DEC-POMDP Stochastic Game Approach for Uncertain MultiAgent Systems
Agency / Branch:
DOD / ARMY
The key innovation proposed in this Phase II effort is a game-theoretic agent-based and simulation-based approach that enables coordinated tactical control of robotic air and ground systems aligned with C2ORE ATD program objectives and supports FCS-equipped Battle Commander's decision-making ability. It allows individual plans and strategies developed for the command and control of separate UMS (UAV and UGV) missions to be compared collectively to determine and resolve resource, time & space conflicts. The proposed Game-theoretic Agent-based and Simulation-based Tool (GAST) consists of: 1) Partially Observable Markov Decision Process (POMDP) modeled agents evaluating best plan of action for each UMS mission given the uncertainty (from UGS) and plans of action for other UMS missions, 2) a fast-time and high-fidelity distributed discrete-event simulation to detect conflicts in the collective plans, 3) a rule-based Mediator that applies constraints on the agent's search algorithms to facilitate in the generation of conflict-free plans. The Phase I results demonstrate that an extended approach can eliminate resource usage conflicts and unnecessary redundancies in the plans. The Phase II effort includes validation of the approach on actual robotic platforms, and development of associated tools for the users to easily and graphically develop scenarios, test and deploy the system.
Small Business Information at Submission:
INTELLIGENT AUTOMATION, INC.
15400 Calhoun Drive Suite 400 Rockville, MD 20855
Number of Employees: