Optimization Study of Intelligent Robot Production Line Scheduling Problem Based on Markov Decision Process

Abstract

With the rapid development of science and technology, industrial technology has continued to advance and improve. Addressing the issue that traditional Markov prediction methods often yield results that deviate from reality, this paper proposes an optimization solution method based on reinforcement learning. It employs a deep deterministic action strategy to compute an infinite-dimensional semi-Markov queuing model (where the queuing system control time is positive infinity), thereby achieving the objective of model optimization. Taking the parts production task O = {2, 12, 19} as an example, the effectiveness of the proposed method is validated by verifying the mold scheduling process solved using the reinforcement learning algorithm. Research shows that the average time spent on each stage of part production using the pre-optimization scheduling strategy was 27.90167, while the average time after optimization was 12.6533. Therefore, applying the reinforcement learning-based optimization method significantly improves the operational efficiency of intelligent robot production line scheduling and ensures reliability.