Adversary Detection and Resilient Control for Multi-Agent Systems (2207.10275v1)

Abstract: This paper presents an adversary detection mechanism and a resilient control framework for multi-agent systems under spatiotemporal constraints. Safety in multi-agent systems is typically addressed under the assumption that all agents collaborate to ensure the forward invariance of a desired safe set. This work analyzes agent behaviors based on certain behavior metrics, and designs a proactive adversary detection mechanism based on the notion of the critical region for the system operation. In particular, the presented detection mechanism not only identifies adversarial agents, but also ensures all-time safety for intact agents. Then, based on the analysis and detection results, a resilient QP-based controller is presented to ensure safety and liveness constraints for intact agents. Simulation results validate the efficacy of the presented theoretical contributions.