A mutli-objective artificial electric field algorithm with reinforcement learning for milk-run assembly line feeding and scheduling problem

Abstract

In automotive mixed-model assembly lines (MMALs), a large number of different parts need to be supplied to the assembly lines on time, which poses significant logistical challenges for manufacturers. However, consistently supplying parts for MMALs is a very complex issue due to factors such as diverse component requirements and logistical coordination in the supply chain. In this paper, we propose a bi-objective optimization problem to minimize the line-side inventory and energy consumption in a milk-run material distribution system. Meanwhile, the number of Kanban and the capacity of the material bin that affect the scheduling is jointly optimized, so that the material distribution scheduling plan is optimized. Considering the character of the problem, a multi-objective artificial electric field algorithm with SARSA mechanism (MOAEFASA) is developed to solve the problem. The algorithm proposed combines the merits of the artificial electric field algorithm (AEFA) and the framework of the non-dominated sorting genetic algorithm (NSGA-II). In addition, several optimization strategies are used to optimize the performance of the algorithm. Finally, the validity of the mathematical model is verified through the Epsilon constraint method and the superiority of the MOAEFASA is illustrated by numerical experiments with four outstanding meta-heuristics.