Ensemble artificial bee colony algorithm with Q-learning for scheduling Bi-objective disassembly line

Abstract

This study addresses a bi-objective disassembly line scheduling problem (Bi-DLSP), considering interference relationships among tasks. The objectives are to optimize the total disassembly time and the smoothing index simultaneously. First, we propose a mathematical model for the concerned problems. Second, improved artificial bee colony (ABC) algorithms are developed to solve the Bi-DLSP, and seven different local search operators are created to strengthen the performance of the ABC algorithms. Third, to further enhance the improved ABC algorithms, we design two Q-learning-based strategies for selecting high-quality local search operators and integrate them into the ABC algorithm during iterations. Finally, we evaluate the effectiveness of the proposed strategies by comparing the classical ABC algorithm, its variants, and two classical multi-objective algorithms for solving 21 instances. We validate the proposed model using the Gurobi solver and compare its results and time efficiency with the proposed algorithms. The experimental results show that the proposed ABC algorithm based on Q-learning (ABC_QL1) performs the best in solving related problems. This study provides a new approach for solving the Bi-DLSP and demonstrates the effectiveness and competitiveness of our method, providing useful insights for research and applications in related fields.