Enabling Re-configurable Multi-Operator Multi-Agent (MOMA) Teams- A Trust Inference and Propagation (TIP) Approach
Abstract
The future vision for the U.S. Air Force includes teams of human and autonomous agents working together to accomplish missions. Rapid, intelligent, and robust reconfiguration of teams has been identified as a core strategy for many mission-critical tasks. For this re-configuration approach to work, swift and robust development of appropriate trust with new, rotating, and evolving team members is essential. Therefore, the high-level goal of this project is to enable multi-operator multi-agent (MOMA) re-configurable teams wherein the human and autonomous agents could dynamically reconfigure to maximize team mission performance. To accomplish this goal, we propose a trust-driven theoretical framework and aim to (1) have a fundamental understanding of team-based trust dynamics in MOMA teams; (2) develop trust prediction algorithms that are able to capture the trust dynamics between any human agent and any new, rotating, and evolving autonomous agents; (3) develop a robust algorithmic framework that enables the optimal teaming and decision-making by fully considering the trust dynamics; (4) incorporate state-of-the-art machine (deep) learning capabilities into our model and testbed (with implementing deep neural networks to approximate value functions).
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 29, 2024
- Source ID
- FA95502310044
Entities
People
- Xi Yang
Organizations
- Air Force Office of Scientific Research
- Board of Regents of the University of Michigan
- United States Air Force