A novel resilient scheduling method based on multi-agent system for flexible job shops

Job shops face heightened uncertainty during the production process due to unforeseen disruptions, emphasising stronger system resilience. To avoid production interruptions caused by both internal and external disturbances, there is a growing demand for resilient systems that can recover quickly and maintain production continuity. This paper proposes a decentralised multi-agent resilient scheduling method with hierarchically distributed multi-agents at three hierarchy structure levels in a flexible job shop to enable each job-scheduling agent capable of responding to disturbances locally in its action space. In this distributed hierarchical way, disturbances are collaboratively dealt by agents at different levels, reducing their impact (ranges) and enhancing optimisation efficiency through collaborative decision-making among agents. Uniquely, each agent is designed and equipped with inventory buffers, excess capacity buffers or lead time buffers, enabling dynamic adjustments to mitigate the effects of disturbances. Distributed decision-making rules are developed to allow agents to make decisions based on local information. Additionally, a multi-agent adaptive collaborative decision-making mechanism is designed. This enables agents to dynamically adjust their strategies in response to emerging disturbances, enhancing the system resilience and optimisation capabilities. Experimental results shows that the proposed method improve resilience by an average of more than 26.7% compared to existing methods.