上海科技大学知识管理系统

In the realm of complex and evolving Multi-Agent Systems (MAS), ensuring system reliability and safety is paramount. These systems, characterized by their partial observability and intricate inter-agent interactions, pose significant challenges for fault management. This paper introduces a novel framework that leverages model checking and timed automata to address these challenges, providing a rigorous solution for fault management in MAS. We formalize the problems of fault identification, diagnosability, and tolerance within the context of partially observable MAS, laying a solid foundation for systematic analysis and interpretation. The framework utilizes timed automata to model the dynamic and non-deterministic nature of MAS, while model checking is employed to ensure comprehensive and rigorous fault management results. Through a case study on an Electric Power System, we demonstrate the framework's efficacy in providing both offline design insights and online decision support, enhancing the system's resilience to faults. The results underscore the framework's ability to navigate the complexities of partially observable MAS, offering clear and actionable guidance for fault mitigation. © 2024 Elsevier Ltd