A Q-learning artificial bee colony for distributed assembly flow shop scheduling with factory eligibility, transportation capacity and setup time

Distributed assembly flow shop scheduling problem (DAFSP) has been considered; however, DAFSP with factory eligibility,transportation capacity and setup time is seldom studied even though these three constraints often exist simultaneously in real-life multi-factory assembly production processes. In this study, DAFSP with Pm-+ 1 layout, factory eligibility, transportation capacity and setup time is proposed and a Q-learning artificial bee colony (Q-LABC) is presented to minimize both makespan and total tardiness. A heuristic is designed to produce initial solution. In employed bee phase, factory set based heuristic is developed, population is divided into two parts by non-dominated sorting and two global search operators are applied in two parts respectively. A Q-learning algorithm has a fixed Q-table when a given condition is met, and is used to dynamically decide the way of selecting food source and the number of the selected food source for onlooker bee. Extensive experiments on 90 instances are conducted to show the effectiveness of new strategies and compare Q-LABC with two comparative algorithms. The computational results demonstrate that the new strategies of Q-LABC are effective and Q-LABC performs better than its comparative algorithms on at least 83.33% and 94.44% instances, respectively.

Automated summary

This study proposes a solution for the distributed assembly flow shop scheduling problem (DAFSP) with Pm-+ 1 layout, factory eligibility, transportation capacity, and setup time. The Q-learning artificial bee colony (Q-LABC) algorithm is introduced to minimize makespan and total tardiness. Experimental results demonstrate the effectiveness of the new strategies and show that Q-LABC outperforms comparative algorithms on at least 83.33% and 94.44% instances, respectively.