A Q-Learning-Based Artificial Bee Colony Algorithm for Distributed Three-Stage Assembly Scheduling with Factory Eligibility and Setup Times

Abstract

The assembly scheduling problem (ASP) and distributed assembly scheduling problem (DASP) have attracted much attention in recent years; however, the transportation stage is often neglected in previous works. Factory eligibility means that some products cannot be manufactured in all factories. Although it extensively exists in many real-life manufacturing processes, it is hardly considered. In this study, a distributed three-stage ASP with a DPm→1 layout, factory eligibility and setup times is studied, and a Q-learning-based artificial bee colony algorithm (QABC) is proposed to minimize total tardiness. To obtain high quality solutions, a Q-learning algorithm is implemented by using eight states based on population quality evaluation, eight actions defined by global search and neighborhood search, a new reward and an adaptive ε−greedy selection and applied to dynamically select the search operator; two employed bee swarms are obtained by population division, and an employed bee phase with an adaptive migration between them is added; a new scout phase based on a modified restart strategy is also presented. Extensive experiments are conducted. The computational results demonstrate that the new strategies of QABC are effective, and QABC is a competitive algorithm for the considered problem.