Optimal Design of Flexible Job Shop Scheduling Under Resource Preemption Based on Deep Reinforcement Learning

Abstract

With the popularization of multi-variety and small-batch production patterns, the flexible job shop scheduling problem (FJSSP) has been widely studied. The sharing of processing resources by multiple machines frequently occurs due to space constraints in a flexible shop, which results in resource preemption for processing workpieces. Resource preemption complicates the constraints of scheduling problems that are otherwise difficult to solve. In this paper, the flexible job shop scheduling problem under the process resource preemption scenario is modeled, and a two-layer rule scheduling algorithm based on deep reinforcement learning is proposed to achieve the goal of minimum scheduling time. The simulation experiments compare our scheduling algorithm with two traditional metaheuristic optimization algorithms among different processing resource distribution scenarios in static scheduling environment. The results suggest that the two-layer rule scheduling algorithm based on deep reinforcement learning is more effective than the meta-heuristic algorithm in the application of processing resource preemption scenarios. Ablation experiments, generalization, and dynamic experiments are performed to demonstrate the excellent performance of our method for FJSSP under resource preemption.