The Rutgers Computer Science Department is excited to share that Associate Professor Jingjin Yu ARL Proposal entitled, Coordinated Maneuver against Adversaries with Imperfect Information: A Game-Theoretic Monte Carlo Tree Search Approach, has been awarded.
The two-year project awarded to Mississippi State and Rutgers aims to develop highly capable AI/ML-driven planning algorithms for deploying multi-agent systems in complex environments executing multi-domain operations (MDO) in the presence of adversaries. Consider the scenario where a squad of human and machine agents is sent out for information gathering under the canopy while recognizing and preparing for adversarial contact. In such operations, complexity arises in many ways, including having only imperfect information about the environments and adversaries, limited/infrequent/asynchronous team communications, etc. To overcome these challenges, we propose a game-theoretic approach to capture complex domain dynamics with multi-agent reinforcement learning (MARL) and further apply advanced search (e.g., Monte Carlo tree search (MCTS)) over the learned knowledge to derive optimal actions to optimize the team's performance against adversaries over extended time horizons.
The anticipated award is earmarked at $800K, with Rutgers receiving $387K. Please join us in congratulating Jingjin Yu.
