Decomposition Method for Multi-Objective Optimization of Batch Order Processing

Abstract

The growth of e-commerce, coupled with the increasing demand for sustainable logistics practices, has heightened the requirements for efficiency and quality in order management. In this context, and with the aim of analyzing the most suitable approaches to the order batching and picking problem, a variant of the JOBPRP is introduced, which seeks to optimize two criteria: the reduction of operational costs and the balancing of workload distribution. Evolutionary algorithms represent a viable alternative for multi-objective optimization; however, they may exhibit challenges in convergence and diversity when dealing with Pareto fronts of irregular structure. Therefore, the performance of the Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D) was evaluated. A comparative study was conducted to analyze different scalarization strategies applied to a large set of experiments involving instances of varying size. The algorithm’s performance was assessed using metrics such as hypervolume, average distance to the ideal solution, and the spread of non-dominated solutions. The results suggest that MOEA/D, when employed with the AASF method, exhibits competitive performance in terms of average hypervolume and the dispersion of solutions along the Pareto fronts.