Security-preserving multi-agent coordination for complex temporal logic tasks

Abstract

This paper investigates the coordination of multiple agents for high-level tasks described by linear temporal logics (LTL). The general purpose for multi-agent coordination is to synthesize a plan such that the LTL task is achieved optimally. In addition to the standard requirement on the correctness of the plan, we further investigate the potential information leakage of each agent during the operating process. Specifically, we consider the scenario where the behavior of each individual agent is partially monitored by a passive intruder modeled as an outside observer or an eavesdropper. The security constraint requires that the intruder can never identify for sure that some specific individual agent is carrying out some sub-tasks of significant importance. To this end, we model the mobile capability of the agent team by a global labeled transition system. To describe information-flow security constraint, motivated by the generic notion of opacity, two different types security requirements are proposed for each individual agent. An effective coordination algorithm is proposed that synthesizes an optimal global plan for the entire agent team such that the global LTL task can be achieved, while the security of each individual agent is preserved. The proposed framework is implemented in real-world experiment and is also demonstrated by several case studies.