The application of actor-critic reinforcement learning for fab dispatching scheduling

Abstract

This paper applies Actor-Critic reinforcement learning to control lot dispatching scheduling in reentrant line manufacture model. To minimize the Work-In-Process(WIP) and Cycle Time(CT), the lot dispatching policy is directly optimized through Actor-Critic algorithm. The results show that the optimized dispatching policy yields smaller average WIP and CT than traditional dispatching policy such as Shortest Processing Time, Latest-Step-First-Served, and Least-Work-Next-Queue.