You are here

Learning Estimates of Aggregate Performance (LEAP)

Award Information
Agency: Department of Defense
Branch: Air Force
Contract: FA8650-09-M-3928
Agency Tracking Number: F083-180-2340
Amount: $100,000.00
Phase: Phase I
Program: SBIR
Solicitation Topic Code: AF083-180
Solicitation Number: 2008.3
Timeline
Solicitation Year: 2008
Award Year: 2009
Award Start Date (Proposal Award Date): 2008-12-19
Award End Date (Contract End Date): 2009-11-19
Small Business Information
12 Gill Street Suite 1400
Woburn, MA 01801
United States
DUNS: 967259946
HUBZone Owned: No
Woman Owned: No
Socially and Economically Disadvantaged: No
Principal Investigator
 Nathan Schurr
 Human-Agent Collaboration Scientist
 (781) 496-2453
 nschurr@aptima.com
Business Contact
 Margaret Clancy
Title: Chief Financial Officer
Phone: (781) 496-2415
Email: clancy@aptima.com
Research Institution
N/A
Abstract

Intelligence, Surveillance and Reconnaissance (ISR) systems are becoming more and more complex, with ever increasing fidelity of data and ever increasing numbers of deployable UAVs and other sensors. Error rates are currently the best predictor of performance in autonomous, human, and sociotechnical systems. However, there are three key issues that arise when trying to determine the error rate of the complete ISR system: 1. learning algorithms must handle the sparse data available; 2. lack of structure in the data; and 3. complexity of overall human-automation system performance. Aptima proposes to address these challenges with a comprehensive human-system approach called Learning Estimates of Aggregate Performance (LEAP). The LEAP approach proposes to leverage techniques in both human modeling and machine learning to arrive at a solution that is both feasible and useful for calculating overall ISR system performance. LEAP leverages Signal Detection Theory to form an accurate model of human error and Relevance Vector Machines in order to reduce the necessary amount of training data. The LEAP approach, once achieved, has the ability to create more accurate performance estimates with much fewer human experiments necessary than with traditional learning approaches. BENEFIT: • Drastic reduction the number of human experiments necessary in addition to leveraging available data to the fullest extent possible. • Groundwork for human operator-based and human operator-model-based experimentation. • Deep understanding of the important human factors that must be considered in the ISR system. • Signal Detection Theory-informed models of human operator error and performance.

* Information listed above is at the time of submission. *

US Flag An Official Website of the United States Government