Guaranteed Capture in Multi-Agent Intercept Problems
Abstract
Multi-player dynamic games using differential game formulations received much attention in the early 2000s in the context of collision avoidance, conflict resolution, and intercept problems. In reach-avoid (RA) games, a type ofdynamic game, an evader is trying to reach a final set P, while avoiding entering a capture set A. Generally, the control policy of the pursuers and evaders in RA games is computed via numerical approximations of Hamilton-Jacobi (HJ) equations. A known shortcoming of these approaches is that they suffer from the curse of dimensionality.This proposal seeks to convert the multi-player RA game, into a multi-agent coverage control problem, where scalable optimal control laws that exploit coordination can be utilized. The proposed scheme coordinates themulti-pursuer team to provide persistent coverage of the evaders reachable set, which is agnostic to evaders future maneuvers. Suitable distributed control laws can be designed to ensure persistent coverage of these time varying domains. Capture can then be guaranteed by ensuring the pursuers reachable sets persistently cover the evaders reachable set.The novelty of this work is twofold. First, by casting the dynamic game as a coverage problem, our work steps away from the computationally intensive dynamic game theory solution, and provides a distributed coverage law toensure capture. Secondly, the size of the pursuer team is explicitly used to reason about the kinematic advantage or disadvantage to the evader, and the number of agents necessary to achieve capture due to kinematic characteristics is provided.In particular, this work impacts the Flight Dynamics and Control area objective of ensuring weapon system coordination success on countering threats for ship defense purposes. It explicitly addresses the challenge area of providing mathematically rigorous success guarantees for a system of vehicles operating with similar objectives, although the framework is abstract enough that it could potentially be applied to a number of different applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 05, 2021
- Source ID
- N000142112410
Entities
People
- Yancy Diaz-Mercado
Organizations
- Office of Naval Research
- United States Navy
- University of Maryland