Game-Theoretic AI for Robust Course of Action (COA) Generation and Wargaming

The OSW26BZ02-DV004 project seeks to develop a game-theoretic Artificial Intelligence (AI) engine for generating and executing optimized courses of action (COAs) in complex, multi-domain wargaming environments. This AI aims to consistently outperform human planners, providing a strategic advantage in planning, doctrine development, and operational analysis. Current military planning is human-intensive, slow, and explores a limited range of COAs, leading to operational risks. The proposed solution must be founded in computational game theory, capable of computing approximate Nash equilibria in large-scale, zero-sum, imperfect-information games through self-play in high-fidelity simulations. Key attributes for the AI include dominant performance, human-interpretability, scalability from tactical to operational scenarios, computational efficiency on modest footprints, and an “anytime” capability. Proposals for this Direct to Phase II project require documentation of a mature prototype demonstrating performance against expert human teams, human-interpretable strategies, and scalability. Phase II will focus on maturing and scaling the AI engine for defense applications, including integration with government M&S environments, validation in complex scenarios, and demonstrating robustness under degraded conditions. Dual-use applications include next-generation wargaming and operational planning for the DoD, and commercial markets such as financial modeling, cybersecurity, and complex business negotiations.