A Learning-Based Multipopulation Evolutionary Optimization for Flexible Job Shop Scheduling Problem With Finite Transportation Resources

• Published in: IEEE transactions on evolutionary computation Vol. 27; no. 6; pp. 1590 - 1603
• Main Authors: Pan, Zixiao, Wang, Ling, Zheng, Jie, Chen, Jing-Fang, Wang, Xing
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.12.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Automated guided vehicles; Cooperation; Evolutionary algorithms; Evolutionary computation; Finite transportation resource; flexible job shop scheduling; Job shop scheduling; Job shops; multipopulation; Optimization; reinforcement learning (RL); Robustness (mathematics); Sociology; statistical learning; Statistics; Task analysis; Transportation
• DOI: 10.1109/TEVC.2022.3219238
• ISSN: 1089-778X

In many practical manufacturing systems, transportation equipment such as automated guided vehicles (AGVs) is widely adopted to transfer jobs and realize the collaboration of different machines, but is often ignored in current researches. In this article, we address the flexible job shop scheduling problem with finite transportation resources (FJSP-Ts). Considering the difficulties caused by the introduction of transportation and the NP-hard nature, the evolutionary algorithm (EA) is adopted as a solution approach. To this end, a learning-based multipopulation evolutionary optimization (LMEO) is proposed to deal with the FJSP-T. First, the multipopulation strategy is introduced and a cooperation-based initialization is designed by combining several heuristics to guarantee the quality and diversity of the initial population. Second, a reinforcement learning (RL)-based mating selection is proposed to realize the cooperation of different subpopulations by selecting appropriate individuals for evolutionary search. Then, a specific local search inspired by the problem properties is designed to enhance the exploitation capability of the LMEO. Moreover, a statistical learning-based replacement is designed to maintain the quality and diversity of the population. Extensive experiments are conducted to test the performances of the LMEO. The statistical comparison shows that the LMEO is superior to the state-of-the-art algorithms in solving the FJSP-T in terms of solution quality and robustness.