Formal Model of Collaborative Fault Tolerant Planning in Multi-Robotic Systems

摘要

Multi-robotic systems are typical examples of com-plex multi-agent systems. The robots - autonomous agents - should cooperate with each other while planning and executing tasks required for achieving various missions. While designing multi-robotic systems, especially for such critical applications as surveillance and rescue operations, we should ensure that the missions could be successfully completed despite robot failures. Therefore, guaranteeing system fault tolerance constitutes an important design goal. However, achieving fault tolerance for multi-agent systems is a notoriously difficult task due to highly decentralized control and asynchronous communication between the agents. In this paper, we propose a formalization of fault tolerance behavior of cooperative mission planning for the multi-robotic system. Our model incorporates such essential fault tolerance mechanisms as the cooperative error recovery and dynamic reconfiguration. To demonstrate how to specify and verify essential fault tolerance requirements, we propose several dedicated mechanisms for knowledge sharing between the robots, which enable cooperative planning. We define the mission as a set of goals that should be accomplished by a collaborative functioning of the robots and derive the conditions for enabling fault tolerant planning and coordinated execution.