Agent-based flow-shop modelling in dynamic environment

This article studies a multiple-objective flow-shop modelling and scheduling problem by multi-agent system (MAS) in a production context characterized by diversified, high-volume production mix. The analysed flow shop is characterised by multi-machine workstations, transfer batches, sequence-dependent setup times and possible re-entrant jobs. The agent-based model is structured from the current flow-shop configuration, defining the main entities and related events to realize the most possible flexible model. After a description of entities and their states, the agent framework with state diagrams is implemented through Jade platform. The system is provided with a simulation-based environment to test it in two main production scenarios including re-entrant jobs and failures. A coordination mechanism between agents and a dedicated scheduling algorithm managed by the MAS allowed to front this kind of events optimising concurrent objectives like work in process, Makespan and buffer queues.