An effective deep actor-critic reinforcement learning method for solving the flexible job shop scheduling problem.

The flexible job shop scheduling problem (FJSP) is a classic NP-hard problem, and the quality of its scheduling solution directly affects the operational efficiency of the manufacturing system. However, the traditional scheduling algorithms suffer from poor generalization when solving FJSP; there are problems such as long computational time and dimensional disasters, especially as the scale of FJSP increases. Therefore, an effective deep actor-critic reinforcement learning (DACRL) method is proposed for solving FJSP. Firstly, the FJSP is modeled as a multi-agent Markov decision process (MMDP), the state space, action space, and reward function in the MMDP are designed. Secondly, a DACRL model is constructed to solve FJSP. The actor network is responsible for choosing the most suitable scheduling rule in different states, while the critic network is responsible for outputting the value function of the actions and providing feedback to the actor network to better adjust the scheduling strategy. Finally, the proposed DACRL method is validated on benchmark FJSP instances of different scales. The experimental results show that the proposed method significantly outperforms the heuristic scheduling rules and double deep Q-network (DDQN) in terms of solution quality. Compared with the meta-heuristic algorithms and the self-learning genetic algorithm (SLGA), the proposed method has higher solution efficiency with the same solution quality. [ABSTRACT FROM AUTHOR]