A Practical Deep Reinforcement Learning Approach to Semiconductor Equipment Scheduling

Abstract

The efficiency of utilizing semiconductor equipment is critical to maximizing profits. The design work of a semiconductor equipment scheduler becomes a difficult task because it requires efficient operation in various situations. In this paper, we propose an approach based on deep reinforcement learning to overcome the difficulties of scheduling. This new approach designs a scheduler that controls the wafer transport robot inside the equipment. A deep neural network applied with a Q-network is used to calculate the benefit of the robot's motion under various conditions. The experimental results show the feasibility of applying deep reinforcement learning to the equipment scheduler. It also shows that pre-trained models can increase productivity by further learning in a variety of production environments.