The application of deep reinforcement learning in flexible job shop scheduling

This paper discusses deep reinforcement learning in the context of application in flexible job shop scheduling. With the purpose to break the limitations in conventional methods in dynamic conditions, a scheduling optimization strategy based on deep Q network (DQN) is proposed. The efficient task allocation and resource allocation are achieved through the design of the state and action space, constructing the DQN model and the integration with the experience replay and target network updating process. The experimental results show that the approach outperforms traditional genetic algorithm in average completion time, machine utilization and computational efficiency and demonstrates the applicability and validity of deep reinforcement learning in flexible job shop scheduling and provides new ideas to dynamic optimization problems in intelligent manufacturing.