision project are implementing and helping in evaluation of human-computer interface prototypes, developing SGML document type definitions for the electronic documents to be stored in the Envision database, and converting those documents into SGML. The undergraduates are gaining an appreciation of current research in human-computer interaction and object-oriented databases as well as hands-on experience in these two important and timely areas of computer science research. IRI-9245622 Goodrich, Michael T.; Salzberg, Steven L. and Ford, Holland C. Johns Hopkins University $177,376 - 12 mos. SDB: A Geometric Framework for the Exploration and Analysis of Astrophysical Data This is the second year funding of a two-year continuing award IRI- 9116843. One of the main challenges facing Astronomy today involves the organization and analysis of the vast volumes of data that has been provided by recent advances in computer and data- collecting technology. This project pools the efforts of researchers in computational geometry, machine learning, and astrophysics in order to develop and apply computational techniques to aid in the analysis and synthesis of astrophysical data. The goal of this project is to provide the astrophysical research community with computational techniques, embodied in a software system called Astro Explorer, for analyzing very large, complex, multi-parameter data sets. Specifically, this project will provide new paradigms for performing general kinds of data extraction, data synthesis across databases, and cluster analysis. All of the methods are based on the unifying framework of viewing data points and the operations that act upon them geometrically. This work extends the current database technology by providing methods for dealing with geometrical data. This approach should significantly improve the productivity of astronomers in their search through the oceans of data that have been produced by the revolution in data collection and storage, for it gives them a computational tool for extracting data that match specified geometric patterns or can be clustered by methods based on natural and intuitive geometric concepts. IRI-9246566 Goodrich, Michael T.; Salzberg, Steven L. and Ford, Holland C. Johns Hopkins University $16,832 - 24 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $29,332) A geometric Framework for the Exploration and Analysis of Astrophysical Data This instrumentation supplement to the award IRI-9116843 provides the necessary computing environment for the project. IRI-9245619 Hachem, Nabil I.; Gennert, Michael A. and Ward, Matthew O. Worcester Polytechnic Institute $165,000 - 12 mos. SDB: Spatio-Temporal Database Management for Global Change Research This is the second year funding of a two-year continuing award IRI-9116988. The goal of this research is the design and development of an integrated system for the management of very large scientific databases, cartographic/geographic information processing, and exploratory scientific data analysis for global change research. The system will represent both spatial and temporal knowledge about natural and man-made entities on the earth's surface, following an object-oriented paradigm. It will apply (user-modifiable) procedures to perform operations on the data, including comparison, derivation, prediction, validation, and visualization. It is an effort to extend the database technology with an intrinsic class of operators, which is extensible and responds to the growing needs of scientific research. Of significance is the integration of many diverse forms of data into the database, including cartography, geography, hydrography, hypsography, images, and urban planning data. Equally important is the maintenance of metadata, that is, data about the data, such as coordinate transformation parameters, map scales, and audit trails of previous processing operations. This project will impact the fields of geographical information systems and global change research as well as the database community. It will provide an integrated database management testbed for scientific research, and a testbed for the development of analysis tools to understand and predict global change. IRI-9246783 Hachem, Nabil; Gennert, Michael A. and Ward, Matthew O. Worchester Polytechnic Institute $9,953 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $15,906) REU: Spatio-Temporal Database Management for Global Change Research This supplement is composed of a REU and an instrumentation supplement to the award IRI-9116988. The REU project involves an undergraduate student developing a high level visual browsing environment for scientific databases. With this additional work, the system supports browsing of a scientific database, inquiring about the type of information it contains, and also it can operate as a browser for extracting the semantics of scientific methods and concepts. The instrumentation supplement provides the necessary computing environment for the project. IRI-9245107 Hansen, Elaine R., Graefe, Goetz and Jouchoux, Alain University of Colorado, Boulder $103,500 - 12 mos. SDB: Parallel Real-Time Scientific Data Collection and Processing through Extensible Database Technology This is the second year funding of a two-year continuing award IRI- 9116547. This work contributes to scientific data processing technology through the development of an extensible toolkit of database operators to capture, clean and manipulate scientific data, and through the study of mechanisms for interactive, real- time scientific processing, such as parallelization and flow control. This work builds on the experience of the research team in gathering and processing spacecraft data and in developing extensible database technologies as part of the Volcano project. The team includes a graduate student in computer science who has worked extensively at the Laboratory for Atmospheric and Space Physics (LASP) of the University of Colorado on spacecraft operations and data management projects. The focus of this project is on the development of new database technologies which can be accomplished by combining real-time and database query processing techniques. The key results of this work will include: (1) techniques for enhancing the responsiveness of database query operations; (2) tools to enable a scientist to capture and process huge amounts of scientific data such as those received from a spacecraft, and to process these data interactively; and (3) an extensible toolkit database operations (such as capture, clean, relate, manipulate, and display scientific data sets) to facilitate analysis of the complex data sets required to understand complex scientific issues such as global ozone depletion. IRI-9116860 Katz, Randy H. and Stonebraker, Michael R. University of California, Berkeley and Gautier, Catherine University of California, Santa Barbara $10,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program - Total Award $140,000) SDB: Design of a Large Capacity Object Server Supporting Earth System Science Researchers This is the first year funding of a three-year continuing award. The goal of this research project is to develop a deeper understanding of how to design and use I/O subsystems for high capacity storage and associated object management systems tailored for the data types encountered by earth system science researchers. A multi-terabyte prototype massive storage management system is constructed and populated with real data derived from actual space based remote instruments. This prototype is tested and made available for use by a community of earth science researchers via a high speed wide-area network. The prototype leverages technologies already pioneered at Berkeley such as extensible database systems (POSTGRES) and high performance storage subsystems (RAID). The work is carried out under the auspices of SEQUOIA 2000, an industry-sponsored project that seeks to find an understanding of global climate change by encouraging the interaction of computer scientists and earth scientists. SEQUOIA 2000 expects that new discoveries, based on uncovering new relationships among collected climate data, will be made possible by giving earth scientists access to a flexible, high capacity storage repository of remotely sensed climate data. IRI-9245148 Kazic, Toni Washington University and Tsur, Shalom Swiss Bank Corporation $82,875 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award $214,034) SDB: Modeling and Simulation Biological Processes as Logical Enterprises This is a remaining first year and the second-year funding of a two-year continuing award IRI-9117005 and an instrumentation supplement that provides the necessary computing environment for the project. A potential use for scientific databases in the biological sciences involves the simulation of in vivo metabolic tracer experiments. The queries of interest in this area are difficult to answer using existing technologies. The proposal suggests two hypotheses: first, that complex biological processes can be modeled as logical enterprises comprising objects, processes and constraint dimensions; second, that deductive database technology is suitable for the computer realization of such models. To test these hypotheses, the proposal focuses on the possible representations of compounds, biochemical reactions, and biochemical and physiological constraints, and the computational consequences of the alternative representations. The reactions come chiefly from intermediary carbohydrate metabolism, and the strength of the computational ideas is evaluated by comparing the results with experimental data from the biological literature. At the conclusion of the project, the strengths and limits of the different representational schemes should be much clearer. The impact on biological sciences will be the development of a generic representational solution for a wide class of biological problems, while the computer science impact will lie in the successful integration of the data and process aspects of the problem in a declarative, symmetric treatment. In the long run it is expected that computer systems of this type will be used for predictive purposes and in the a prior assessment of in vivo experiments. IRI-9245819 Maier, David; Stanley, James; Walpole, Jonathan and Wolfe, Michael Oregon Graduate Institute of Technology $215,727 - 12 mos. (Jointly funded with the Physical Chemistry Program - Total Award $245,727) SDB: Database Support for Scientific Computing This a second year funding of a two-year continuing award IRI- 9117008 and an instrumentation supplement to this award that provides the necessary computing environment for the project. Scientific data types show promise for support of scientific computing at all computer system levels, from programming and visualization tasks, through compiler and database technology, to operating systems. Rather than having a distinct data type mechanism at each level, this project concentrates on a coordinated approach across the levels of compiling, data management and operating systems, centered on database support for scientific data types. Key aspects of the project are the development and deployment of a Hybrid Data Manager as a software substrate for rapid experimentation in coordinated data type support across these layers. The HDM uses object-oriented database technology to build a powerful front end for data type instances residing in other storage systems. Existing datasets, as well as newly created objects are to be supported using a staged approach that can lend some support to existing codes while providing full data type support for newly created applications. An important phase of the project is the deployment of the HDM on applications in current use in two scientific domains, computational molecular chemistry and microstructure property predictions in materials science. The experience gained in this project will provide input in the design of a new generation of database systems that will better support the space and modeling demands of scientific applications as well as integrate tightly with compiler and operating system components for increased performance. IRI-9117011 Ryan, William B.F. Columbia University $24,977 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Marine Geology and Geophysics Program - Total Award $59,977) and IRI-9120358 Tyce, Robert C. University of Rhode Island $24,996 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Marine Geology and Geophysics Program - Total Award $59,996) SDB: Visualization Tools for Large Seafloor Databases Accessed over Networks This is an interinstitutional collaborative research of William Ryan and Robert Tyce. This is the first year funding of a two-year continuing award. Visualization tools are being developed to allow the scientists, teachers and students to view and manipulate representations of the ocean floor directly on the screens of their desktop computers. Many different types of pictures can be examined, including maps, perspective views, pseudocolor imagery, shaded relief diagrams and 3-D stereograms. The raw data from which the views are created include bathymetry from multibeam and phase interferometric sonars and backscatter from side-looking sonars. Using electronic communication over Internet, the users connect from their computer (the server) into a client computer at Columbia University to begin the visualization session. The client returns to the user with menus and dialog boxes for the purpose of selecting the display type, map boundaries, color tables, fonts, data files, etc. Moments later the representation appears in a window on the user's screen with computations run in the client computer. The user, given a responsive, intuitive and interactive interface, is liberated from the chores of locating and maintaining large complex databases and software code. The users find their work accelerated by an immediate response to a query and by a greater depth of understanding from advanced graphic representations of very large datasets. IRI-9116770 Segev, Arie University of California, Berkeley $35,988 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award $69,988) SDB: Processing Heterogeneous Data in Scientific Databases This is the first year funding of a three-year continuing award. Many important information systems applications require access to data stored in multiple autonomous databases. These databases often use different hardware, software, database management systems and are distributed among various geographical locations. Heterogeneity among the databases, in the systems, semantics and data distribution poses challenge to the conventional data retrieval techniques. In this project, key issues related to the processing of heterogeneous data are investigated. In particular, the focus of the work is on joining of the records of the same instances across disparate databases. A new operator, the Entity Join, has been defined. It uses the common attributes present in the joining data to probabilistically estimate the correctness of the result. The model is extended to cover all types of data heterogeneity problems including temporal heterogeneity. Additional operators are defined to enhance the data manipulation capability of the model, and their optimization explored. This project also examines the use of rules in the process of matching objects. The different algorithms are first simulated to assess their performances, and then prototypes built to further evaluate the optimization concepts. The results of this work will facilitate data analysis and exploration in many statistical and scientific applications including census, surveys, epidemiology, biology, astronomy and other areas. IRI-9116809 Shapiro, Linda G.; Tanimoto, Steven L. and Brinkley, James F. University of Washington $70,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $120,000) SDB: Visual Database System for Computer Vision Research This is the first year funding of a two-year continuing award. A database system for computer vision must be able to manipulate data in numerous forms, running a full spectrum from values of scalar features, through arrays of light intensities, to complex linked data structures representing networks of partially and tentatively recognized shapes and models. This work entails the design and partial implementation of a prototype vision database system based on a hierarchical, relational data structure that can handle a wide variety of types of data. The design also includes a novel visual user interface that allows the user to examine and comprehend the entities being processed. In this work, the organization and access functions for the database system are developed through a study of the kinds of structures and operations used in two different vision application domains: robotics and medicine. Both primitive access functions and some higher-level vision operations are implemented in the prototype system. The flexibility and efficiency of the system are evaluated for vision algorithms being developed in robot vision and medical image processing research projects. The result will be a general scientific database system that can be accessed via both graphical and application program interface and that will satisfy the needs of computer vision systems. IRI-9116451 Sirovich, Lawrence; Bisshopp, Frederic E.; Everson, Richard and Vitter, Jeffrey Brown University $55,718 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division and the Fluid, Particle and Hydraulic Systems Program - Total Award $151,435) SDB: Management, Analysis and Representation of Large Scientific Databases This is the first year funding of a two-year continuing award. The project focuses on the management storage and analysis of large scientific databases. A cornerstone of the activity is the Karhunen-Loeve procedure which in a mathematically well defined sense provides an intrinsic and ideal representation for many forms of scientific data. The development of the snapshot method allows this procedure to be applied to hitherto unmanageably-large databases. This method is applied to the following cases: optical imaging of cortical activity using fluorescent dyes; a new simulation of the turbulent jet; and sea surface measurements obtained by telemetric means. Each of these cases leads to gigabyte or larger datasets. Lossy and lossless data encoding techniques are developed and melded in this project. New methods, including the wavelet and Wigner transform are developed for the visualization and presentation of large databases. Application of these methods to on-the-fly data acquisition and compression of large datasets is investigated. Results of this project will provide a set of tools which will make possible the collection, compaction, analysis, rapid exploration, visualization, progressive browsing and communication of exceedingly large databases of the sort encountered in engineering, pattern recognition and electrophysiology. IRI-9245447 Smith, Terence R.; Su, Jianwen; El Abbadi, Amr; Agrawal, Divyakant and Dozier, Jeff University of California, Santa Barbara and Dunne, Thomas University of Washington and Ramakrishnan, Raghunath University of Wisconsin, Madison $156,103 - 12 mos. (Jointly funded with the Biological Instrumentation and Resources Division - Total Award $206,103) SDB: Toward a System that Supports Conceptual Modeling in Data Intensive Scientific Investigation This is the second year funding of a three-year continuing award IRI-9117094. The goal of the proposed research is to design, analyze, and perform feasibility studies of key components of systems that support scientific modeling in data-intensive and numerically-intensive applications, particularly in the area of large-scale environmental science. The two major sets of issues under investigation are explicit support for the complex spatio- temporal objects that are required in the computational modeling of scientific phenomena (as well as their temporal and spatial relationships) and transparent support for data access from large, heterogeneous and distributed data sources. As products of this research, the following tools will be developed: languages for the definition and manipulation of complex, spatio-temporal objects and for the mathematical and statistical modeling of scientific phenomena; algorithms for the support of schema updates and evolution; optimization techniques for spatial and temporal manipulations; protocols for storage and operation abstraction; and platforms for collaborative scientific research. The particular domain of scientific application involves database support for investigating the hydrology of the Amazon basin, which is chosen as a typical Earth Observation System (EOS) project. The research, however, is designed to benefit many areas of the natural sciences by providing a set of tools for increasing the productivity of scientists who need to integrate complex modeling activities with the use of large datasets, which is a problem encountered in most EOS projects. IRI-9117084 Soloway, Elliot and Martin, William University of Michigan, Ann Arbor $42,720 - 12 mos. (Jointly funded with the Interactive Systems Program and the Information Technology and Organizations Program - Total Award $77,720) SDB: Computer-based Support for Scientific Data Analysis This is the first year funding of a three-year continuing award. The "human-computer interface" is paramount if scientists are to take full advantage of the vast quantities and types of data now becoming available. Attention needs to be focused on more than the query language; rather, the interface needs to integrate scientific databases into the everyday work practices of scientists, e.g., data exploring, hypothesis generating and testing, report and chart making. The Task/Artifact Methodology is being employed; it consists of cycles of cognitive task analysis, system building, and testing in ecologically valid contexts, e.g., practicing scientists using the system on a daily basis. The resultant computer-based environment, ReV (Representations for Visualization) is user- and task-centered, as opposed to being technology-driven. ReV serves as the scientist's "notebook" permitting him/her to move among databases, hypotheses, reports, charts, etc. ReV is being designed for scientists in nuclear engineering and for scientists engaged in global change research. As computers become integral to the moment- by-moment work practices of scientists, well-designed tools become critical. This project, then, serves as a model for how scientist- centered computing environments can be developed and deployed. IRI-9117153 Sparr, Ted M.; Bergeron, R. Daniel and Meeker, Loren D. University of New Hampshire $69,999 - 12 mos. (Jointly funded with the Geology & Polar Program, the Polar Earth Sciences Program, and the Information Technology and Organizations Program - $149,999) SDB: Integrating Data Management, Analysis and Visualization for Collaborative Scientific Research This is the first year funding of a three-year continuing award. The goal of this multidisciplinary project is to design and prototype a new approach in database environments to support collaborative scientific research. The prototype integrates scientific data visualization and mathematical and statistical analysis tools with database support in a highly interactive environment. A new model for scientific data is founded on the notion that a query of a scientific database conceptually creates new derived data whose relationship to the parent database is defined by the query. Each query, in principle, leads to the discovery of additional structure in the data that is either explicit in the results of the query, or hypothesized by the scientist(s) from the results of queries. The system uses a process flow graph to represent queries. The project, carried out by a team of two computer scientists, an applied mathematician, and scientists from earth science, civil engineering and atmospheric science, contributes to ongoing research in the fields of environmental biology and chemistry, oil reservoir analysis, and polar ice core study. This work will produce a new data model for scientific data and will design and prototype an integrated data management, analysis and visualization environment to support interdisciplinary scientific investigation. IRI-9245625 Ullman, Jeffrey D.; Howard, Craig ; Law, Kincho and Teicholz, Paul Stanford University $100,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Structures and Building Systems Program - Total Award $200,000) SDB: Integrated Data Exchange and Concurrent Design for Engineered Facilities This is the second year funding of a three-year continuing award IRI-9116649. The architecture-engineering-construction industry is highly fragmented, both vertically (between project phases, e.g., planning, design, and construction) and horizontally (between specialists at a given project phase, e.g., design). The industry needs an integrated computing environment that will support concurrent design with fast data exchange and powerful change management capabilities. The long-term potential benefits are improved designs, reduced construction time, fewer design errors and omissions, minimization of costly rework, and better life-cycle facility management. The key contributions of this research project are methods to (1) translate and communicate project data dynamically, overcoming both logical barriers (differing views of overlapping data) and physical barriers (data distribution); and (2) detect, analyze, and manage changes efficiently during concurrent design processes. These contributions are demonstrated in prototype software that integrates systems for CAD, drafting, analysis, and simulation along with relational and object-oriented databases. The first software component is an integrated set of object-oriented, multi-machine utilities to provide fast, flexible, intelligent data translation and transfer. The second component is a set of constraint handling and change management utilities that will support declarative definition of data dependencies, constraints, and conflict resolution strategies. The project will take advantage of industry links through Stanford University's Center for Integrated Facility Engineering to develop a test bed of industrial-strength databases to demonstrate the software. IRI-9245620 Walther, Sandra S. and Peskin, Richard L. Rutgers University $138,000 - 12 mos. SDB: Object-Oriented Data Management for Interactive Visualization of Scientific Data This is the second year funding of a two-year continuing award IRI- 9116558. This project continues development of a data management strategy to organize scientific datasets into online (i.e. "in memory") object-oriented numerical databases. The strategy uses hashing maps based on the spatio-temporal coordinates of the dataset itself and addresses a key issue in interactive data analysis -- maintaining a retrieval route for the user between the visual representation on the screen and the actual data. This is accomplished through a set of data structures that manage computed data as "computational objects" at the nodes of a mesh organized as a single plane, a volume of planes, or a timeseries. All references to the data are resolved into queries about the maps. The hashing maps can be resident in a workstation while managing a database that is itself distributed across numerous facilities. Current extensions of this research include: (1) design of parallel storage and searching techniques of the distributed data objects; (2) interactive user pacification of filter functions (local or non-local) for graphics and diagnostics; and (3) generalization of management utilities to support conversion of any user specified data set to object-oriented database organization. This research seeks to alleviate the continuing user problem of "data glut" in science and engineering. Equally important are its implications for the development of software strategies that can utilize the complex and diverse hardware developments promised for the next decade in massively parallel computing. IRI-9245924 Walther, Sandra and Peskin, Richard L. Rutgers University $20,000 - 12 mos. Object-Oriented Data Management for Interactive Visualization of Scientific Data This instrumentation supplement to the award IRI-9116558 provides the necessary computing environment for the project. SPECIAL PROJECTS BIR-9112037 Brunt, James W. University of New Mexico $5,000 - 24 mos. and BIR-9119824 Michener, William K. University of South Carolina, Columbia $5,000 - 24 mos. (Jointly funded with the Biological Instrumentation and Resources Division, the Geography and Regional Science Program and the Long Term Projects in Environmental Biology Program - Total Award $46,660) Workshop: Environmental Information Management and Analysis -- Ecosystem to Biosphere Scales This is an interinstitutional joint project of Drs. Brunt and Michener. The award supports a symposium for the exchange of information and technology among computer scientists, Geographic Information System/Data Management specialists, and environmental scientists (atmospheric, marine, and terrestrial) engaged in scientific database development and management of complex long-term and broad area data sets, and addressing questions at ecosystem through biosphere scales. The symposium is organized along four related themes: (1) ecosystem regional and global scale environmental research; (2) scientific databases for broad scale research; (3) data/information management technical aspects; and (4) management and analysis of broad scale spatial databases. The knowledge exchange and cross-fertilization of ideas among scientists from a wide spectrum of disciplines are essential for the development of the scientific databases, data management, and analytical tools which will be necessary to address the problems facing environmental scientists at a variety of scales. A high quality volume of the proceedings will be produced and preliminary manuscripts will be available electronically. IRI-9120664 Hopcroft, John E. Cornell University $5,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $10,000) Workshop: Information Access and Capture in Engineering Design Environments Computing hardware, digital memory systems, and high speed computer networks have progressed to the point where building very large digital document collections and providing facile distributed access to these collections is practical and economical. However, many difficult problems still stand in the way of implementing such systems. The time is ripe to begin building a broad science base to support this activity and to begin academic programs at universities to produce the PhD's needed to advance progress in this area. This award funds a three-day workshop to bring together about thirty key researchers in the area of information capture and access. Participants represent broad coverage of the research area and a balance between industrial and academic, and between project implementors and more science-base-oriented individuals. The coordinators of the workshop are John Hopcroft, Computer Science Department, Cornell University and Gregory Zack, Design Research Area, Xerox. The primary objectives are to identify key research issues in information capture and access, and to highlight research directions where progress is likely. In addition, the workshop identifies common infrastructure that would foster cooperative research efforts. IRI-9123156 Hunter, Lawrence E. National Library of Medicine and Shavlik, Jude W. University of Wisconsin, Madison $5,000 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Biological Infrastructure and Resources Division - Total Award $19,000) Workshop: Creating an Infrastructure for Intelligent Systems in Molecular Biology Research applying intelligent systems technology to problem in molecular biology is expanding rapidly. Techniques such as artificial intelligence, neural networks, object-oriented databases, large-scale computer modeling, and robotics are being successfully applied to problems in genetics, protein structure, development, evolution, and many other aspects of biology. The rapid growth in this interdisciplinary research has created a need for infrastructure development and program planning. The workshop brings together about two-dozen active researchers in the field with representatives of NLM and NSF to discuss issues such as coordination of conferences and publication venues, distribution of research support, interdisciplinary training programs, and the sharing of data and other resources. In addition to facilitating communication, the goal of the meeting is to produce a report describing community needs and a set of standing committees to help address those needs. Intelligent systems technology is relevant to several key problems in the HPCC Program. Directly, advanced software technology and algorithms require intelligent systems technology; and indirectly, high performance computing systems will be developed based on novel intelligent software algorithms. This workshop addresses some of the key issues in the problem domain of molecular biology, which have implications in many other problem domains of the HPCC Program. IRI-9106560 Kraft, Donald M. Louisiana State University $33,750 - 6 mos. Travel to the ACM/SIGIR International Conference on Research and Development in Information Retrieval (ICRDIR): Copenhagen, Denmark; June 21-24, 1992 The continuing interest in information retrieval systems is based upon the need to cope with the massive problems associated with storage, access, and retrieval of information. It is clear that text processing is a vast area, which encompasses information retrieval, is of prime importance in computing and beyond, and that also relates to many aspects of computing, including architecture, data and file structures, and, of late, graphics and interface design. The ACM/SIGIR International Conference on Research and Development in Information Retrieval (ICRDIR) is the premier conference concerned with the vital area of information access and retrieval as it effects the important issues in text processing. The topics at ACM/SIGIR are highly interdisciplinary. At least ten graduate students from different disciplines in various universities across the country are invited to participate in the conference, based on submitted papers, in order to be able to present their work and be exposed to scholars from all over the world doing parallel work. The students' participation offers them the opportunity of getting wide feedback on their ideas, and on the other hand enriches the conference by their innovative ideas. IRI-9217185 Leggett, John Texas A&M $10,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program - Total Award $23,930) Workshop: Hyperbase Systems The need for high quality, directed research on hyperbase systems (hypermedia database systems) has become very apparent to researchers in many fields. This workshop brings together approximately 25 participants from the areas of hypermedia, database, collaborative and information retrieval systems for an intensive two-day workshop, held in Washington, D.C. in October 1992. The workshop has two primary goals. The first goal is to bring synergy to the hyperbase system research area through understanding and building of research relationships in this diverse and multidisciplinary group. The second goal is to establish an agenda for research in this critical area of next- generation information systems. The result of the workshop is a report identifying: critical research issues facing hyperbase system researchers, current progress on the issues, potential methods of approach (including literature citations) and a general research agenda for the field. The results will be widely disseminated in the refereed and professional society literature and through the appropriate conferences. IRI-9201138 Motro, Amihai George Mason University $35,750 - 18 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $39,750) Workshop: Uncertainty Management in Information Systems -- from Needs to Solutions Uncertainty pervades real world scenarios, and must therefore be incorporated into every information system that attempts to provide a complete and accurate model of the real world. Yet, present generation information systems have very limited capabilities in this regard. On the other hand, many theoretical models have been proposed for the management of uncertainty, but often without thorough understanding of the actual problems faced by researchers, developers, designers and users of information systems. This project concentrates on a planning workshop that brings together leading researchers in the scientific communities of information systems (e.g., database systems, information retrieval systems, expert systems, office information systems) and uncertainty modeling (often working within frameworks such as mathematical logic, probability theory, fuzzy set theory, possibility theory, and evidential models) to study the needs of the information systems community and to tap the expertise of the uncertainty modeling community, for solutions that respond to these needs. This workshop will promote true dialogue between these two scientific communities; it will establish the state of the art in this field, and will set the course for future research. The project, which is co-sponsored by the European Esprit program, will foster collaborations between researchers in the United States and Western Europe. IRI-9208743 Rusinkiewicz, Marek University of Houston $1,997 - 12 mos. (Jointly funded with DARPA - Total Award $13,997) Workshop: Interoperability in Heterogeneous Database Systems This workshop investigates application and research issues of interoperating data and knowledge bases. Multidatabase interoperability and resolution of semantic heterogeneity constitute enabling technologies for the development of complex applications requiring access to multiple information resources. The main objective is to bring together researchers from industry and academia to discuss how the basic research addresses problems encountered by applications in the heterogeneous database environment. Participants have an opportunity to study real user problems identified by the industry representatives. The collaboration between basic and application researchers can then be based on a solid foundation created by basic research. The main topics to be discussed include: semantic integration of information from diverse sources; methods to resolve semantic incompatibilities; query optimization techniques and distributed transaction processing in multidatabase environments; descriptive modeling of a synchronous data processing; and incorporation of emerging technologies (such as active and object-oriented databases) into multidatabase systems. Results of the workshop will be summarized in a final report that will evaluate the progress made in multidatabase interoperability and identify areas in need of further research. IRI-9244976 Russell, Lucian Argonne National Laboratory $8,548 - 12 mos. Experimental Use of Electronic Media to Speed Up the Development of a New Area of Research This is a supplemental award to IRI-9121721 to add funds not available in FY91. This project's goal is to accelerate the rate of maturation of the new field of scientific databases. MESSAGE FROM THE PROGRAM DIRECTOR FY 1992 was a very good year for the Information Technology and Organizations Program. Along with its budget growth to $4.68 million, there were a number of notable highlights. These include the funding of a second large collaboratory experiment. This one is a cooperative agreement with the University of Michigan to build a collaboratory for upper atmospheric and space physics research. While the collaboratory will integrate many electronic resources for the research community, its feature attraction will be the remote control of a suite of instruments located in Greenland, and the sharing of data and analysis, in real-time, by researchers in many U.S. locations. The second workshop for grantees of the Program whose research is in coordination theory and collaboration technology was held in Washington, D.C. in July. Exciting research results were enthusiastically shared by this growing community of scholars. We are proud that two new NYI awards were made for researchers funded by the Program. These awards and the four new RIA awards are helping to build a community of new researchers in the fields funded by the Program. Growing interest in the Program's funding areas is further evidenced by the increased participation of other IRIS programs in the joint funding of awards. Thirteen joint awards were made by other IRIS programs with IT&O this year. We all believe that these joint activities are good for the research community. They help break down disciplinary barriers and provide extra insurance that proposals do not "fall between the cracks" of our program areas. INFORMATION TECHNOLGOY AND ORGANIZATIONS PROGRAM 1. Scientific Scope. IT&O is the primary NSF program providing research support for creating new knowledge about the integration of computer and communications technology into group activities. Special attention is paid to distributed computing aspects of electronic collaboration. The footprint of the research funded by this Program is its multi-disciplinary focus on phenomena of coordination and collaboration. A major area of funding in the Program is for research to provide the knowledge base for development of the "Collaboratory", a national resource that uses networking and computer technology to support scientific collaboration independent of distance by allowing robust, remote, interaction with colleagues, instruments, data and archived knowledge. The development of "electronic libraries" is part of this funding area. Another major area is the analysis of the social and economic impacts of the technology. In these pursuits the Program funds: 1) development of mathematical theory of coordination (including both quantitative theories like those of economics and discrete theories like the knowledge formalisms underlying computer communication protocols; 2) Distributed AI; 3) empirical studies of existing coordination technologies and groupware; 4) the design of experiments in which coordination phenomena can be generated and controlled; 5) exploratory design of collaboratory-like systems which integrate invention and implementation; and 6) studies on the impact of computer and communications technology. Research supported in this Program is expected to contribute to the knowledge-base about processes of coordination and cooperation among autonomous units in human systems, in computer and communications systems and in hybrid organizations of both systems. The problem domains can focus on: o The principles underlying how people collaborate and coordinate work efficiently and productively in environments characterized by a high degree of decentralized computation and decision-making; o Principles and design of multi-user interfaces; o Evaluation of collaboratory testbeds; o The structure and outputs of organizations, industries and markets which incorporate sophisticated, decentralized information and communications technology as important components of their operations; o Information technology impacts and policy. Research topics include: computer technology diffusion; telecommunications policy; standards design process; regulation effects on design and use of computerized information systems; and, social impacts of information technology. 2. Relation to Other Programs. The IT&O Program funds multi- disciplinary research on the design of technology for computer supported cooperative work, including groupware. This part of the Program also relates to the domain of the Interactive Systems Program which is concerned with individual user interfaces. The IT&O Program funds research on distributed AI, while the Knowledge Models and Cognitive Systems Program funds work on other AI problems. The theory development funding generates some jointly considered proposals with the CCR Division's Theory program, with Economics and with Mathematics Division programs. Both the experimental design and exploratory systems development research funding will depend on close cooperation with a broad spectrum of NSF and other agency programs as these efforts lead closer to design of a real "collaboratory". INFORMATION TECHNOLOGY PROGRAM FISCAL YEAR 1992 RESEARCH PROJECTS COORDINATION THEORY/COLLABORATION TECHNOLOGY IRI-9100149 Arkin, Ronald C. Georgia Tech Research Corporation $20,000 - 24 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $119,901) Operation and Communication in Multi-Agent Reactive Robotic Systems This research is to study communications and control requirements for multiple autonomous robots to cooperate in accomplishing a task, where the robots are organized heterarchically instead of using the more common hierarchical (master/slave) control. Each behaves reactively, in accordance with sets of relatively primitive motor schemas. The resulting organization does not require an explicit world model, and should be more robust and hence more suitable for operations in remote and/or hazardous environments. A key research issue is to evaluate the efficacy of such systems operating under constrained inter-agent communications. Such agents could be more economical to operate. The goal of the research is to understand and develop control and communication mechanisms that produce robust yet efficient cooperative robot behavior. IRI-9216848 Atkins, Daniel; Clauer, Robert C.; Weymouth, Terry E. and Olson, Gary M. University of Michigan, Ann Arbor $598,512 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office and the Magnetospheric Physics Program Office - Total Award $708,512) A Scientific Group Communications & Collaborative Testbed for Upper Atmospheric Research This is a cooperative agreement to fund a multidisciplinary effort linking research in computer science, behavioral science, and upper atmospheric and space science to build a prototype system for a distributed but shared working environment: the vision of a collaboratory. This effort conceives, develops, deploys, tests, evaluates, and integrates a high performance group centered computing environment into the collaborative experimental and modeling activities ongoing in the upper atmospheric research community. The upper atmospheric researchers here are a collaborating group engaged in observational activities using a variety of instruments located at the Sondre Stromfjord, Greenland upper atmospheric research facility. Many of these activities are directed at rare or intermittent phenomena requiring real time control of instruments by the scientists observing the changing conditions. This is presently accomplished by visits to the remote facility. To prototype test and evaluate the distributed tools for collaboration, research under this agreement will develop a user- oriented, rapid prototyping testbed built around the Sondre Stromfjord facility and its user community. Testing and evaluation of the prototype tools will involve measurements of human behavior. This research will add to the understanding of effective use of collaboration tools by performing studies of use and effectiveness of these tools among the testbed users. The Sondre Stromfjord researchers expect to obtain greater efficiency in joint experimental operations, analysis, and discovery by guiding the requirements for the collaboration environment which they will utilize to support their research. IRI-9123840 Clauer, Robert C. University of Michigan, Ann Arbor $13,141 - 12 mos. Workshop on Scientific Collaboration and Interaction with Remote, Unique Facilities This is a small invitational planning workshop on Scientific Collaboration and Interaction with Remote, Unique Facilities held on the campus of the University of Michigan on January 21-24, 1992. A major function of the workshop is to bring together members of the scientific research community who are engaged in collaborative experimental research using a unique, remote facility, with technologists and social scientists working on collaborative technology. The researches are engaged in upper atmospheric and space research utilizing the Sondre Stromfjord, Greenland Upper Atmospheric Research Facility. The goal of the workshop is to develop a plan to implement and testbed collaborative tools to be utilized by scientists engaged in collaborative and observational research using this facility. Included in these plans are design of a system for accessing the instruments at this facility from remote locations, such as at the University of Michigan. It is expected that the output of this workshop will be a proposal for establishing the experimental testbed for this collaborative effort. IRI-9208319 Dewan, Prasun Purdue University $56,095 - 12 mos. Flexible Collaborative Software Engineering This is collaborative research with John Riedl, University of Minnesota (IRI-9208546). Much of software engineering-design, programming, debugging, testing, code reviews, program demonstrations, and program management-requires collaboration among multiple users, possibly geographically dispersed. In this project the researchers propose to investigate flexible support for the activities of cooperating software engineers. In particular they propose to investigate an approach that offers (i) a wide range of concurrency control mechanisms including serializable transactions and interactive transactions; (ii) fine-grain access control for collaborative applications that allows a user to recover from other users' mistakes and explore various alternatives with them, (iv) a multidimensional inheritance model that allows collaboration parameters to be specified independently in hierarchies of groups of objects and groups of users; and (v) multimedia support integrated with concurrency control, access control, and undo/redo. The unique features of this research are its focus on the relationships of these five goals, on flexible mechanisms for achieving each of the goals, and on structured specification of options by end-users. The researchers will use results from previous research done in flexible coupling, long transactions, adaptive concurrency control systems. inheritance, single-user undo/redo, capability-based protection, collaborative tools, and software engineering tools, They plan to implement the approach by extending an existing system developed at Purdue and Minnesota. They will evaluate the performance of the various collaboration methods and study how a particular method can be automatically chosen by the systems based on a minimum performance level requested by the user. In addition to an understanding of the applicability of collaborative technology to software engineering, the research will produce prototype software engineering tools to demonstrate the novel aspects of the research. IRI- 9244253 Fischer, Gerhard; Lemke, Andreas; and McCall, Raymond University of Colorado, Boulder $240,000 - 12 mos. (Jointly funded with the Interactive Systems Program - Total Award $250,000) Supporting Collaborative Design with the Integrated Knowledge-Based Design Environments This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. The goal of this project is to develop a conceptual framework and a prototype system for collaboration in an synchronous mode among members of design teams. The proposed design environments include knowledge-based and graphic construction components with issue-based hypermedia systems designed to support collaboration. The application domain for the prototype system is the design of communications network within buildings. The complexity of such projects forces large and heterogenous groups to work together over long periods of time. The large and growing discrepancy between an amount of potentially relevant knowledge for the design task and the amount any one designer can know and remember puts limits on progress in design. Overcoming this limit is a central challenge for developers of systems that support both individual and collaborative design efforts. The work in this project builds on previous developments embodied in the FRAMER system for the design of human-computer interfaces and the JANUS system for architectural design. IRI-9246975 Fischer, Gerhard and Lemke, Andreas University of Colorado, Boulder $10,000 - 12 mos. REU: Supporting Collaborative Design with Integrated Knowledged- Based Design Environments This is an REU supplemental award. This supplement adds two undergraduate students to the project. One of the students will do research to develop domain constraints and to build a knowledge base of rules. The other will gather data on network layouts that are needed to track existing equipment and capabilities. Both will be exposed to experimental research techniques involved in the design of computer supported cooperative work environments. See IRI-9244253, above, for a description of the larger proposal. IRI-9246063 Fischer, Gerhard University of Colorado, Boulder $15,000 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $30,000) Supporting Collaborative Design with Integrated Knowledge-Based Design Environments This is a small supplement to IRI-9244253, above. The supplement requests instrumentation matching funds from our Cross Directorate Activities Division and from the Information Technology and Organizations Program, IRIS Division. The equipment sought enables the group to move to a new computational platform that will reduce the costs and maintenance; provide broad functionality in terms of user interface tools and prototyping environment, and will add broader availability of the systems designed in the project, and will also allow access to lower cost multi-media interface technology. IRI-9116860 Katz, Randy H. and Stonebraker, Michael R. University of California, Berkeley $130,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $140,000) SDB: Design of a Large Capacity Object Server Supporting Earth System Science Researchers This research is to design I/O subsystems for high capacity storage and associated object management systems tailored for the kinds of data types encountered in scientific research on global climate change related to the physical interaction of the ocean, land and atmosphere. The researchers are constructing a prototype system, populating it with real data and making it available to earth system science researchers on the National Research and Education Network (NREN). The prototype object management system is based on significant extensions of the dual technologies of extensible database systems (POSTGRES) and redundant arrays of inexpensive disks (RAID), developed in part under previous NSF support. This research relates to Digital Equipment Corporationþs (DEC) project Sequoia 2000. DEC has provided the researchers funded under this grant with the equipment necessary to perform the research. IRI-9217185 Leggett, John Texas A&M Engineer Experiment Station $13,930 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $23,930) Workshop on Hyperbase Systems The need for high quality, directed research on hyperbase systems (hypermedia database systems) has become very apparent to researchers in many fields. This workshop brings together approximately 25 participants from the areas of hypermedia, database, collaborative and information retrieval systems for an intensive two-day workshop, held in Washington, D.C. in October 1992. The workshop has two primary goals. The first goal is to bring synergy to the hyperbase system research area through understanding and building of research relationships in this diverse and multidisciplinary group. The second goal is to establish an agenda for research in this critical area of next- generation information systems. The result of the workshop is a report identifying: critical research issues facing hyperbase system researchers, current progress on the issues, potential methods of approach (including literature citations) and a general research agenda for the field. The results will be widely disseminated in the refereed and professional society literature and through the appropriate conferences. IRI-9122541 Levitt, Raymond E. Stanford University $145,475 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Cross Disciplinary Activities Office - Total Award $179,975) The Virtual Design Team: Simulation Decision-Making and Information Flow in Concurrent, Multi-disciplinary Design This research integrates ideas from artificial intelligence (AI) and coordination theory to develop a Virtual Design Team VDT)- an object-oriented, discrete event, computer simulation model of the flow of decisions and information among participants in complex, multi-disciplinary design projects. Conducting controlled experiments in real organizations to assess the performance of alternative coordination structures is both expensive and fraught with practical difficulties. Compounding this, powerful computer simulation tools to analyze the performance of organization structures by simulation-analogous to finite element models for simulating structural behavior-have not yet been developed. The VDT addresses this need by using AI techniques to model how the organizational structure of, and the information processing technologies used by, managers in a multi-disciplinary project design team effect the richness, volume and timing of information passing among team members. The long range goal of this work is to provide a computerized analysis tool that can help researchers and practitioners analyze the performance of organizations. The first phase of the project will model the impact of organizational structure and information processing technologies on task duration. The second phase will extend the model to capture the impacts of organization structure and information technologies on design quality and reliability. IRI-9122508 Lewis, Clayton; Zigura, Ilze and Reitsma, Rene University of Colorado, Boulder $113,132 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $118,132) Model-based Group Decision Support: The Impact of Shared Simulation Models and Tailorable Information Viewing on Group Decision Making Outcomes and Processes Simulation models can be useful decision aids for many kinds of problems, and have been incorporated into Decision Support Systems for single users. But many important problems require group decision making, and effective ways of making simulation models available to groups need to be developed. This research evaluates the usefulness of computer simulation models in group policy negotiations, using laboratory and real-world tasks in the domain of river basin management. It compares five different modes of model access during negotiations, ranging from no access at all to free access controlled by the group of negotiators as a whole. Since different negotiators in a group typically have interests in different aspects of a policy decision this research explores the usefulness of permitting negotiators to create tailored views of model results, so that they can quickly evaluate the impact of a policy alternative on their particular interests. Tailored views may exclude unimportant data and may compute and present useful transformations of raw model results. The results of the proposed studies should be of direct value in exploiting advances in computer simulation technology and user interface design in support of group policy negotiations. IRI-9120074 Marschak, Thomas; Varaiya, Pravin; Lawler, Eugene; Oren, Shmuel; and Vazirani, Umesh University of California, Berkeley $300,053 - 12 mos. Coordination Mechanisms: Theory and Applications This award is for continued support of an interdisciplinary research program on problems of coordination. The research began in 1989 when the Principal Investigators were successful competitors in the first NSF Special Initiative on Coordination Theory and Technology. The Principal Investigators' disciplines are Computer Science, Electrical Engineering, Economics and Operations Research. Some of the problems under study are primarily theoretical; some of them require computational experimentation; and some of them model the tasks of real organizations. The problems may be viewed in the following way: a designer is given a organization-a group of persons/processors each of whom has private information about some aspect of the organization's changing environment. The organization's task is to compute a specified function of its current environment; often the value computed can be interpreted as an ACTION that is the organization's desired response to its environment. The designer has to construct a MECHANISM (sometimes called a "protocol") which (1) tells each person/processor what to do (what computations to perform, what messages to send to whom) at each of a sequence of steps, and (2) achieves the task when the sequence stops. The designer wants the mechanism to be quick and cheap-according to several interesting cost measures-as compared to the other mechanisms that also achieve the task. The investigators also conduct an interdisciplinary Seminar on Problems of Coordination-in which the investigators, graduate students and visitors take part. The investigators also guide dissertations on coordination problems which often take interdisciplinary approach. IRI-9204861 Morris, James H. and Chandok, Ravinder Carnegie Mellon University $311,642 - 12 mos. Computer-Support for Collaborative Writing The goal of this project is to continue development of a "work in preparation" (PREP) editor to study collaborative writing relationships. The proposed work builds on previous work that resulted in design of a prototype editor that supports some important cognitive needs of social Interaction such as (1) the recognition of individual contributions to a collaboration and (2) the detection of change from one version to another. While the project continues to address further issues in supporting the cognitive needs of writers working together, the project also begins to work on supporting the social needs of writing groups, needs such as flexible sharing of partial products and managing parts of the writing task. For example, to support flexible sharing of partial products, the project defines a set of parameters that will allow users to control what they exchange (character, paragraph, column, document), when they exchange it (automatically, upon request) and the speed of exchange. The work on this front makes a theoretical and practical contribution to computer science. The work on supporting the cognitive needs of collaborative writers leads to embedding in text and graphic editors more general, coherent an consistent models for annotations. The work on supporting the social needs of collaborative writings groups leads to the development of new protocols for collaborative documents, modeled as distributed, disconnectable databases. IRI-9270323 Morris, James H. and Chandok, Ravinder Carnegie Mellon University $27,572 - 12 mos. Computer-Support for Collaborative Writing This is a small supplement to award (IRI-9204861) for the purpose of adding a graduate student and some computer equipment to aid in the development and testing of audio components for the PREP Editor system being studied and developed at Carnegie Mellon University. See above for the project description. IRI-9248671 Olson, Gary and Soloway, Elliot M. University of Michigan, Ann Arbor $100,159 - 12 mos. Technology Support for Collaborative Workgroups This is a supplement to award IRI-8902930, "Technology Support for Collaborative Workgroups." This supplement is to allow the research group to conclude data collection and analysis that has been underway during the grant and to write results and submit articles for publication. This project examines how a small group of collaborators works together to design software requirements, and what impact--good or bad--the use of groupware (integrated hardware/software for group processes) might have on this process. Scientists from the University of Michigan will cooperate with colleagues from the Software Technology Program at Microelectronics and Computer Technology Corporation (MCC) and Arthur Anderson & Co. The specific projects represent the blending of skills in the building of computer systems, the empirical analysis of human behavior, and the application of theory, both from technology and from cognitive science, to the design and analysis of technology augmentation of work. The approach stresses the need to use a vision of future technology environments to navigate the design space of collaborative technology, but uses the science base of behavioral science on the one hand and computer science and information systems on the other to assist in the development of technology experiments. The research will pursue the following specific activities: Studies of current practice-- 1. An analysis of collaboration from interview data already collected by MCC 2. Observational studies of collaboration as they occur now; Studies of groupware supported collaboration-- 3. Studies of collaboration using MCC groupware prototypes 4. Studies of collaboration using University of Michigan groupware prototypes. The project will use observational methods to study group behavior in situ, and will use successive generations of groupware prototypes as experimental tools for better understanding the collaborative process itself. The goal of the project is to contribute to the science base of collaboration theory, which is a blend of the behavioral sciences and the computer and information sciences interested in understanding how autonomous agents coordinate their behavior in the pursuit of a common task. IRI-9217527 Olson, Gary M. and Malone, Thomas P. University of Michigan, Ann Arbor $81,896 - 12 mos. NSF Coordination Theory and Collaboration Technology Grantee Workshop This award is to conduct a workshop composed of NSF grantees whose awards were made in the new area of inter-disciplinary research known as Coordination Theory and Collaboration Technology (CTCT). The purposes of the workshop are: First, to provide a forum for the exchange of knowledge in the form of perspectives, findings, methods and theory for this young, interdisciplinary field. Second, to provide an opportunity for the field to coalesce and grow. While certain subsets of the community of the researchers mingle at various professional meetings, as a whole, these investigators do not have regular contact. Third, this provides an opportunity for the research from these projects to be made more visible to the Washington based research funding and policy community. This workshop represents a planning opportunity for the IT&O Program. The future research agenda for CT2 funding and community building is an important part of the workshop program. The workshop is held on July 8 through July 10, 1992 in Washington, DC. A workshop report will be prepared and made available through publications and in response to requests. IRI-9249060 Perlin, Kenneth H. New York University $62,500 - 12 mos. PYI: Graphical Interface Alternatives to Window Systems The PI works on design of interfaces for collaborative environments. The current research is on designs that allow one to see the layout of many other collaborator's surfaces at a glance. This research aims to provide a recursive system of indefinitely zoomable magnifying glasses through which one can read, write, or create cross- references on an indefinitely enlargeable (zoomable) surface. A number of algorithms for dynamically displaying such information in "real time" on unenhanced bit-mapped workstations have already been developed by the author. The research will continue by defining and implementing the basic collection of functions in the interactive zoomable window system. Then, the system will be extended across a network, thereby making available a shared information space which will allow an arbitrary number of users to work in a common virtual environment, each possibly enjoying a different view of this environment. The researcher will also build applications, including multiscale graphical "yellow pages" and shared spreadsheet systems, to test the real benefit of the implemented system. IRI-9208319 Riedl, John University of Minnesota $48,936 - 12 mos. Flexible Collaborative Software Engineering This is collaborative research with Prasun Dewan, Purdue University (IRI-9208319). Much of software engineering-design, programming, debugging, testing, code reviews, program demonstrations, and program management-requires collaboration among multiple users, possibly geographically dispersed. In this project the researchers propose to investigate flexible support for the activities of cooperating software engineers. In particular they propose to investigate an approach that offers (i) a wide range of concurrency control mechanisms including serializable transactions and interactive transactions; (ii) fine-grain access control for collaborative applications that allows a user to recover from other users' mistakes and explore various alternatives with them, (iv) a multidimensional inheritance model that allows collaboration parameters to be specified independently in hierarchies of groups of objects and groups of users; and (v) multimedia support integrated with concurrency control, access control, and undo/redo. The unique features of this research are its focus on the relationships of these five goals, on flexible mechanisms for achieving each of the goals, and on structured specification of options by end-users. The researchers will use results from previous research done in flexible coupling, long transactions, adaptive concurrency control systems. inheritance, single-user undo/redo, capability-based protection, collaborative tools, and software engineering tools, They plan to implement the approach by extending an existing system developed at Purdue and Minnesota. They will evaluate the performance of the various collaboration methods and study how a particular method can be automatically chosen by the systems based on a minimum performance level requested by the user. In addition to an understanding of the applicability of collaborative technology to software engineering, the research will produce prototype software engineering tools to demonstrate the novel aspects of the research. IRI-9122447 Rohrbaugh, John and Reagan-Cirincione, Patricia SUNY at Albany $99,693 - 15 mos. Improving the Accuracy of Forecasts: A Process Intervention Combining Social Judgement Analysis and Group Facilitation Considerable empirical evidence suggests that groups seldom do as well as their best members on judgement tasks. To address this problem, a Group Decision Support System (GDSS) has been developed that incorporates techniques for reducing both the problems with interaction processes and the problems with cognitive processing that groups face when making collective judgments. A preliminary evaluation of this GDSS demonstrated that the process enabled groups to perform significantly better than their most capable members on cognitive conflict tasks. However, experimental conditions were not sufficiently representative of environmental conditions to warrant broad generalization of the results. This research extends the previous work by using judgment tasks frequently faced by managerial decision makers, improving the quality of the GDSS, and increasing the motivation of the participants. In the initial phase of the study, two expert forecasting tasks will be developed for use in the experiment and professional facilitators will be hired and trained to apply the process correctly. Next, an experiment, involving 24 five-member groups, will be conducted. To ensure that participants are motivated to perform well, they will be paid differentially according to the quality of their group judgments. In the final phase of study, the process intervention will be introduced into several organizational settings. IRI-9117011 Ryan, William B.F. Columbia University $20,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Marine Geology & Geophysics Program - Total Award $59,977) Visualization Tools for Large Seafloor Databases Accessed Visualization tools are being developed to allow the scientists, teachers and students to view and manipulate representations of the ocean floor directly on the screens of their desktop computers. Many different types of pictures can be examined, including maps, perspective views, pseudocolor imagery, shaded relief diagrams and 3-D stereograms. The raw data from which the views are created include bathymetry from multibeam and phase intererometric sonars and backscatter from side-looking sonars. Using electronic communication over Internet, the users connect from their computer (the server) into a client computer at Columbia University to begin the visualization session. The client returns to the user with menus and dialog boxes for the purpose of selecting the display type, map boundaries, color tables, fonts, data files, etc.. Moments later the representation appears in a window on the user's screen with computations run in the client computer. The user, given a responsive, intuitive and interactive interface, is liberated from the chores of locating and maintaining large complex databases and software code. The users find their work accelerated by an immediate response to a query and by a greater depth of understanding from advanced graphic representations of very large datasets. IRI-9015407 Schatz, Bruce R.; Peterson, Larry L.; Hudson, Scott E.; and Ward, Samuel University of Arizona $82,468 - 12 mos. (Jointly funded with the Numeric and Computation Program, the Software Systems Program, the NSFNET Program and the Division of Biological Instrumentation and Resources Program - Total Award $400,099) Systems Technology for Building a National Collaboratory This project is to develop the systems technology necessary to build a nationwide information infrastructure for the scientific community and to demonstrate this technology by applying it to a mini-collaboratory for a specific community of molecular biologists that studies the genetic structure of the C. Elegans nematode. The goals are to collect the community knowledge into a digital library, develop the technology to manipulate the library, and to learn how to facilitate effective utilization of this technology for sizeable communities. This mini-collaboratory will take advantage of the underlying communication support provided by NSFNET to support users nationwide. The researchers address several research problems necessary to the successful design of the mini-collaboratory. These include designing bulk transfer protocols that facilitate the rapid movement of data across wide area networks, discovering efficient data clustering and catching strategies, providing a uniform interface for displaying, editing, searching and grouping a wide range of complex objects, and supporting directory services that can be used to locate many distributed resources. IRI-9247049 Schatz, Bruce R. University of Arizona $10,000 - 12 mos. REU: Systems Technology for Building a National Collaboratory This is an REU supplemental award to IRI-9015407, above. This proposal will add two students who will work closely with the research team. One student will do research on information retrieval and/or network protocols while the other will continue research started as an undergraduate last year on representing the literature of the research community in electronic form. IRI-9257252 Schatz, Bruce R. University of Arizona $25,000 - 12 mos. NYI: NSF Young Investigator This project is building a model "electronic community system", a digital library containing all the knowledge of a community of molecular biologists and a software environment to manipulate this library across the NSFNET, with the first release currently running in 20 biology laboratories. Several studies are being pursued in collaboration to investigate technology and sociology issues in constructing electronic communities. An "information space" is the representation used to support uniform transparent retrieval from many distributed data sources; it consists of uniform objects which package items from the sources and interconnect them to form a relationship graph. There are data model issues in providing uniformity across multiple data types and network caching issues in providing transparency of interaction across wide-area networks. There are also information retrieval issues in supporting semantic-based retrieval from an automatically generated representation which includes a domain- specific thesaurus. Finally, there are sociological issues in how the communication patterns of the community change with significant electronic support. IRI-9123720 Sklansky, Jack University of California, Irvine $240,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence and the Knowledge Models and Cognitive Systems Programs - Total Award $330,000) Biologically Inspired Intelligent Classifiers This project will advance knowledge about design of intelligent agents. It has two objectives. The first objective is the development of mathematical insights and procedures for automating feature discovery and learning by evolution -- two biologically inspired forms of long-term learning -- in the construction of automatic pattern classifiers. The second objective is the incorporation of automated feature discovery and evolution in automatic analyzers of large-scale image data banks and time-varying visual signals. This project will work to achieve these objectives by building on the Principal Investigator's and his students' techniques of adaptive hyperplane placement, window training (an extended form of stochastic approximation), relaxed branch-and-bound search, and genetic feature selection. The project will implement and test the resulting mechanisms of long-term learning in three-stage neural classifiers of handwritten numerical characters, tree classifiers of medical images in a large data bank, and adaptive segmenter of moving objects in digitized video. This project will contribute to the technology of long term learning in intelligent machines and to multi-sensor robotic vision by automating the discovery of highly discriminating combinations of features from diverse sensors. IRI-9241467 Smith, John B.; Smith, F. D.; Calingaert, Peter; Jeffay, Kevin and Holland, Dorothy University of North Carolina, Chapel Hill $90,000 - 12 mos. (Jointly funded with the Interactive Systems Program, the Software System Program, and the NSFNET Program - Total Award $300,000) Building and Using a Collaboratory: A Foundation for Supporting and Studying Group Collaborations This is the second year funding of a three-year continuing award IRI-9015443. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. The central focus of this research is experimental observation of groups doing real collaborative work and using systems and communications media explicitly designed to aid in their tasks. Expected results are to expand understanding of how people collaborate and how to design systems that augment collaborative activities. To this end the project has five interdependent components: 1) a theoretical foundation for observing and understanding the social and cognitive aspects of group collaborations; 2) tools for rapid prototyping and reconfiguration of application environments for use by working groups, and multi-media communications to support multi-person interactions; 3) protocol analysis tools to record and study how individuals and groups interact through the networked computer environment; 4) application testbed systems (generic and domain specific) that can be used by groups engaged in real work; and 5) group studies and experiments to test system, social and cognitive hypotheses. IRI 9117084 Soloway, Elliot M. and Martin, William University of Michigan, Ann Arbor $35,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Interactive System Program - Total Award $77,715) Developing Guidelines to Providing Computer-based Support for Scientific Data Analysis The "human-computer interface" is paramount if scientists are to take full advantage of the vast quantities and types of data now coming available. Attention needs to be focused on more than the query language; rather, the interface needs to integrate scientific databases into the everyday work practices of scientists, e.g., data exploring, hypothesis generating and testing, report and chart making. The Task/Artifact Methodology is being employed; it consists of cycles of cognitive task analysis, system building, and testing in ecologically valid contexts, e.g., practicing scientists using the system on a daily basis. The resultant computer-based environment, ReV (Representations for Visualization) is user- and task-centered, as opposed to being technology-driven. ReV serves as the scientist's "notebook" permitting him/her to move among databases, hypotheses, reports, charts, etc. ReV is being designed for scientists in nuclear engineering and for scientists engaged in global change research. As computers become integral to the moment-by-moment work practices of scientists, well-designed tools become critical. This project, then, serves as a model for how scientist-centered computing environments can be developed and deployed. IRI-9117153 Sparr, Ted M.; Meeker, Loren D. and Bergeron, R. Daniel University of New Hampshire $50,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $149,999) Integrating Data Management, Analysis and Visualization for Collaborative Scientific Research The goal of this multidisciplinary project is to design and prototype a new approach in database environments to support collaborative scientific research. The prototype integrates scientific data visualization and mathematical and statistical analysis tools with database support in a highly interactive environment. A new model for scientific data is founded on the notion that a query of a scientific data base conceptually creates new derived data whose relationship to the parent database is defined by the query. Each query, in principle, leads to the discovery of additional structure in the data that is either explicit in the results of the query, or hypothesized by the scientist(s) from results of queries. The system uses a process flow graph to represent queries. The project, carried out by a team of two computer scientists, an applied mathematician, and scientists from earth science, civil engineering and atmospheric science, contributes to ongoing research in the fields of environmental biology and chemistry, oil reservoir analysis, and polar ice core study. This work will produce a new data model for scientific data and will design and prototype an integrated data management, analysis and visualization environment to support interdisciplinary scientific investigation. IRI-9209880 Yun, Xiaoping University of Pennsylvania $52,500 - 36 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $100,000) Coordination of Mobile Manipulators This research investigates coordination of multiple mobile manipulators for grasping and transporting large and irregularly shaped objects. A mobile manipulator considered in this study is composed of a mobile platform for locomotion and a manipulator for grasping and manipulation. A large object without special features such as handles cannot easily be grasped by the conventional end effectors such as jaw grippers or multifingered hands. A new approach based on the concept of enveloping grasp and manipulation by multiple mobile manipulators is proposed. The major contribution of the proposed study will be control algorithms for coordination of locomotion and manipulation of a mobile manipulator with the motion of the manipulator being constrained, and coordination of two mobile manipulators for transporting large objects. An experimental system will be developed to test and evaluate the proposed coordination algorithms. DISTRIBUTED COMPUTING/SHARED ENVIRONMENTS IRI-9248188 Boland, Richard Case Western Reserve University $27,376 - 12 mos. Coordination of Distributed Decision Making in a Corporate Planning Environment This is a small supplement to award IRI-9015526 for the purpose of allowing field research at the Digital Equipment Corp. Maynard, MA site to continue so empirical results from the project, described below, can be forthcoming. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. In distributed decision environments, intelligent agents at network nodes interact with private databases, common databases, and each other to coordinate their independent actions. This project explores how certain human, organizational, and technological elements in a distributed corporate planning setting, condition the level of coordination and integration of the agents' individual plans. The researchers work with the Management Systems Research Group of Digital Equipment Corporation (DEC) to develop, test and evaluate decision support facilities for their individual, autonomous planning decisions. The researchers evaluate the outcomes of systems use among a set of interdependent business unit managers in terms of the coordination and integration of their individual plans taken as a whole, as well as the level of trust, commitment, understanding a cooperation they experience in this distributed planning setting. The distributed decision making of these autonomous, yet interdependent, agents is a setting of very high equivocality and uncertainty. The research is action-based, selectively adding structure and feedback capabilities too increase the richness of communication supported by the systems. It is a theory driven, empirical, investigation of how increased communication richness of a decision support facility affects the experience and outcome of a distributed planning systems. IRI-9211418 Chen, Hsinchun University of Arizona $63,804 - 24 mos. RIA: Building a Concept Space for an Electronic Community System Researchers in a scientific community or organizational members in a company are often overwhelmed by the amount of current information, the subject and computer knowledge required to access this information, and the constant influx of even larger amounts of new information. Using an electronic community system as a testbed for the aim of the research is to address specifically the information sharing, retrieval, and filtering problems through the use of a variety of AI and statistical techniques, including knowledge acquisition, automatic thesaurus generation, and neural network computing. A Concept Space is designed to represent high- level community knowledge such as topics, subjects, techniques, and people in terms of nodes and their weighted relationships in a structure that is akin to a semantic network. This Concept Space can be perceived as a high-level conceptual description of the basic information items that reside in an Information Space. It is created mainly from a knowledge acquisition process involving subject experts and from an automatic cluster analysis of documents stored in the community's online, textual databases. Two kinds of system interface are considered both making extensive use of the Concept Space. A browsing interface permits users to traverse this weighted concept network themselves, select concepts of interests, and access its underlying information. An activation interface based on a Hopfield network spreading activation algorithm traverses the network automatically and identify concepts most relevant to searchers' needs, with very little searcher intervention. Both types of interface are designed to provide user-friendly access, to facilitate learning, and to avoid the need for extensive user computing knowledge. This research is designed to provide insights for the development of more timely, comprehensive, and accessible electronic community systems and computer-based organizational memory. IRI-9249062 Durfee, Edmund H. University of Michigan, Ann Arbor $62,500 - 12 mos. PYI: Real-Time AI, Cooperative Problem Solving, and Intelligent Systems This PYI will continue his research on distributed artificial intelligence (AI) with an emphasis on building intelligent computing systems for dynamic, multiagent applications. The research uses blackboard systems and other integrating architectures that allow an agent's perception, reasoning, communication, and motor components to collectively control an agent's behavior. The PI is also pursuing work in real-time artificial intelligence. He is working on developing preliminary theories of real-time reasoning that account for both meeting deadlines and for negotiating to change deadlines. This work on techniques for negotiating ties into his additional research interest in uncovering the general principles underlying coordination between intelligent agents. The goal is to build interdisciplinary theories of coordination, drawing on concepts from management science and operations research as well as AI, and to embody these theories in practical mechanisms for coordinating multiple intelligent computing systems. The near term plan is to develop a core of interdisciplinary concepts using a hierarchy of behavioral specifications as a common representation, with the long-term objective of developing fundamental theories of a science of coordination. The researcher is also interested in applying techniques for coordinating AI systems to problems in human-human and human-computer interactions. IRI-9216094 Hendler, James and Agrawala, Ashok K. University of Maryland, College Park $5,000 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Database and Expert Systems Program - Total Award $15,000) Workshop on Artificial Intelligence in Real Time This workshop on "Artificial Intelligence in Real-Time" is aimed at examining how AI systems can both be supported and can help to support real-time operating systems. This area is of critical importance as AI systems need to function in the support of such critical applications as nuclear power plant control, aircraft operation, hospital life support systems, and military command and control, among others. The workshop, to be held at the University of Maryland in the Fall of 1992, brings together members of both AI and real-time communities to explore issues of mutual interest. A report, produced by this workshop, will aid in the planning activities of the KMCS, RMI, and ITO Programs. according to NSF Manual 10,122.1.3 (b), such proposals are exempt from peer review. IRI-9246974 Hiltz, Starr R. New Jersey Institute of Technology $5,000 - 12 mos. REU: Distributed Group Support Systems This is an REU supplemental award to IRI-9015236. This supplement adds one undergraduate students to the project. The students will assist in the full range of project activities involving the design, testing and evaluation of new group decision support tools under development in the research. The larger project is described below. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. This project involves an integrated program of theory building, software tool development and assessment, and controlled experiments in distributed group support systems. Such systems embed group decision support systems tools and procedures within a computer-mediated communication system. The primary objective of the project is to build a general theory, supported by empirical evidence, to understand how variations in group structures and software tools affect the process and outcome of decision making. This multi-disciplinary effort focusses on synchronous computer conference, in which participants are distributed locations, and on a synchronous computer conference, in which participation is distributed in time as well as space. IRI-9210918 Jones, Patricia University of Illinois $89,722 - 36 mos. RIA: Cooperative Support for Distributed Supervisory Control Human-machine interaction is a critical component of complex, highly automated systems such as space shuttle ground control and flexible manufacturing systems. In such systems, typically a team of human operators acts as supervisory controllers: monitoring data adjusting parameters, and detecting, diagnosing, and compensating for failures. In the past, most research efforts in this area have focused on a single human operator: modeling one operator's interaction with a complex system; designing single-user displays, and so forth. With the advent of knowledge-based technology, more"intelligent" tools are appearing in the context of monitoring and supervisory control. Still, such tools are designed for use by a single operator. The intent of this project is to build models and methods to support cooperation between human supervisory controllers and between humans and "intelligent associate systems" in supervisory control. Specifically, the three objectives of the proposed are to: 1) Develop a framework for modeling distributed supervisory control; 2) Develop prototype knowledge-based tools based on the modeling framework; and 3) Apply the modeling framework and collaborative knowledge-based tool design concepts to supervisory control environment: satellite ground control at NASA Goddard Space Flight Center. IRI-9211165 Little, Thomas Boston University $100,000 - 36 mos. RIA: Enabling Technologies for Multimedia Computing Audio, video and graphics will play a major role in shaping the evolution of computers and communications. To this end, the formulations of new mechanisms for multimedia computing and visualization are proposed, with emphasis on time-dependent data. Specific research objective are (1) the development of protocols for multi-media communications. (2) the modeling of time-dependent data for storage, communication, and interfaces; (3) the development of new paradigms for interactive digital video; and (4) the provision of time-dependent information retrieval in a distributed database and shared-object environment. An operational distributed multi-media information system will be developed to demonstrate the work. The prototype system will be designed to support scale-up for general-purchase multi-media computing including computer-aided instruction and collaborative work. Current technology will be used for data compression, communication, and interface toolkits; tailored to implement the theoretical results. The proposed research consists of the development of communications protocols and other mechanisms to enable interactive multimedia computing systems. Although many application-specific systems have been developed, effective support for multimedia applications in a general distributed environment is not widely studied. This work emphasizes the enabling technology rather than applications per se; however, an application will be developed that may lead to innovative approaches to learning, human-computer interaction, and computer-aided education. IRI-9249487 Pasquale, Joseph University of California, San Diego $62,500 - 12 mos. PYI: Decentralized Control in Large Distributed Systems This PYI is performing decentralized resources control, i.e., collections of decision-making agents which reside on a geographically distributed set of computers and which control resources so that work can be carried out in an integrated and coordinated fashion. Finding good methods for adaptive decentralized control addresses the question of how to make task and resource allocation choices correctly and efficiently in light of the formidable problems which arise as a result of distributing control. Probably the most difficult of these problems is that multiple agents must make good fast coordinated decisions based on uncertain and differing views of the global systems state. The research approach is experimental. The Principal Investigator will create a laboratory to work on problems of load balancing, network routing and distributed sensing and interpretation. The laboratory will consist of multiple workstations with a file server, interconnected by a variety of different networks for work on decentralized coordination with notes acting as controlling agents numbering in the hundreds of thousands, and development software, including prototype object-oriented programming environments and expert system environments. IRI-9258392 Pollack, Martha E. University of Pittsburgh $25,000 - 12 mos. NYI: NSF Young Investigator This research project addresses the design of autonomous agents that are capable of intelligent behavior in dynamic environments with multiple agents. The focus is on the development of effective techniques for managing computational resources, with particular attention being given to meta-level control strategies, i.e., procedures that allocate and monitor the agent's base-level reasoning resources. Such procedures trade optimality of performance for time-efficiency, a trade-off that is essential in dynamic or multiagent environments. Experimentation is being conducted, using a simulation system, to determine the relative advantages that result from the deployment of alternative reasoning strategies in various environments. IRI-9120358 Tyce, Robert C. University of Rhode Island $20,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Marine Geology and Geophysics Program - Total Award $59,967) Graduate School of Oceanography Visualization tools are being developed to allow the scientists, teachers and students to view and manipulate representations of the ocean floor directly on the screens of their desktop computers. Many different types of pictures can be examined, including maps, perspective views, pseudocolor imagery, shaded relief diagrams and 3-D stereograms. The raw data from which the views are created include bathymetry from multibeam and phase intererometric sonars and backscatter from side-looking sonars. Using electronic communication over Internet, the users connect from their computer (the server) into a client computer at Columbia University to begin the visualization session. The client returns to the user with menus and dialog boxes for the purpose of selecting the display type, map boundaries, color tables, fonts, data files, etc.. Moments later the representation appears in a window on the user's screen with computations run in the client computer. The user, given a responsive, intuitive and interactive interface, is liberated from the chores of locating and maintaining large complex databases and software code. The users find their work accelerated by an immediate response to a query and by a greater depth of understanding from advanced graphic representations of very large datasets. IRI-9213460 Wilkenfield, Jonathan and Kraut, Sarit University of Maryland, College Park $98,972 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $103,038) A Strategic Model of Negotiation for Autonomous Agents Distributed Artificial Intelligence (DAI) has been concerned with the design of automated agents which can interact effectively in order to satisfy their goals. This project will contribute to both the theory and design of efficient autonomous agents which are able to communicate and negotiate in such a way as to enhance the possibility of reaching mutually beneficial agreements. The authors are developing a strategic model of negotiation that takes the passage of time during the negotiation process itself into consideration. This model is used in the creation of a prototype of an autonomous agent that participates in situations where negotiations are required, and where time constraints exist. A preliminary model has been constructed, based on the differential impact of time on the parties, the effects of two-party and restricted N-party environments, and full information. The researchers are extending the model to deal with situations where the information available to the agents is incomplete, to more complicated N-party situations, and to agents with limited rationality. Parallel to these theoretical extensions, the design of the automated negotiator is being developed and a prototype created. The strategic model is used to identify efficient strategies for the automated negotiator, for various negotiation situations. A library of meta-strategies in which the strategies will be presented using the finite-automata framework is being built. The automated negotiator identifies the appropriate parameters of its situation, searches for the appropriate strategy, and negotiates according to that strategy. IMPACT/POLICY IRI-9210398 Barua, Anitash University of Texas, Austin $66,250 - 36 mos. RIA: Economic Models of Strategic Information Technology Investments Information Technology (IT) is considered as a strategic weapon that can be used to secure competitive advantage. Most of the relevant studies in this domain seek to make the case through the development of conceptual frame works, and ex post examples of successful deployment of strategic IT. Empirical evidence, however, suggest that many strategic IT investments became unfortunate strategic necessities rather than means of gaining competitive advantage. Thus, many questions regarding the economics of strategic IT remain unanswered. The research analyzes IT investments under various competitive situations through the development of formal economic models. The research consists of three studies in distinct but related areas within strategic IT. The first study involves IT related quality competition in a setting where two firms have the opportunity to improve the quality of their services through IT investment. The research analyzes the division of technology related benefits between the firms and the consumers, and the impacts of technological improvements on investment levels. The second study focuses on competition through IT innovation. It analyzes the IT investment incentives of a market leader and a challenger, and characterizes situations in which there may be a reversal of such incentives. Adoption of inter-organizational systems is the topic of the third study. In particular, the optimal adoption times of an Electronic Data Interchange network by supplier firms are studied. The research is based on the Industrial Organization (IO) approach with both theoretical and empirical components. Game theory is used as a primary theoretical research tool. The studies enhance the understanding of many economic aspects of IT investments in competitive environments. IRI-9249140 Flamm, Kenneth Brookings Institute $15,500 - 12 mos. Participation in International Study of High Performance This is a small supplement to an existing grant IRI-9120646. The grant supports U.S. participation in a study of high performance computing in Japan, Germany and the U.S. This award is for U.S. Researchers participation in a comparative multinational study of networking and high performance computing initiatives. The study is being conducted under the aegis of the Organization for Economic Cooperation and Development (OECD). This proposal is to support U.S. researchers in the initial planning, design and early data collection phase of the project. Project members from Europe and Japan will be supported by their national governments. This project provides an excellent opportunity for the U.S. to learn about programs of other nations that relate to our high performance and also help us maintaining international competitiveness. IRI-9209321 Greenstein, Shane University of Illinois, Urbana-Champaign $89,808 - 36 mos. RIA: Evaluating the Economic Benefits from Improvements in Computer Equipment It has long been the goal of many researchers to measure the effects of the extraordinary technical improvements in computing markets. Yet is has been very difficult to place an economic value on such improvements. One means for doing so measures a buyer's satisfaction with a product, or "buyer's surplus" -- the difference between what a buyer was willing to pay and what he actually paid. A proper measure of buyer's surplus provides an index of the value to users from a cheaper system or a technically more proficient system. What has prevented research on surplus in the computer market is a absence of sufficiently detailed data on the micro- behavior of computer users. This research will fill this gap. It will develop and use a detailed and rich new data set containing about 70 percent of all medium to large computer users in the United States. A related goal of this research is to shed light on the many factors influencing the diffusion on new computer equipment. Since much technical improvement is embodied in new generations of hardware, the economy can only benefit from technical advance when that new generation of equipment is purchased. This research underlying the demand for computing. It will analyze the factors determining the decision to adopt technically advanced computing equipment. HRD-9250110 Gutek, Barbara A. California Institute of Technology $20,000 - 12 mos. (Jointly funded with the Division of Human Resource Development - Total Award $200,000) VPW: Computer Literacy in Workgroup This research assesses the effect of learning resources (e.g., presence of a local expert in a work group, information centers, corporate-sponsored courses) on computer literacy. The assessment will be accomplished through a follow-up study of 30 of 89 work groups studied in three waves from 1987 to 1989. Computer literacy was first assessed in these groups in late fall 1989, with the development of an instrument for measuring computer literacy. The main question now to be addressed is; in work groups in which computer use is already well-established, will computer literacy of the work group show an increase, decrease, or no change from 1989 to 1992? We anticipate finding instances of all three, we generally focus on the conditions which foster increased and decreased computer literacy over time. We will do this by collecting data from 30 work groups which vary in the computer literacy score of the group. We will select 15 groups that perform primarily clerical functions, from the large set of work groups which have previously been studied. Everyone in each of the select work groups will fill out short questionnaire and a two-week log of their interactions with other in the work group. This research focuses on the workplace as an environment for increasing the computer literacy of American society. The proposed research will explore the factors associated with computer literacy and computer expertise in ongoing work groups. The project furthers VPW program objectives to provide opportunities for women to advance their careers in science or engineering through research, and to encourage other women to pursue careers in these areas through the investigator's enhanced visibility as a role model on the host campus. The proposed activities which contribute to the second objective include; teaching a course, "Women and Men at Work;" giving lectures in other courses and/or public lectures; meeting with women's groups student, faculty, and staff); and mentoring students. IRI-9247052 Kling, Rob University of California, Irvine $5,000 - 12 mos. REU: Real Time Interactive Information Systems This is an REU supplemental award to award IRI-9015497. This supplement adds one undergraduate student to the project who will participate in data collection and interview activities associated with the project. In addition, the student will learn about the research concepts involved in the larger project through both a course of readings and through participation. The selection criteria are meritorious. The larger project is described below. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. This study investigates certain aspects of the role of computerized information systems as instruments of coordination in complex organization. Questions addressed in this research include: What kinds of coordination problems do computerized systems actually resolve, and to what extent? What social and economic impacts result from the use of such systems? The research gathers empirical data through cross sectional studies, comparative case studies and a longitudinal survey of manufacturing firms that have relatively structured systems that can be characterized as Computer Integrated Manufacturing. Results expected from this research include empirically grounded accounts of the role of computerized information systems in coordinating activities of complex organizations and increased theoretical understanding of the economic and sociological impacts of these information systems. IRI-9200205 Lucas, Henry C. and Baroudi, Jack New York University $101,518 - 12 mos. Information Technology and the Productivity of Professional Workers The purpose of this study is to 1) provide a large scale organizational test of an extended version of the Davis, et al Technology Acceptance Model (TAM) (Davis, F., Bagozzi, Warshow, P., "User Acceptance of Computer Technology": A Comparison of Two Theoretical Models.: Management Science, vol. 35, no. 9 (Aug. 1989, pp.982 \1003) and 2) to determine the extent to which the deployment and use of personal computer workstations leads to improved individual and organizational productivity. The research involves a study of the antecedents use of a stock broker's workstation system and the subsequent impacts on individual and branch office productivity as related to the workstation's use or non use. The research model is drawn and extended from a model proposed by Davis, et al (1989). The research design includes a longitudinal study with groups of offices implementing the new system compared against control groups with the system. The controlled research design helps to determine the impact of the new technology on individual and group productivity using economic outcome variables. HRD-9252943 Marshall, Catherine R. Oregon Graduate Institute $20,000 - 12 mos. (Jointly funded with the Division of Human Resource Development - Total Award $187,292) VPW: New Methodologies for Evaluating Interactive Technologies Since we lack fundamental means for evaluating the merits of non- traditional technologies, this research looks at extending currently impractical analytic techniques, such as conversation analysis, to more easily used measures of interaction effectiveness. In a laboratory experiment, face-to-face communication will be compared with two forms of mediated communication: audio-only and audio-plus-video. Two types of tasks will be studied: one with a strong visual-spatial component and one with a strong verbal-logical component. Task performance will be measured and questionnaires given. Using the results of the objective measures of effectiveness, systematic, non-anecdotal methods for conversation analysis will be developed using the videotaped protocols. The research will make a substantial new corpus of task-related interaction available to the scientific community. This project explores and evaluates the use of new techniques for measuring the effectiveness of interactive technologies, including teleconferencing systems, multimedia computing systems and portable office environments. This product furthers VPW program objectives to provide opportunities for women to advance their careers in science or engineering through research, and to encourage other women to pursue careers in these areas through the investigator's enhanced visibility as a role model on the host campus. The proposed activities which contribute to the second objective include: mentoring and counseling women students (in department with no women faculty); teaching one course in interactive systems; co-leading the Interactive Systems Reading Group; advising students, and OGI as an institution, on policies and mechanisms necessary for the development of careers for women in science and engineering. IRI-9214683 Panzar, John C. Northeastern University $25,000 - 12 mos. Twentieth Annual Telecommunications Policy Research Conference This award is for partial support for the Twentieth Annual Telecommunications Policy Research Conference (TPRC) in Solomons, Maryland, September 12-14, 1992. The conference provides a forum for interactions between national and international scholars in basic and applied research (from engineering, communications and information science, operations research, economics, law and public policy) and leaders in communications and information industries, government policy makers and specialists from public-interest organizations. A major function of the conference is to provide a forum for the presentation of preliminary studies to alert policy makers to new work relevant to their interests and to provide an opportunity for early critical review by peers. After revision, based in part on comments from their TPRC presentation, the papers are published in professional journals or elsewhere. All TPRC papers are collected and made available upon request from the TPRC office in Washington, D.C. IRI-9213628 Rule, James B. SUNY Stonybrook $88,133 - 12 mos. Computing in Organizations, 1985-1992 This study develops and analyzes comparative data on change in the uses of computing by a representative selection of greater New York private sector firms. Returning to the sample studied by Attewell and Rule in the late 1980's, the proposed research will document changes in computing use and in other basic features of the original 184 establishments. Detailed telephone interviews will be carried out with as many of these establishments as possible, and site interviews will take place in a sub-sample of about fifty. Resulting analyses should illuminate the form and extent of change in computing activities over roughly seven intervening years, as well as the determinants of such change. INFORMATION/DECISION THEORY IRI-9248672 Ballard, Dana H. University of Rochester $80,000 - 24 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Knowledge Models and Cognitive Systems Program - Total Award $225,000) Animate Robotics Vision This award is a two-year creativity extension to a current three-year award (IRI-8903582), based on creative accomplishments in active vision and reinforcement learning. The PI has proposed extensions to his work in combining deictic primitive behaviors (primitives that dynamically reference the world rather than depending on models) to produce the complex behaviors needed for intelligent robotic agents to perform practical tasks in active vision. Topics will include economical search of three- dimensional space for small objects, speedup of the learning algorithms, and development of a set of qualitative grasping primitives, integrated with visual feedback, as a natural extension of the vision primitives already developed by the PI. This last is a promising alternative to the fine, open loop control used in conventional robotic systems. IRI-9021437 Cowan, George and Simmons, Leonard M. Santa Fe Institute $63,857 - 12 mos. A Broad Research Program on the Science of Complexity The Santa Fe Institute (SFI), a non-profit research and graduate teaching organization, proposes to continue its programs of interdisciplinary research and education on the principles that determine the dynamic behavior of complex systems. The Institute is founded on the premise that there are common principles that determine the behavior of various complex systems, especially systems that are sufficiently complex to learn and adapt. Their approach is to understand these principles by studying specific systems in the context of more general themes. Since 1987 it has operated a center in Santa Fe where leading scholars in the physical, computational, life, and social sciences come together. The project includes: a) A program on the basic physics, mathematics, information sciences, and computational aspects of complexity including work on the shared properties of complex systems, and innovative computer modeling. b) A residential research program on adaptive computation employing the techniques of general algorithms and classifier systems and including work on developing computational and visualization tools. c) A residential research program on economics as a complex, adaptive system. d) A program on life sciences, that includes modeling of the immune system, generic data analysis, and models of protein folding. e) A Summer School on the Sciences of Complexity in Santa Fe, managed by SFI in collaboration with a number of other institutions. IRI-9207633 Freuder, Eugene C. University of New Hampshire, Durham $91,562 - 36 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Cross Disciplinary Activities Office - Total Award $186,250) Constraint-Based Reasoning: Computation and Representation Constraint-based reasoning has been used in many areas of artificial intelligence: vision, language, planning, diagnosis, scheduling, configuration, design, temporal reasoning, defeasible reasoning, truth maintenance, qualitative physics, logic programming, and expert systems. This research focusses on the constraint satisfaction problem paradigm, which underlies many of these applications. The research objectives are: (1) characterizing tractable problem classes, (2) developing new algorithms, (3) addressing knowledge representation issues, (4) addressing knowledge acquisition issues, (5) studying extensions of the basic constraints satisfaction problem paradigm. IRI-9216941 Kumar, Vipin University of Minnesota $48,264 - 12 mos. SGER: Parallel Multi-Agent Planning This is a Small Grant for Exploratory Research (SGER) to apply existing techniques in parallel processing to the problem of planning with multiple agents. A model is designed which allows a view of the problem as a sequence of interleaved tree search and satisfaction steps. The research investigates various factors influencing the performance of this model when the search and constraint satisfaction are performed on parallel computers. These factors include partitioning the plan generation process among various processors, the nature of various distributed data structures and the effect of ordering heuristics on complexity of search space. The new models developed are implemented and validated on the new generation CM5 and the nCUBE2 parallel computers. IRI-9116399 Nilsson, Nils J. Stanford University $32,905 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program, the Knowledge Models and Cognitive Systems Program and the Database and Expert Systems Program - Total Award $98,715) Research On Autonomous Agents This is the first year funding of a three-year continuing award. This research is concerned with developing autonomous, cooperating, adaptive computational agents. Expected applications of this research are in the control of mobile robots capable of performing delivery, maintenance, and/or construction tasks and of agents that gather and manipulate symbolic information over computer networks. The research will concentrate on bounded-time action computation and learning mechanisms and on combining these components in integrated agents able to function cooperatively in dynamic, uncertain environments. IRI-9246701 Smith, Vernon L.; McCabe, Kevin and Rassenti, Stephen University of Arizona $5,000 - 12 mos. REU: Experiments in Computer Coordinated Decentralized Resource Allocation Systems This is an REU supplemental award to IRI-8921141. This supplement adds an undergraduate student to the project. The student will learn how to run experiments and to assist in the handling of subjects. The student will also help to chart results, prepare databases and learn software development as it applied to the experimental environment. The larger project is described below. This project is to conduct laboratory experiments to produce data for analysis of computer coordinated decentralized resource allocation systems. The work concentrates on a new research initiative to develop computer coordinated bidding models of (1) process allocation and (2) capacity planning for central processing (computer) unit (CPU) services. (1) is formulated as linear program, while (2) is an integer program, for maximizing system surplus value: the inputs are bides for processing time, or component values and budgets in the capacity problem; the outputs of the reported surplus maximizing algorithms are service time allocations in the process allowed problem and prices, or components to be purchased, and cost allocations in the capacity problem. Allocative efficiency will be evaluated by the ration of actual monetary surplus values and budgets. Some research will also continue on some what related problems in networks-- particularly networks of natural gas pipelines and of electric power supply. This work is relevant to understanding how to design and to efficiently allocate time and equipment in decentralized computer networks. SPECIAL PROJECTS IRI-9120756 Hoffman, Lance George Washington University $44,907 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $54,907) Partial Support for the ACM Second Conference on Computers, Freedom, and Privacy This is a multi-disciplinary conference that brings together leading researchers, commentators and policy makers concerned with issues of the impact of computer technology on society. The conferees survey the implications for privacy and freedom of the interaction of computers, computer data banks, and computer networks with electronic mail, computerized personal information direct marketing information and government data, including data related to human genomes. A book with ten articles on the subjects of the conference will be published. The Conference is scheduled for March 18 - 20, 1992 at the L'Enfant Plaza Hotel in Washington, D.C. IRI-9220985 Mitchell, Tom M. Carnegie Mellon University $5,926 - 12 mos. Symposium on Cognitive and Computer Science: Mind Matters This Symposium entitled, Mind Matters, held at Carnegie Mellon University on October 25-27, 1992, features unpublished papers of a technical nature describing original research in the areas of computer science and cognition. It is held in honor of Allen Newell, who has made seminal contributions to computer science and cognitive science . He is generally regarded as one of the founding fathers of the field of artificial intelligence. Dr. Newellþs research for the past decade has focussed on the development of SOAR, as both an integrated architecture for building intelligent systems and as a unified theory of cognition. Soar is the unifying thread for this symposium. A member of the Soar community follows each invited speakerþs talk, discussing the implications of the results just presented for the Soar architecture. The speakers, leading scholars in computer science and cognitive research, had their topics are chosen to reflect the diversity of Professor Newellþs interests. The proceedings of the symposium are to be published as a book by Lawrence Erlbaum Publishers, as part of their Carnegie Symposium on Cognition series. INT-9123796 Tripathi, Satish K. University of Maryland, College Park $30,000 - 12 mos. (Jointly funded with the Division of International Programs - Total Award $45,250) Cooperative Research in Computer Science, Indo-U.S. Workshop August,4-6, 1992, Bangalore, India This project supports participation of a U.S. team of computer scientists and engineers in a Indo-U.S. workshop on "Cooperative Research in Computer Science" to be held in Bangalore, India, August 4 to 6, 1992. The areas of expertise to be represented include: robotics and computer vision, computer systems an networking, software engineering, and artificial intelligence. The objectives of the workshop are to: present papers by U.S. and Indian participants about the recent advances in the various topical areas, identify research needs in each of these areas, identify suitable areas for Indo-U.S. collaborative research, and establish firm collaborative research, and establish firm collaborations between U.S. and Indian scientists. Scope: This is the second Indo-U.S. workshop in the area of computer science. The first was held in 1989 in the city of Hyderabad, India with participation from India's main governmental and private research organizations with interest in computer research and applications. Since that time a number of collaborative activities have resulted. The present project is designed to continue and intensify this mutually beneficial interaction and collaboration. The project meets the objectives of the U.S.-India Cooperative Science Program. IRI-9246173 Tripathi, Satish K. University of Maryland, College Park $4,380 - 12 mos. Assignment and Allocation of Processors in Parallel Processing Systems This supplement provides travel support for a trip to India in January 1992. Dr. S.K. Tripathi (University of Maryland) will be visiting Indian institutions to identify and contact Indian researchers and discuss and plan a workshop described above. MESSAGE FROM THE PROGRAM DIRECTOR This past year the program was active in a variety of new issues affecting the field of human-computer interaction. In particular, three program planning workshops took place producing reports intended for wide dissemination. The topics were: Research Directions in Virtual Environments: This workshop took place during March 23 and 24, 1992, at the University of North Carolina at Chapel Hill. The report was published in Computer Graphics, vol. 23, No. 3, August, 1992. Spoken Language Understanding: This workshop was held in February, 1992 and a report has been issued, although at this time it has not yet been published in a journal. It is currently referenced as Technical Report No. CS/E 92-014, Oregon Graduate Institute, September 1, 1992. Facial Expressions: This is the latest in the series of planning workshops, and its report has not yet been issued. It took place in Crystal City, VA at the end of July, 1992. The report was completed early 1993. In all cases a number of participants, in addition to the steering committees shown in the workshop awards below, contributed significantly to the proceedings and to the reports, and their assistance is gratefully acknowledged there. On programmatic issues, in addition to the four major program elements shown below, we noticed an upsurge in the Small Business Research Innovation (SBIR) awards in the human-computer interface area. We also actively participated in Research Experiences for Undergraduates (REU), Research Initiation Awards, and other NSF programs. A significant new thrust of the program is the advent of virtual environments as a broad methodology for visualization of scientific data and possibly as an experimental testbed for a wide range of cognitive science experiments. Also to be noted is the integration of different modalities in multimedia interaction and, in particular, the emphasis in combining speech analysis and recognition with natural language research into the area of spoken language understanding. The latter approach will be evident in the coming year since a good part of the natural language activity has gravitated to this program. INTERACTIVE SYSTEMS PROGRAM Program Description. The Interactive Systems (HCI) Program considers scientific and engineering research oriented towards the enhancement of human-computer communications and interactions in all modalities. These modalities include speech/language, sound, images and, in general, any single or multiple mode, sequential or concurrent, human-computer input, output, or action. Research topics encompass but are not limited to: visualization and interactive computing (including scientific research on virtual environments, and manipulation with visualization), speech and language interfaces (including speech and language recognition, understanding, analysis, and synthesis), a variety of interactive communication modalities such as facial expression and physiological interfaces, gesture, stylus, sound and auditory, tactile, haptic, etc., and interfaces that adapt to the user such as intelligent adaptive and autonomous agents, intelligent information-handling interfaces, learning-, educational- and decision-environment intelligent interfaces. This program recognizes and encourages the emergence of new approaches and the use of novel and realistic environments instrumented to capture human expression and signals in order to explore and validate hypotheses about the laws governing human- computer interactions and the principles of their use in various domain tasks. It encourages the discovery or refinement of computer models of perceptual/sensory and cognitive human-computer interactions. The program scope includes also visualization of and interaction with real or virtual, complex, high-dimensional numeric, symbolic, or pictorial knowledge in computer-assisted environments, as well as measurement and evaluation of the performance of the models and of the methodologies used in human- computer interactive systems. The objectives of the program also include upgrading the human resources and infrastructure of human- computer interaction education and research, and encouragement to women, minorities, and persons with disabilities to participate in scientific research in this area. The major program elements are: o Visualization and Interactive Computing. This element supports research on understanding and representing complex, high- dimensional, physical or abstract systems, models, or objects for the enhancement, explanation, or elucidation of human-computer interactions. This area includes studies on the scientific basis for visualization of scientific and statistical information, semiotics in interfaces, visualization of programming objects, visualization of task decomposition or synthesis in qualitative representations of quantitative information, various aspects of virtual environments, principles of human exploration, comprehension, and understanding of images and other numeric, symbolic, or pictorial knowledge representation. The visualization and interactive manipulation aspects include research on dynamic mappings of functional interactions with real or abstract objects or systems of objects such as machines, robots and physical or chemical objects. Some examples of domains of differing or changing scales are interactions at a distance, atomic to microscopic scale interactions, or very large scale systems. Some examples of investigative questions are those which address efficiency, transparency, the human sense of naturalness, fidelity, clarity, or other characteristics of the interactions. o Speech and Natural Language Understanding. This is one of the Grand Challenge areas, naturally related to high performance computing and communications. This challenge has as its eventual goal for the human to interface with the computer reliably and robustly through spontaneous, user-independent, natural language in real time for understanding, possibly in a multilingual exchange when collaboration takes place. Clearly, many aspects of basic human-computer interaction research are involved in this element of the program: Semantic aspects of speech and natural language; spontaneous speech and language, recognition, analysis, and synthesis; syntactic, semantic, pragmatic, and prosodic factors; numeric and signal processing, symbolic and connectionist representations and architectures; models of the auditory and vocal tracts and related cognitive functions as they are associated with machine recognition and synthesis of speech, and the automation of the processes of speech/language acquisition and adaptation; dialogue models and response generation to queries; and finally, their place in multimodal systems. o Communication Modalities. This element focuses on determining and understanding basic principles of human expression input modalities and computer facial animation modalities. It includes studies of human-generated or human-controlled sounds and vibrations, tones, music, handwriting and stylus interfaces, gestures, posture, body language, facial expression, tactile, haptic and other motor channels, chemical senses (olfaction, taste) and effectors, and electromagnetic input-output (e.g. electric, magnetic, or optical measurements) to assess or influence human commands, human intent, human states of perception, cognition or affective states (e.g., attention, confusion, satisfaction, etc.) or their use to guide computer simulations or processes. o Adaptive Human Interfaces. This element supports basic research with the ultimate aim of making computers adapt dynamically to human users or their environment to facilitate or to enhance interactive tasks. The focus is on the human physical, physiological, psychological, perceptual, or cognitive interactive behavior including their use in dynamically adaptive models of human-computer interactions. One example is the intelligent automatic sequencing and spatial organization of visual or auditory information, to match the expressed or implied needs and goals of the user, dynamically during rapid interactions based on the discovery of those decision and performance strategies that humans use in stressful or constrained situations. Other environments for intelligent agent research involve the retrieval of information from a characterization of the user habits in database search, in learning and educational activities, and in a variety of decision- making tasks. INTERACTIVE SYSTEM PROGRAM FISCAL YEAR 1992 RESEARCH PROJECTS VISUALIZATION AND INTERACTIVE COMPUTING VIRTUAL REALITY IRI-9213822 DeFanti, Thomas A.; Sandin, Daniel J. and Kenyon, Robert V. University of Illinois $172,947 - 12 mos. Prototyping and Quantitative Assessment of an Intuitive Virtual Reality Environment and Its Application to Grand Computational Science Virtual reality includes 3-D display of views which track the user's perspective viewpoint in real time. Two major existing modes are head-mounted displays (HMD) and boom-mounted displays. This work is on an alternative environment: a room ("CAVE") constructed from 7'x7' screens on which graphics are projected. This work will add engineering enhancements to improve the performance and sonic feedback of this room. Quantitative assessment of the benefits of this type of display for a number of tasks will be done. Collaborative effort will take place with discipline scientists working on grand challenge problems in computational science. IRI-9213777 Fuchs, Henry and Bishop, Gary University of North Carolina, Chapel Hill $26,614 - 6 mos. NSF Workshop on Research Directions in Virtual Reality This workshop is a planning workshop to determine research problems in the area of virtual reality that deserve future attention. "Virtual Reality" is thought by many to be a poor and inaccurate description of an enormously popular field. It's various descriptions include a set of capabilities to deliver a tightly coupled, highly interactive, principally 3-D visual experience in which the user has the feeling of being immersed in a 3-D computer- generated environment. Other descriptions describe a mixture of simulated objects with customary objects, such as images, from real environments. Even other descriptions include warping of the perceived visual space and time, or emphasize direct manipulation of objects in the virtual reality or the coordinated experience of two or more persons in a shared simulation scenario. The main goals of the workshop are to identify the most important areas of basic research in virtual reality, to determine the relationship of NSF-funded basic research to other programs, and to produce a set of recommendations on the infrastructure required for sustained basic research. The invitational workshop participants represent a balance of researchers in the area of virtual reality including researchers with NSF grants and researchers in government and industrial laboratories. The workshop focused on four research areas: perception, human-machine interaction, hardware, and applications. (A report was published in Computer Graphics, v. 26, no. 3, pp. 153-177, August, 1992.) IRI-9248697 Pausch, Randy F. University of Virginia $62,675 - 12 mos. PYI: Software Architectures for Multi-Modal Human-Computer Interaction This is the second year base, first year matching, and second year partial matching support of the five-year continuing Presidential Young Investigator (PYI) award IRI-9157583. This research aims at integrating multimodal input streams from a human to a computer into higher level semantic actions. This project is important because of the imbalance of information flow between a human and a computer in the current state-of-the-art. Computer output rate is typically several orders of magnitude greater than the user input rates permitted by current input modalities. Improvements may be achieved by examining new software architectures that deal with probabilistic inputs and with combinations of simultaneous probabilistic input streams. Higher level semantics of multi-modal inputs will be addressed by research on specification languages which properly incorporate time, the sharing of semantic information at various levels of synthesis, and the development of a general framework to support the combination of probabilistic input streams. IRI-9247050 Pausch, Randy F. University of Virginia, Charlottesville $10,000 - 12 mos. REU: Software Architectures for Multi-Modal Human-Computer Interaction This is an REU supplemental award to IRI-9157583 which is a PYI project. This supplement adds two undergraduate students to the project. The students will help in the development of a kernel Virtual Reality platform which supports rapid prototyping of three- dimensional graphical simulations with glove input, voice input, voice output, and sound output. The emphasis is on creating an environment where authors can write small Virtual Reality applications quickly in a high-level, object-oriented interpreted language. In this project the graphics update rates are separated from the simulation computation and a library of object classes is provided for rapid prototyping. See award IRI-9248697 above for additional details on this project. IRI-9212976 Reed, Daniel A. University of Illinois, Champaign-Urbana $250,000 - 12 mos. HPCC: Virtual Reality: Understanding Massively Parallel Computer Systems This is the first year funding a three-year continuing award. Understanding the dynamics of massively parallel computer systems is critical to advancing the state of the scientific art. Increasing numbers of scientists use high-performance computer systems as their main research tool. However, it is impossible to accurately predict performance of application programs of massively parallel processor systems and it is difficult to identify the reasons for poor performance. This work is to use a head-mounted display and virtual reality technology to develop virtual environment representations of the time varying state of a parallel computer system. Visual data-immersion models will be developed for data representation and process control. Much of the work will focus on the development of appropriate visual and aural idioms for computer and application scientists, making performance analysis by immersion in the data readily understandable. IRI-9202424 Wright, William V. University of North Carolina, Chapel Hill $30,000 - 12 mos. SGER: Micro-Teleoperation at Atomic Scale This is a Small Grants for Exploratory Research (SGER) effort. The work is a pilot experiment to determine the feasibility of using a virtual reality environment with a scanning tunneling microscope (STM) for real-time viewing and manipulation of atoms and molecules. Recent advances make it possible to link a STM to force feedback and head-mounted display system for new types of scientific instruments. These instruments will give the user the subjective experience of presence at the atomic scale on a physical surface. Scanning can be at video rates for real-time visualization. Force feedback provides the capability for tactile exploration of the contours of the electronic environment presented at the atomic, molecular and substrate surfaces. Atoms and molecular fragments can be manipulated interactively with the probe. The same probe can cut molecular bonds and bring fragments together for reactions. The products of resulting reactions can be explored to determine their morphology and other characteristics. The STM collaborators are Dr. R. Stanley William and graduate students from the Chemistry Department of UCLA. VISUALIZATION AND MANIPULATION IRI-9160466 Aye, Tin Physical Optics Corporation $0 - 6 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $49,998) SBIR: Real-Time Autostereoscopic 3-D Holographic Display This is an SBIR Phase I award. This work is to make available a high resolution, full color autostereoscopic 3-D display. With this design there is no need for the glasses, shutters or polarizers which in the past have caused viewer discomfort and dimmed images. The display is scalable in size, lightweight and environmentally stable. The approach is based on a stationary multiplexed volume reflection holographic screen and a simple conventional projection system. Three-color images are received by each eye separately because of the construction of the screen. It is intended to be efficient and compatible with current output devices. IRI-9249746 Boult, Terrance E. Columbia University $420 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program - Total Award $50,000) PYI: 3-D Computer Vision This is the second year matching of a five-year continuing award and third-year base award IRI-9057951. This research will support Dr. Boult's investigations in 3-D computer vision. His planned research includes stereo algorithms, surface estimation, segmentation, material identification, error modeling, and information fusion. Dr. Boult is also interested in complexity theory and in the psychology of vision. IRI-9161250 Eichenlaub, Jesse B. Dimension Technologies, Inc. $0 - 6 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $49,983) SBIR: An Autostereoscopic Display for Telerobotic Applications An Autostereoscopic Display for Telerobotic Applications. This is an SBIR Phase I award. This work is to make available an autostereoscopic display for telerobotic applications. User dissatisfaction with awkward 3-D display technologies tends to negate the benefits of 3-D visualization. This effort uses an autostereoscopic LCD display to provide a compact and portable systems for 3-D display. It avoids the use of special glasses or head-mounted display devices to achieve 3-D perspective. The user's left and right eyes receive binocular images, which produce a visual perception of objects at the display screen. The use of this device in telerobotic systems will enable the user to view the end-effector and manipulated objects with greater utilization of human perception and with increased precision and safety. IRI-9210324 Eisenberg, Michael University of Colorado, Boulder $90,000 -12 mos. RIA: Programmable Applications for Computational Physics: Integrating Programming Environments with Direct Manipulation Interfaces This proposal is funded under the Research Initiation Awards (RIA) Program Announcement, NSF 88-99. Current computational environments for scientists suffer from various deficiencies. They are generally non-interactive and difficult to program. The provide little in the way of direct manipulation interfaces, domain-specific constructs or facilities for symbolic (non- numerical) programming. This work will build a suite of programmable applications for computational physics in three domains: oscillators, diffusion-limited aggregation, and chemical kinetics. These applications will be designed to maximize expressiveness. The will have interactive, learnable interfaces and a domain-enriched dialect of Scheme to provide programmability. This will provide a powerful and expressive computational media for scientists. The experience will be articulated in terms of broad, practical guidelines for programmable application design. IRI-9258684 Eisenberg, Michael University of Colorado, Boulder $25,000 - 12 mos. NYI: NSF Young Investigator This is the first year of a five-year NSF Young Investigator (NYI) award. The aim of this project is to develop an empirically and theoretically motivated framework for the design of powerful, expressive software applications that integrate programming languages and direct manipulation interfaces. The programming environments for these applications will be designed to be learnable, domain-specific (or "domain-enriched") and interactive. The strategies for combining language and interface constructs will emphasize opportunities for creative symbiosis. Sample applications will include those for the domains of physics, mathematics and graphics design. The long term goals of this project include empowerment of the end users of applications by providing them with the capability for interactive programming. IRI-9106389 Hanson, Andrew J. Indiana University $79,390 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $85,077) Interactive Mathematical Visualization This is the first year of a two-year continuing award. One long-term goal of this research is to develop and exploit techniques to make it possible for scientists to study and solve problems in mathematics and physics that could not be undertaken without automated assistance. Another long term goal is to deduce general principles about the nature of human capabilities and limitations in mental engagement with abstract data spaces. Intelligent constraint-based interaction (rule-based) methods will be combined with semi-automated optimization techniques to develop a domain-independent interactive system of constraints. This system will be designed to allow manipulation of objects that are transforms of non-observables which are conceptually difficult mathematical properties. Simulations that transform non-observables into manipulatable observables are useful to exploit human learning processes to master abstract problems. The abstract concepts range from issues of relative size and speed estimation, unfamiliar dimensions, and chaotic processes in physics and interactions in many-body problems. Domain-dependent constraints on representations are used to form a model of a desired class of spatial configurations. This leads to a rational automation methodology. Direct, context-free manipulation of objects, rather than manipulation of corresponding surrogate "control"objects, will also be examined. This effort seeks to lay a foundation for understanding human capacity for mental engagement with visual representations of abstract concepts. This will be useful as a basis for subsequent experiments on psychophysical theories of visualization. IRI-9246595 Hartson, Rex H. and Hix, Deborah Virginia Polytechnic Institute $8,900 - 12 mos. Building Bridges and Interfaces: Developing User-Centered Representation Techniques for Communicating User Interface Designs This is a supplement for travel to IRI-9023333. This research attacks the problem of usability by fundamental improvements to the interface development process. It aims at a new generation of behavior-oriented techniques for representing the details for the interaction between users and systems. Interactions are represented by means of a User Action Notation (UAN) which is developed for communication among persons with various professional backgrounds involved in various phases of the development cycle of a specific interface. Experiments to evaluate an existing UAN are being performed and the criteria for new features are being studied. A theoretical analysis is being done based on cognitive theory, relating the approach to other work in human-computer interaction. This supplement will provide $100 each for 9 students in the human-computer interaction program at Virginia Polytechnic Institute to assist them to attend the SIGCHI þ92 conference in Monterey, California, May 1 through 7, 1992, and adds two undergraduate students to the project. This is an important professional experience for potential young researchers. The potential of computers is limited by how effectively users can access their computing power via the user interface. This research attacks the problem of usability by fundamental improvements to the interface development process. It aims at a new generation of behavior-oriented techniques for representing the details for the interaction between users and systems. The results will be integrated into a set of coherent requirements for the next generation of a UAN. The requirements will then form the basis for the development of tools which use the revised UAN. The tools include an editor and tools to generate code and skeletal user documentation from a UAN description of a task. IRI-9245172 Hopcroft, John E. Cornell University $220,000 - 12 mos. (Jointly funded with the Division of Computer and Computation Research - Total Award $230,000) A Program of Research in Environments for Scientific Computation This is the third year funding of a three-year continuing award IRI-9006132. The research is on graphical user interfaces and their integration into a simulator environment for exploration of model systems in the physical sciences and in engineering. The approach integrates geometric modeling, symbolic mathematics, object oriented programming, numerical analysis and code generation to build a new type of environment for engineering analysis and simulation. A simulator generator permits scientists to investigate new methodologies, to quickly create unique simulators to solve particular problems, and to take full advantage of rapidly changing hardware technology. This project impacts science and engineering methodology and education. IRI-9247646 Hopcroft, John E. Cornell University $20,000 - 12 mos. REU: A Program of Research in Environments for Scientific Computation This is an REU supplemental award to IRI-9006137. This supplement adds two more students to the three undergraduate students already on the project. The research is on graphical user interfaces and their integration into a simulator environment for exploration of model systems in the physical sciences and in engineering. One student is working on the problem of writing code for parallel architectures with the goal of finding a small set of high level construction primitives which can be compiled to various parallel architectures. Two students are working on the problem of how to develop tools that will allow non-computer experts to access information easily. One of these students is developing algorithms for classifying components of a text file to determine its structure for automatic information extraction algorithms. The other student is looking at the question of automatically finding e-mail addresses using intelligent agents. The added students will work on similar projects. The larger project is described below. The fundamental goal of this research is the development of the scientific base necessary to reduce the time and effort required to assemble and use computer tools for simulation and analysis of physical systems. Most existing simulators solve a limited range of problems, are difficult to modify, and are difficult to integrate because each uses its own representation scheme. The research will address these issues and build an environment that automates many of the steps in creating a simulator. Defining in simulator requires; choosing equations to model the behavior of the physical systems, choosing and implementing computational methods for solving the equations defining mechanisms to handle the complex control flow of simulations, and choosing representation schemes for physical objects, their properties, equation, etc. Currently, these steps are carried out implicitly, without computer aid. The researchers will design a software architecture that allows simulators to be described in terms of equations and mathematical properties naturally suited to the problem. The research investigates: a model representation system for representing the quantities and laws that describe the physical behavior in a given analysis model; an analysis specification system, a language for defining the behavior of the simulation in terms of the laws and quantities; a scene editor; and, a low level interface for integrating existing modules such a numerical packages or finite element codes. IRI-9161274 Johnson, David A. Tini Alloy Company $0 - 6 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $50,000) SBIR: Tactile Feedback Output Device for Scanning Tunneling Microscope This is an SBIR Phase I award. This work is to make available a multiple-stimulator tactile array device mounted on a mouse. This device will provide a capability for the finger to sense patterns, texture or variations in the surface of an image. It is designed specifically to provide tactile perception of surface texture in images derived from a scanning tunneling electron microscope (STM). The electrical potentials generated by the scanning tip of an STM are translated into multiple-colored pictures portraying surface geometry. This device will augment visual output to enable multi- sensory perception in STM and it will be useful in many other human-computer interaction environments. IRI-9246496 Kaufman, Arie E. and Giacalone, Allessandro SUNY Stony Brook $29,224 - 12 mos. ROA: Interactive Prototyping of 2-D and 3-D User Interfaces This is a Research Opportunity Award (ROA) supplement to IRI- 9008109. Dr. Reuven Bakalash is conducting research at SUNY Stony Brook for the summer of 1992 in Dr. Kaufman's laboratory. Dr. Bakalash is currently an Assistant Professor of Computer Science at Hofstra University, New York, a teaching institution. Dr. Bakalash has been involved in the Cube project focusing on the development of 3-D volume visualization methods and its application to cell biology. The volume visualization efforts are closely linked with domain scientists in biophysics, neuroscience and biology. The efforts also contributes to the development of real-time interactive environments. IRI-9020603 Lewis, Michael University of Pittsburgh $100,000 - 12 mos. Ecological Cognition in Visualization This is the first year funding of a three-year continuing award. This effort contributes to the theory of a user model and a measurement methodology designed, in part, to permit quantitative assessment of visualization interfaces. Ecological theory states that the state of affairs in the world can be represented directly through cognitive "states of affairs" (corresponding to ecological processing aspects). This work is to test the hypothesis that ecological processing includes cycles of "envisioning" and "inspection" information processing operations. The difficulty of cognitive tasks can then be characterized by the allocation of processing resources (e.g., the allocation of attention) between the (relatively) memory-independent ecological operations and controlled, memory-intensive symbolic processing. Thus, this approach allows consideration of the task context as well as viewing logical cognitive operations. It makes possible testable predications about both abstract and concrete tasks. It can be folded into existing GOMS models for quantitative assessment of interface metaphors allowing subsequent evaluation and minimization of the difficulty of interacting with direct manipulation and visualization interfaces. IRI-9209576 Lohse, Gerald University of Pennsylvania $70,000 - 12 mos. RIA: Modeling Graphical Perception and Cognition This proposal is funded under the Research Initiation Awards (RIA) Program Announcement, NSF 88-99. The research builds on a cognitive model to provide a scientific basis for understanding how people perceive and process information from graphic displays. Phase 1 proposes to model ocular scanning behavior using knowledge about the graphic display and the goal structure of the task. The goal is to measure the similarity among sequences of eye movements used to answer questions posed to a range of graphic displays. Phase 2 of this research explores errors people make processing information from complex information displays. The studies examine when working memory overload occurs, what information is lost, and how overload influences performance accuracy. The resulting model will provide a research tool for understanding visual cognition and perception. It will also provide designers a metric for measuring display quality, working memory constraints, and information processing time. IRI-9123468 Olsen, Dan R. Brigham Young University $57,790 - 12 mos. Pattern-Driven User Interfaces This is the first year funding of a three-year continuing award. Direct manipulation user interfaces are difficult to build. User Interface Management Systems (UIMS) have attempted to alleviate this problem by providing models of the interactive dialog which can drive the interface implementation. Such models, however, have failed to directly express the changes of information which comprise most operations in direct manipulation interfaces. A data pattern and transformation language based on matching will be developed as a model of such additional processes. Techniques will be developed for using this language as a user interface design tool. These techniques will include approaches for visually expressing the patterns and transformations interactively. This data transformation language can then be used for creating visualization of data. Such a language can express attribute modifications, style sheets, critics (which monitor and advise users) and searches. Unifying the interaction in such a transformation language provides a foundation for more intelligent tools to reason about the behavior of the interface. SPEECH AND NATURAL LANGUAGE UNDERSTANDING IRI-9248596 Beckman, Mary E. Ohio State University $46,462 - 12 mos. PYI: Computer Research This is the fourth year matching and fifth year base amount of a five-year continuing award IRI-8858109. This Presidential Young Investigator Award is for support of Prof. Mary E. Beckman at the Ohio State University. Her research is aimed at understanding and modeling the prosodic patterns which organize speech into coherent units of sound. Earlier work, including most existing speech synthesis systems and all speech recognition systems, is based on linguistic theories from the 1960's, which posited only a single type of organizational unit, the phonetic segment, which constituted the information primitives. More recently, however, linguists have recognized that the prosodic patterns of intonation and stress organize speech into units of various sizes. A computational model based on those theories would take advantage of the way prosodic patterns encode a hierarchy of phonological works and phrases. Such a model would make constructive use of the variation in spectral organizational structure to recover the higher-level informational structure. The significance of this work is that it should yield a computational model for timing patterns that will be implemented in an articulatory synthesis program and ultimately will be incorporated into an automatic speech recognition system, thus bringing a new set of tools to bear on a difficult problem area. IRI-9202458 Brennan, Susan E. SUNY Stonybrook $81,458 - 12 mos. Interactive Processes of Language Use in Human-Computer Interfaces This is the first year funding of a three-year continuing award. The goal of this research is to better understand the interactive process of human language use with a computer system. The theory underlying this research is that human-computer interaction, like human communication, falls into a class of coordinated action that proceeds by the systematic exchange of evidence of understanding. Two processes will be examined: (a) entrainment which is hypothesized to influence lexical choice, and (b) grounding, which is hypothesized to set the placement and level of appropriate feedback. Psychological experiments will use simulated language and voice interfaces to examine (1) what variables influence the lexical choice people make and how such a system might adapt to idiosyncratic language input, (2) what kind of text output should the system generate to present itself as a coherent dialog partner, (3) what kinds of context-sensitive feedback should a speech or language interface provide. This research contributes to practical goals concerning interactive systems,including: designing adaptive user interfaces, enabling more robust error handling, and making natural language, command and voice interfaces easier for people to use. IRI-9208831 Cole, Ronald A.; Zahorian, Stephen A. Oregon Graduate Institute $31,762 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $35,762) NSF Workshop of Spoken Language Understanding This is a planning workshop to determine critical research problems in the area of spoken language understanding. The main goals of the workshop are: (a) it identifies the most important areas of research in speech and natural language understanding, with particular attention to those not currently addressed; (b) to determine how NSF can benefit from other programs, such as the DARPA speech and natural language program, and yet distinguish itself from these programs; and (c) to produce a set of recommendations to NSF to help guide future research opportunities. The workshop participants represent a balance of researchers in the areas of speech recognition, natural language understanding and human-computer interaction. Participants included researchers with NSF grants and researchers in government and industrial laboratories. Discussion took place in plenary sessions and in breakout groups, which focused on three research areas: robust systems; the interaction of speech and natural language; and human-computer interaction. IRI-9104984 Cole, Ronald A. Oregon Graduate Institute $119,994 - 12 mos. Spoken Letter Recognition This is the first year funding of a three-year continuing award. The ability to perform and find phonetic distinctions is a fundamental unsolved problem in computer speech recognition. Spoken letter recognition has practical applications and requires discrimination of acoustically similar words. An existing system will be improved in the following areas: (a) isolated letter recognition, (b) multiple letter recognition (sequences separated by brief pauses), and (c) continuous letter recognition (naturally spelled words). The result of this research will be a highly accurate, speaker-independent system that retrieves words from spellings. Strategies for improving isolated letter recognition include improving the signal representation and the broad category segmentation algorithm. Better letter segmentation strategies will be developed for those cases when the speaker does not pause as requested. Two-level classification will allow consideration of all possible letter sequences in conjunction with discrimination of whole-letter classification as in isolated letter recognition systems. Coarticulation models will be introduced to recognize between phoneme transitions, including dynamically adjusted duration models and boundary detection. IRI-9161057 Hutchins, Sandra E. Emerson & Stern Associates $0 - 6 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $49,805) SBIR: Remote Voice Interface for Computer Control This is an SBIR Phase I effort. This work is to determine the feasibility of developing remote voice interface which addresses the problem of handling users' spontaneous corrections of their own errors or of those made by a listener. In this case the listener is a speech recognition system. The objectives include building a linguist model for error correction using the taking of telephone numbers as an example. Based on these results, vocabulary and grammar models for error correction will be created, using existing tools, and integrated in their Soliloquy voice interface software system and then an evaluation will be performed. Applications include remote data entry, database inquires, computer control, call sorting and other environments. IRI-9247145 Novick, David Oregon Graduate Institute $5,375 - 12 mos. REU: Computational Models of Dialogue: Speech-Act Theory Meets Real Utterances This is an REU supplemental award to the Research Initiation Award (RIA) award IRI-9110797. This supplement adds one undergraduate student to the project. The student will help in the investigation of problems with the use of speech-act theory in computational systems for understanding spontaneous language. The goal is to provide speech recognition systems with a reasonable method of interpreting unconstrained discourse. The focus is on dialogue models that include meta-knowledge about the state of conversation control. The student will investigate the following hypotheses: (1) Are there contextual factors that distinguish filled pauses from pauses, and (2) are there conversational features for which preceding filled pauses are predictive? The larger project is described below. The research aims at resolving long-standing problems with the use of speech-act theory in computational systems for natural language understanding and to provide methods of interpreting unconstrained discourse. The integration of verbal and non-verbal speech acts are still very poorly understood and is critical for human-computer interactions. This research seeks to integrate many modalities with the study of speech acts, and important conceptual framework in speech and dialogue processes. The goals are (1) to determine how the theory of meta-acts can be extended and refined to model negotiations of reference and confirmation of mutuality of knowledge, and (2) to determine if meta-act models can be validated predictively. Meta-knowledge models include information about conversations that are non-verbal, such as turn-taking. The project derives meta-acts as a rational strategy of action, given attributions of the participants' beliefs, goals and expectations at the point in the dialogue in which the acts actually occur. The acts are then expressed in a rule-based simulation of similar conversations. A byproduct of the research is a database of analyzed and transcribed data useful for other researchers. It is expected that there will be substantial impact in human-computer interactions research and aspects of research in speech and natural language processing. IRI-9248730 Ostendorf, Mari and Shattuck-Hufnagel, Stefanie Boston University $150,000 - 12 mos. (Jointly funded with the DARPA - Total Award $300,000) Evaluating the Use of Prosodic Information in Speech Recognition and Understanding This is the fourth year funding of a five-year continuing award IRI-8905249. This grant was extended from a three-year award on the basis of the special creativity of the work. Speech provides natural language with prosodic information which may be used to simplify the resolution of ambiguities that are difficult from text alone. This research investigates the representation, detection and integration of prosodic information in a spoken language system (SLS). The approach of this research combines linguistic theory, speech knowledge, statistical modeling and natural language processing techniques. The current work, aimed at evaluation the potential use of prosodic phrase boundary information in speech understanding, is based on a database of FM radio newscasting speech. The use of prosodic information offers great potential for improving spoken language systems. The current work will also include an extension of the algorithms to spontaneous speech. IRI-9247174 Silverman, Harvey F. Brown University $4,000 - 12 mos. REU: A Microphone Array System for Speech Recognition Input This is a REU supplemental award to IRI-8901882. This supplement adds one student to the project which is part of a joint NSF-DARPA initiative. The goal of the research is a real-time, talker- independent, connected (continuous) speech recognizer for a hard, moderate-sized vocabulary. The input for the recognizer is obtained from a tracking microphone-array system. The student will work on a project of acoustic field plotting for the microphone array project. The project includes the enhancement of the accuracy of a cartesian robot which covers approximately a 3 meter by 4 meter area. The acoustic field plotter is to be increased from two degrees of freedom to four degrees of freedom. IRI-9249517 Steedman, Mark J. University of Pennsylvania $37,110 - 12 mos. (Jointly funded with the Information Technology and Organization Program - Total Award $74,220) Computer Synthesis of Contextually Appropriate Intonation of Spoken Language This is the second year funding of a two-year continuing award IRI- 9018513. The structure imposed upon spoken sentences by intonation seems frequently to be orthogonal to their traditional surface- syntactic structure. The involvement of two apparently uncoupled levels of structure in natural language grammar appears to complicate the path from speech to interpretation unreasonably, and to thereby threaten a number of computational applications in speech recognition and speech synthesis. This work will complete the specification of a generative theory of intonation in relation to syntax and discourse function, and instantiate the theory by constructing a "discourse-model driven" utterance generator and speech synthesizer. Such a device is to be contrasted with "text- to-speech" generation. The input to the characteristic of a simple natural language query system. On the other basis of such representations, the systems will be capable of constructing a variety of contextually appropriate intonation contours for any given sentence covered by its grammar. Such a device is expected to considerably improve comprehensibility and naturalness of the synthesized speech. IRI-9014829 Weintraub, Mitchel SRI International, Menlo Park, CA $133,334 - 12 mos. Model-Based Speed-Sound Separation for Speech Recognition in Noise This is the second year funding of a three-year continuing award IRI-9014829. This research is to develop a new, model-based, spectral estimation algorithm for recognition of noisy speech. The new algorithm represents a significant improvement over previous estimation algorithms because it incorporates more information about speech spectral distribution. The proposed work improves the estimation by incorporating information about its dynamic properties and its quasi-periodic nature. To capture the dynamic properties the research will study several types of hidden Markov models (HMM). An estimation of the position of the harmonics will be made to address the periodicity properties. From these will be made an estimate of the spectral energy at any given frequency dependent on its proximity to the nearest harmonic. Incorporation of dynamics and periodicity should improve speech recognition performance in noisy environments. COMMUNICATION MODALITIES FACIAL EXPRESSION INTERFACES IRI-9117110 Badler, Norman I. and Steedman, Mark University of Pennsylvania $80,000 - 12 mos. Communication, Co-articulation, and Dialog Gesture in Facial Animation This is the first year funding of a three-year continuing award. The goal of this research is 3-D animation of facial expressions including those conveying emotion correlated with the intonation of the voice. This includes differentiation of timing, pitch and emphasis that are related to semantic distinctions of discourse as "given" and "new" information. Until now, existing systems have not taken into account the link between these two features. The algorithm to be used will embody rules that control these various modes of expression. The choice of accents and their placement will be examined in conjunction with whether the message is new or old. The facial model integrates the action of each muscle or group of muscles as well as the propagation of their movement. A two-person conversation model will also be investigated to explore the roles of the listener and the speaker and how turn-taking is established facially and by gestures. BNS-9120868 Ekman, Paul University of California, San Francisco $10,000 - 12 mos. (Jointly funded with the Language, Cognition and Social Behavior Program - Total Award $111,922) Automating Facial Expression Coding This project will develop a neural network system to dynamically code facial expression, automating a process which, though widely used is not highly time, labor, and cost intensive. Automation will make this metric accessible to many more researchers at less cost, while eliminating residual interscorer discrepancies. Measurement of facial expression is valuable in many domains, and has lead to unique insights in problems as diverse as human development, sociology, developmental and social psychology. From a practical standpoint, different aspects of expression elucidate whether a listener is empathetic or hostile (important in politics and business), distinguish abusive from non-abusive caretakers (social work), predict divorce in dysfunctional married couples, and may incriminate dissembling witnesses (forensics). Reliability of facial expression as a gauge of nervous system function has been affirmed through autonomic nervous system indices as well as electromyography and electroencephalography. Initially, automating measurement of facial movement will involve static images of faces of a single subject, which will be pre-processed to normalize size and brightness, compressed with a back-propagation image compressor, and fed to a second neural network whose output will correspond to the "Action units" of the widely used "Facial Action Coding System." Recent techniques such as weight-sharing will be used where necessary. Recurrent networks will be used to allow processing of dynamical information, with extension to implementation to multiple subjects. This methodological advance has the potential for a significant contribution to the infrastructure of science on many domains. IRI-9217831 Ekman, Paul; Huang, Thomas C. and Sejnowski, Terrence University of California, San Francisco $11,932 - 7 mos. (Jointly funded with the Robotics and Machine Intelligence Program, the Social Psychology Program and the Division of Behavioral and Cognitive Sciences - Total Award $47,725) Workshop on Facial Expression Understanding This project is a planning workshop to determine those research problems in the area of facial expression understanding that deserve future attention. The main goals are (a) to identify the most important areas of research on how to extract information from facial activity relevant to a person's emotional, cognitive and physical state; (b) to enhance communication between human and machine through the development of emerging computer vision capture techniques for facial processing and categorization of facial expression; (c) to consider how we can facilitate the training of new investigators in the relevant fields. The workshop will produce a set of recommendations to the investigative community and to NSF to help guide research opportunities. The report will be submitted for publication in standard journals. The workshop participants represent a balance of researchers in the areas of facial expressions understanding, computer vision and human computer interaction. Participants include researchers with NSF grants and researchers in government and industrial laboratories. The workshop will consist of plenary sessions and breakout groups focused on several research areas. The steering committee includes Dr. Paul Ekman of the University of California at San Francisco, Dr. Thomas Huang of the University of Illinois at Champaign-Urbana and Dr. Terrence Sejnowksi of the Salk Institute. GESTURE AND STYLUS INTERFACES IRI-9245958 Girson, Andrew D. Digital Video Process, Inc. $0 - 12 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $309.00) SBIR: An American-Sign Language-to-English Recognition & Translation System This is an award for Phase II of the development of an American Sign Language (ASL) acquisition, recognition and translation system. Such a system could eventually allow deaf signers and English speaking individuals to communicate without the aid of a human interpreter. The Phase II research focuses on several critical capabilities: (1) the quantification of ASL as a spatial/visual language; (2) the acquisition of ASL for internal use by a computer; (3) the recognition of dynamic ASL gestures; and (4) the context-sensitive translation between ASL and English. A proof-of-concept prototype will be built. Machine translation algorithms for ASL-to-English will be developed. Potential intermediate products include: gestural interfaces for computers and home appliances, telerobotic controllers, virtual reality and simulations tools, and ASL research tools. IRI-9247352 Girson, Andrew D. Digital Video Process Inc. $0 - 12 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $8,901) SBIR: An American-Sign Language-to-English Recognition and Translation System This is a supplemental award for Phase II of the development of an American Sign Language (ASL) acquisition, recognition and translation system. Such a system could eventually allow deaf signers and English speaking individuals to communicate without the aid of a human interpreter. The Phase II research focuses on several critical capabilities: (1) the quantification of ASL as a spatial/visual language; (2) the acquisition of ASL for internal use by a computer; (3) the recognition of dynamic ASL gestures; and (4) the context-sensitive translation between ASL and English. A proof-of-concept prototype will be built. Machine translation algorithms for ASL-to-English will be developed. Potential intermediate products include: gestural interfaces for computers and home appliances, telerobotic controllers, virtual reality and simulations tools, and ASL research tools. IRI-9213472 Oviatt, Sharon L.; Cohen, Phillip R. SRI International $182,213 - 12 mos. Writing and Talking to Future Interactive Systems This is the first year of a three-year continuing award. As computing moves in the direction of small portable devices, speech and pen will become input modalities of choice. This research addresses a critical issue of the identification of the structural, linguistic and performance characteristics of handwriting and speech and issues of their interaction when used together in an application. A series of subject-placed experiments will assess writing and speech when they are: (1) channeled by different types of system prompts, (2) resolving typical system recognition errors, and (3) experiencing system response delays differing in magnitude and predictability. This research will provide a more complete understanding of how people will write and speak to future interactive systems and how this can be altered by fundamental parameters of system interactivity. It will also provide a more principled basis for the design of complex multimodal systems. IRI-9247051 Pausch, Randy F. University of Virginia $10,000 - 12 mos. REU: Creating Custom User Interfaces Based on Gesture This is an REU supplemental award to the Research Initiation Award (RIA) grant IRI-9009881. This award adds two students to the project. The students will be responsible for extending the gesture mapping system to control two-dimensional tasks. The goal is to track user gestures and interpret them as control signals where continuous motion are mapped into a set of analog device control signals rather than into discrete control signals. The system allows disabled children to use their best range of physical motion. The students will also participate in studies to evaluate the system. IRI-9113787 Quek, Francis and Jain, Ramesh C. University of Michigan $86,839 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and the Cross Disciplinary Activities Office - Total Award $151,829) A Gesture Interpretation and Voice Recognition Multi-Modal Human Interface This is the first year funding of a three-year continuing award. This research is to development computer vision methods for tracking hand gestures. Gestures are followed by video tracking of a glove studded with light emitting diodes. Other glove marking methods will also be examined. Gesture interpretation will be performed by a dynamic vision system using the correspondence via event detection (CED) algorithm which segments and tracks the motion of the LEDs over multiple video frames. The primary focus of the research will be to extend existing algorithms to handle occlusion. Constraints inherent in hand anatomy, function and gesture categories will be exploited. Voice input will be implemented using commercial disconnected- speech recognition technology. The voice commands will be used segment gesture epochs and to disambiguate the gesture categories. Rules for combining gesture interpretation and voice for gestures stream segmentation will be developed. A gesture library constituting taxonomy of gestures will be used for temporal coordination and conflict resolution. Both monocular and stereo camera environments will be used. Parallelization of the algorithm will be done to assure real-time capability and to support subsequent research in which bare hands might be tracked without the use of markers. IRI-9021270 Wickens, Christopher D. University of Illinois, Urbana $71,965 - 12 mos. Analysis of the Principles Underlying 3 Dimensional Graphics Perception This research has the objectives of formulating and validating theory-based principles for the visual representation of higher dimensional scientific data. Two theoretical perspectives are joined in this work. One addresses the similar characteristics of the display of two or more data points that facilitate their cognitive integration, but disrupts the focussing of attention on single data points. This perspective emphasizes the similarity induced by contours and color. The second addresses biases and characteristics of depth perception. Subjects view multidimensional data bases on a workstation that characterize the constraints between dimensional values in different scientific domains (e.g., meteorology, electrophysiology, economics), in various static and dynamic, color and monochrome, 3-D formats that allow test of the hypotheses. They answer a range of questions of the displayed data, and the speed and accuracy of the response is used to evaluate graphical formats. While a wealth of display technology is now available for scientific visualization, there are very few empirical validations of the techniques offered, and those are not described within the framework of general principles of human perception and cognition, in a way to generalize to other applications. This research will advance knowledge by providing such generalization. PHYSIOLOGICAL INTERFACES IRI-9202100 Anderson, Charles W. and Aunon, Jorge I. Colorado State University $91,968 - 12 mos. Alternate Modes of Human-Computer Interaction: EEG Recognition with Neural Networks This is the first year funding of a three-year continuing award. This research is to study the feasibility of human-computer interaction through the use of on-line recognition of electroencephalogram (EEG) patterns. A system that can identify, on-line, which of several mental tasks a person is performing could provide an alphabet with which a severely disabled person can compose commands to devices like a wheel chair. Most prior research has focused on recognition accuracy rather than on fast, real-time responses. The goals of this research are to characterize the limitations of current EEG recognition methods, to increase the accuracy and decrease the computation time of EEG recognition by use of a neural network approach, and to develop a practical real-time method. The performance of conventional methods and of neural network learning methods designed for high-dimensional data will be compared. Evaluations of feature and classifier characteristics will be based on classification accuracy, robustness and computation speed. This project has relevance to the field of pattern recognition in general, and to the neural network and human-computer interaction fields in particular. IRI-9213257 Gevins, Alan EEG Systems Laboratory $138,961 - 12 mos. Neural Signals of Cognition During Computer Use This is the first year funding of a three-year continuing award. This is research to test the feasibility of obtaining images of brain electrical activity related to information processing and learning while people use different types of computer interfaces. The first part of the effort is to extract electro-encephalographic (EEG) signals related to cognitive aspects of computer use from contaminating potentials generated by eye movements and muscle activity. A formal experiment will then be conducted in which subjects acquire skill in using both a graphical and non-graphical user interface to perform two text-editing tasks (correcting typographical or semantic errors). Neural-network pattern discrimination will be used to find the differences between the EEG spatial patterns for the typographic and semantic tasks, as well as how the patterns change as subjects learn to use each type of user interface. Successful completion of this project should pave the way for using EEG measures to investigate problems in human- computer interaction. ACOUSTIC INTERFACES IRI-9214233 Duda, Richard O. San Jose State University $64,785 - 12 mos. A Computational Model for Sound Localization This is the first year funding of a two-year continuing award. The goal of this research is to create and evaluate a model of the process by which people locate sounds in three dimensions. Emphasis will be placed on explaining well established but still incompletely understood psycho-acoustical phenomena. One example is our ability to locate sounds coming from above or below, despite the fact that there are no binaural differences between the sounds that reach the ears. Another example is our ability to locate sounds in reverberant environments that contain multiple sound sources, where echoes and reflections act as additional, virtual sources. A third example is our ability to judge distance, since loudness alone is not an adequate cue. The model is based on the physical effects of sound propagation, and on neurophysiological studies that have traced the auditory cortex. It extends an existing computational model of the cochlea by incorporating monaural and binaural, temporally-based correlation methods to extract the information needed for source localization. If successful, this work should significantly improve the abilities of computers to recognize speech or other sounds as they occur in everyday, multisource environments, thereby extending the range of effective human-machine interaction. IRI-9214694 Smith, Stuart; Kevkowitz, Haim and Pickett, Ronald University of Massachusetts-Lowell $88,988 - 12 mos. Auditory Representation of Scientific Data This is the first year funding of a two year continuing award. The long-term goal of this work is to develop and evaluate auditory representations of multidimensional data. For some forms of data (e.g., time series) auditory displays may be especially well suited to human perceptual analysis and possibly much more effective than current visual displays. Auditory data representation is fundamentally a mapping from a data variable to a property of a sonic event. Where the variable is multidimensional, the measure on each dimension of the data sample determines the value of the corresponding sound property of the event, and the values of all the measures together determine the overall character of the event. For example, if the event is a simple tone, one data dimension could be mapped to pitch, another to loudness, and a third to duration. A variety of auditory displays can be created depending on the kind of data and the kind of analysis to be conducted. Research in auditory data representation has been severely hampered by the lack of an appropriate computing environment. The Principal Investigators are building a prototype computing environment for research in auditory data representation and, using this environment, to develop several different auditory data representations and to conduct preliminary evaluations of these auditory data representations with both real and synthetic data. TACTILE AND HAPTIC INTERFACES IRI-9160591 Heusinkveld, Paul A. HEUS, Incorporation $0 - 6 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $49,862) SBIR: The Keygrip - A Device for Data Input and Output This is an SBIR Phase I effort. This work is to develop a tactile device that could be an alternative to conventional keyboards. The device had five keys and is held and operated by one hand using a chording method. Because is held by one hand it may be used in fixed or mobile applications and while performing other main tasks. An output mode will be added to the device using a tactile feedback mechanism and possibly a visual panel. ADAPTIVE INTERFACES INTELLIGENT INTERFACE AGENTS IRI-9205668 Maes, Patricia Massachusetts Institute of Technology $199,376 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $207,376) Learning Interface Agents This is the first year funding of a three-year continuing award. Intelligent interface agents are computer programs that employ artificial intelligence and other techniques in order to provide expert, autonomous assistance to a user dealing with particular computer applications. This research addresses the question of learning as a method by which an agent can acquire sufficient amounts of information about the user, the application, and the communication and collaboration strategies to be able to empower the user. The goal of this project is to find the necessary and sufficient conditions for the hypothesis to hold that an interface agent can be given access to the information it needs to program itself. The agent is given a minimum of background knowledge and learns appropriate situation-action rules by observing the user's interaction with the application, by receiving feedback from the user about certain rules, and by copying the situation-action rules employed by agents assisting other users in the same application. The project will first develop a generic architecture for building learning interface agents. This will then be used in concrete examples to test the main hypothesis. IRI-9216401 Prueitt, Paul S. Georgetown University $28,206 - 12 mos. SGER: Simulation Models in Adaptive Human Interfaces This is a Small Grant for Exploratory Research (SGER) award. The research is contribution to the theory of non-linear simulation, to the theory of adaptive human interfaces and intelligent interface agents, and to the theory of human-computer interaction. Episodic transitions occur during human-computer interaction and in systems in related field including biology, the physical science and the behavioral sciences. For example, the transition from one focus of attention to another is an episodic transition between homeostatic states of selective attention. These "phase transition" behaviors may be triggered by a variety of events such as completion of a cognitive subtask or by external interrupts. The present research is to develop tests for the hypothesis that critical phase transitions are governed by a conservation law called the law of "conservation of affordance." Affordances are expressions of system needs which are dynamically created and annihilated through the perturbation and subsequent restoration of homeostatic conditions. Affordances constitute the linking of behaviors of various parts of the system that normally operate at different time scales in order to achieve behavioral goals. Affordance is modeled mathematically by two opposing sources in a non-conservative energy flow on embedded Hamiltonian manifolds. This mathematical formulation provides an explicit theoretical basis for dynamic engines in simulation models, for the testing of the conservation law, and for the derivation of principles for use in the domain of human-computer interaction and elsewhere. In particular, it provides a rationale for the development of an interface specification methodology for generic, isolatable and portable computational modules that support the conservation law. A set of simulations will be developed using this mechanism which can support modeling of higher order functions such as goal formation, selective attention, choice, intention and various concepts underlying cognitive and perceptual simulation issues. An important anticipated result is that the correct use of these mechanisms will make possible sensory encoding and data fusion for artificial systems to permit the capture and understanding of rare and novel events. The interface and modularity issues have consequences for interoperability, software capitalization and the flexible sharing of research results between groups of investigators. INFORMATION RETRIEVAL ENVIRONMENTS IRI-9215085 Marcus, Richard Massachusetts Institute of Technology $89,187 - 12 mos. Retrieval Assistance Mediated by Conceptual Views and Multiplex Searching Previous research on assistance for users of database systems has demonstrated a number of techniques that have high potential for making the search process easier and more effective for inexperienced as well as expert searchers. This research enhances an experimental search assistance system so as to make it a more effective tool for research investigations. Two enhancements are: (1) to replace the textually oriented user interface with a graphical user interface (GUI); and (2) to extend the underlying search models so as to serve a basis for assistance for searching in general databases with a verbal indexing character as well as in bibliographic databases. The enhanced system assistance is tested on a variety of experimental users to determine how well the GUI can aid the user by encapsulating the conceptual representation of the user's problem and guide in the generation, execution, evaluation, and modification of the multiplex search strategies necessary to facilitate efficient and effective searching. Testing compares results for bibliographic and full-text and other non- bibliographic databases as well as for different modalities of user interfaces, ranging from highly user-directed interactions to highly automated, or system-directed, modes where user initiatives may be limited to an initial problem statement and a minimal amount of relevance feedback on retrieved documents. IRI-9117084 Soloway, Elliot M. University of Michigan, Ann Arbor $0 - 12 mos. (Jointly funded with the Database and Expert Systems Program and Information Technology and Organizations Program - Total Award $77,715) Developing Guidelines To Providing Computer-Based Support for Scientific Data Analysis The "human-computer interface" is paramount if scientists are to take full advantage of the vast quantities and types of data now becoming available. Attention needs to be focused on more than the query language; rather, the interface meeds to integrate scientific databases into the everyday work practices of scientists, e.g. data exploring, hypothesis generating and testing, report and chart making. The Task/Artifact Methodology is being employed; it consists of cycles of cognitive task analysis, system building, and testing in ecologically valid contexts, e.g,. practicing scientists using the system on a daily basis. The resultant computer-based environment, Rev (Representation for Visualization) is user- and task-centered, as opposed to being technology-driven. Rev serves as the scientist's "notebook" permitting hem/her to move among databases, hypotheses, reports, charts, etc. Rev is being designed for scientists in nuclear engineering and for scientists engaged in global change research. As computers become integral to the moment-by-moment work practices of scientists, well-designed tools become critical. This project, then, serves as a model for how scientist-centered computing environments can be developed and deployed. LEARNING, EDUCATIONAL AND DECISION ENVIRONMENTS IRI-9116640 Lewis, Clayton and Polson, Peter G. University of Colorado, Boulder $129,389 - 12 mos. Exploration and Learning in Interactive Systems This is the first year funding of a three-year continuing award. Users at all levels of expertise prefer to learn software tools by exploration. However, our knowledge of how users explore systems, and how exploration can lead to effective learning, remains limited. In this work studies are made of learning by exploration looking at factors that lead to retention as measured by the time saved in performing similar tasks repeatedly and for different task instructions intended to promote different modes of exploration. A theoretical model of the cognitive processes involved in exploration is extended to account for the differing patterns of explorations and to model the consequences if the various kinds of exploration for retention. The results will be incorporated into a software design methodology using the cognitive walkthrough approach for tools that can be effectively mastered through exploration. MDR-9154059 Mostow, David J. Carnegie-Mellon University $10,000 - 36 mos. (Jointly funded with the Applications of Advanced Technologies Program - Total Award $425,066) Using Automatic Speech Recognition to Improve Reading Comprehension Deficiency in reading comprehension has become a critical national problem. Workplace illiteracy costs billions of dollars in corporate retraining, industrial accidents, and reduced competitiveness. Although intelligent tutoring systems could help, their inability to see or hear students limits their effectiveness in diagnosing and remediating deficits in comprehension. This pilot project will test the feasibility of addressing this fundamental limitation in high-speed computing, automated speech processing, reading research, and artificial intelligence. Children's oral reading of elementary science material will be monitored using automated recognition of connected speech and prosodic features (such as hesitation and intonation) to extract information identified by teachers and reading experts in interactive instructional environments. IRI-9241467 Smith, John B.; Smith, F.D.; Calingaert, Peter; Jeffrey, Kevin and Holland, Dorothy C. University of North Carolina, Chapel Hill $10,000 - 12 (Jointly funded with the Information Technology and Organization Program, the Division of Computer and Computation Research Division and the Networking and Communications Research Program - Total Award $300,000) Building and using a Collaboratory: A Foundation for Supporting and Studying Group Collaborations This is the third year funding of a three-year continuing award IRI-9015443. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. The central focus of this research is experimental observation of groups doing real collaborative work and using systems and communications media explicitly designed to aid in their tasks. Expected results are to expand understanding of how people collaborate and how to design systems that augment collaborative activities. To this end the project has five interdependent components: 1) a theoretical foundation for observing and understanding the social and cognitive aspects of group collaborations; 2) tools for rapid prototyping and reconfiguration of application environments for use by working groups, and multimedia communications to support multi-person interactions; 3) protocol analysis tools to record and study how individuals and groups interact through the networked computer environment; 4) application testbed systems (generic and domain specific) that can be used by groups engaged in real work; and 5) group studies and experiments to test system, social and cognitive hypotheses. IRI-9244462 Strong, Gary W. Drexel University $64,485 -12 mos. Relationship Between Task Structure and Choice of Navigational Aid Navigating complex computer applications has become a critical problem. Understanding what the innate human navigational abilities are and how they relate to specific types of environmental complexity will facilitate the design of assistive devices for the navigation of complex computer applications. Two different techniques for using spatial information to help computer users navigate complex applications have been proposed by the authors. These techniques are derived from studies that have shown how the spatial distribution of information constrains attentional processes in vision. In this project the researchers are undertaking an examination of the relationship between the structure of computer search tasks and the spatial structure of assistance information. They are constructing four computer applications. In two cases the task and guidance information structures are matched in terms of spatial information needs and spatial information presentation, and in the other two they are mismatched. Use of these applications by subjects will be compared on the basis of speed and accuracy. The research is expected to produce a deeper understanding of human-computer task performance under periods of stress and the need for rapid decision-making. IRI-9148602 Thompson, William B. University of Utah $10,000 - 12 mos. (Jointly funded with the Robotics and Machine Intelligence Program and DARPA - Total Award $250,000) Vision-Based Navigation in Large-Scale Space This is the second year funding of a three-year continuing award IRI-8901888. This Principal investigator has transferred from the University of Minnesota to the University of Utah. The award includes a subcontract to the University of Minnesota for continuation of their role in studying how human experts use maps. This award in the Joint NSF/DARPA Initiative on Image Understanding and Speech Recognition is for an interdisciplinary study of computer vision in navigational reasoning. Computational and psychological approaches to spatial cognition will be combined to model expert map interpretation and perception of large-scale environments. The investigators will also develop representations and algorithms for navigational problem solving, with special attention to the role of vision in map-based navigation. Much of the work will be integrated with research projects by industrial collaborators. SPECIAL PROJECTS IRI-9244253 Fischer, Gerhard University of Colorado, Boulder $10,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program - Total Award $250,000) REU: Supporting Collaborative Design with Integrated Knowledge- Based Design Environments This is an REU supplement award. This supplement adds two undergraduate students to the project. One of the students will do research to develop domain constraints and to build a knowledge base of rules. The other will gather data on network layouts that are needed to track existing equipment and capabilities. Both will be exposed to experimental research techniques involved in the design of computer supported cooperative work environments. The larger project is described below. This research is funded under the Special Initiative on Coordination Theory and Collaboration Technology. This is one of eleven winners under that competition. The goal of this project is to develop a conceptual framework and a prototype system for collaboration in an synchronous mode among members of design team. The proposed design environments include knowledge-based and graphic construction components with issue- based hypermedia systems designed to support collaboration. The application domain for the prototype system is the design of communications network within heterogenous groups to work together over long periods of time. The large and growing discrepancy between an amount of potentially relevant knowledge for the design task and the amount any one designer can know and remember puts limits on progress in design. Overcoming this limit is a central challenge for developers of systems that support individual and collaborative design efforts. The work in this projects builds on previous developments embodied in the FRAMER system for the design if human-computer interfaces and the JANUS system for architectural design. IRI-9222783 Garcia, Oscar N. George Washington University $0 - 12 mos. (Jointly funded by the Division of Information, Robotics, and Intelligent Systems - Total Award $96,941) IPA Mobility Assignment Dr. Oscar N. Garcia is an IPA assigned to the IRIS Division. His Program Directorship is in the areas of Speech and Languages Interfaces, Visualization and Interactive Computing, Adaptive Human Interfaces and Communication Modalities. IRI-9245172 Stevens, Kenneth N.; Shattuck-Hufnagel, Stefanie and Manuel, Sharon Y. Massachusetts Institute of Technology $99,861 - 12 mos. (Jointly funded with the Language, Cognition and Social Behavior Program - Total Award $159,861) Word Recognition Based on Phonetic Features and Acoustic Properties This is the third year funding of a three-year continuing award IRI-8910561. The aim is to develop and evaluate procedures for representing words in the lexicon in a way that permits direct access of the words based on acoustic properties extracted from the speech signal. The basic unit for lexical representation is the phonetic feature of the type used by linguists for phonological descriptions. This representation indicates the relative locations of events expected in the sound, specifies additional features found relative to these landmarks, and indicates the degree of context-conditioned variability that is expected for individual features. The research phases include (1) the study of principal kinds of fluent speech variability, (2) the development of a small lexicon based on event-related features, (3) the examination of the performance of access of words based on event-related features, and (4) the development of an lexical access system with automatic extraction of acoustic properties. IRI-9212592 Wah, Benjamin W. University of Illinois, Urbana $0 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $35,395) HPCC: NSF Workshop on HPCC: Vision, Natural Language and Speech Processing and Artificial Intelligence This proposal is concerned with a workshop on High Performance Computing and Communications (HPCC) for Grand Challenge Applications - Vision, Natural Language and Speech Processing, and Artificial Intelligence. This workshop brings together invited experts from academia and industry, with the goal of identifying near-term and long term problems in supporting these grand challenge application problems. Many traditional results in these grand challenge applications involving vision, natural language and speech processing, and artificial intelligence were developed without the availability of HPCC systems. On the other hand, computer architects design HPCC systems without considering requirements in vision, natural language and speech processing, and artificial intelligence applications. In this workshop, key issues and potential approaches/research directions for the next five to ten years are identified, with the goal of answering the following problems: (1) What are grand challenge applications in vision, natural language and speech processing, and artificial intelligence that can benefit by the availability of HPCC systems? (2) How should HPCC systems be designed so they can support grand challenge applications in these areas? The interdisciplinary HPCC initiative has created a needed for infrastructure development and program planning. The meeting brings together about 22 experts in these fields with representatives of NSF to discuss current research issues and formulate interdisciplinary research frameworks. The goal of the meeting is to report those issues and frameworks and disseminate the conclusions in a document to the research community. MESSAGE FROM THE PROGRAM DIRECTOR The Robotics and Machine Intelligence Program (RMI) funded a number of research planning workshops in FY92, and those are listed in the General Research and Special Projects section. In FY92, the Program continued its collaboration with other NSF programs in funding of awards in the Scientific Databases Initiative and the Intelligent Material Handling Initiative, and participated in the High Performance Computing and Communications Initiative's Grand Challenge Application Group competition. Awards in those initiatives are also listed in the General Research and Special Projects section. Also in FY92, the Program participated in the NSF/EPRI (Electric Power Research Institute) Intelligent Control Initiative. The resulting RMI awards will be made in FY93 and listed in next year's Summary of Awards. Given the transfer of the Speech Recognition research area to the Interactive Systems Program, the RMI Program's budget for the remaining areas had a significant increase compared to FY91. New emphasis is being given in the Program to research on manipulation and motion planning with geometric uncertainty, to research on micro-scale robotic systems including microelectromechanical systems (MEMS), and to investigations of active sensing and image understanding techniques. Research advances in these topics are essential to the ultimate achievement of more intelligent and robust robotic and sensing systems, which will enhance the acceptance of robotics by end-user application communities. The RMI Program will continue to cooperate with other NSF programs and other agencies in encouraging the exploration of fresh ideas and approaches that result from the interaction of many scientific and engineering disciplines. ROBOTICS AND MACHINE INTELLIGENCE PROGRAM FISCAL YEAR 1992 RESEARCH PROJECTS COMPUTER VISION AND PATTERN RECOGNITION DMS-9245456 Abhyankar, Shreeram and Bajaj, Chanderjit Purdue University $20,000 - 12 mos. (Jointly funded with the Computational Mathematics Program - Total Award $120,000) Mathematical Sciences: Algorithmic Algebraic Geometry This is the second year funding of a three-year continuing toward DMS-9101424. This research is planned in algorithmic algebraic geometry and its applications to geometric modeling, symbolic computations over finite fields, and computational group theory, with special emphasis on the following topics: 1. Unambiguous solid model representations; 2. Robust surface intersections; 3. Implicitization and parameterization; 4. Multi-polynomial resultants; 5. Resolution of singularities in characteristic; and 6. Galois groups of algebraic curves. Algebraic geometry is a venerable, deep discipline in pure mathematics, involving symbolic abstractions of geometric objects. This is an area that is currently seeing diverse and exciting applications. IRI-9246448 Ahuja, Narenda and Anderson, George J. University of Illinois, Urbana $125,001 - 12 mos. (Jointly funded with DARPA - Total Award $250,000) Image Analysis, Synthesis and Perception of Dynamic 3-D Scenes for Tactical Navigation This is third year funding of a three-year continuing award IRI-8902728. This award in the Joint NSF/DARPA Initiative on Image Understanding and Speech Recognition is for a study of the visual cues that humans need to guide vehicles and manipulate objects remotely. Telerobotic operators may be hampered by response delays, limited visual bandwidth, and insufficient time to attend to critical details. This study will identify essential visual cues that must be extracted or preserved to enable competent navigation, obstacle avoidance, landing, docking, and manipulation. Theories will be tested by human capabilities in synthesized dynamic visual environments. IRI-9115280 Ben-Arie, Jezekiel Illinois Institute of Technology $68,864 - 12 mos. Real Time Shape Description Using Gaussian Wavelet Groups With Parallel Implementation This is the first year of a two-year continuing award. The research focuses on the development of novel image description operators based on the notion of Gaussian Wavelet Groups (GWG). These groups are synthesized in clusters that have high co-joint spatial-frequency resolution and are also efficient in detection and localization of shapes directly from gray-scale images. Specialized GWGs will be developed that are capable of measuring various shape descriptors such as location, orientation, size and boundary curvature. The GWGs will be generated in real time by novel lattice networks. To describe whole shapes, a novel topological structure called a scale-spinal-graph (SSG) will be developed, along with a network that can generate the SSG with parallel operations. The project will also develop an efficient image representation scheme using clusters of GWGs, for real time image compression. IRI-9249746 Boult, Terrance E. Columbia University $50,000 - 12 mos. PYI: 3-D Computer Vision This is the second year matching and the third-year base funding of a five-year continuing Presidential Young Investigator Award IRI- 9057951. This research supports Dr. Boult's investigations in 3-D computer vision. His research includes stereo algorithms, surface estimation, segmentation, material identification, error modeling, and information fusion. Dr. Boult is also interested in complexity theory and in the psychology of vision. DIR-9245592 Bovik, Alan C. and Diller, Kenneth R. University of Texas, Austin $20,000 - 12 mos. (Jointly funded with the Database Software Development Program, the Instrumentation and Instrument Development Program and the Computational Biology Program - Total Award $150,000) Numerical Instrumentation for Shape and Shape-Change Analysis via Three-Dimensional Microscopy The investigators will focus their research on fully developing the automated instrumentation capabilities of microscopic systems that offer the potential for quantifying the 3-D shape parameters, and changes in shape parameters over time, of microscopic-scale biological specimens. They plan to continue their program of studying diversity of complementary imaging modalities, including confocal microscopy, stereo microscopy, and conventional optical sectioning microscopy. The program of study will encompass microscopic data acquisition via various methods. The images obtained will be subjected to appropriate filtering schemes based on the physical principles of image formation to sharpen and improve clarity in individual micrographs and to remove information spurious to that of the local specimen structure. The filtering techniques, which will require considerable analytic study in their development, will be specific to each optical system used. Finally, the aggregate local optical data will be assembled and analyzed via sophisticated numerical instrumentation techniques to reconstruct the 3-D structure of a given imaged specimen and to quantify its size and shape characteristics as it evolves over time. The methods developed will be implemented and evaluated on a broad range of specimen types, including stained and unstained single and multiple celled subjects. Several different tissue preparations will be used to study the time evolution of specimen morphology over time during various biological processes. IRI-9247388 Chelberg, David M. Purdue University Research Foundation $4,000 - 12 mos. REU: Automatic Strategy Generation for Computer Vision This award provides supplemental funding for undergraduate participation in this research project IRI-9011421. The student will integrate the Programmer's Hierarchical Interactive Graphics Standard (PHIGS) into an existing geometric modeling system to accelerate the rendering of 3-D models. He will also develop a menu-driven user interface for model specification. This Research Initiation Award supports Dr. Chelberg's application of sensor modeling, Bayesian inference, utility theory, and real-time planning to problems of recognizing objects in range imagery. This approach to active computer vision is particularly appropriate for manufacturing, where object models can be derived automatically from design-database information. Possible applications include sensor selection (or design), part recognition, orientation determination, and bin picking. MIP-9245398 Chellappa, Ramalingam R. University of Southern California $10,000 - 12 mos. (Jointly funded with the Circuits and Signal Processing Program - Total Award $68,604) Representation and Recovery of Discontinuities in Some Image Processing Problems This research deals with the general problem of recovery of discontinuities in image processing problems such as image restoration, edge detection, and surface interpolation from sparse depth data. It uses a natural generalization of the basic Geman and Geman line process to handle arbitrary real-valued directions and magnitudes, and methods such as Besag's pseudo-likelihood methods to estimate parameters. IRI-9246449 Cooper, David B. Brown University $125,000 - 12 mos. (Jointly funded with DARPA - Total Award $250,000) Image Understanding for Service Robots This is the third year funding of a three-year continuing award IRI-8905436. This award in the Joint NSF/DARPA Initiative on Image Understanding and Speech Recognition is for development of sensing and navigational algorithms for service robots. The ultimate goal is to develop robots for servicing vehicles,loading munitions, scraping paint on shipboard, scrubbing surfaces, etc. Indoor robotics will be emphasized, although the studied techniques of passive stereo vision and continuous-contact tactile sensing should also be of use for outdoor robotics. The multidisciplinary team (including ties with several industrial groups) will also study problem-solving techniques in machine learning, planning, geometric reasoning,and robot control. IRI-9243659 Dyer, Charles R. University of Wisconsin, Madison $108,294 - 12 mos. Understanding the Dynamics of 3-D Shape Appearance This is the second year funding of a two-year continuing award IRI-9022608. As a monocular observer moves in a 3-D world, the parts which can be seen and their geometry change with the observer's vantage point. This project undertakes to quantitatively describe and model the relationships between motion and the induced change in the appearance of 3-D shape. Because motion is such an integral part of this information gathering and processing process, the investigators are developing a uniform spatio temporal approach to 3-D object representation and intermediate-level motion description for dynamic model-based 3-D object recognition. To model explicitly the dynamic appearance of objects as the viewpoint moves, the approach uses "aspectspace" which is the cross product of the image plane and viewpoint space. New 3-D object representations are being developed in this multidimensional space, including the asp and the rim appearance representation. These representations describe how visible, geometric features of projected shape can change over time due to egomotion or object motion. Complementing this work on dynamic viewer-centered representations, the research will also derive intermediate-level motion descriptions from the motion of features in spatio temporal image volumes prior to object recognition and description. Methods will be developed for intermediate-level motion description including spatiotemporal surface flow and flow line detection, segmentation of spatio temporal surfaces, and the detection of properties such as cyclic motion and T-junction motion. IRI-9209728 Fleck, Margaret M. University of Iowa $35,201 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $49,792) RIA: Finding Boundaries in Variable-Scale Images This is the first year funding of a three-year continuing award. Computer vision systems typically begin by finding boundaries in their input images, i.e. locations at which properties change sharply (often indicating the edges of objects). Boundary-finding algorithms require an estimate of how much values vary within each image region. Currently-available algorithms assume that this scale of variation is constant across the image and, typically, that it is known a priori. This assumption holds only for the simplest images and fails for many common sorts of inputs, including images containing significant texture, functions describing image texture (e.g. texture orientation at each image location), and surface depths computed by stereo matching. This research will develop an algorithm for estimating scale within individual image regions and, using it, a boundary finder that can operate on a much wider range of inputs. The new scale estimator relies on two keys ideas. To eliminate contamination from a few "bad" values, the standard deviation (the traditional scale estimator), is replaced by an estimator from robust statistics (e.g. the alpha-trimmed standard deviation). To avoid peaks in scale estimates near boundaries, the algorithm derives its scale estimate from the minimum-scale neighborhood of each image location. IRI-9209212 Flynn, Patrick, J. Washington State University $34,211 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $39,209) Feature Group Utility in Model-Based 3-D object Recognition Systems This is the first year funding of a three-year continuing award. One of the most demanding applications in computer vision is model-based 3-D object recognition. Such systems often use correspondence techniques to identify objects in an image; sets of bindings are formed between groups of scene features (e.g., points, curves, surfaces, or volumes) and compatible groups of model features. The discrimination ability (or utility) of individual feature groups in these systems is a challenging and important problem in computer vision research. When recognition systems are expected to select the correct model(s) out of a database of hundreds or thousands, indexing methods (which quickly reject a subset of the model database from further consideration by the system) are useful and often essential if objects are to be recognized efficiently. The use of feature groups offers a promising technique for performing this indexing task. IRI-9209729 Forsyth, David A. University of Iowa $43,783 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $58,783) Modeling and Recognizing Curved Surfaces from Image Outlines This is the first year funding of a three-year continuing award. Recognizing objects from a single view is hard, because their appearance depends crucially on the direction from which they are viewed. Curved surfaces are particularly subject to this effect. This research concentrates on recognizing curved surfaces from their outline in a single perspective image, using both theoretical studies and practical implementations, along two lines of investigation. The problem has already been solved for generic algebraic surfaces by the Principal Investigator. The result indicates that an image outline contains substantial amounts of information about an object's shape. The research will investigate extending this result to a wider range of popular surfaces, and will also investigate the practicalities of applying the resulting algorithms. Recent work by the PI and colleagues suggests that there is a large range of surface models which are intrinsically easy to recognize. These results suggest building a surface modeler oriented specifically towards computer vision applications, where the resulting models would be a fortiori easy to identify from any view. The research investigates theoretical issues involved in constructing such a modeler, and will implement a simple modelling system with these properties. IRI-9257298 Forsyth, David, A. University of Iowa $25,000 - 12 mos. NYI: Young Investigator Award This is the first year base funding of a five-year Presidential Young Investigator continuing award. Shape is an important cue in recognizing objects in pictures. Shape is, however, difficult to describe and use in computer programs for object recognition, because objects appear different in images taken from different view points. This effect is particularly pronounced for curved three-dimensional objects. Recent work has shown that, for some surfaces, it is possible to compute descriptions that draw only on information from a single picture, and are the same for every possible view of a given surface. These invariant descriptors instantly identify an object, up to a known ambiguity. This research investigates advanced descriptors and color information to recognize curved three-dimensional objects, drawn from a large number of known examples, in pictures taken from an unknown viewing position. IRI-9244932 Gauch, John M. Northeastern University $8,000 - 12 mos REU: Deformable Models for Computer Vision This award provides supplemental funding for undergraduate participation in this research project IRI-9109431. The student will work on computer graphics tools for visualizing deformable models and will develop interactive tools for initializing such models to fit anatomical structures in 3-D medical images. The student will also participate in the implementation and evaluation of the models themselves and related computer vision research tasks. The principal theme of this research is the development and application of deformable models for computer vision. In contrast to traditional, purely geometric object models, deformable models simulate elastic materials which change shape in response to externally applied forces. The main advantage of this approach is that it produces coherent object models where low level vision techniques often suffer from pixel artifacts and may not yield closed object boundaries. The work begins by extending existing surface-based models in two fundamental ways. First, models are developed which can dynamically change to reflect the underlying topology of the object being modeled. This makes it possible to model object deformations with less complete prior knowledge. Second, the deformable models are generalized so objects in higher dimensions can be modeled. This will enable scientists to study object behavior in time-sequence images and gain an understanding of the mechanisms involved in object deformation. In order to accurately model complex objects in nature, computational methods are needed that are capable of handling models with hundreds of thousands of elements. The approach combines two optimization techniques, simulated annealing and multiresolution analysis, to reduce both the time and space complexity of fitting deformable object models. To evaluate effectiveness of the deformable models, the object modeling techniques are applied to several challenging biomedical computer vision problems: image segmentation, shape description and object recognition. IRI-9245060 Hager, Gregory D. Yale University $80,645 - 12 mos. Resource-Bounded Sensor-Based Decision Making in Unconstrained Environments This is the second year funding of a two-year continuing award IRI- 9109116. A central problem in sensor data fusion is the recovery of complex, multi-component models from various sensory modalities. An essential component of this process is a description of what decision must be made from the recovered model. This description is referred to as the sensing task. The goal of this research project is to develop and analyze techniques for recovering the minimal or least detailed model required to make a satisfactory decision for a given sensing task. In particular, the focus is on the task-directed recovery of composite models (models composed of several components). The model recovery method is based on the numerical solution of systems of nonlinear constraints using interval bisection. Previous work, based on making decisions about objects modeled by a single parametric form, showed that these methods are general, natural, simple to implement, and computationally effective. In addition, the investigators were able to incorporate notions of the cost of computation and the value of information into the recovery process, and to terminate the recovery process when the model with the highest net value (decision payoff minus computational cost) was reached. They are now extending the underlying recovery method by incorporating model refinement and data segmentation in a manner that also exploits information about the sensing task. The results are expect to have particular impact in application areas where good a priori models ar not available. Examples of these domains include classification and sorting of irregular (e.g. naturally occurring) objects, supervisory control, and ultimately, sensor-based control of partially or fully autonomous vehicles. IRI-9257990 Kriegman, David J. Yale University $25,000 - 12 mos. NYI: Young Investigator Award This is the first year base funding of a five-year Presidential Young Investigator continuing award. This research addresses problems in machine perception and robotics. The work is focused in two areas: recognizing curved objects in images, and mobile robot navigation. Recognizing complex curved 3-D objects, whose image features depend strongly on viewpoint, is one of the fundamental problems of computer vision. Objects are modelled by collections of algebraic surfaces and their intersection curves, and the geometric constraints involved in predicting and interpreting the images of these models are represented by sets of polynomial constraints. Techniques from algebraic geometry and robust numerical methods can be used to solve these constraints. Viewpoint dependent features are predicted from an object model and represented as an aspect graph which enumerates all topologically distinct views. Elimination theory is used to relate image measurements to objects, and optimization techniques can be applied to estimate an object's pose; objects are then recognized from a library of models. Problems to be addressed include automatic generation of object models, new object representations, prediction for sensors with limited resolution, computational efficiency. and indexing large data bases. Mobile robots provide a dynamic test-bed for robust computer vision algorithms in an unstructured world. In addition to developing new sensors, research focuses on obtaining the necessary information to perform a task. While some activities are readily achieved through visual serving, others require developing a 3-D representation of the environment. Stereo and structure-from-motion provide 3-D information from image measurements which can be incorporated into a relational map. This map forms the basis for motion planning; paths to revisit previously seen places and strategies for exploring new areas can be determined. Additionally, robot tasks and cultural constraints are described more naturally when objects are recognized within this map. IRI-9244479 Levinson, Robert University of California, Santa Cruz $47,461 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $94,922) Adaptive Pattern-Oriented Chess This is the first year funding of a three-year continuing award. Although chess computers now are competitive at master and grandmaster levels, that is where their resemblance to human players ends. Psychological evidence indicates that human chessplayers search very few positions, and base their positional assessments on structural/perceptual patterns learned through experience. Morph is a computer chess program that has been developed to be more consistent with the cognitive models. The learning mechanism combines weight-updating, genetic algorithms, and explanation-based and temporal-difference learning to create, delete, generalize and evaluate graph patterns. An associative pattern retrieval system organizes the database for efficient processing. The main objectives of the project are to demonstrate capacity of the system to learn, to deepen our understanding of the interaction of knowledge and search, and to build bridges in this area between AI and cognitive science. To strengthen the connections with the cognitive literature the system's knowledge is to come from its own playing experience, no sets of pre-classified examples are given and beyond its chess pattern representation scheme little chess knowledge such as the fact that having pieces is valuable (let alone their values) has been provided to the system. Further, the system is limited to using only one ply of search. IRI-9249141 Malik, Jitendra University of California, Berkeley $62,500 - 12 mos. PYI: Computer Vision This is the fourth year matching and base funding of a five-year continuing award IRI-8957274. This Presidential Young Investigator award is for the support of Dr. Malik's work in computer vision. He has brought deep mathematical insight to innovative problems in shape from shading, curved-object recognition using aspect graphs, and scale-space approaches to edge detection and early vision. He is continuing work in image segmentation, shape recovery, and object recognition. IRI-9244469 Medioni, Gerard University of Southern California $78,314 - 12 mos. Structural Hashing: Efficient Object Recognition This is the second year funding of a two-year continuing award IRI- 9024369. The PI has developed an object recognition method based on hash tables for indexing and rapid retrieval of encoded models. The plan is to perfect the method for flat (2D) objects, implement it on a Connection Machine, and move to noisy data, texture, and highly cluttered scenes; and also to generalize the method to 3-D objects, using the splash as the basic data representation. The goal of this research is more efficient and robust object recognition. IRI-9210861 Meer, Peter Rutgers University Busch Campus $34,860 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $39,860) RIA: Robust Object Recognition This is the first year funding of a three-year continuing award. Performance of computer vision algorithms is contingent upon the quality of the input. With uncorrupted data most of the algorithms perform well, however, once noise is present the results may become unreliable. The noise can appear as uncertainty about the input and/or incorrect assumptions about the underlying model structure. A robust computer vision algorithm should tolerate both types of noise. The methods borrowed from statistics or estimation theory were not designed for visual data where resistance to data uncertainty and tolerance of incorrect assumptions are of paramount importance. Good performance for noisy visual data can be achieved through robust detection of the consensus among the outputs of a large number of identical non-robust processes, with a methodology called the Consensus Paradigm. A similarity exists between the computational steps of recently proposed 3-D object recognition algorithms and the robust estimation methods. The research will exploit this similarity and develop object recognition algorithms with improved noise resistance. Results of the research will contribute to the development of techniques for autonomous decision making in unfamiliar visual environments. IRI-9244455 Miller, Russ SUNY at Buffalo $85,120 - 12 mos. Parallel Algorithms for Image Processing and Computational Geometry This is the second year of a three-year continuing award IRI- 9108288. The research consists of the design, analysis, and implementation of efficient algorithms and paradigms to solve problems in image analysis and computational geometry on massively parallel computers. The focus of the algorithm development will be on fine-grained distributed memory topologies for which commercially available machines or prototypes already exit. These include the mesh, hypercube, pyramid, and reconfigurable mesh. The concentration will be on problems with application to robotics and image processing. Initially, the research will consider 2-and 3- dimensional problems in intermediate-level image analysis, as well as 2- and 3-dimensional problems in computational geometry. The optimalities of all of the problems considered are currently open problems on the appropriate architectures. Further, it does not appear that for the problems considered, algorithms for 2- dimensional objects can be extended to solve the problems for 3- dimensional objects in optimal time. Rather, it appears that radically different approaches will be required. The problems considered involve connectivity, proximity, area, intersection, and minimal-area enclosing polygons, to name a few. IRI-9247044 Miller, Russ SUNY at Buffalo $4,000 - 12 mos. REU: Parallel Algorithms for Image Processing and Computational Geometry This award provides supplemental funding for undergraduate participation in this research project. The student will work on minimization techniques on massively parallel computers, for use in solving crystal structures using X-ray and electron diffraction data. IRI-9210786 Nabet, Bahram Drexel University $65,000 - 24 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $70,000) RIA: Optoelectronic and Preattentive Vision In accomplishing some tasks, biological visual systems can outperform even the most sophisticated image processing systems. The goal of this research is to build optoelectronic circuits which perform some important image sensing and processing tasks in a somewhat similar manner biological visual systems. The circuits are then not only useful in performing these tasks better than conventional methods, but also provide a tool of analysis, a level of abstraction, for the original biological systems. This goal is methodically pursued by proposing elements of a visual system that are based on a basic model of ionic conduction in nerve membranes. Processing capabilities of each of the proposed elements is demonstrated. Next, more sophisticated tasks are proposed which can be performed by combining these basic elements together. It is argued that this intelligent control of acquisition of visual data is essential for real time operation in a complex dynamic world. IRI-9246341 Negahdaripour, Shahriar University of Miami $4,595 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $9,190) Equipment Supplement: RIA: Relaxing the Brightness Constancy Assumption in Motion Vision This award provides supplemental funds for replacement of equipment following the PI's transfer from another institution. This award continuation is for analysis of time-varying imagery when the camera and light source are both moving relative to objects in a scene. This extends the optic flow techniques developed for camera or object motion with a constant light source. Applications include passive vision in outer space and active vision under water or in reconnaissance situations. Industrial robotic applications may also be developed. IRI-9240182 Pinter, Robert B. University of Washington $68,608 - 12 mos. Neural Networks Modeling Cortical Cells for Machine Vision This is the second year funding of a three-year continuing award IRI-9113148. Biological visual systems perform many demanding tasks which are far beyond the present capabilities of electronic hardware. The success of many biological systems can often be attributed to a large degree of preprocessing at the peripheral sensory levels. Mathematical statements of relevant preprocessing include multiplicative nonlinear lateral inhibition, center-surround receptive fields and Gabor function receptive fields. The objectives of this project are to design and implement in digital simulations such biologically-motivated receptive fields, for automated visual preprocessing. The design will be suitable for future implementation in analog parallel integrated circuitry. The capability of priming the visual preprocessor to concentrate on certain features via Adaptive Resonance Theory will be investigated. These preprocessing schemes will be tested on an industrially relevant task: in-process inspection of laminated composites. IRI-9240551 Ponce, Jean University of Illinois, Urbana and Kriegman, David J. Yale University $139,998 - 12 mos. Representations and Algorithms for Recognizing and Locating Three- Dimensional Curved Objects from Monocular Images This is the second year funding of a two-year continuing award IRI- 9015749. This award supports research on ways for computer vision systems to exploit the Computer Aided Design models available for many manufactured objects. The goal is to learn how to build a vision system capable of recognizing and locating instances of curved 3-D objects modeled by parametric surface patches and their intersection curves in the imperfect line drawing obtained by edge detection from a single image. Ongoing implementation of research system building blocks will be completed, including specialized components of a solid modeler, algorithms for constructing exact aspect graphs of object models, and techniques for locating model instances from image contours. Novel recognition strategies will be investigated, based on both object- and viewer-centered representations. New data structures and algorithms will be developed that efficiently implement these strategies; these will be integrated into a working recognition system. A notable aspect of this research is the systematic characterization of geometric constraints by polynomial equations, which may then be solved by application of available computational and mathematical tools. IRI-9243658 Preston, Kendall J. Carnegie Mellon University $62,787 - 12 mos. SC: Research on Design Evaluation of Image Processing Systems for Pattern Recognition This is the second year funding of a two-year continuing award IRI- 9100243. This two year continuing award supports software capitalization for the continued development of benchmarks for evaluation of image processing systems. The tasks include investigation of the performance of new system designs and design modifications, concentrating on massively parallel architectures and the "software" image processing systems which run on high-performance work stations under UNIX, MS/DOS, and Apple Macintosh operating systems; expansion of the Preston-Seigart benchmark research to include a larger segment of the image processing community; and conduct of a trend analysis on hardware and software architectures to determine what areas of image processing research are likely to produce the most cost-effective platforms. The results of this project will be fed back to the participating organizations and also disseminated through professional publications, conference presentations, and liaison with professional society and industrial standards committees. IRI-9113787 Quek, Francis and Jain, Remesh C. University of Michigan, Ann Arbor $49,233 - 12 mos. (Jointly funded with the Interactive Systems Program and the Cross Disciplinary Activities Office - Total Award $151,839) A Gesture Interpretation and Voice Recognition Multi-Modal Human Machine Interface This is the first year funding of a three-year continuing award. This research is to development computer vision methods for tracking hand gestures. Gestures are followed by video tracking of a glove studded with light emitting diodes. Other glove marking methods will also be examined. Gesture interpretation will be performed by a dynamic vision system using the correspondence via event detection (CED) algorithm which segments and tracks the motion of the LEDs over multiple video frames. The primary focus of the research will be to extend existing algorithms to handle occlusion. Constraints inherent in hand anatomy, function and gesture categories will be exploited. Voice input will be implemented using commercial disconnected speech recognition technology. The voice commands will be used to segment gesture epochs and to disambiguate the gesture categories. Rules for combining gesture interpretation and voice for gestures stream segmentation will be developed. A gesture library constituting taxonomy of gestures will be used for temporal coordination and conflict resolution. Both monocular and stereo camera environments will be used. Parallelization of the algorithm will be done for to assure real-time capability and to support subsequent research in which bare hands might be tracked without the use of markers. IRI-9249350 Samet, Hanan University of Maryland, College Park $70,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $95,000) Spatial Data Acquisition and Processing This is the second year funding of a three-year continuing award IRI-9017393. The efficient processing of spatial data plays an important role in solving problems in computer vision, robotics, and computer graphics. The acquisition of spatial data as well as its processing will be investigated. The acquisition will be concerned with data useful for cartographic applications (e.g., terrains) and maps in general. In the case of maps, the interest is in the acquisition of relative spatial information (e.g., based on a map's legend) rather than precise information (e.g., locations of cities, roads, etc.). An investigation into the concept of a hypermap and the related representation issues will be conducted. Other problems include the investigation of hierarchical surface representations, large spatial databases (especially those including line segments). Attempts will be made to parallelize any algorithms that are developed and they will be tested on a Connection Machine. DBS-9213246 Sanocki, Thomas and Bowyer, Kevin W. University of South Florida $8,733 - 12 mos. (Jointly funded with the Human Cognition and Perception Program - Total Award $69,864) Time Course of Object Recognition This is the first year funding of a three-year continuing award. Despite the ease with which humans recognized everyday objects, the processes by which this occurs have become a central concern in cognitive science, because object recognition is very difficult; for example, computer scientists have only been partially successful in creating computer programs for recognizing objects, and even those programs stretch the computational limits of the most sophisticated machines. Since humans are "good at" object recognition, this research will focus on human abilities, with emphasis on how computational complexities might be handled. Human recognition takes place over a time span of about 100 ms., and the research will use newly developed techniques to control exactly what information is available during that period of time. The working hypothesis underlying this research is that for humans computational burdens are reduced because the brain uses certain types of information early in that time period and other types of information later in that time period. For example, early in time general information about the shape or orientation of an object might be used to narrow down the possibilities, whereas later in time information about more idiosyncratic details would be used to arrive at the proper identification. The research will examine a number of possible types of information that have been suggested in the psychological and computer science literatures. The research will also use several related techniques, because each has its own strengths and weaknesses. Initial work will use relatively simple objects for exploration; later work will use large samples of common objects. Later work will also seek to extend principles of earlier work to the perception of objects within scenes, and to the process of word perception. The knowledge gained will help in the design of robots for a variety of purposes and of aids for visually disabled persons. IRI-9243418 Shapiro, Linda G. and Haralick, Robert M. University of Washington $155,000 - 12 mos. Automated Model-Based Computer Vision This is the second year funding of a three-year continuing award IRI-9023977. This project addresses a key problem which currently limits the success of machine vision in new applications: design of algorithms. Methods are explored to partially automate the design of vision algorithms for recognizing and localizing industrial parts and other man-made objects. The project suggests a unified methodology for partially automating the generation of vision algorithms using methodologies of search, reasoning under uncertainty, perspective geometry analysis, and especially mathematic morphology. The work also addresses the important software-related issue of testing and evaluation of vision algorithms. The results should find use in industrial, military, and space applications. IRI-9247045 Shih, Frank Y. New Jersey Institute of Technology $4,000 - 12 mos. REU: Development of New Back-Propagation Morphological Algorithms This award provides supplemental funding for undergraduate participation in this research project IRI-9109138. Under the guidance of graduate students, the undergraduate student will develop and implement an integrated graphics, image, and text package for creating, displaying, manipulating, and storing various types of graphical objects such as lines, rectangles, and circles. Mathematical morphology an increasingly important and often-used technique in image processing and machine vision applications. Nevertheless, the iterations of morphological operations are a bottleneck in implementation on a pipeline parallel architecture. The objectives of this project are to propose a set of new morphological operators, called back-propagation morphology, which is different from the traditionally defined morphology for solving time-consuming iteration problems, and to develop its underlying theorems, algorithms and architectures for various vision applications. First-stage experiments show that the back-propagation morphology has the advantages of deriving a root of a signal which only requires two scans without numerous iterations and being suited for parallel architectures. It is anticipated that the proposed research program will lead to fundamental advances in the theoretical understanding of the new back-propagation morphological operations. The algorithms and architectures developed and the research findings produced by the project will also have substantial utility for industry in advancing machine vision inspection and recognition technologies. IRI-9123720 Sklansky, Jack University of California, Irvine $45,000 - 36 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Information Technology and Organizations Programs - Total Award $330,000) Biologically Inspired Intelligent Classifiers This project will advance knowledge about design of intelligent agents. It has two objectives. The first objective is the development of mathematical insights and procedures for automating feature discovery and learning by evolution -- two biologically- inspired forms of long-term learning -- in the construction of automatic pattern classifiers. The second objective is the incorporation of automated feature discovery and evolution in automatic analyzers of large-scale image data banks and time- varying visual signals. This project will work to achieve these objectives by building on the Principal Investigator's and his students' techniques of adaptive hyperplane placement, window training (an extended form of stochastic approximation), relaxed branch-and-bound search, and genetic feature selection. The project will implement and test the resulting mechanisms of long- term learning in three-stage neural classifiers of handwritten numerical characters, tree classifiers of medical images in a large data bank, and adaptive segmenter of moving objects in digitized video. This project will advance the technology of long-term learning in intelligent machines, leading to machines that can detect and explain coherence buried deeply in enormous amounts of noisy multidimensional data. This project will also advance the technology of multiple-sensor robotic vision by automation the discovery of highly discriminating combinations of features from diverse sensors. IRI-9102860 Sloan, Kenneth R. University of Alabama, Birmingham $46,573 - 12 mos. (Jointly funded with the Office of the Cross- Disciplinary Activities Office and the Division of Computer and Computation Research - Total Award $81,605) Contours, Surfaces, and Volumes This is the first year funding of a two-year continuing award. This award addresses the problem of reconstructing a three-dimensional object from a set of planar contours (cross-sections). This is an important problem in many research areas. For example, biologists, medical clinicians, and quality control inspectors all try to understand the shape of objects, based on serial sections. There are two basic approaches to the problem: volume-based and surface-based reconstruction. If the spacing between the sections is relatively small, then the volume-based approach is preferred and the intermediate contours may not be necessary. However, when the spacing between sections is large, a surface-based approach is necessary. This research will continue the Principal Investigator's investigations into reconstruction of surfaces from sparsely-spaced planar contours, using a surface-based approach and comparing it with volume-based approaches. The algorithms and methods developed should be useful in both scientific and industrial applications. IRI-9243837 Srihari, Sargur N. State University of New York $104,381 - 12 mos. Knowledge-Based Document Image Understanding This is the second year funding of a two-year continuing award IRI- 9014110. This research is about automating one aspect of analyzing a document image and deriving a high-level representation of its visual content. Documents contain photographs and accompanying text. This effort is concerned with arriving at an integrated interpretation of the communicative unit consisting of photographs and their captions. When text describes salient aspects of a photograph, it is possible to use the text to direct a vision system in understanding the photograph. There are two components to this research: the first deals with language issues and the second with development of a vision subsystem. Methods of extracting visual information from text, specifically cues required to identify salient objects, are to be studied; such information may be present in a variety of forms, based on both syntax and semantics. The role of textually extracted visual cues in performing visual object recognition is also to be studied. As a test of the theory, it is proposed to develop a system where the result of parsing a caption of a newspaper photograph is used to identify human faces in the photograph. The face location subsystem will incorporate scale invariant techniques, and filters that characterize faces based on the presence of distinguishing visual features. IRI-9112267 Thompson, William B. University of Utah $56,380 - 12 mos. Structure-From-Motion Based on Information at Surface Boundaries This is continuation funding of a project originated while the PI was at the University of Minnesota. The project is in its first year and the PI has transferred to the University of Utah. Existing computational models of structure-from-motion are based on variations of optical flow or feature point correspondences within the interior of single objects. This project investigates a new, boundary-based cue for determining motion and shape. An approach is outlined for determining that an object is rotating in depth. Only simple, qualitative computations are required, avoiding the need to accurately determine how flow patterns are varying over the surface of interest. Generalizations involving integration of boundary and surface cues will be explored. Ultimately, the approach will be extended to other types of motion such as translation of a wide field-of-view cameras. To support this boundary-based analysis improved methods will be evaluated for estimating optic flow in the presence of flow discontinuities. The proposed algorithm integrates flow/surface analysis with the structure-from-motion determination while addressing certain deficiencies in previously proposed techniques. Results from this study should provide more general reliable and economical methods for dealing with the important problems involved in analyzing visual motion. IRI-9247374 Thompson, William B. University of Utah $0 - 12 mos. (Jointly funded with the Interactive Systems Program and DARPA - Total Award $250,000 Vision-Based Navigation in Large-Scale Space This is the third year funding of a three-year continuing award IRI-8901888. This award in the Joint NSF/DARPA Initiative on Image Understanding and Speech Recognition is for an interdisciplinary study of computer vision in navigational reasoning. Computational and psychological approaches to spatial cognition will be combined to model expert map interpretation and perception of large-scale environments. The investigators will also develop representations and algorithms for navigational problem solving, with special attention to the role of vision in map-based navigation. Much of the work will be integrated with research projects by industrial collaborators. The award includes a subcontract to the University of Minnesota for continuation of their role in studying how human experts use maps. IRI-9247374 Thompson, William B. University of Utah $4,000 - 12 mos. REU: Vision-Based Navigation in Large-Scale Space This award provides supplemental funding for undergraduate participation in this NSF/DARPA research project. The student will be responsible for synthesis of test data for evaluation of a system being developed to integrate high-level problem solving and lower-level perception to solve robot navigation problems. This will involve the generation of synthetic views of terrain from given locations. IRI-9201751 Tomasi, Carlo Cornell University $85,000 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $100,000) The Factorization Method for Image Sequence Analysis This is the first year funding of a three-year continuing award. The long-term goal of this research is the recovery of accurate camera motion and dense three-dimensional shape information from television images at video rate. In prior work, the Principal Investigator developed a matrix-based factorization method for this task, based on a set of feature points tracked from frame to frame of a dense image sequence. Experiments with the method demonstrated a dramatic performance improvement over existing shape and motion recovery systems. This research will now develop a characterization of the noise sensitivity of image sequence analysis, explore efficient and incremental numerical methods for factorization, and reformulate the method for perspective projection, multiple motions, and dense shape results. As an exploratory advance into a new area of research an investigation of the link between motion analysis and multi-frame of object recognition will also be conducted. This research should contribute to a better theoretical understanding of visual motion analysis, and produce a vision module that would let a robot localize itself in the environment, draw a map of its own surroundings for navigation and obstacle avoidance, and perceive the shape of objects in order to recognize or manipulate them. IRI-9103143 Tuceryan, Mihran and Jain, Anil Michigan State University $233,401 - 24 mos. Integration Of Perceptual Grouping Modules With 3-D Interpretation Modules In Computer Vision The intermediate stage of visual processing consists of general purpose perceptual organization modules and 3-D interpretation modules. This research is concerned with computational and algorithmic models of the intermediate processing modules, and the interaction of these modules with each other. It studies the integration of a number of perceptual grouping modules with 3-D interpretation modules. If the integration is done across levels as proposed here, the specific knowledge about the physical world provides a powerful set of constraints with which to combine the outputs of the grouping modules. The project involves the development and implementation of a subset of the possible visual processing modules in an integrated manner. The goal of this project is to demonstrate that when the results of lower and intermediate level modules are interpreted within the context of 3- D modules, that is, when feedback is provided from the higher-level modules, the results obtained are more robust and well defined. The specific modules involved in this research are: (1) the proximity and smoothness-grouping modules integrated with symmetry detection modules; (2) the grouping modules integrated with 3-D line-labeling modules; (3) the symmetry-detection module integrated with shape-from-contour modules. Algorithms will be tested on real images of cluttered scenes of opaque, piecewise- smooth objects, and the usefulness of the grouping algorithm and the integration process in analyzing these images will be assessed. IRI-9243427 Udupa, Jayaram K. and Herman, Gabor T. University of Pennsylvania $105,286 - 12 mos. Surfaces and Objects, and Their Borders in Multidimensional Images: Theory and Algorithms This is the second year of a two-year continuing award IRI-9013341. The objectives of this research are (i) to develop a unified geometric theory of objects and their boundaries in n-dimensional discrete spaces, and (ii) to develop algorithms for detecting objects and boundaries in n-dimensional images. The approach to theory development uses an orthogonal tessellation of the n-space into hypercubes, leading to results which are discrete counterparts of some essential results of continuous topology. Thus, objects and boundaries can be represented as graphs, and their detection translates to graph traversal. The algorithm development focuses on devising efficient traversal algorithms, both when the images are locally segmentable and when they may not be. In the former case there is always a unique spanning tree which is easy to traverse. In the latter case, there is no unique solution and hence the approach is to develop optimal (e.g., using dynamic programming) and suboptimal traversal strategies. The development of common theory and algorithms for visualization and analysis of images will be of use in many disciplines using multidimensional image data. IRI-9249178 Ullman, Shimon; Grimson, W. Eric L. and Richards, Whitman Massachusetts Institute of Technology $142,653 - 12 mos. Image Partitioning and Selection in Human and Machine Vision This research was originally funded for a three year period under award IRI-8900267. On the basis of creative accomplishments to date in this work, the award has been extended to five years. This award funds the fourth year. The work has focused on the role of feature selection and saliency in both human and machine vision, and has produced novel results in saliency nets, complexity analyses for recognition methods, formal models for analyzing effectiveness of recognition, and criteria for goodness of features, and foundations for automating model acquisition from experience. The research continues and now focuses on explorations in automatic active model acquisition; methods for matching of point features and smooth contours for determining and optimizing image-to-model matching; formal analyses of robustness to noise and uncertainty; and relative effectiveness of feature types; and exploration of saliency-based visual attention and human approaches to coordinate frame selection. This could lead to advances in both understanding of human visual recognition performance and the design of improved machine vision systems. The promise is for both greater efficiencies for recognition in complex scenes, and more robust recognition, with less dependence on pre-defined models. IRI-912204 Weiman, Carl F.R. Transitions Research Corporation $0 - 24 mos. (Jointly funded with the Small Business Innovation Research Program - Total Award $249,895) SBIR: Robot Vision System Based on Log-Polar Image Plane Coordinates The objective of this SBIR Phase II project is to design and build a dynamic camera control system and processing hardware to exercise high-performance real time vision algorithms. High performance is based on reformulating algorithms in log-polar image plane coordinates, which significantly reduces pixel count and computations per pixel. Capabilities of this system will include depth perception based on binocular stereo and optic flow in dynamic environments. This system can be duplicated as a testbed for vision researchers. The binocular camera head will be available as a self-contained module which can be attached to any processing architecture. Phase I research has demonstrated a complete spectrum of image processing algorithms formulated in log-polar coordinates, proving the feasibility and efficiency of such a system. Orders-of-magnitude reduction in computational requirements suggest dramatic benefits for robot vision in general and mobile robots in particular. The proposed technology could improve the cost/performance ratio for robot vision systems by several orders of magnitude. This opens commercial niches unattainable by conventional Cartesian-coordinate-based computer vision. Markets include autonomous floor cleaners, transporters, security robots, and service robots. IRI-9157260 Wolberg, George The City College of CUNY $36,960 - 12 mos. PYI: Algorithms for Image Manipulation This is a second year base of a first year matching funding of a five-year Presidential Young Investigator Award. The research aims to develop algorithms for image manipulation, including new techniques for image reconstruction, antialiasing, and generalized image warping. This investigation is developing and evaluating new nonlinear algorithms for image reconstruction which are efficient, amenable to hardware implementation, and superior to popular techniques of equivalent cost. Such algorithms play an important role in warping systems. The research also investigates the role of image warping in scientific visualization, and examines the consequences of separable image warping for elastic matching in object recognition. IRI-9112649 Zhuang, Xinhua University of Missouri, Columbia $75,952 - 12 mos. Robust Image - Motion Analysis This research evaluates image motion analysis using a model-fitting estimator developed by the Principal Investigator. This estimator, which fits the unknown likelihood function with a function selected from a multi-scale family of model likelihood functions, is expected to be highly robust, less dependent on initial guesses, and resistant to data confusion in multi-object tracking. The method has so far been tested with simulated data; tests would be expanded to real data from both laboratory and outdoor scenes, including mixed modalities with range and visual images. ROBOTIC PERCEPTION AND ACTIVE SENSING IRI-9248975 Aloimonos, Yiannis J. University of Maryland, College Park $50,000 - 12 mos. PYI: Purposive and Qualitative Active Vision This is the second year matching and third-year base funding of a five-year PYI award IRI-9057934. This award supports research in active vision and navigational reasoning. The principle goal is to build robotic systems with robust real-time visual capabilities for accomplishing specific navigational tasks such as moving-object detection by a moving observer, object tracking, and obstacle avoidance. Other research in robust scene analysis, visual learning, parallel computation, path planning, and other related topics will also continue. IRI-9248672 Ballard, Dana H. University of Rochester $80,000 - 24 mos. (Jointly funded with the Information Technology and Organizations Program and the Knowledge Models and Cognitive Systems Program - Total Award $225,000) Animate Robotic Vision This award is a two year creativity extension to a current three- year award (IRI-8903582), based on creative accomplishments in active vision and reinforcement learning. The PI has proposed extensions to his work in combining deistic primitive behaviors (primitives that dynamically reference the world rather than depending on models) to produce the complex behaviors needed for intelligent robotic agents to perform practical tasks in active vision. Topics will include economical search of three-dimensional space for small objects, speedup of the learning algorithms, and development of a set of qualitative grasping primitives, integrated with visual feedback, as a natural extension of the vision primitives already developed by the PI. This last is a promising alternative to the fine, open-loop control used in conventional robotic systems. IRI-9247810 Cooper, Paul Northwestern University $4,000 - 12 mos. REU: Parallel Recognition and Manipulation of Structure This award provides supplemental funding for undergraduate participation in this research project IRI-9110482. The student will develop software for simulation of connectionist networks, and will interface the software to hardware controlling the position of a camera. This research project examines some of the basic high- level computer vision problems associated with the task of manipulating and assembling structural frames under autonomous visual control. The approach treats high-level vision as uncertain inference. Uncertainties arising from the sensing process and scene itself are explicitly represented and resolved with prior knowledge at a high level of abstraction. The uncertain inference processes are realized as a parallel network, based on a Markov Random Field formalism. The structure assembly domain focuses attention on geometric reasoning about spatial relationships between parts, particularly in the presence of uncertainty. Finally, the assembly task provides a functional goal that can define criteria for deciding when too much uncertainty is present. This provides a rationale for planning and acquiring new images in an active framework. This research could have significant impact on the flexibility of future computer vision systems, and may also increase our understanding of human high level vision. IRI-9210560 Crisman, Jill D. Northeastern University $28,304 - 12 mos. RIA: Instruction of a Robot Using Visual Commands This is the first year funding of a three-year continuing award. Discovering the fundamental vision-motion primitives that are used by people to navigate and manipulate objects would lead to a natural human-machine interface for programming robotic systems especially in unstructured environments. This research explores the definition of a set of natural robotic interface commands used for navigation. A more natural navigational interface would allow robots to be more easily integrated into applications such as material handling for flexible manufacturing, planetary or underwater exploration, or automated wheelchairs. The programmer issues a command such as "go there" to the robot by specifying "there" as a location on a video screen. The robot then navigates to the desired location, avoiding any obstacles along the way. As a tool for discovering a complete set of navigational commands, project will implement and experiment with an initial set of commands. This experimentation should show where the initial command set is redundant or lacking and is essential for insuring that the commands are natural for the user. As part of this research, a vision algorithm will be developed that can extract visually distinctive features from a wide variety of objects. The command set and the algorithms for feature extraction and tracking will be the largest contribution of this work. Future work involves further testing of the navigational commands on a convenient test platform and extending the methodology derived for discovering the primitive navigation commands to deriving manipulation commands. IRI-9242077 Davis, Ernest S. New York University $53,150 - 12 mos. Perception and Planning This is the third year funding of a three-year continuing award IRI-9001447. This research supports further development of the PI's logic-based representation for common sense spatial and perceptual reasoning. Past efforts have developed a theory of qualitative spatial and physical reasoning for predicting interactions of moving objects and behaviors of kinematic chains. New work focuses on a strategic planner for simulated and real robots, with emphasis on the use of knowledge in planning of perceptual actions. IRI-9248259 Huttenlocher, Daniel P. Cornell University $62,500 - 12 mos. PYI: Recognition and Robotic Assembly This is the third year base and matching funding of a five-year PYI award IRI-9057928. This award supports Dr. Huttenlocher's research in model-based object recognition. Robotic recognition in complex scenes tends to be slow and of uncertain reliability. Dr. Huttenlocher has been developing methods that are efficient and accurate, with predictable or controllable error probabilities. He will extend his characterization of Hough-based methods to interpretation-tree search and other object recognition approaches, and plans to develop new algorithms using techniques of computational geometry. These perceptual capabilities will be used in a robotic planning and assembly system being developed jointly with Dr. Bruce Donald. IRI-9244616 Huttenlocher, Daniel P. Cornell University $4,000 - 12 mos. REU: Recognition and Robotic Assembly This award provides supplemental funding for undergraduate participation in this research project IRI-9108610. The student will participate in implementation of a massively-parallel version of this laboratory's rasterized approximation methods for computing the Hausdorff distance between points of a model and corresponding points of an image set. IRI-9246470 Rao, Nageswara Old Dominion University $4,000 - 12 mos. REU: Multi-Sensor Systems for Robot Navigation in Partially Known Terrains This award provides supplemental funding for undergraduate participation in this research project. The student will implement and test a multi-sensor navigation algorithm using the TRC Labmate mobile robot, equipped with 32 ultrasonic and infrared distance sensors. Navigation methods using multi-sensor systems are considered for terrains whose geometric models are known only in certain parts. Exact navigation algorithms, as well as those that obtain approximate paths with lesser computation time while compromising the optimality are studied. Navigational methods for terrains, where polygonal approximations are inefficient, are explored based on fractal interpolation and approximate path planning. The navigation process involves the problem of extracting information about the geometry of the terrain by suitably combining the sensor outputs. Techniques for (a) extracting integrated information, and (b) fusing the new information into the existing terrain model, are investigated. Two new paradigms for sensing: (a) active sensing where sensor operations are repeated using sequential statistical methods, and (b) learning fusion rules from examples where the system is trained with examples using the methods of empirical and structural risk minimization, are studied. These two methods, together with the existing Bayesian-based methods, provide the design paradigms for the multi-sensor system. For computational purposes, a general distributed sensor system specified by a set of first order propositional sentences is studied. This system encompasses, as special cases, the distributed versions of Bayesian inference, and geometric reasoning with logic and algebra. In the case of sensors whose outputs are real-valued vectors, the entire process of sensor integration is shown to be implementable in hardware using combinational circuits and comparators. This implementation is particularly suitable for real-time applications. Also implementations on several other computational systems are studied. IRI-9115939 Raviv, Daniel Florida Atlantic University $55,508 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $61,158) Optical-Flow-Based Active Visual Navigation This is the first year funding of a three-year continuing award. This research program is focused on basic theoretical aspects of ego- moving observer (e.g. a camera), a huge amount of visual information is captured. The aim of this project is to extract relevant visual information from a sequence of images and to use it as part of feedback control loops. In other words, to characterize changing visual data using only a few variables which are essential for the dynamic response of the observer relative to the scene, and use these to control the motion of the observer. The vision-based algorithms will be integrated in a real-time closed loop control system that has been developed at FAU. In addition, the Robot Systems Division at the National Institute of Standards and Technology (NIST) is collaborating in this research and has agreed to provide FAU with all the available hardware and software that are necessary for the successful completion of this project. IRI-9210763 Swain, Michael University of Chicago $31,298 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $35,298) RIAl: Active Vision: Low Resolution Cues, Eye Movements and Visual Routines This is the first year funding of a three-year continuing award. Recent research in active vision has shown that creating a visual system for a robot interacting with an unstructured, dynamic world differs vastly from interpreting a static image. On one hand, the constraints of real-time performance force optimizations such as a focus of attention and spatially-variant sensing. On the other hand, the availability of a large number of views, whose acquisition is controlled by the robot, has proven to simplify problems enormously. In addition, the processing necessary to permit intelligent behavior has been found to be easier to achieve than the difficult goal of reconstruction and full interpretation of a scene from a single image. This research explores two open problems suggested by ongoing research in active vision. One problem is the study of control of a spatially variant (foveal) sensor using the low-resolution formation available in the periphery. In particular, this research will examine the use of color information, which has been shown to be well-preserved under low resolution. The other problem is the integration of set a of visual modules, each capable of sustaining a particular task, with a suitable robot architecture so that the resulting system is capable of timely, flexible interaction with a dynamic world. IRI-9246898 Tarn, Tzyh-Jong Washington University $4,000 - 12 mos. REU: Design of Sensor-Referenced Action Planning and Control for Robotic Systems in Workstations This award provides supplemental funding for undergraduate participation in this research project IRI-9106317. The student will participate in development of a graphics package to simulate adual PUMA 560 manipulator system. This research project supports investigation of the problem of sensor-referenced action planning and control for robotic systems in a workstation setting, superimposed on model-based planning and control which has been the norm for robotic system control. The paradigm is to model the known and sense the unknown in the task space and implement the planning and control accordingly. Tasks are formulated in terms of events sensed in the task space, which in many cases are represented by a pattern of fused information from different sensors. Ultimately, this machine intelligence approach will provide high-level decision-making capability for automation, promote system integration, and lead to more user-friendly planning and decision/control systems for robotic workstations. To accomplish this, a number of problems must be addressed. Initially, the research will address a new type of motion description and efficient algorithmic formulation of kinematic and dynamic action capabilities of robot arms. The motion description uses a phase-space approach (velocity vs. position) instead of the usual state-space (position vs. time) approach. Preliminary results indicate that motion description and planning in phase-space is more compatible with sensor-referenced control than is time-based planning and control. The algorithmic formulation of action capabilities uses the known parameters both of the arms and the task environment in order to provide a model-based foundation for sensor-referenced intelligent action planning and control. ROBOTIC, REASONING, LEARNING, PLANNING AND CONTROL IRI-9208920 Adrion, Richard W.; Stankovic, John A; Riseman, Edward; Ramamritham, Krithi and Grupen, Roderic A. University of Massachusetts, Amherst $125,000 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program, the Division of Computer and Computation Research and DARPA - Total Award $210,000) Intelligent, Real Time Complex Computing Systems This is the first year funding of a three-year continuing award. The effort focuses on coordinated research in robotics, vision, real-time AI, and real-time software systems, in the context of a robotic assembly testbed. The emphasis is on flexible manufacturing automation involving active perception, planning, and cooperative activities among agents in real time. The project investigates integration of dextrous manipulation and vision, with cooperation among static and mobile robots and humans, and addresses current limitations of robotic systems in dealing with uncertainties and rapidly changing environments and tasks (such as occur in short-run production). Activities include development of multiple resolution representations which permit arbitration between local reflexive and global combinatoric strategies; studying tradeoffs between solution quality and computational speed in real-time systems; modeling cooperation and communication to achieve goal-oriented behavior; integrating architectures and algorithms for extracting relevant environmental information for control of distributed robotic manipulators; implementation of active perception to support model-based, goal-oriented sensing for manufacturing assembly operations; constructing high-level symbolic approaches to reasoning about geometry; and implementing learning mechanisms which model the environment based on experience over a general class of tasks to guide perception, planning, and multi- agent cooperation. IRI-9244615 Allen, Peter K. Columbia University $62,500 - 12 mos. PYI: Extending the Capabilities of Robotic Systems This is the fifth year base and matching funding of a Presidential Young Investigator award IRI-8657151. The focus of this research is to explore the integration of vision and tactile sensing in complex robotic tasks such as manipulation and assembly. The research is being carried on in two separate projects. The first is the creation of a multi-arm robotic test bed that utilizes high- speed, frame-rate vision sensing to track a moving part. The vision system is mounted on a robotic arm and the servo control of the arm is modified by the vision sensing. The vision system utilizes pyramids and scale space in order to segment and track the objects in real-time. Attached to the robot arm is an intelligent tactile sensor/gripper that can monitor contact forces and moments, resulting in the secure and stable grasping of the moving part by the visually-servoed gripper. Once the object is grasped, it can be used in coordinated assembly tasks using the two robotic arms. The second project involves the Utah/MIT dexterous hand. The first task is to build a set of higher-level primitives, once they have been defined, they can be combined into a grasping language to create programs of grasping/manipulation sequences. The next task is to develop and evaluate new tactile sensors to be used with the hand in order to support the requirements of intelligent grasping. With the addition of tactile sensory feedback, a hybrid control algorithm will be developed for the hand that utilizes position, force and tactile feedback for hand control. IRI-9023395 Asada, Haruhiko Massachusetts Institute of Technology $119,736 - 18 mos. Nonlinear Robot Compliance Control Using Neural Networks The goal of this research is to establish a new method for nonlinear compliance control using neural nets to explore possibilities in machine learning and control for robots and telemanipulators. Here, compliance is treated as nonlinear mapping from a measured force to a corrected motion and is represented by a multi-layer neural network, as well as by Gaussian networks. The objectives of the proposed research are three-fold. One is to develop a new method for representing "compliance" to deal with highly nonlinear, such as by stiffness and damping matrices. The second objective is to develop learning methods for the generation and teaching of compliance or force feedback strategies. The neural network approach allows us to teach a desired compliance using data acquired from a human operator. It does not need explicit feedback laws and detailed task models such as those required for conventional analytic methods. It is hoped that the new approach will also allow the transfer of human skill in compliant motion control to robots and telemanipulators. The third objective of the proposed project is to develop a real-time, neural net controller that is involved directly in the feedback loop a of robot control system. Efficient methods must be developed to analyze and design the nonlinear feedback system in order to accomplish smooth, stable responses. IRI-9248974 Atkeson, Christopher G. Massachusetts Institute of Technology $62,500 - 12 mos. PYI: Learning in Humans and Robots This is the fifth year base and matching funding of a five-year continuing award IRI-8858719. The research focuses on motor control and learning, using tools and techniques from biological, computer, and engineering sciences to study how humans control and improve their movement and how machines might attain human levels of performance. This research has three main thrusts: computational studies of motor learning in general; experimental studies of robot learning; and psychophysical studies of human motor learning. The significance of this work lies in the implications it has for both humans and robots: in the former, better understanding of how we learn new arm movements, better understanding of how we learn new arm movements, compensate for motor disorders, and adapt to prosthetic and assistive devices; in the latter, more useful and usable robots. IRI-9247192 Atkeson, Christopher G. Massachusetts Institute of Technology $8,000 - 12 mos. REU: Computational and Experimental Studies of Motor Learning in Humans and Robots This award provides supplemental funding for undergraduate participation in this research project. One student will continue her work on development of a computer vision system for object tracking based on illumination or color. The other student will develop an infrared tracking system for mobile robots, based on an on-board infrared emitter and an array of tracking stations. BCS-9009934 Bekey, George A. University of Southern California $25,000 - 12 mos. (Jointly funded with the Bioengineering and Aiding the Disabled Program - Total Award $75,000) Human and Robot Models for Intelligent Prosthetic Hands This is the third year funding of a three-year continuing award. This proposal is to develop a novel approach to the design of a hand prosthesis. The PIs first develop a formal model of grasping behavior based upon experiments on humans. They then formulate a minimal set of control inputs that is necessary to control grasping movements. Finally, both neural network and knowledge-based approaches are investigated as possible methods of implementing intelligent control mechanisms for an already-developed hand prosthesis. The proposed work should lead to better ways of controlling artificial hands; it also may have an impact on the design of robotic manipulators. IRI-9249226 Canny, John F. University of California, Berkeley $62,500 - 12 mos. PYI: Motion Planning This is the fourth year base and third-year matching funding of a five-year PYI continuing award IRI-8958577. This Presidential Young Investigator award recognizes the outstanding productivity and potential of Dr. Canny as a researcher and educator in computer vision and robotics. While a graduate student, he published influential work in edge detection and in collision avoidance for robot motion planning, solving long-standing problems by new techniques. Dr. Canny is continuing his work in robot motion planning and computational algebra and geometry, including research in compliant motion planning and grasp planning. IRI-9247335 Carpenter, Gail A. Boston University $40,102 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $80,203) Analysis of Neural Networks for Adaptive Pattern Recognition This award begins a two year creativity extension to award IRI- 9000530. The PI has made creative accomplishments in the field of artificial neural systems, relating to the development of the Adaptive Resonance Theory (ART) and a resulting series of neural networks for computationally-efficient and/or fast real-time learning and pattern recognition. These results have gained wide recognition and have quickly found their way into interesting applications, including a very large CAD-based group technology program at the Boeing Company that is expected to streamline the design process by reducing parts inventory. Other applications include pattern recognition in seismic, medical, military (target recognition), and robot navigation applications at Lincoln Labs; and a number of machine learning/expert system applications, including application to development of the Medicare Uniform Clinical Data Set at the University of Nevada School of Medicine. A new approach to supervised learning, involving combinations of self-organizing modules, has recently been developed, and shows promise of performance exceeding that of backpropogation and genetic algorithms. Other work is exploring coding and recognition of temporal sequences of events, with applications to speech perception, motion planning, and object recognition, and a new exploration of noise-resistant invariant filtering for rapid image recognition. In the extended period, research will focus on enhancing the ability of ART networks to learn new patterns rapidly while preserving the recognition capabilities for previously- learned patterns. This will include methods and structures to combine information from multiple sources, multi-scale (coarse- fine) recognition, hierarchical architectures that reduce overall connectivity requirements, and robustness to noisy data. The project will also explore local-logic implementation of fuzzy ART algorithms, and study implementations of medium-term memory, as well as continuing pursuit of temporal processing and rapid invariant filtering. IRI-9247046 Dawson, Darren M. Clemson University $4,000 - 12 mos. REU: Position/Force Control of Robot Manipulators in the Presence of Uncertainty This award provides supplemental funding for undergraduate participation in this research project IRI-9111258. The student will be responsible for interfacing of a PC to a DC motor, the writing of control software, and experimentation with robust nonlinear control of a robot. A new position/force control decomposition is suggested for the formulation of control strategies for constrained robot manipulation. Separate position and force controllers can be developed from the corresponding decoupled model for a single-joint nonredundant manipulator or two cooperating robot arms. The position controllers can compensate for the associated uncertainty naturally present in the robot arm dynamics. This new control decomposition is utilized to investigate: 1) the advantages and disadvantages of different position/force decompositions, 2) the effects of surface uncertainty in conjunction with any manipulator dynamic uncertainty, 3) the advantages and disadvantages of different control schemes through computer simulation and experimentation, 4) the use of multiple sensors in conjunction with advanced position/force controllers, 5) the effects of actuator and sensor dynamics on position/force controller performance. If the above problems can be solved, the integration of robot manipulators into the manufacturing of automobiles, planes, surface-ships, spacecraft, appliances, etc, will become much easier. IRI-9247453 Dean, Thomas L. Brown University $25,000 - 12 mos. PYI: Integrated Planning and Control This is the fourth year base funding of a five-year PYI continuing award IRI-8957601. Dr. Dean continues his work in temporal reasoning under uncertainty for which he won a best-paper prize at the 1988 AAAI Conference as well as time-dependent planning, mobile robotics, computer vision, and spatial reasoning. Dr. Dean's background combines artificial intelligence, control theory, and operations research. IRI-9247042 Dean, Thomas L. Brown University $4,000 - 12 mos. REU: Integrated Planning and Control This award provides supplemental funding for undergraduate participation in this research project. The student will extend some existing algorithms for learning finite state automata to handle more realistic performance criteria, and will test the algorithms on a mobile robot trying to learn the structure of its environment for navigation purposes. IRI-9249399 Donald, Bruce R. Cornell University $62,500 - 12 mos. PYI: Robotics, Computational Geometry and Artificial Intelligence This is the fourth year base and matching funding of a five-year Presidential Young Investigator award IRI-8957316. The research is developing new algorithmic techniques for robot motion planning, including a new theory of planning in highly uncertain and unstructured environments. The goal is to develop a task-level geometrical planning theory for representing, reasoning about, and manipulating physical systems. A specific objective is the generation of assembly strategies from task-level descriptions. The investigation will concentrate on the following problems: (1) Algorithms and techniques for modeling geometric constraints, and the planning of robot motions using these constraints. (2) Synthesis of compliant-motion programs using geometric constraints imposed by the task. (3) Explicit modeling of uncertainty and error in sensing, control, and the shape of the manipulated part. (4) Development of a precise theory of error diagnosis and recovery, and a method for generating sensor-based plans with built-in detection and recovery. IRI-9244995 Donald, Bruce R. Cornell University $4,000 - 12 mos. REU/PYI: Robotics, Computational Geometry and Artificial Intelligence This award provides supplemental funding for undergraduate participation in this research project. The student will work on constructing robots and programming them in a new real-time, concurrent, embedded scheme developed in this laboratory. The student will also assist in development of an infrared modem for communication among cooperating robots, will work on algorithms for synchronizing their behavior, and will participate in further development of the Cornell Generic Controller. IRI-9246468 Donald, Bruce R. Cornell University $4,000 - 12 mos. REU: Algorithmic Techniques for Task-level Robot Planning This award provides supplemental funding for undergraduate participation in this research project IRI-9000532. The student will conduct research on planning and execution of robot strategies in the presence of noise, error, and uncertainty in actuation, sensing, and modeling. The student will also conduct experiments on gripping, grasping, and manipulation strategies for mobile robots, will assist in the design and construction of a new mobile robot that exploits single-line CCD and infrared cameras, and will develop manipulation strategies based on computational geometry. This research parallels Dr. Donald's Presidential Young Investigator research in kinodynamic approximation and related areas of robotic theory. His work in compliant motion planning will emphasize handling of uncertainty and errors in sensing, real time control, and the modeling of parts. Algorithms will be tested on an experimental, force-controlled assembly system using the Cornell robot arm. The ultimate goal is to permit high level specification of force-controlled assembly tasks, without requiring humans to specify the required forces and sensing strategies. IRI-9249142 Erdmann, Michael A. Carnegie Mellon University $62,500 - 12 mos. PYI: Information Requirements of Robot Manipulation Tasks This is the second year base and first year matching funding of a five-year Presidential Young Investigator Award for research into the information requirements of manipulation tasks. The aim is to develop a high-level representation of tasks that facilitates the decomposition into subtasks and the translation of high-level specifications into low-level robot commands. Such translation is complicated by uncertainties which may propagate through the system. A manipulation language will be developed to support task decomposition and abstraction, and several types of manipulation tasks will be studied, including parts orienting, assembly by implosion, and large-scale mounting operations. The project will study information flow among the decomposed subtasks. This research should lead to a better understanding of the fundamentals of manipulation, a better characterization of the relations among information, speed of assembly, and task solvability, and a better understanding of how to design robots for operation in a world with uncertainties. IRI-9247043 Erdmann, Michael A. Carnegie Mellon University $4,000 - 12 mos. REU: Information-Based Task-Level Planning This award provides supplemental funding for undergraduate participation in this research project. The student will assist in developing special purpose robotic grippers to house radial sensors being designed in this laboratory, and will develop an X-interface for interactively controlling movements of a Zebra Zero robot. IRI-9248973 Fearing, Ronald S. University of California, Berkeley $55,787 - 12 mos. PYI: Dextrous Robotic Manipulation with Multi-finger Hands This is the second year base and first-year matching funding of a five-year continuing Presidential Young Investigator Award, IRI- 9157051. The research addresses dextrous manipulation with tactile sensor-equipped multifinger robot hands, micromanipulation with millimeter-sized cooperating mobile modules, and understanding of nonlinear dynamics to control an acrobatic robot. The goal of these robotic systems is to manipulate and explore environments that are currently inaccessible due to size, distance, or cost. IRI-9248046 Fearing, Ronald S. University of California, Berkeley $8,000 - 12 mos. REU: Dextrous Robotic Manipulation with Multi-finger Hands This award provides supplemental funding for undergraduate participation in this research project IRI-9157051. One student will build a small, dexterous, tendon-driven three-fingered robotic manipulator prototype for surgical applications. The other student will employ media similar to blood vessel walls to detect holes and inclusions in mock-up blood vessels. CCR-9209793 Gelsey, Andrew Rutgers University, Busch Campus $32,326 - 24 mos. (Jointly funded with the Numeric, Symbolic and Geometric Computation - Total Award $69,652) RIA: Robust Simulation of Physical Systems Numerical simulation is an important tool for predicting the behavior of physical systems. Many powerful numerical simulation programs exist today. However, using these programs to reliably analyze a physical situation requires considerable human effort and expertise to set up a simulation, determine whether the output makes sense, and repeatedly run the simulation with different inputs until a satisfactory result is achieved. Automating this process is not only of considerable practical importance but will also significantly advance basic AI research in reasoning about the physical world. This project will address these problems of setup, analysis, quality assurance, and feedback for numerical simulation. IRI-9123747 Goldberg, Kenneth University of Southern California $95,000 - 12 mos. (Jointly funded with the Cross Disciplinary Activities Program - Total Award $110,000) Reduced-Complexity Manipulation with the Parallel-Jaw Gripper This is the first year funding of a three-year continuing award. This research explores emerging concepts of minimalist robotics, by analyzing compliant motion planning for manipulators with reduced sensing and actuation complexity such as the parallel-jaw gripper, as alternatives to more complex multifingered, multijointed robot hands which can be more difficult to program. The work investigates means to program grasp, recognition, orientation, and positioning of an industrially-important class of parts whose cross-sections can be represented by a combination of curved arcs and straight polygonal edges. The research combines computational analysis of part geometry with methods from stochastic estimation theory. A programmable parts kitting system will be developed as an experimental testbed. This system will automatically generate programs from CAD models of parts, for robot recognition and manipulation of those parts. IRI-9211366 Gottschlich, Susan N. Rensselaer Polytechnic Institute $39,139 - 12 mos. RIA: Automated Assembly with a Robotics Manipulator/Hand System This is the first year funding of a two-year continuing award. This research investigates the use of robotic hands in automated robotic assembly. In many assemblies, the clearances allowed between mating parts are smaller than the uncertainties associated with a robotic assembly cell. Though it is possible to reduce these uncertainties by using special purpose mechanisms and high precision fixturing devices, this limits flexibility and increases cost. It is preferable for the manipulator to use force/torque feedback during the execution of assembly mating operations to reduce alignment errors between parts which are caused by the uncertainties. By incorporating a robotic hand into the robot manipulator system, the performance of the assembly cell will be improved. The accuracy of force/torque measurements will be increased by putting sensors on both the manipulator and the hand. The hand will serve both as a general fixturing device so that the grasped part can be held rigidly during assembly, and as a micromanipulator so that fine adjustments to the position of the grasped part can be made. The central goal of this project is a general and automated method for synthesizing force/torque guided motion strategies which operate on a manipulator/hand system and are capable of reducing uncertainty-induced errors during assembly execution. IRI-9116297 Grupen, Roderic A. and Weiss, Richard S. University of Massachusetts, Amherst $95,999 - 12 mos. Sensor-Based Incremental Planning for Multifingered Manipulators This is the first year funding of a three-year continuing award. The research focuses on planning of multifingered robotic grasping of objects in situations where there is incomplete information. The grasping behavior is continuously and incremetally refined based on sensed information during task execution and reasoning about object and manipulator geometry. The grasping operation combines reflexive behavior based on simple stimulus-action rules and local information only, with reactive behavior which provides robust capabilities for dealing with uncertainties It minimizes the amount of preplanning resourcefully without a priori models of the environment or the task. Such systems will be needed for autonomous robotic operation in unstructured, unfamiliar, or unpredictably changing situations such as space, underwater, or environmental operations, and may also be easier and more economical to deploy in more structured environments such as some manufacturing processes. IRI-9242207 Jordan, Michael and Jacobs, Robert Massachusetts Institute of Technology $34,729 - 12 mos. A Modular Connectionist Architecture for Control This is the third year funding of a three-year continuing award IRI-9013991. This grant supports research in machine learning for control of nonlinear dynamical systems. Drs. Jordan and Jacobs are studying multinetwork connectionist architectures for discovering piecewise control tasks, and learning multiple tasks simultaneously. They expect multinetworks to show faster learning rates, better generalization, more interpretable representation, and more efficient use of hardware than single neural networks. IRI-9123266 Koditschek, Daniel E. Yale University $98,058 - 12 mos. Dynamical Dexterity in Robotic Manipulation This is the first year funding of a three-year continuing award. The research focuses on analytical and experimental study of dynamical dexterity in robotic manipulation, addressing fundamental issues in planning, coordination, and effective use of limited sensory capabilities. The research builds on the PI's previous research on robot juggling of objects in two dimensions; this will now be extended to three dimensions. The paradigm of juggling demands a holistic, task-oriented approach to planning and control, involving tracking of one or several objects in space, trajectory matching, making compliant intermittent contact or impact, and coupling of sensing and manipulation in real time. The results should provide a rich addition to the knowledge base for design and execution of a wide variety of robotic tasks including dynamical manipulation of objects on the fly for rapid assembly-line operations, and some forms of dynamic legged locomotion. IRI-9112717 Koren, Yoram and Borenstein, Johann University of Michigan $219,844 - 24 mos. Sensor-Based Control of a Mechanical Snake This research explores sensor-based control of a mechanical snake through development of obstacle-avoidance and motion-planning algorithms. The mechanical snake is a robot-like mechanism, a prototype of which is now being tested in the laboratory. It consists of a sensorized head link (that detects obstacles) and six identical regular links (that provide motion). As an obstacle is detected, the computer in the head link will be able to recommend a motion of each link so that when that link arrives in the neighborhood of the obstacle it will not collide with it. Developing the algorithms to make this system work is a far more challenging problem than the state-of-the-art, real-time collision avoidance strategies currently in use for mobile robots and robot arms. When completed, the snake will be able to move into areas that people cannot enter (e.g., small cracks in collapsed highway after an earthquake) or are too dangerous for people (e.g., a radioactive environment). At these locations, the will perform surveillance (e.g., look for survivors through remote video, check radioactivity levels) or other task (e.g., bring medicine to survivors). IRI-9112531 Kraft, L. Gordon and Glanz, Filson H. University of New Hampshire $79,797 - 12 mos. (Jointly funded with the Neuroengineering Program - Total Award $159,594) Stability Analysis of Neural Control This is the first year funding of a two-year continuing award. Stability analysis for neural control systems is a difficult and important open research question. The problem is difficult because neural control systems are inherently non-linear and may involve large-scale plants with unknown parameters and structure. The problem is important because practical application of neural control demands that the closed-loop system operate in a stable manner. In this project several ideas from control system fields are integrated to generate stability arguments for neural control systems. The concepts of Lyapunov functions from adaptive control, performance measure minimization and dynamic programming from optimal control and the adaptive critic from reinforcement learning are combined to formulate a methodology for stability and convergence rate analysis for a neural controller. IRI-9248905 Kreutz-Delgado, Kenneth University of California, San Diego $62,500 - 12 mos. PYI: Sensor Based Robotic Assembly This is the third year base and second-year matching funding of a five-year Presidential Young Investigator Award IRI-9057631. Funding supports Dr. Kreutz-Delgado's development of a real-time testbed for sensorbased control of a dual-arm robotic manipulation system. The current research, begun at the Jet Propulsion Laboratory, involves application of a spatial operator algebra to real-time task-level planning and use of neural-network approaches for manipulator and camera coordination. IRI-9246687 Kuipers, Benjamin J. University of Texas, Austin $97,781 - 12. (Jointly funded with the Cross Disciplinary Activities Office - Total Award $101,681) Qualitative Methods for Robot Exploration This research was originally funded for a three-year period. On the basis of creative accomplishments to date, the award has been extended to five years. This award funds the fourth year. The research has focused on autonomous robotic navigation methods that avoid the need for detailed metric information. Achievements include derivation of spatial/geometric representations from topological rather than metric information, thereby controlling the accumulation of position error and rendering robotic navigation more robust to sensor errors and incomplete metric data; development of a method for incremental improvement of navigational performance by construction locally-optimal control laws for path planning with incomplete knowledge; and a method to generate a global control law from a weighted average of local laws, combining fuzzy and classical control. The work now turns to integration of these individual developments to produce a robotic system that can (1) learn about its own sensorimotor capabilities and about the metrics and geometry of its environment, (2) autonomously and incrementally develop robust intelligent control laws, and (3) generally improve its performance through experience and practice. DDM-9241441 Luo, Ren C. North Carolina State University $10,000 - 12 mos. (Jointly funded with the Operation Research and Production Systems Program - Total Award $60,00) Development of a Complex Manufacturing Floor-Material Handling System Using Guide-Path Independent Autonomous Vehicles This is the first year funding of a two-year continuing award. The objective of this research is to study high-level global path planning and dispatching of multiple vehicles using statistical models and graph representation techniques and a low-level multi- sensor based local path planning and navigation system. These efforts will support the development of a complex manufacturing floor material handling system using guide-path-independent autonomous vehicles. Particular efforts to be emphasized in this project include development of a geometric representation of the factory floor and a graph representation of the paths in a factory; development of vehicle dispatching techniques and a collision avoidance strategy; and the development of an on-board Hierarchical Multiprocessor Computer System to support real-time operation of an intelligent mobile robot. The design and implementation of an experimental system to test the proposed approaches will also be pursued. ECS-9245737 Lang, Jeffrey H. Massachusetts Institute of Technology $20,000 - 12 mos. (Jointly funded with the Solid-State and Microstructures Engineering Program - Total Award $286,796) Basic Research on Microelectromechanical Systems This is the first year funding of a three-year continuing award. This research includes a collaborative effort between the Massachusetts Institute of Technology and Case Western Reserve University to continue research on flat silicon micromotors. The research will extend the studies of pattern definition and final release. Fabricated motors will be used to study friction and wear, and to study new materials to reduce both friction and wear. A One goal of the proposed research will be the development of an improved set of micromotors. The research will integrate sensors together with micromotors to produce a fully - actuated, and possibly closed-loop controlled, micromechanism. IRI-9114208 Mason, Matthew T. Carnegie Mellon University $269,888 - 36 mos. Task-Level Force Constraints The purpose of this research is to explore force constraints as a model of manipulator action. Robotic manipulation is usually modeled in terms of programmed motion-goals are specified as the motions of task objects, and robots are programmed by a sequence of arm motions. The programmed motion model works fairly well, provided that objects move only when rigidly grasped by the robot. However, many robot operations violate this assumption, manipulating objects without a rigid grasp. For example, when placing an object, the object may slip in the fingers. Or, when grasping an object, the object may be re-oriented and centered by pushing and squeezing. Other examples include striking, tilting, and the use of passive compliances. For these operations, and many others, the programmed motion model is awkward, because there is no direct transformation from desired object motions to robot motions. The research explores the use of task level force constraints, as an intermediate level model of action. A task-level force constraint is any constraint on the force applied to a task object. For the operations where programmed motion is awkward, force constraints often provide a natural and effective way to transform desired object motions to corresponding robot commands. The approach is to build a planner, applying methods of classical mechanics to derive force constraints, and then using a variety of geometrical and analytical techniques to derive robots command parameters. Experimental evaluation will use a industrial manipulator and vision system to perform positioning and assembly tasks in a planar task domain. IRI-9257269 Newman, Wyatt Case Western Reserve University $25,000 - 12 mos. NYI: Control in Autonomous Systems This is the first year base funding of a five-year Young Investigator Continuing Award. This research emphasizes interaction control and reflex control in autonomous systems. Physical interaction with the environment (e.g., opening doors; moving obstructions; complying with mechanical constraints) is a key ingredient of reactive autonomous systems, and impedance control will be furthered to provide such behavior. In particular, stable and "gentle" manipulation behaviors will be investigated and constructed using principles of passive physical equivalents. The target resulting behavior should achieve stable interaction among any number of independently-controlled machines. Such physical behavior will provide a suitable platform for investigation of reflex-like, higher-level controllers. Reflex controllers, like conventional control systems, generate commands based on sensory inputs. Unlike traditional control, though, reflex-like responses invoke logic-based and switched-mode actions that cannot be analyzed in the traditional terms of linear superposition of inputs or outputs. Use of such nonlinear control constructions has proven highly valuable in understanding control in biological creatures, though such control is inadequately understood to be applied to machines. In this research, biological stimulus/response behaviors will be mimicked in constructing computer controls for the purpose of experimentally investigating stability properties and design techniques for behavior-based systems. IRI-9210332 Passino, Kevin M. Ohio State University Research Foundation $30,000 - 24 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $60,000) RIA: Modeling, Analysis and Design of Expert Control Systems This is the first and second years of a three-year continuing award. Recently, there has been a significant increase in the use of techniques from applied Artificial Intelligence (AI) such as expert systems to implement the control functions for complex industrial processes. The research funded is showing that the design of such expert controllers (controllers implemented via expert systems) can be accomplished using the same design philosophy as that used for fuzzy controllers. Moreover, to ensure that such expert controllers can be trusted in critical environments (e.g., aircraft control, spacecraft control, process control) this research is developing a discrete event system (DES) theoretic framework for the modeling and analysis of expert control systems. In particular, a mathematical model is introduced that can represent "rule-based" expert systems and a wide class of processes. Techniques from DES theory are being developed for the analysis of reachability (to study inference chains), cyclic behavior (to verify that the expert system will not get stuck in circular reasoning), and stability properties (to verify critical properties related to the safe operation of expert control systems). The application of the approach to the solution of a load balancing problem in flexible manufacturing systems and process control problems is being investigated. Moreover, implementation issues for expert controllers are being investigated via the implementation of an expert controller for a flexible link manipulator. IRI-9208429 Pu, Pearl University of Connecticut $30,000 - 24 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program - Total Award $59,433) RIA: An Efficient Case-based Assembly Sequence Generation System Automatic assembly sequence generation (ASG) is important to efficient manufacturing and concurrent engineering. There are two main difficulties with developing tools for ASG: (i) the combinatorics makes solutions by blind search intractable and (ii) criteria for optimal assembly sequences are difficult to formalize. Earlier work of the PI investigated the use of case-based reasoning techniques to simultaneously address both problems with encouraging results, especially concerning the extensibility to covering large sets of problems. The practical objective of this project is to verify these results on a larger prototype, and thus demonstrate the usefulness of case-based assembly sequence generation for practical applications. The theoretical objective is to investigate new methods for adaptive learning and case-matching which will form part of the new prototype. The result of this work will be a fully implemented case-based reasoning system for a large class of assembly devices. The conceptual results involved in building such a system should prove useful to research in intelligent design and manufacturing systems and in the case-based reasoning field. IRI-9100681 Reif, John H. Duke University $72,874 - 12 mos. (Jointly funded with the Computer and Computation Theory Program - Total Award $117,874) Toward Autonomous Robots: Robust, Adaptive & Dynamic Motion Planning This is the first year funding of a three-year continuing award. The research is to study computational complexity of adaptive and robust robot motion planning algorithms which can be effective in situations with incomplete information or in dynamically changing environments. The planner's performance is to adapt as new information is acquired or the environment changes. The research will also consider movement problems with inherent errors in the sensing and movement control, so that the motion planning algorithms must be robust to lack of information on the position of the robot. The emphasis is on the design of algorithms that give provable optimal or near-optimal performance, focusing especially on some less-investigated motion planning problems. The use of massive parallelism will be explored, to speed up the algorithms. The results should contribute to a more fundamental understanding of the nature of motion planning problems, and ultimately to more robust planning approaches for less-structured environments. IRI-9113370 Roy, Asim Arizona State University $54,078 - 12 mos. A Polynomial Time-Algorithm for the Construction and Training of a Class of Multi Layer Perceptrons This award funds research on a class of multilayer perceptron-like neural networks for classification problems. The algorithm uses linear programming methods to incrementally generate hidden layers in a restricted higher-order perception, which is trainable in polynomial time. The research will investigate extensions of the algorithm to address classification problems with large training sets, which normally take a long time to coverage. Behavior of the algorithm will be evaluated on representative problems from computer vision and robotics. IRI-9248976 Russell, Stuart J. University of California, Berkeley $25,000 - 12 mos. PYI: Architectures and Algorithms for Autonomous Intelligent Systems This is the third year base funding of a five-year continuing Presidential Young Investigator award IRI-9058427. This award supports research in optimal decision making under resource constraints (the theory of bounded rationality) and studies of bias control for effective concept induction in probabilistic domains. A special thrust is the incorporation of existing background knowledge in neural-network learning systems. Dr. Russell will apply this work to robotic manipulator learning, path planning, and control architectures for simulated and physical robotic systems. His research in bounded rationality and utility theory may have additional application in other areas of decision making. IRI-9248673 Sastry, Shankar S. and Canny, John F. University of California, Berkeley $82,801 - 12 mos. Nonholonomic Motion Planning for Robots This research was originally funded for a two year period under award IRI-9014490. On the basis of creative accomplishments to date in this seminal work, the award has been extended to four years. This award funds the third year. This grant supports research in differential geometry and nonlinear control theory to advance our understanding of motion planning for nonholonomic systems. Nonholonomic problems arise in mobile robot navigation, rolling contacts and robotic grasping, free-floating robots in space or underwater, and control of legged robots. The work has focused on theoretical investigations of motion planning in nonholonomic situations (where momentum must be conserved). Explorations have included mobile robotics, space and undersea robotics, motion planning for redundant link manipulators, and grasp planning. Planned further explorations include deepening the understanding of motion planning for space robot operations, flight control, submarine steering, and adaptation of potential field based planning for collision avoidance in nonholonomic situations. IRI-9246469 Sethi, Ishwar K. and Jain, Anil K. Wayne State University $4,000 - 12 mos. REU: Artificial Neural Networks and Decision Trees This award provides supplemental funding for undergraduate participation in this research project IRI-9002087. The student will participate in research in robot localization using entropy networks and ultrasonic sensing. The research is aimed at developing a systematic artificial neural network (ANN) design and training methodology and exploring its applications in robotics. The salient feature of the research is the use of statistical pattern recognition techniques to solve some key aspects of ANN and learning procedures. One major limitation of present ANNs operating under the supervised mode of learning is that these networks are unable to self-configure the architecture for a given classification problem. A solution to this problem is to establish a direct relationship between a class of nonparametric classifiers, i.e. decision trees, and the multilayer neural networks. It can be shown that a decision tree can be restructured as a three-layer neural net. Exploiting this restructuring allows neural network design and training methodology to have self-configuration capability. The tree-to-network mapping also provides a solution to the credit assignment problem thus making it possible to train each layer separately and progressively. The research will explore automatic tree generation, incorporation of incremental learning, and adaptability in such networks. The proposed methodology will be applied to robot learning where networks are required to perform regression rather than classification. The main expected benefit of the proposed research is that it will make available an ANN design and training procedure that is systematic. This should result in many more applications for artificial neural networks. MSS-9114674 Whittaker, William L. Carnegie Mellon University $37,500 - 12 mos. (Jointly funded with the Division of Mechanical and Structural Systems - Total Award 150,000) Autonomous Retrieval of Buried Objects This is the first year funding of a two-year continuing award. The objective of this research is to develop and integrate the perception, planning, and manipulation technologies necessary for autonomously detecting, uncovering, and retrieving fragile buried objects. Such technology has application in the retrieval of waste containers from landfills, recovery of munitions from shallow burial, exposure of cables beneath the ocean floor, and acquisition of extraterrestrial soil/rock samples. The proposed work will advance current knowledge in the fields of software architectures and reasoning systems for autonomous robots working with representations of buried objects, and manipulation for extrication of objects from geo or diffuse material. IRI-9113491 Widrow, Bernard Stanford University $58,321 - 12 mos. (Jointly funded with the Instrumentation, Sensing and Measurement Program - Total Award $116,641) Neural Networks for Adaptive Nonlinear Control This is the first year funding of a three-year continuing award. Many basic scientific issues remain to be addressed in order to facilitate the wide use of neural networks in control systems. This research focuses on trainable state estimation for neural networks in order to deal with effects of plant and sensor noise and incomplete availability of state measurements. It will also explore neural network implementation of a self-tuning regulator for adapting a controller to track changes in a nonlinear plant; techniques for controller weight initialization that can decrease network training time and also reduce the probability of convergence of the weights to undesirable local minima; and adaptation of networks for navigational obstacle avoidance to robotic manipulator obstacle avoidance. IRI-9241711 Williams, Ronald J. Northeastern University $62,420 - 12 mos. Connectionist Learning Algorithms for Temporal Processing and Multi-Scale Search This is the third year funding of a three-year continuing award IRI-8921275. This grant supports theoretical and experimental work in two important areas of artificial neural computation: recursive networks for processing time-varying signals, and multiscale networks for exploiting information at multiple resolutions. Recursive networks will be important for control and signal- processing problems with feedback delays of unknown duration. Multiscale architectures and learning strategies are needed for large-scale problems of practical importance, including those with complex spatial and temporal components. Proposed target applications are in speech recognition and adaptive sensorimotor control. IRI-9210423 Xiao, Jing University of North Charlotte $37,004 - 12 mos. (Jointly funded with the Division of Cross Disciplinary Activities Office - Total Award $47,004) RIA: Spatial Knowledge Reasoning and Robotic Assembly in the Presence of Uncertainties This is the first year funding of a three-year continuing award. This project addresses automatic systems capable of reasoning about a robotic environment based on imperfect sensory data and recovering errors of robot motions caused by intrinsic and inevitable system uncertainties. Specifically, the project studies how to acquire sufficiently accurate information about the spatial state of an object in the presence of sensing and modeling uncertainties and develops practical strategies for automatic assembly motions based on the spatial states of the parts involved. The project is expected to result in the implementation of a prototype system capable of automatic recognition of contacts and generation of contact-based plans, as well as proper execution of the plans, all in the presence of uncertainties. IRI-9024499 Zheng, Yuan F. and Hemani, Hooshang Ohio State University $93,358 - 12 mos. Neural Gait Synthesis for Autonomous Biped Robots This award funds research on integration of methods for reflexive gait synthesis and adaptation to difficult terrains, for a biped walking robot. The research will explore neural-network-based mechanisms for training a central pattern generator to generate a new gait that is learned by a terrain-monitoring learning and adaptation unit. It will also study the convergence properties of backpropagation and reinforcement learning mechanisms for use in the learning and adaptation unit. The long-term goal of this research in biped robotics is to achieve more agility and mobility than is possible in multilegged walking machines or other forms of machine locomotion, and ultimately to provide capability for mobile robots to augment or replace humans in hazardous operations or in rough terrain. IRI-9247143 Zheng, Yuan F. Ohio State University $8,000 - 12 mos. REU: Neural Gait Synthesis for Autonomous Biped Robots This award provides supplemental funding for undergraduate participation in this research project. One student will use her electrical engineering training to design, analyze, and build a breadboard model of coupled oscillators to control the walking motions of a biped robot. The other student will develop a computer actuator musculo-skeletal system for adaptive control of biped walking machine. MULTI-ROBOT COORDINATION AND COOPERATION IRI-9100149 Arkin, Ronald C. Georgia Technical Research Corporation $69,901 - 24 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program, the Information Technology and Organizations Program, and the Cross Disciplinary Activities Office - Total Award $119,901) Cooperation and Communication in Multi-Agent Reactive Robotic Systems This research is to study communications and control requirements for multiple autonomous robots to cooperate in accomplishing a task, where the robots are organized heterarchically instead of using the more common hierarchical (master/slave) control. Each robot behaves reactively, in accordance with sets of relatively primitive motor schemes. The resulting organization does not require an explicit world model, and should be more robust and hence more suitable for operations in remote and/or hazardous environments. A key research issue is to evaluate the efficacy of such systems operating under constrained inter-agent communications. Such agents could be more economical to operate. The goal of the research is to understand and develop control and communication mechanisms that produce robust yet efficient cooperative behavior. IRI-9202423 Bay, John S. Virginia Polytechnic Institute $50,000 - 12 mos. SGER: Behavioral Control of Teams of Mini-Robots This award is in the Small Grants for Exploratory Research mode. The objective of this project is the design of behavioral control techniques for a simulated network of "army ant" robotic devices. The devices envisioned are small, inexpensive, and identical mobile platforms which cannot transport material individually, but can cooperate to lift and carry arbitrarily large objects. The behavioral control approach seeks to minimize the complexity of the units and the required communications by designing their collective function into their group behavior rather than through individual performance. Each "robot" will be programmed to respond to sensory stimulus with simple reactions. The collective effect of many such reactions will be the desired maneuvering task, in much the same way that teams of humans can transport heavy objects with minimal communications. Sensory inputs will include applied forces and the attitude of the load being carried. The set of behavioral reactions will be implemented in a computer simulation. The simulation will also model the dynamics of the load, the force and attitude sensors, "beacon" signals emitted by the units, and the possibility of uneven or rough terrain. The speed, stability and accuracy of the material transport system will be evaluated in order to further develop the controls and to suggest behavioral approaches to broader problems, such as team-forming, reorganization, navigation, and message-passing. The result of the work will be a paradigm for material transport using modular networks of small machines. BCS-9216691 Kumar, Vijay and Yun, Xiaoping University of Pennsylvania $0 - 12 mos. (Jointly funded with the Bioengineering and Aiding the Disabled Program - Total Award $80,000) BAC: Cooperation and Coordination of Two Arms in Biological and Robotic Systems This is the first year funding of a three-year continuing award. This research project will investigate the coordination of two physically-coupled human arms or two artificial arms in cooperative manipulation tasks. The project will focus on the interaction when there is physical contact between the two arms. The mechanisms underlying the generation of trajectories and the interaction forces and moments of two human arms will be studied. The investigators also will develop control algorithms that would specify the motion and forces of two artificial manipulators during the performance of similar coordinated tasks. This research is important to the understanding of coordinated reaching and manipulation by humans. In addition, the research will improve the control algorithms for coordinated multi-robotic reaching and grasping. IRI-9116399 Nilsson, Nils J. Stanford University $32,905 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Information Technology and Organizations Programs - Total Award $98,715) Research On Autonomous Agents This is the first year funding of a three-year continuing award. This research is concerned with developing autonomous, cooperating, adaptive computational agents. Expected applications of this research are in the control of mobile robots capable of performing delivery, maintenance, and/or construction tasks and of agents that gather and manipulate symbolic information over computer networks. The research will concentrate on bounded-time action computation and learning mechanisms and on combining these components in integrated agents able to function cooperatively in dynamic, uncertain environments. DDM-9121430 Tanchoco, Jose M.A. and Cipra, Raymond J. Purdue University $15,000 - 12 mos. (Jointly funded with the Design and Computer- Integrated Engineering and the Operations Research and Production Systems Programs - Total Award $47,971) SGER: Distributed Flow Control of Intelligent Pallets The objective of this research is to investigate the concept of an intelligent pallet. The idea is to provide intelligence to the media on which parts are physically carried through the various manufacturing resources. The pallets will be assumed to be enslaved in the system, each equipped with its own central processing unit (CPU) and memory devices. In this scenario, the manufacturing resources are merely agents which will service the commands from the Intelligent pallet to process, move, and/or assemble the parts(s) it carries. This project includes the analysis of technological barriers and economic feasibility. Evaluation of the structure of the control system and its implementation including issues such as the feasibility of a heterarchical control architecture, problems associated with wireless communication, and information flow will also be addresses at both theoretical and experimental levels. Applicability and generalizability of the results of this effort is enhanced through industrial collaboration and matching support from the Purdue Engineering Research Center. IRI-9209880 Yun, Xiaoping University of Pennsylvania $47,500 - 36 mos. (Jointly funded with the Information Technology and Organizations Program and the Cross Disciplinary Activities Office - Total Award $100,000) RIA: Coordination of Mobile Manipulators This research investigates coordination of multiple mobile manipulators for grasping and transporting large and irregularly shaped objects. A mobile manipulator in this study is composed of a mobile platform for locomotion and a manipulator for grasping and manipulation. A large object without special features such as handles cannot easily be grasped by the conventional end effectors such as jaw grippers or multifingered hands. A new approach based on the concept of enveloping grasp and manipulation by multiple mobile manipulators is proposed. The major contribution of the proposed study will be control algorithms for coordination of locomotion and manipulation of a mobile manipulator with the motion of the manipulator being constrained, and coordination of two mobile manipulators for transporting large objects. An experimental system will be developed to test and evaluate the proposed coordinationalgorithms. SPECIAL PROJECTS IRI-9247144 Arkin, Ronald C.; Book, Wayne J.; Lawton, Daryl T. and Vachtsevanos, George Georgia Technical Research Corporation $4,000 - 12 mos. REU/IMHS: Intelligent Sensor-based Robotic Strategies for Material Handling This award provides supplemental funding for undergraduate participation in this research project. The student will assist in the design and integration of a small robot manipulator with a mobile robot base platform, will produce computer codes to support control data communication between a base station host and the vision algorithms for the mobile robot. This award is in the joint ENG/CISE research thrust in Intelligent Material Handling Systems. This research effort studies mobile manipulation for material handling, in the form of integrated control of a mobile platform and robotic arm. Such an integrated system can offer major advances in material handling, providing a wide range of new capabilities for such systems. This includes the ability to function in a partially modelled environment, the ability to carry out a diverse set of operational tasks relevant to material handling, and the ability to carry out those tasks in real time. The highly interdisciplinary research team brings together a wide-ranging collection of enabling technologies including system architecture, robotic modeling, sensing, navigational planning, reactive control, micromanipulation, and others. The involvement of Georgia Tech's existing Materials Handling Research Center will provide an appropriate venue for both testing of ideas and facilitation of technology transfer to interested industrial partners. ASC-9217041 Berwick, Robert C.; Bizzi, Emilio; Bulthoff, Heinrich H.; Jordan, Michael; Wexler, Kenneth; Poggio, Tomaso; Rivest, Ronald L.; Winston, Patrick and Yang, Woodward Massachusetts Institute of Technology $125,000 - 12 mos. (Jointly funded with the New Technologies Program, the Division of Behavioral and Cognitive Sciences, the Division of Advanced Scientific Computing, the Knowledge Models and Cognitive Systems Program, and the Experimental Systems Program - Total Award $600,00) HPCC: High Performance Computing for Learning This is the first year funding of a five-year continuing award in the High Performance Computing and Communications (HPCC) Initiative's Grand Challenge Application Groups competition. In this award to Berwick, Bizzi, Bulthoff, Jordan, Wexler, Poggio, Rivest, Winston, and Yang at MIT, the research project has been designed explicitly to push the High Performance Computing algorithmic and architectural envelope via a CM-5 and VLSI testbed and to address many of the HPCC goals. It will advance new algorithms and software for a broad class of optimization and learning problems, tested on and directly driving operations system and architectural changes on the CM-5 (working with one of the CM- 5's key architects). The learning problems addressed are essentially an entire class of modeling/optimization problems that intersect with nearly all HPCC Grand Challenge Problems. IRI-9114446 Canny, John; Malik, Jitendra; Fearing, Ronald S.; Adiga, Sadashiv and Mendel, Max University of California, Berkeley $125,000 - 12 mos. (Jointly funded with the Dynamic Systems and the Control Program - Total Award $150,000) IMHS: Intelligent Distributed Control of Material Handling This is the first year funding of a two-year continuing award in the joint ENG/CISE research thrust in Intelligent Material Handling Systems. This research effort involves simulation, control, and implementation of an experimental integrated material handling system. The following research issues, felt to be critical to the success of intelligent material handling systems, are studied: Distributed software architecture using a variety of expert software packages, with robust communication protocols; Object-oriented software for simulation of material handling systems at various levels of abstraction; Planning of safe motion and part acquisition using a variety of sensing methods; Real-time sensor processing algorithms for intelligent control of operations; Methods for dealing with uncertainty due to a variety of sources; Testing and validation of concepts in two "microfactories;" and Rapid prototyping of custom end-effectors and fixtures to handle particular parts. The research involves integration of a number of specialist programs, running on a variety of processors, using a versatile interface protocol. The specialists include a package for handler modeling and motion simulation and control, two real-time multiprocessors for control and sensor processing, a vision package, a mechanical design package, and an NC machine code generator. These systems will communicate via the CPU interprocess communication protocol, which has been successfully demonstrated with many of the systems. In addition, the BLOCS object-oriented simulation package, which was developed at Berkeley, will be used to model the actual system at several levels and to validate and improve the design in terms of information flow, material flow, and error tolerance. Eventually, BLOCS will perform some high-level monitoring and control of the actual material handling systems. IRI-9247048 Canny, John; Malik, Jitendra; Fearing, Ronald S.; Adiga, Sadashiv and Mendel, Max University of California, Berkeley $8,000 - 12 mos. REU/IMHS: Intelligent Distributed Control of Material Handling This award provides supplemental funding for undergraduate participation in this research project. The two students will develop an intermediate-level interface to Robot World control software, to support sensor-based motion strategies and force and compliance control of the manipulators. IRI-9245629 Chu, Wesley W.; Cardenas, Alfonso F. and Taira, Ricky University of California, Los Angeles $50,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Division of Instrumentation and Resources - Total Award 206,781) SDB: A Knowledge-Based Multi-Media Distributed Database System for Radiology This is the second year funding of a two-year continuing award IRI-9116849 in the Scientific Database Initiative. This research deals with the use of domain and application knowledge to merge and manage scientific multi-media data from multiple sources. The primary goal of this research is to develop methodologies to integrate knowledge with biomedical image databases and to provide approximate, summary and conceptual query answering via generalization, specialization, and association operations. Further, logical indexing techniques for retrieving images by semantic contents and the requirements of the query language constructs for supporting such cooperative query answering will be developed. Such features are important in aiding scientists to extract new scientific knowledge from the raw data. These concepts will be validated in a testbed linked with the biomedical image databases. The joint research between the investigators from the Computer Science Department and Medical School will assure the prototype system and experiments used are of direct interest to biological and medical applications. The new methodology enables access to the vast storehouse of images by content and features rather than by artificial keys, such as a person's ID. Such capabilities enhance the image databases by characterizing the objects functions or dysfunctions and behavior which will lead to improved image analysis and diagnosis techniques. The outcomes of this research should be extensible to other image and multi-media database applications and result in an experimental knowledge-based image database system that will be made available to the research community. ASC-9217384 Crutcher, Richard M. University of Illinois, Urbana $11,998 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the New Technologies Program and the Centers Program and the Computational Mathematics Program and the Division of Advanced Scientific Computing - Total Award $450,000) HPCC: Radio Synthesis Imaging: This is the first year funding of a five-year continuing award in the High Performance Computing and Communications Initiative's (HPCC) Grand Challenge Application Groups competition. This award is for the direct implementation of three computing recommendations of the Astronomy and Astrophysics Survey Committee. It will implement a prototype of the next generation of astronomical telescope systems - remotely located telescopes connected by high speed networks to very high performance, scalable architecture computers and on-line data archives. The very computationally intensive algorithms for calibration and imaging of radio synthesis array observations will be optimized and new algorithms which utilize the massively parallel CM-5 will be developed. MSS-9113802 Donath, Max and Ramlingam, Subbiah University of Minnesota, Twin Cities $100,000 - 12 mos. (Jointly funded with the Dynamic Systems and Control and the Computer Integrated Engineering Programs - Total Award $150,000) IMHS: Behavior-Based Control for Autonomous Tending of Manufacturing Workcells This is the first year funding of a two-year continuing award in the joint ENG/CISE research thrust in Intelligent Material Handling Systems. This research addresses behavior based control for autonomous tending of manufacturing workcells. Automation systems assume that all objects are where they are supposed to be and that sensors and machines function as they are meant to. Failure to satisfy these assumptions can often result in catastrophic failure. Methodologies that facilitate the integration of sensors, computers and machines in order to autonomously accomplish complex tasks, i.e. sensing and dealing with the unexpected, are not well developed. This project explores a behavioral control strategy to accomplish these tasks. The idea is based on observations of naturally occurring examples of autonomous systems, such as animal behavior. The strategy is based on the implementation of multiple behaviors taking advantage of multiple sensors and actuators functioning concurrently. Such ideas have been tested on robots, and the result has provided a successful strategy for robust operation in dynamically changing, unstructured environments. This project will use this approach in the tending of machine tools used for low volume batch manufacturing, including the transfer of tools and parts among stations on the shop floor. With the experience gained in this research, progress is expected towards the long term goal of developing a general methodology for designing and building an hierarchical structure of integrated behaviors which can be use to address many generic material handling issues. IRI-9217831 Ekman, Paul; Huang, Thomas G. and Sejnowski, Terrence University of California, San Francisco $11,932 - 7 mos. (Jointly funded with the Interactive Systems Program, the Social Psychology Program and the Division of Behavioral & Cognitive Sciences - Total Award $47,725) Workshop on Facial Expression Understanding This award funds a planning workshop to be held in the Washington, DC area in July 1992 to determine those research problems in the area of facial expression understanding that deserve future attention. The main goals are (a) to identify the most important areas of research on how to extract information from facial activity relevant to a person's emotional, cognitive and physical state; (b) to enhance communication between man and machine through the development of emerging computer vision capture techniques for facial processing and categorization of facial expression; (c) to consider how we can facilitate the training of new investigators in the relevant fields. The workshop will produce a set of recommendations to the investigative community and to NSF to help guide funding opportunities. There will be publications in standard journals of a report. The workshop participants represent a balance of researchers in the areas of facial expressions understanding, computer vision and human computer interaction. Participants include researchers with NSF grants and researchers in government and industrial laboratories. The workshop will consist of plenary sessions and breakout groups focused on several research areas. The steering committee includes Dr. Paul Ekman of the University of California at San Francisco, Dr. Thomas Huang of the University of Illinois at Champaign-Urbana and Dr. Terrence Sejnowski of the Salk Institute. IRI-9208151 Goldberg, Kenneth University of Southern California $7,165 - 12 mos. (Jointly funded with the Dynamic Systems and Control Program - Total Award $14,330) Workshop on Geometric Uncertainty in Motion Planning This award funds a research planning workshop on geometric uncertainty in motion planning, to be held June 15-17, 1992 at the USC Marine Science Center on Catalina Island in California. The organizers are Profs. Ken Goldberg and Ari Requicha of the University of Southern California and Matthew Mason of Carnegie- Mellon University. Robotic operations in most practical settings must cope with uncertainties arising from sensor noise, control error, mechanical tolerances, and incomplete or inaccurate models of the environment. Effective motion-planning methods are critical to successful robotic implementations. Researchers from academia, government, and industry laboratories will come together to review past work, work in progress, and limitations and alternatives to existing theory that suggest priorities for future research. IRI-9216094 Hendler, James and Agrawala, Ashok K. University of Maryland, College Park $5,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Knowledge Models and Cognitive Systems Programs - Total Award $15,000) Workshop on Artificial Intelligence in Real Time This workshop on "Artificial Intelligence in Real Time" is aimed at examining how AI systems can both be supported and can help to support real-time operation systems. This area is of critical importance as AI systems need to function in the support of such critical applications as nuclear power plant control, aircraft operation, hospital life support systems, and military command and control, among others. The workshop, to be held at the University of Maryland in the Spring of 1993, brings together members of both AI and real-time communities to explore issues of mutual interest. A report, produced by this workshop, will aid in the planning activities of the KMCS, RMI, and ITO Programs. DDM-9212644 Hocken, Robert J. University of North Carolina, Charlotte $25,000 - 12 mos. (Jointly funded with the Manufacturing Systems Program, the Division of Design and Manufacturing, and the Design and Computer- Integrated Engineering Program - Total Award $97,532) National Science Foundation - Division of Design and Manufacturing Systems Grantees Conference; Charlotte, North Carolina; January 6- 8, 1993 This award is to support the annual grantees conference of the Division of Design and Manufacturing Systems. This year the conference has been expanded to include participation by grantees of the Information Robotics, and Intelligent Systems Division and Engineering Centers Division who are doing manufacturing-related research. By encouraging interaction among the grantees of these Divisions, this conference will help the individual researchers to become informed about the ongoing activities of their colleagues. In addition, the conference program organization allows for ample time to discuss manufacturing research in detail with the collective research community at the meeting, with feedback and input from the National Science Foundation. IRI-9120664 Hopcroft, John E. Cornell University $5,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program - Total Award $10,000) A Workshop on Information Access and Capture in Engineering Design Environments Computing hardware, digital memory systems, and high speed computer networks have progressed to the point where building very large digital document collections and providing facile distributed access to these collections is practical and economical. However, many difficult problems still stand in the way of implementing such systems. The time is ripe to begin building a broad science base to support this activity and to begin academic programs at universities to produce the PhD's needed to advance progress in this area. This award funds a three day workshop to bring together about 30 key researchers in the area of information capture and access. Participants represent broad coverage of the research area and a balance between industrial and academic and between project implementors and more science base oriented individuals. The workshop will be held November 12-14, 1991 in Ithaca, New York. The coordinators are John Hopcroft, Chair, Computer Science Department, Cornell University and Gregory Zack, Manager, Xerox Design Research Area. The primary objectives of the workshop are to determine the state of the art in information capture and access, to identify key research efforts and projects, and to highlight research direction where progress is likely. In addition the workshop will identify common infrastructure that would foster cooperative research efforts. The Xerox Corporation is a co-sponsor of the workshop. ECS-9106811 Jacobsen, Stephen C. and Wood, J. E. University of Utah $5,000 - 12 mos. (Jointly funded with the Division of Electrical and Communications Systems - Total Award $50,000) Workshop on Microelectromechanical Systems for NSF/NASA/DOD/NIH/DOC Applications This grant will support a multi-agency workshop on microelectromechanical systems (MEMS). Three specific objectives of the workshop are (1) to introduce MEMS to representatives of several Federal agencies, (2) to identify MEMS applications relevant to these agencies, and (3) to provide a high-quality report on the status of MEMS technology and a prediction of future developments and applications. The workshop will be held in the Washington, DC area in February 1993. IRI-9247047 Levinson, Robert University of California, Santa Cruz $4,000 - 12 mos. REU: Adaptive Pattern-Oriented Chess This award provides supplemental funding for undergraduate participation in this research project. The student will continue work on a scheme for monitoring the performance (prediction strength) of a chess-playing program and dynamically altering its internal parameters to improve the program's learning. IRI-9122541 Levitt, Raymond E. Stanford University $30,000 - 12 mos. (Jointly funded with the Information Technology and Organizations Program and the Cross Disciplinary Activities Office - Total Award $179,975) The Virtual Design Team: Simulation Decision-Making and Information Flow in Concurrent, Multi-disciplinary Design This is the first year funding of a three-year continuing award. This research integrates ideas from artificial intelligence (AI) and coordination theory to develop a Virtual Design Team (VDT)-an object-oriented, discrete event, computer simulation model of the flow of decisions and information among participants in complex, multi-disciplinary design projects. Conducting controlled experiments in real organizations to assess the performance of alternative coordination structures is both expensive and fraught with practical difficulties. Compounding this, powerful computer simulation tools to analyze the performance of organization structures by simulation - analogous to finite element models for simulating structural behavior - have not yet been developed. The VDT addresses this need by using AI techniques to model how the organizational structure of, and the information processing technologies used by, managers in a multi-disciplinary project design team effect the richness, volume and timing of information passing among team members. The long range goal of this work is to provide a computerized analysis tool that can help researchers and practitioners analyze the performance of organizations. The first phase of the project will model the impact of organizational structure and information processing technologies on task duration. The second phase will extend the model to capture the impacts of organization structure and information technologies on design quality and reliability. IRI-9208947 Pachowicz, Peter W. and Michalski, Ryszard S. George Mason University $9,659 - 12 mos. (Jointly funded with DARPA - Total Award $26,000) NSF/DARPA Workshop on Machine Learning and Vision This award funds a research planning workshop on machine learning and computer vision, to be held in Harper's Ferry, West Virginia, on October 15-17, 1992. Research in some laboratories is beginning to make progress toward introduction of learning capabilities into computer vision systems. There is a growing interest in interaction between the learning and computer vision research communities. This workshop will provide a forum for leading researchers from both fields to come together to identify concepts and approaches in machine learning that have promise for application to vision, and to develop a research agenda to advance the use of learning in computer vision systems. A report summarizing the current state of the art and outlining research opportunities and priorities will be produced and widely distributed. DDM-9245484 Palekar, Udatta and Ferreira, Placid University of Illinois, Urbana $50,000 - 12 mos. (Jointly funded with the Operations Research and Production Systems Program, and the Structures and Building Systems Program - Total Award $151,871) IMHS: An Enabling Environment for Design and Control of Intelligent Material Handling Systems This is the first year funding of a two-year continuing award in the joint ENG/CISE research thrust in Intelligent Material Handling Systems. The object is to develop an enabling environment which will integrate analysis and synthesis techniques from Operations Research, Artificial Intelligence, Controls Engineering, Geometry, Robotics and Vision, and Parallel Computing into a seamless, extendable system for design and control of material handling systems. The development of such an environment requires that a number of fundamental problems in each of the areas mentioned above be solved or adapted to make them useful in material handling. Further, integration of results from these areas into a single system raises a number of questions involving communications, resource allocation, interfacing protocols etc. A proof-of-concept system demonstrating the concept of an intelligent material handling enabling environment will be developed. This system will be configured to control material handling systems in two different domains: factory floors and building construction sites. MSS-9203431 Sanderson, Arthur C. Rensselaer Polytechnic Institute $8,000 - 8 mos. (Jointly funded with the Dynamic Systems and Control, Production Systems, and Engineering Systems Programs - Total Award $31,490) Group Travel grant to 1992 IEEE International Conference on Robotics and Automation in Nice This is a group travel grant award for U.S. participation in the IEEE International Conference on Robotics and Automation to be held in Nice, France during the period of May 11-15, 1992. Traditionally held in the U.S., it has become the largest annual conference in the field with more than 20 countries participating. This will be the first time that it has been held outside of the U.S. and it is important that the U.S. remain prominent in this activity. A competitive procedure will be used to select 48 U.S. participants for partial airfare reimbursement. IRI-9116809 Shapiro, Linda G.; Tanimoto, Steven L. and Brinkley, James F. University of Washington $50,000 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program and the Database and Expert Systems Program - $120,000) SDB: A Visual Database System for Computer Vision Research This is the first year funding of a two-year continuing award in the Scientific Databases Initiative. A database system for computer vision must be able to manipulate data in numerous forms, running a full spectrum from values of scalar features, through arrays of light intensities, to complex linked data structures representing networks of partially and tentatively recognized shapes and models. This work entails the design and partial implementation of a prototype vision database system based on a hierarchical, relational data structure that can handle a wide variety of types of data. The design also includes a novel visual user interface that allows the user to examine and comprehend the entities being processed. In this work, the organization and access functions for the database system are developed through a study of the kinds of structures and operations used in two different vision application domains: robotics and medicine. Both primitive access functions and some higher-level vision operations are implemented in the prototype system. The flexibility and efficiency of the system are evaluated for vision algorithms being developed in robot vision and medical image processing research projects. The result will be a general scientific database system that can be accessed via both graphical and application program interfaces and that will satisfy the needs of computer vision systems. ASC-9217091 Taylor, D. Lansing and Fahlman, Scott E. Carnegie Mellon University $100,000 - 12 mos. (Jointly funded with the Knowledge Models and Cognitive Systems Program, the New Technologies Program, the Long-Term Projects in Environmental Biology Program and the Division of Advanced Scientific Computing and the Division of Environmental Biology - Total Award $599,213) HPCC: High Performance Imaging in Biological Research This is the first year funding of a five-year continuing award in the High Performance Computing and Communications Initiative's (HPCC) Grand Challenge Application Groups competition. This award is to research and develop an Automated Interactive Microscope (AIM). The AIM will combine the latest technologies in light microscopy and reagent chemistry with advanced techniques for computerized image processing, image analysis, and display, implemented on high-performance parallel computers. This combination will produce an automated, high-speed, interactive tool that will make possible new kinds of basic biological research on living cells and tissues. While one milestone of the research will be to show the proof-of-concept of AIM, the ongoing thrust will be continued development as new technologies arise and the involvement of the biological community. INT-9123796 Tripathi, Satish K. University of Maryland, College Park $2,500 - 12 mos. (Jointly funded with the Division of International Programs - Total Award $45,250) Cooperative Research in Computer Science, Indo-U.S. Workshop August 4-6, 1992, Bangalore, India This award supports participation of a U.S. team of computer scientists and engineers in an Indo-U.S. workshop on "Cooperative Research in Computer Science" held in Bangalore, India August 4 to 6, 1992. The areas of expertise to be represented include: robotics and computer vision, computer systems and networking, software engineering, and artificial intelligence. The objective of the workshop is to present papers by U.S. and Indian participants about the recent advances in the various topical areas, identify research needs in each of these areas, identify suitable areas for Indo-U.S. collaborative research, and establish firm collaborations between U.S. and Indian scientists. This is the second Indo-U.S. workshop in the area of computer science. The first was held in 1989 in the city of Hyderbad, India with participation from India's main governmental and private research organizations with interest in computer research and applications. Since that time a number of collaborative activities have resulted. The present workshop is designed to continue and intensify this mutually beneficial interaction and collaboration. The project meets the objectives of the U.S.-India Cooperative Science Program. IRI-9245625 Ullman, Jeffrey D. and Law, Kincho H. Stanford University $50,000 - 12 mos. (Jointly funded with the Database and Expert Systems Program and the Structures and Building Systems Program - Total Award $150,000) SDB: Integrated Data Exchange and Concurrent Designs for Engineered Facilities This is the second year funding of a three-year continuing award IRI-9116646 in the Scientific Database Initiative. The architecture-engineering-construction industry is highly fragmented, both vertically (between project phases, e.g., planning, design, and construction) and horizontally (between specialists at a given project phase, e.g., design). The industry needs an integrated computing environment that will support concurrent design with fast data exchange and powerful change management capabilities. The long-term potential benefits are improved designs, reduced construction time, fewer design errors and omissions, minimization of costly rework, and better life-cycle facility management. The key contributions of this research project are methods to (1) translate and communicate project data dynamically, overcoming both logical barriers (differing views of overlapping data) and physical barriers (data distribution); and (2) detect, analyze, and manage changes efficiently during concurrent design processes. These contributions are demonstrated in prototype software that integrates systems for CAD, drafting, analysis, and simulation along with relational and object-oriented databases. The first software component is an integrated set of object-oriented, multi-machine utilities to provide fast, flexible, intelligent data translation and transfer. The second component is a set of constraint handling and change management utilities that will support declarative definition of data dependencies, constraints, and conflict resolution strategies. The project will take advantage of industry links through Stanford University's Center for Integrated Facility Engineering to develop a test bed of industrial-strength databases to demonstrate the software. INDEX OF PRINCIPAL INVESTIGATORS A Abhyankar, Shreeram Purdue University Abola, Enrique Columbia University Adiga, Sadashiv University of California, Berkeley Adrion, Richards W. University of Massachusetts Agrawal, Divyakant University of California, Santa Barbara Agrawala, Ashok K. University of Maryland, College Park Ahlswede, Thomas Central Michigan University Ahuja, Narendra University of Illinois, Urbana Allen, Peter K. Columbia University Aloimonos, John Y. University of Maryland, College Park Andersen, George J. University of Illinois, Urbana Anderson, Charles W. Colorado State University Arkin, Ronald C. Georgia Tech Research Corporation Asada, Haruhiko Massachusetts Institute of Technology Ashley, Kevin University of Pittsburgh Atkeson, Christopher G. Massachusetts Institute of Technology Atkins, Daniel University of Michigan, Ann Arbor Aunon, Jorge I. Colorado State University Aye, Tin Physical Optics Corporation B Baclawski, Kenneth Northeastern University Badal, Dushan Z. University of Colorado, Colorado Springs Badler, Norman I. University of Pennsylvania Baird, Bridget Connecticut College Bajaj, Chaderjit Purdue University Ballard, Dana H. University of Rochester Baral, Chitta University of Texas, El Paso Barnden, John New Mexico State University Barnes, Julie Bowling Green State University Baroudi, Jack New York University Barto, Andrew G. University of Massachusetts, Amherst Barua, Anitash University of Texas, Austin Batali, John University of California, San Diego Bateman, John University of Southern California Bay, John S. Virginia Polytechnic Institute Beckman, Mary E. Ohio State University Bekey, George A. University of Southern California Ben-Arie, Jezekiel Illinois Institute of Technology Bernstein, Arthur J. SUNY, Stonybrook Bergeron, R. Daniel University of New Hampshire Berwick, Robert Massachusetts Institute of Technology Bishop, Gary University of North Carolina, Chapel Hill Bisshopp, Frederic E. Brown University Bizzi, Emilio Massachusetts Institute of Technology Bjorson, Robert D. Scientific Computer Associates Boggess, Lois Mississippi State University Boland, Richard Case Western Reserve University Book, Wayne J. Georgia Tech Research Corporation Borgida, Alex Rutgers University Borenstein, Johann University of Michigan Boser, Bernhard University of California, Berkeley Boult, Terrance E. Columbia University Bourne, Philip Columbia University Bovik, Alan C. University of Texas, Austin Bowyer, Kevin W. University of South Florida Brennan, Susan E. SUNY Stonybrook Brinkley, James F. University of Washington Brunt, James W. University of New Mexico Bulthoff, Heinrich H. Massachusetts Institute of Technology C Calingaert, Peter University of North Carolina, Chapel Hill Canny, John F. University of California, Berkeley Carberry, Mary Sandra University of Delaware Carbonell, Jaime G. Carnegie Mellon University Cardenas, Alfonso F. University of California, Los Angeles Carey, Michael J. University of Wisconsin, Madison Carlson, Gregory University of Rochester Carpenter, Gail A. Boston University Chandok, Ravinder Carnegie Mellon University Chelberg, David M. Purdue University Research Foundation Chellappa, Ramalingam R. University of Southern California Chen, Hsinchun University of Arizona Chen, Weidong Southern Methodist University Chomicki, Jan Kansas State University Chrysanthis, Panos K. University of Pittsburgh Chu, Wesley W. University of California, Los Angeles Cipra, Raymond J. Purdue University Clancey, William J. Institute for Research on Learning Clauer, Robert C. University of Michigan, Ann Arbor Cohen, Fred E. University of California, San Francisco Cole, Phillip R. SRI International Cole, Ronald A. Oregon Graduate Institute Coombs, Michael New Mexico State University Cooper, Paul Northwestern University Cooper, David B. Brown University Cooper, Gregory University of Pittsburgh Cottrell, Garrison University of California, San Diego Cowan, George Santa Fe Institute Crisman, Jill D. Northeastern University Croft, W. B. University of Massachusetts, Amherst Cruickshank, Alexander Gordon Research Conferences Crutcher, Richard M. University of Illinois, Urbana D D'Ambrosio, Bruce Oregon State University Davis, Ernest S. New York University Davis, Karen C. University of Cincinnati Dawson, Darren M. Clemson University Dean, Thomas L. Brown University Dechter, Rina University of California, Irvine DeFanti, Thomas A. University of Illinois Dewan, Prasun Purdue University Dietterich, Thomas Oregon State University Diller, Kenneth R. University of Texas, Austin Donald, Bruce R. Cornell University Donath, Max University of Minnesota Dorr, Bonnie University of Maryland, College Park Dozier, Jeff University of California, Santa Barbara Duda, Richard O. San Jose State University Dunne, Thomas University of Washington Durfee, Edmund H. University of Michigan, Ann Arbor Dyer, Charles R. University of Wisconsin, Madison E Eich, Margaret H. Southern Methodist University Eichenlaub, Jesse B. Dimension Technologies Inc. Eisenberg, Michael University of Colorado, Boulder Ekman, Paul University of California, San Francisco ElAbbadi, Amr University of California, Santa Barbara Elgot-Drapkin, Jennifer Arizona State University Elkan, Charles University of California, San Diego Ellman, Thomas P. Rutgers University, New Brunswick Elmagarmid, Ahmed K. Purdue University Elman, Jeffrey L. University of California, San Diego Epstein, Susan L. CUNY Hunter College Erdmann, Michael A. Carnegie Mellon University Etzioni, Oren University of Washington Everson, Richard Brown University F Fahlman, Scott E. Carnegie Mellon University Faloutsos, Christos University of Maryland, College Park Fearing, Ronald S. University of California, Berkeley Ferreira, Placid University of Illinois Ferrin, Thomas E. University of California, San Francisco Fischer, Gerhard University of Colorado, Boulder Flamm, Kenneth Brookings Institution Fleck, Margaret M. University of Iowa Flynn, Patrick J. Washington State University Ford, Holland C. Johns Hopkins University Forrest, Stephanie University of New Mexico Forsyth, David A. University of Iowa Fox, Edward A. Virginia Polytechnic Institute Freuder, Eugene C. University of New Hampshire, Durham Fuchs, Henry University of North Carolina, Chapel Hill Futrelle, Robert P. Northeastern University G Garcia, Oscar N. George Washington University Gauch, John M. Northeastern University Gautier, Catherine University of California, Santa Barbara Gelernter, Herbert SUNY, Stonybook Gelfond, Michael University of Texas, El Paso Geller, James New Jersey Institute of Technology Gelsey, Andrew Rutgers University Gennert, Michael A. Worcester Polytechnic Institute Gevins, Alan EEG Systems Laboratory Ghandeharizadeh, Shahram University of Southern California Giacalone, Alessandro SUNY Stony Brook Girson, Andrew D. Digital Video Process, Incorporated Glanz, Filson H. University of New Hampshire Gleitman, Lila R. University of Pennsylvania Goel, Ashok Georgia Institute of Technology Goldberg, Kenneth University of Southern California Goodrich, Michael T. Johns Hopkins University Gordon, Peter C. Harvard University Gottschlich, Susan N. Rensselaer Polytech Institute Graefe, Goetz University of Colorado, Boulder Grant, John Townson State University Greenstein, Shane University of Illinois, Urbana- Champaign Grimson, W. Eric L. Massachusetts Institute of Technology Grosz, Barbara J. Harvard University Gruenwald, Le University of Oklahoma Grupen, Roderic A. University of Massachusetts, Amherst Guthrie, Louise New Mexico State University Gutek, Barbara A. California Institute of Technology H Hachem, Nabil I. Worcester Polytechnic Institute Haddawy, Peter University of Wisconsin Hager, Gregory D. Yale University Hamburger, Henry J. George Mason University Hanks, Steven University of Washington Hansen, Elaine R. University of Colorado, Boulder Hanson, Andrew J. Indiana University Haralick, Robert M. University of Washington Hartson, Rex H. Virginia Polytechnic Institute Haussler, David University of California, Santa Cruz Heath, Lenwood S. Virginia Polytechnic Institute Heckerman, David Institute of Decision Systems Research Hemani, Hooshang Ohio State University Hendler, James University of Maryland, College Park Henrion, Max Institute of Decision Systems Research Henschen, Lawrence J. Northwestern University Herman, Gabor T. University of Pennsylvania Heusinkveld, Paul A. HEUS, Incorporation Hiltz, Starr R. New Jersey Institute of Technology Hirschman, Lynette Oregon State Institute Hirsh, Haym Rutgers University, Busch Campus Hix, Deborah Virginia Polytechnic Institute Hocken, Robert J. University of North Carolina, Charlotte Hodges, Julia Mississippi State University Hoffman, Lance George Washington University Holland, Dorothy C. University of North Carolina, Chapel Hill Hopcroft, John E. Cornell University Horvitz, Eric Institute of Decision Systems Research Howard, Craig Stanford University Huang, Thomas G. University of California, San Francisco Hudson, Scott E. University of Arizona Hull, Richard B. University of Southern California Hunter, Lawrence E. National Library of Medicine Hutchins, Sandra E. Emerson and Stern Associates Huttenlocher, Daniel P. Cornell University I Ide, Nancy M. Vassar College Indurkhya, Bipin Boston University Ioannidis, Yannis E. University of Wisconsin, Madison J Jacobs, Robert Massachusetts Institute of Technology Jacobson, Stephen C. University of Utah Jain, Ramesh C. University of Michigan Jain, Anil K. Michigan State University Jeffrey, Kevin University of North Carolina, Chapel Hill Johnson, David A. Tini Alloy Company Jones, Patricia University of Illinois Jordan, Michael Massachusetts Institute of Technology Joshi, Aravind K. University of Pennsylvania Jouchoux, Alain University of Colorado, Boulder K Kaelbling, Leslie P. Brown University Kagan, Herbert Gordon Research Conferences Kalita, Jugal University of Colorado Kambhampati, Subbarao Arizona State University Katz, Randy H. University of California, Berkeley Kaufman, Arie E. SUNY Stonybrook Kazic, Toni Washington University Kent, Robert E. University of Arkansas, Little Rock Kenyon, Robert University of Illinois Kevkowitz, Haim University of Lowell Kibler, Dennis University of California Kifer, Michael SUNY at Stonybrook Kling, Rob University of California, Irvine Klir, George SUNY Binghampton Koditschek, Daniel E. Yale University Koetzle, Thomas Columbia University Kolodner, Janet L. Georgia Institute of Technology Koren, Yoram University of Michigan Korf, Richard E. University of California, Los Angeles Kraft, Donald M. University of Louisiana Kraft, L. Gordon University of New Hampshire Kraut, Sarit University of Maryland, College Park Kreutz-Delgado, Kenneth University of California, San Diego Kriegman, David J. Yale University Kuipers, Benjamin J. University of Texas, Austin Kulkarni, Sanjeev R. Princeton University Kumar, Akhil Cornell University Kumar, Vijay University of Pennsylvania Kumar, Vipin University of Minnesota Kwasny, Stan C. Washington University L Lang, Jeffrey H. Massachusetts Institute of Technology Langridge, Robert University of California, San Francisco Law, Kincho H. Stanford University Lawler, Eugene University of California, Berkeley Lawton, Daryl T. Georgia Tech Research Corporation Leggett, John Texas A&M Engineer Experiment Station Lehnert, Wendy University of Massachusetts, Amherst Lemke, Andreas University of Colorado, Boulder Lesser, Victor R. University of Massachusetts, Amherst Levinson, Robert University of California, Santa Cruz Levitt, Raymond E. Stanford University Lewis, Michael University of Pittsburgh Lewis, Clayton University of Colorado, Boulder Lifschitz, Vladimir University of Texas, Austin Little, Thomas Boston University Lobo, Jorge University of Illinois, Chicago Lohse, Gerald University of Pennsylvania Loui, Ronald P. Washington University Lucas, Henry C. New York University Luo, Ren C. North Carolina State University M Madnick, Stuart E. Massachusetts Institute of Technology Maes, Patrica Massachusetts Institute of Technology Maguire, Bassett J. University of Texas, Austin Maier, David Oregon Graduate Institute of Technology Malik, Jitendra University of California, Berkeley Malkani, Hogan J. Tennessee State University Malone, Thomas P. University of Michigan, Ann Arbor Mamrak, Sandra A. Ohio State University Manual, Sharon Y. Massachusetts Institute of Technology Marcus, Richard S. Massachusetts Institute of Technology Marschak, Thomas University of California, Berkeley Marshall, Catherine R. Oregon Graduate Institute Martin, William University of Michigan, Ann Arbor Martin, James H. University of Colorado, Boulder Mason, Matthew T. Carnegie Mellon University McAloon, Ken CUNY at Brooklyn College McCabe, Kevin A. University of Arizona McCall, Raymond University of Colorado, Boulder McCartney, Robert University of Connecticut McKeown, Kathleen R. Columbia University McLeod, Dennis University of Southern California Medioni, Gerard University of Southern California Meeker, Loren D. University of New Hampshire Meer, Peter Rutgers University Busch Campus Mendel, Max University of California, Berkeley Michalski, Ryszard S. George Mason University Michener, William K. University of South Carolina, Columbia Miller, Russ SUNY at Buffalo Minker, Jack University of Maryland, College Park Mitchell, Tom M. Carnegie Mellon University Moore, Johanna University of Pittsburgh Mooney, Raymond J. University of Texas, Austin Morris, James H. Carnegie Mellon University Mostow, David J. Rutgers University, New Brunswick Motro, Amihai George Mason University Mozer, Michael C. University of Colorado, Boulder Mumford, David Harvard University Musen, Mark A. Stanford University N Nabet, Bahram Drexel University Naughton, Jeffrey F. University of Wisconsin, Madison Negahdaripour, Shahriar University of Hawaii Monoa Newman, Wyatt Case Western Reserve University Nilsson, Nils J. Stanford University Nirenburg, Sergi Carnegie Mellon University Novick, David Oregon Graduate Institute of Science and Technology O Olson, Dan R. Brigham Young University Olson, Gary M. University of Michigan, Ann Arbor Oren, Shmuel University of California, Irvine Ostendorf, Mari Boston University Ouksel, Aris University of Illinois Oviatt, Sharon L. SRI International P Pachowicz, Peter W. George Mason University Palekar, Udatta University of Illinois, Urbana Panzar, John C. Northeastern University Paris, Cecile University of Southern California Parker, Stotts D. University of California, Los Angeles Pasquale, Joseph University of California, San Diego Passino, Kevin M. Ohio State University Research Foundation Passoneau, Rebecca Columbia University Pausch, Randy F. University of Virginia Pearl, Judea University of California, Los Angeles Perlin, Kenneth H. New York University Perskin, Richard L. Rutgers University Pescitelli, Maurice Jr. Northeastern University Peterson, Larry L. University of Arizona Pickett, Ronald University of Lowell Pinter, Robert B. University of Washington Pirolli, Peter University of California, Berkeley Poggio, Tomaso Massachusetts Institute of Technology Pollack, Martha E. University of Pittsburgh Polson, Peter G. University of Colorado, Boulder Ponce, Jean University of Illinois, Urbana Porter, Bruce W. University of Texas, Austin Preston, Kendall J. Carnegie Mellon University Prieditis, Armand E. University of California, Davis Provan, Gregory University of Pennsylvania Prueitt, Paul S. Georgetown University Pu, Calton Columbia University Pu, Pearl University of Connecticut Q Quek, Francis University of Michigan R Ramakrishna, M. V. Michigan State University Ramakrishnan, I. V. SUNY Stonybrook Ramakrishnan, Raghunath University of Wisconsin, Madison Ramamritham, Krithi University of Massachusetts, Amherst Ramlingam, Subbiah University of Minnesota Rao, Nageswara Old Dominion University Rassenti, Stephen University of Arizona Raviv, Daniel Florida Atlantic University Reagan-Cirincione, Patricia SUNY at Albany Reed, Daniel A. University of Illinois, Champaign- Urbana Reif, John H. Duke University Reitsma, Rene University of Colorado, Boulder Rendell, Larry A. University of Illinois, Urbana Richards, Whitman Massachusetts Institute of Technology Riseman, Edward University of Massachusetts, Amherst Rissland, Edwina L. University of Massachusetts, Amherst Ristad, Eric S. Princeton University Rivest, Roland L. Massachusetts Institute of Technology Rohrbaugh, John SUNY at Albany Ross, Kenneth A. Columbia University Rounds, William C. University of Michigan, Ann Arbor Rowe, Lawrence A. University of California, Berkeley Roy, Asim Arizona State University Rule, James B. SUNY Stonybrook Rumelhart, David E. Santa Fe Institute Rusinkiewicz, Marek University of Houston Russell, Lucian Argonne National Laboratory Russell, Stuart J. University of California, Berkeley Ryan, William B.F. Columbia University S Saton, Gerard Cornell University Salzberg, Steven L. Johns Hopkins University Salzberg, Betty J. Northeastern University Samet, Hanan University of Maryland, College Park Sanderson, Arthur C. Resselaer Polytechnic Institute Sandin, Daniel J. University of Illinois Sanocki, Thomas University of South Florida Sastry S. Shankar University of California, Berkeley Schank, Roger C. Northwestern University Schatz, Bruce R. University of Arizona Schlimmer, Jeffrey C. Washington State University Schubert, Lenhart University of Rochester Sechrest, Stuart University of Michigan, Ann Arbor Sedelow, Walter A. University of Arkansas, Little Rock Sedelow, Sally Y. University of Arkansas, Little Rock Segev, Arie University of California, Berkeley Sejnowski, Terrence University of California, San Francisco Sellis, Timoleon University of Maryland, College Park Sethi, Ishwar K. Wayne State University Shapiro, Linda G. University of Washington Shattuck-Hufnagel, Stefanie Massachusetts Institute of Technology Shavlik, Jude W. University of Wisconsin, Madison Shenoi, Sujeet University of Tulsa Shieber, Stuart N. Harvard University Shih, Frank Y. New Jersey Institute of Technology Shortliffe, Edward H. Stanford University Sieg, John C., Jr. University of Massachusetts Siegel, Michael Massachusetts Institute of Technology Silberschastz, Abraham University of Texas, Austin Silverman, Harvey F. Brown University Simmons, Leonard M. Santa Fe Institute Sirovich, Lawrence Brown University Sklansky, Jack University of California, Irvine Sloan, Kenneth R. University of Alabama, Birmingham Smith, F. D. University of North Carolina, Chapel Smith, Stuart University of Lowell Smith, Terence R. Universtiy of California, Santa Barbara Hill Smith, John B. University of North Carolina, Chapel Hill Smith, Vernon L. University of Arizona Soloway, Elliot M. University of Michigan, Ann Arbor Sparr, Ted M. University of New Hampshire Srihari, Sargur N. University of New York, State Srivastava, Jaideep University of Minnesota Stankovic, John A. University of Massachusetts, Amherst Stanley, James T. Oregon Graduate Institute of Science and Technology Steedman, Mark J. University of Pennsylvania Stevens, Kenneth N. Massachusetts Institute of Technology Stonebraker, Michael R. University of California, Berkeley Stong, Gary W. Drexel University Storey, Veda C. University of Rochester Su, Jainwen University of California, Santa Barbara Subrahmanian, Venkatraman University of Maryland, College Park Swain, Michael University of Chicago Swartout, William University of Southern California T Taira, Ricky University of California, Los Angeles Talburt, John R. University of Arkansas, Little Rock Tanchoco, J.M.A. Purdue University Tanimoto, Steven L. University of Washington Tarn, Tzyh-Jong Washington University Taylor, D. Lansing Carnegie Mellon Univeristy Tecuci, Gheorghe George Mason University Teicholz, Paul Stanford University Thomason, Richmond Carnegie Mellon University Thompson, William B. University of Utah Tomasi, Carlo Cornell University Tomita, Masaru Carnegie Mellon University Tong, Christopher Rutgers University Touretzky, David Carnegie Mellon University Towsley, Don F. University of Massachusetts, Amherst Tretkoff, Carol CUNY at Brooklyn College Tripathi, Satish K. University of Maryland, College Park Truszczynski, Miroslaw University of Kentucky Tsotras, Vassilis Polytechnic University, New York City Tsur, Shalom Swiss Bank Corporation Tuceryan, Mihran Michigan State University Tyce, Robert C. University of Rhode Island U Udupa, Jayaram K. University of Pennsylvania Ullman, Jeffrey D. Stanford University Ullman, Shimon Massachusetts Institute of Technology V Vachtsevanos, George Georgia Tech Research Corporation Van Gelder, Allen University of California, Santa Cruz Varaiya, Pravin University of California, Berkeley Veronis, Jean Vassar College Vijayashanker, K. University of Delaware Vitter, Jeffrey Brown University W Wah, Benjamin W. University of Illinois, Urbana Walpole, Jonathan Oregon Graduate Institute of Technology Walther, Sandra S. Rutgers University Wang, DeLiang Ohio State University Ward, Matthew O. Worcester Polytechnic Institute Ward, Samuel University of Arizona Warmuth, Manfred University of California, Santa Cruz Warren, David S. SUNY Stonybrook Wehrmeirster, Robert M. Data Parallel Systems, Inc. Weiman, Carl F.R. Transitions Research Corporation Weintraub, Mitchel SRI, International, Menlo Park, CA Weiss, Richard S. University of Massachusetts, Amherst Weld, Daniel, S. University of Washington Weymouth, Terry E. University of Michigan, Ann Arbor Wexler, Kenneth Massachusetts Institute of Technology Whittaker, William L. Carnegie Mellon University White, Halbert University of California, San Diego Wickens, Christopher D. University of Illinois, Urbana Widrow, Bernard Stanford University Wilensky, Robert University of California, Berkeley Wilkenfield, Jonathan University of Maryland, College Park Wilks, Yorick New Mexico State University Williams, Ronald J. Northeastern University Winslett, Marianne University of Illinois, Urbana Winston, Patrick Massachusetts Institute of Technology Wolberg, George The City College of Cuny Wolfe, Michael Oregon Graduate Institute of Science and Technology Wood, J. E. University of Utah Wright, William V. University of North Carolina, Chapel Hill X Xiao, Jing University of North Carolina, Charlotte Y Yager, Ronald Iona College Yang, Woodward Massachusetts Institute of Technology Ydstie, B. Erik University of Massachusetts, Amherst Yen, John Texas A&M University Yu, Clement T. University of Illinois, Chicago Yun, Xiaoping University of Pennsylvania Z Zahorian, Stephen A. Oregon Graduate Institute Zheng, Yuan F. Ohio State University Zhuang, Xinhua University of Missouri, Columbia Zigura, Ilze University of Colorado, Boulder OMB: PT: KW: NSF 93-133 (Supersede NSF 91-141)