A Fault-Tolerant Scheme of Multi-agent System for Worker Agents

The paper introduces a fault-tolerant scheme of MAS (Multi-Agent System) for worker agents using the checkpointing and rollback recovery mechanism. To discuss the fault-tolerance of working agents in MAS, we consider the extended MAS model based on task delegation and observation which can be independent of node failures. In the proposed MAS, to preserve global consistency, the facilitators maintain a task plan in a stable storage by using checkpoints taken when either completing a task plan or receiving a response of a subtask within the task plan. In this paper, we present a fault-tolerant scheme which takes over blocked worker agent problem of MAS by using communication-induced checkpointing.