Algoritmo evolutivo multiobjetivo basado en descomposición para la optimización del procesamiento por lotes de pedidos

Abstract

The demand for sustainable logistics practices, coupled with the rise of e-commerce, has led to greater requirements for efficiency and quality in order processing. Within this framework, and with the aim of studying the most suitable methods to address the problem of order grouping and preparation, a variant of the JOBPRP is presented with two objectives: operational costs and balanced workload distribution. In this context, evolutionary algorithms are strong alternatives for multi-objective search, yet they may face challenges related to convergence or diversity when dealing with irregular Pareto fronts. Therefore, the performance of the Multi-Objective Evolutionary Algorithm Based on Decomposition (MOEA/D) was studied. A comparative analysis of its performance was conducted using different scalarization methods across an extensive set of experimental tests applied to instances of various sizes of the problem under consideration. Performance indicators such as hypervolume, the average distance to the ideal solution, and the dispersion of non-dominated solutions were used. The results indicate that the MOEA/D based on the AASF method demonstrates strong performance in terms of average hypervolumes and solution dispersion across the fronts.