Intelligent approach for distributed flexible job-shop scheduling with deep reinforcement learning and quality-diversity optimization

In the field of intelligent manufacturing, integrated scheduling of automated guided vehicles (AGVs) and machines significantly impacts makespan and energy consumption. However, existing scheduling methods struggle to efficiently coordinate AGV transportation and task execution, especially in distributed flexible job-shop scheduling (DFJSP) scenarios, where this challenge is particularly pronounced. To address this issue, this paper proposes a deep reinforcement learning-enhanced quality-diversity (QD) optimization algorithm that effectively leverages transportation and machine behavioral features to generate high-quality and diverse Pareto-optimal solutions. First, a knowledge-assisted collaborative heuristic strategy is designed to optimize the scheduling of jobs, machines, and factories by considering AGV transportation characteristics, thereby improving overall solution quality. Second, to address the low utilization rate of search operators, an intelligent selection mechanism based on deep reinforcement learning is introduced to overcome the limitations of random selection, enhancing both search efficiency and optimization performance. Simulation experiments demonstrate that the proposed algorithm significantly outperforms existing methods in optimizing makespan and energy consumption, validating its effectiveness.