Lightweight bi-level optimization algorithm for synchronized scheduling of production and transportation in a reconfigurable flexible job shop

In reconfigurable flexible job shops, jobs have flexible processing routes and machine configurations change dynamically. This makes the transportation scheduling of Work-in-Progress (WIP) between machines highly complex and crucial to the production process. Coordinating production and transportation can significantly boost workshop efficiency. Thus, the synchronization of production and transportation scheduling is addressed. A bi-level scheduling model is developed. The upper-level minimizes production costs in production scheduling, considering flexibility and reconfigurability. The lower-level minimizes transportation costs in transportation scheduling, considering speed-adjustable Automated Guided Vehicles (AGVs). To solve this complex problem efficiently, a lightweight bi-level optimization algorithm is designed. In it, an accurate surrogate model and an improved metaheuristic are performed sequentially to achieve the lightweight evaluation and high-fidelity evaluation of the lower-level optima. Ten rules that contain problem-related knowledge are discovered by rule mining technologies including gene expression programming and Q-learning. Since these rules can better reflect problem characteristics, rule-based features are extracted to improve the accuracy of the surrogate model. Experimental results show that all discovered rules, especially the dynamic adaptive rule, are highly effective in generating high-performance solutions. The rule-based surrogate model can quickly and accurately estimate the lower-level optima. By incorporating this surrogate model, the proposed lightweight algorithm cuts down on computing budget without sacrificing accuracy, outperforming other bi-level optimization algorithms.