A novel Q-learning based variable neighborhood iterative search algorithm for solving disassembly line scheduling problems

Abstract

This paper addresses disassembly line scheduling problems (DLSP) to minimize the smoothing index with the workstation number threshold. First, a mathematical model is developed to formulate the concerned problems. Second, seven novel neighborhood structures are designed based on the feature of the DLSP and the corresponding local search operators are designed. Third, a novel Q-Learning based variable neighborhood iterative search (Q-VNIS) algorithm is first proposed to solve the DLSP. Q-learning is employed to select the premium local search operator in each iteration. Finally, the effectiveness of Q-learning in the proposed Q-VNIS is verified. To test the performance of the proposed Q-VNIS, 20 cases with different scales are solved and the Friedman test is executed. The experimental results and discussions show that the proposed Q-VNIS competes strongly for solving the DLSP.