AI Enabled Game Theoretic Approach for Hypersonic Swarm on Swarm Engagement Scenarios

The Missile Defense Agency (MDA) seeks innovative game theoretic technologies for future exo-atmospheric hypersonic combat scenarios. The distinguishing features of the problem are (i) distributed decision architecture of adversarial swarm-vs-swarm engagement and (ii) long flight ranges, hypersonic velocities, and high maneuverability of the vehicles in both swarms. ControlX-UIUC team proposes solutions for the problem of intercepting adversarial hypersonic swarms by exploiting the game-theoretic nature of large-scale pursuit-evasion games. The proposed solution will build on game theoretic methods enhanced by artificial intelligence (AI) / machine learning (ML) tools to develop optimal profiles for target assignment and interception in high-speed swarm-on-swarm scenarios. We consider scenarios where vehicles are allowed to retarget or divert their flight paths, subject to energy and momentum limitations. The team also proposes computationally efficient means to streamline this process. We develop algorithms based on a hierarchical/hybrid approach consisting of: (i) high-level multi-vehicle assignment based on cooperative game theory, non-cooperative game theory, and discrete optimization; (ii) low-level multi-vehicle trajectory optimization based on computational differential games, centralized convex programming relaxations, and approximate online optimization for real-time computations, and (iii) enhancing the algorithms using AI/ML techniques for pre-launch training. In Phase II, ControlX-UIUC team will expand on Phase I results by producing software/hardware components and demonstrating technology through prototype testing. ControlX will optimize designs for integration into future MDA hypersonic interceptor systems. By the end of Phase I, we intend to increase the TRL to 4. By the end of Phase II, we aim for TRL 6-7. Approved for Public Release | 24-MDA-11673 (8 Jan 24)