Multi-objective scheduling optimization for dynamic parallel batch job shops based on hybrid IPSO-DQN algorithm

The dynamic job shop scheduling problem is crucial in modern manufacturing industries, especially in parallel batch processing environments where achieving multi-objective optimization still faces great challenges. In this paper, we propose a hybrid scheduling algorithm that incorporates the Improved Particle Swarm Optimization (IPSO) algorithm and Deep Q-Network (DQN). IPSO enhances the global search capability and local refinement exploration by introducing adaptive inertia weights and chaotic perturbation strategies for generating initial scheduling plans and optimising the hyper-parameters of the DQN. The DQN, on the other hand, takes advantage of the deep reinforcement learning to construct an intelligent scheduling agent that can effectively handle high-dimensional state spaces and is responsible for making real-time decisions and adjustments in dynamic environments. This combination gives full play to the global search capability of IPSO and the learning adaptation capability of DQN, and is able to better handle the dynamic job shop scheduling problem in parallel batch processing environments while achieving multi-objective optimization. In addition, this paper develops a multi-objective optimization method based on migration learning, which achieves more flexible objective balancing through domain adaptation and dynamic weight adjustment. The results show that the hybrid IPSO-DQN algorithm proposed in this paper achieves significant improvements in several metrics, such as maximum completion time, total processing time, and energy consumption, compared with the traditional methods. In addition, the method demonstrates good generalisation ability in dynamic job shops of different sizes and complexities, providing a new solution for the intelligence and greening of the manufacturing industry.