| NSF Org: |
CCF Division of Computing and Communication Foundations |
| Recipient: |
|
| Initial Amendment Date: | September 11, 2012 |
| Latest Amendment Date: | September 11, 2012 |
| Award Number: | 1239229 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Nina Amla
namla@nsf.gov (703)292-7991 CCF Division of Computing and Communication Foundations CSE Directorate for Computer and Information Science and Engineering |
| Start Date: | October 1, 2012 |
| End Date: | September 30, 2016 (Estimated) |
| Total Intended Award Amount: | $310,000.00 |
| Total Awarded Amount to Date: | $310,000.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
1600 SW 4TH AVE PORTLAND OR US 97201-5508 (503)725-9900 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
Portland OR US 97207-0751 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): |
Information Technology Researc, CPS-Cyber-Physical Systems |
| Primary Program Source: |
|
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.070 |
ABSTRACT
Effective engineering of complex devices often depends critically on the ability to encapsulate responsibility for tasks into modular agents and ensure those agents communicate with one another in well-defined and easily observable ways. When such conditions are followed, it becomes possible to detect where problems lie so they can be corrected. It also becomes possible to optimize the agents and their communications to improve performance. Cyber-physical systems (like robots, self-piloting aircraft, etc.) modify themselves to improve performance break those conditions in that some agent modules negotiate their own communications and decide their own actions, sometimes taking advantage of the physics of the world in ways we did not anticipate. This renders difficult application of standard engineering tools to accomplish critical fault diagnosis and design optimization. This project will produce analysis methods address the specific needs of cyber-physical systems that, by their natures, break the rules of convention. We will apply these new methods to the design and analysis of self-improving controllers for flapping-wing micro air vehicles.
This work will provide advances in both model-checking related formal design methodologies and in module-based self-adaptive control in computationally resource constrained cyber-physical systems. The formal methods advances will significantly expand our ability to properly design and verify systems that tightly couple computation, sensors, and actuators. The specific test application addressed is significant to a number of nationally important security and defense efforts and will directly impact identified national priorities.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external
site maintained by the publisher. Some full text articles may not yet be available without a
charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from
this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
Biomimetic flapping-wing micro-aerial vehicles (FWMAV) have been the focus of much recent research due to their potential for both civilian and military applications. Because of heir insect-like size and relative simplicity in comparison with more traditional unmanned aerial systems (such as quadrotors or fixedwing airplanes), they can be made relatively cheaply and be a part of personal equipment of soldiers, first responders, law enforcement members etc.
A FWMAV qualifies as a Cyber Physical System (CPS). A CPS is a system where a controlled physical process and information processing is tightly coupled. Working with CPS introduces unique challenges. Our apporach was to design a multi-agent system (MAS) where control laws are directly adapted at a higher level of abstraction in the control law hierarchy. In this system, agents are responsible for collecting and estimating vehicle pose, recording waypoint locations for trajectory following, generating inputs needed by the split-cycle oscillator, monitoring vehicle behaviour and, when necessary, conducting diagnostics and adapting the control rule base. This required developing fault detection and fault recovery mechanisms based on a combination of extrinsic and intrinsic evolution. The objective was to create an autonomous working FWMAV capable of trajectory following.
The FWMAV was mounted on a styrofoam disk that floated in a water tank (see figure). This effectively removed one dimension of movement. That is, the flapping wings were now only required to move the vehicle in an X-Y plane. Trajectories could now be defined for testing by specifying a series of waypoints, via (x,y) coordinates, on the surface of the water. The ultimate goal was to see if the FWMAV could autonomously move from waypoint to waypoint. We were able to meet this goal (details below.)
The FWMAV used throughout this research effort was constructed by the Co-PI at Wright State University (see attached figure). The MAS developed at Portland State University is shown in the second figure. Each agent is described below:
a) collection agent-- collects camera data and outputs estimated pose (positional) data
b) controller agent -- based on pose information calculates 4 parameters (two for each wing) sent to the onboard oscillator. These parameters are used by the oscillator to control wing movements. (The oscillator design was done by the Co-PI at Wright State University.)
c) strategy agent -- contains a list of the waypoints along a trajectory for the FWMAV to follow
d) monitor agent -- receives movement information from the controller agent and pose information from the collection agent and determines if FWMAV is operating properly. If not, it initiates diagnostic operations.
e) diagnostic agent -- initiates tests to determine if faults exist. If so, initiates fault recovery operations.
The most important of the agents is the controller agent since it issues the four parameters to an onboard oscillator that determines wing movements. We developed a subsumption architecture rule base to specify the behavior of this agent.
A subsumption architecture consists of a series of layers where lower layer rules produce simple, critical behavior such as avoiding obstacles while higher levels produce more sophisticated behavior needed for trajectory following. Higher level behavior subsumes lower level behavior. A subsumption architecture is ideal for navigation control in dynamic physical environments. It permits reactive behavior without resorting to prior path planning because there is no world model required.
The MAS was able to obtain the pose information and the controller agent would issue four parameter values to the onboard oscillator as the vehicle followed a trajectory. The problem was how to determine what those four parameter values should be. Here we employed online learning.
The vehicle will be placed in its operational environment – i.e. a water tank – and object parameter values will be intrinsically evolved for each basic movement (move forward, turn left, etc.). The initial object parameter values will be evolved using an in-house developed simulator. Online learning phase was performed continuously. Each time a rule fires in the subsumption architecture rule base the monitor agent is informed, so it knows what maneuver was commanded. The monitor agent observes the pose parameters and determines if the performance is within limits or is degrading. Thus, the monitor agent continuously learns about how the vehicle is performing. If the observed behavior deviates too much from the expected behavior, then diagnostics are run. If the diagnostics confirm the behavior has degraded below some threshold, then the rule base is adapted.
In summary, we were able to achieve the following outcomes:
1) We showed that a MAS is suitable for the control of a FWMAV with many advantages over conventional control systems. (A MAS had never before been used to control a FWMAV.)
2) Using this MAS, a high degree of autonomous FWMAV behavior (including trajectory following and obstacle avoidance) can be done.
Last Modified: 03/03/2017
Modified by: Garrison Greenwood
Please report errors in award information by writing to: awardsearch@nsf.gov.
