WCSE

Reconfigurable Flexible Assembly Line Balancing Problem Based on Reinforcement Learning

Abstract— Because of the variability of models, constraints and objectives of reconfigurable flexible assembly line, there are few methods specially applied to reconfigurable flexible assembly line. The universality and learning ability of reinforcement learning give it the potential to solve such problems. This paper focuses on the number of assembly workstations required to form an assembly line and the cycle time of the assembly line to complete the task, and takes the average assembly workstation rate as the optimization goal. The paper attempts to solve this problem through reinforcement learning, which is rare in this field, and experiments are carried out with some cases. The results show that reinforcement learning is feasible to improve the work efficiency of reconfigurable flexible assembly line.

Index Terms— reconfigurable flexible assembly line, maximize average workstation rate, reinforcement learning

Jingzhao Gan
Intelligent Manufacturing and Machine Vision Center, Tsinghua University, China
Long Zeng
Intelligent Manufacturing and Machine Vision Center, Tsinghua University, China
Fengyuan Shi
Intelligent Manufacturing and Machine Vision Center, Tsinghua University, China

[Download]

Cite: Jingzhao Gan, Long Zeng, Fengyuan Shi, " Reconfigurable Flexible Assembly Line Balancing Problem Based on Reinforcement Learning, " WCSE 2022 Spring Event: 2022 9th International Conference on Industrial Engineering and Applications, pp. 968-979, Sanya, China, April 15-18, 2022.