| NSF Org: |
CNS Division Of Computer and Network Systems |
| Recipient: |
|
| Initial Amendment Date: | September 17, 2015 |
| Latest Amendment Date: | September 17, 2015 |
| Award Number: | 1543656 |
| Award Instrument: | Standard Grant |
| Program Manager: |
David Corman
CNS Division Of Computer and Network Systems CSE Directorate for Computer and Information Science and Engineering |
| Start Date: | October 1, 2015 |
| End Date: | September 30, 2020 (Estimated) |
| Total Intended Award Amount: | $1,100,000.00 |
| Total Awarded Amount to Date: | $1,100,000.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
240 FRENCH ADMINISTRATION BLDG PULLMAN WA US 99164-0001 (509)335-9661 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
Pullman WA US 99164-2752 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | 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
This project aims to design algorithmic techniques to perform activity discovery, recognition, and prediction from sensor data. These techniques will form the foundation for the science of Activity- Prediction Cyber-Physical Systems, including potential improvement in the responsiveness and adaptiveness of the systems. The outcome of this work is also anticipated to have important implications in the specific application areas of health care and sustainability, two priority areas of societal importance. The first application will allow for health interventions to be provided that adapt to an individual's daily routine and operate in that person's everyday environment. The second application will offer concrete tools for building automation that improve sustainability without disrupting an individual's current or upcoming activities. The project investigators will leverage existing training programs to involve students from underrepresented groups in this research. Bi-annual tours and a museum exhibit will reach K-12 teachers, students and visitors, and ongoing commercialization efforts will ensure that the designed technologies are made available for the public to use.
Deploying activity-predictive cyber-physical systems "in the wild" requires a number of robust computational components for activity learning, knowledge transfer, and human-in- the-loop computing that are introduced as part of this project. These components then create cyber physical systems that funnel information from a sensed environment (the physical setting as well as humans in the environment), to activity models in the cloud, to mobile device interfaces, to the smart grid, and then back to the environment. The proposed research centers on defining the science of activity-predictive cyber-physical systems, organized around the following thrusts: (1) the design of scalable and generalizable algorithms for activity discovery, recognition, and prediction; (2) the design of transfer learning methods to increase the the ability to generalize activity-predictive cyber-physical systems; (3) the design of human-in-the-loop computing methods to increase the sensitivity of activity-predictive cyber-physical systems; (4) the introduction of evaluation metrics for activity-predictive cyber-physical systems; and (5) transition of activity-predictive cyber-physical systems to practical applications including health monitoring/intervention and smart/sustainable cities.
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.
The overall mission of this project is to create a new science of activity-predictive cyber-physical systems, ApCPS. In support of this mission, we designed machine learning methods to model and track activities from sensor data and designed three activity-predictive applications with societal impact. First, we designed algorithms to discover, recognize, and predict occurrences of activities from ambient and mobile sensor data. We collected over 100 aggregated years of smart home sensor data and reported 98% recognition accuracy of 12 activities for the 22+ million labeled instances. We also collected over 69 thousand labeled instances of 33 activities from mobile data and reported 48% accuracy for these activities. We then created methods to make the algorithms more scalable and robust. These include domain adaptation strategies to adapt labeled data from one domain (e.g., one home or person) to another, without supervision. We also designed an algorithm to track multiple residents in a single smart home and make time series learning more robust with adversarial strategies. Finally, we designed clinician-in-the-loop methods to visualize digital behavior markers and detect clinically-relevant behavior anomalies.
We deployed our methods in three applications. The first is a digital memory notebook app that provides a ubiquitous memory aid for individuals experiencing memory limitations. The app partners with a smart home to track daily activities, maintain a list of completed activities, and prompt residents to initiate activities that have not yet been performed at their normal time. The second is an automated home and employs a data driven, activity-based HVAC controller. This home maintains specified comfort settings for each detected activity while minimizing energy consumption. The automated home yielded 5.14% energy savings over a baseline on/off controller in four apartments. The third application is a robotic activity support system. A mobile robot partners with a smart home to track daily activities. When the resident has difficulty initiating or completing the activity, the robot approaches the resident and plays videos of the activity or the missed step. The robot also maintains a map of activity-critical items in the home and leads the resident to items that are needed for the current activity.
Our ApCPS technologies have been described in a two-semester Gerontechnology course sequence. They are also highlighted in a series of youtube videos and displayed at a local museum.
Last Modified: 10/02/2020
Modified by: Diane J Cook
Please report errors in award information by writing to: awardsearch@nsf.gov.
