Research on the Application Model of Reinforcement Learning in Mechanical Automation Production Line Scheduling

This paper proposes a production line scheduling optimization model based on DRL, aiming to improve production efficiency, shorten production cycle and optimize resource allocation through reinforcement learning algorithm. The model models the production line scheduling problem as MDP and realizes intelligent scheduling decision of production tasks through DQN. A dynamic reward mechanism is introduced, which is used to adjust the scheduling decisio-nmaking based on the real time feedback. Moreover, PER technique is adopted to improve the learning efficiency of this model. The results show that the proposed model is better than the conventional GA and HEURISTIC in the production period, the time to complete the task and the resource utilization. The scheduling system using deep reinforcement learning can make scheduling decisions more quickly and accurately in a complex production environment, improve production efficiency and reduce production time. In the experimental setting, through the scheduling of different production tasks, the production cycle of the reinforcement learning model is shortened by an average of 18%, the task completion time is reduced by 12 %, and the resource utilization is improved by 15 %. These results show that deep reinforcement learning provides an effective solution to solve complex multi-task scheduling problems that traditional scheduling methods cannot cope with.