Adaptive and automated fault-tolerance for multi-agent systems

摘要

Fault tolerance in MASs is usually implemented with dynamic replication strategy, which demands real time replication of all agents. This adds to unnecessary replicas, more cost and also increases system complexity. Therefore the challenge of balanced replication is still prevailing. The major thrust of this work is to overcome the mentioned challenge and hence provide an equilibrium between active and passive replication of agents. The proposed framework relies on the fact that only critical agents need to be replicated, wherein criticality of an agent is assessed on the basis of its cardinality. The proposed framework has been implemented in JADE and provides promising results. The solution proposed adapts well to the transient environment situations of the agents and provides fault tolerance by making optimum utilization of the system resources.