Improved meta-heuristics with Q-learning for solving distributed assembly permutation flowshop scheduling problems

This study addresses a distributed assembly permutation flowshop scheduling problem, which is of great sig-nificance in practical manufacturing systems. We aim to sequence products and jobs, and assign jobs to the appropriate factory to minimize the total flowtime of products. First, a mathematical model is developed to describe the concerned problems. Then, four meta-heuristics, e.g., artificial bee colony, particle swarm optimi-zation, genetic algorithm, and Jaya algorithm, and their variants are proposed. Three initialization strategies are developed to generate high-quality initial solutions. Four local search operators are designed to improve the performance of the algorithms. Q-learning is embedded to select the premium local search strategy during it-erations. Based on 81 large-scale benchmark instances, comprehensive numerical experiments are carried out to evaluate the effectiveness of the proposed algorithms. The experimental results show that the proposed Jaya with Q-learning-based local search has strong competitiveness, and it updates optimal solutions for 51 out of 81 benchmark instances.