Predictive Regret-Matching for Cooperating Interceptors to Defeat an Advanced Threat

Abstract

Threats combining kinematic superiority, high-g maneuvering and evasive capabilities are in development. These advanced threats can reduce the survivability of high-value assets (HVAs). Here, we demonstrates that it is possible to defeat such an advanced threat with cheaper lower-performance interceptors using an alternative approach to traditional optimal control. These interceptors harness the knowledge of the forecasted regions that the threat can access, referred to as threat reachability. Applying reachability, the interceptors can be organized to block the passage of the threat to the HVAs as well as to defeat it. Here, we have developed a reachability calculator that is scalable to accommodate multiple interceptors and combined it with an on-line regret-matching learner derived from game theory to produce the self-organization and guidance for the interceptors. Numerical simulations are provided to demonstrate the validity of the resulting solution. Furthermore, some comparison is provided to benchmark our approach against a recently published differential game solution on the same scenarios. The comparison shows that our algorithm outperforms the optimal control solution.