A Computational Study of Genetic Crossover Operators for Multi-Objective Vehicle Routing Problem with Soft Time Windows

TL;DR

This study evaluates how different genetic crossover operators perform in solving multi-objective vehicle routing problems with soft time windows, considering problem structure and comparing with local search methods.

Contribution

It provides a systematic analysis of genetic crossover operators' effectiveness in MOCO for vehicle routing with soft time windows, highlighting the influence of problem structure.

Findings

Genetic algorithms' performance varies with problem structure.

Certain crossover operators outperform others depending on problem class.

Local search approaches are competitive with genetic algorithms.

Abstract

The article describes an investigation of the effectiveness of genetic algorithms for multi-objective combinatorial optimization (MOCO) by presenting an application for the vehicle routing problem with soft time windows. The work is motivated by the question, if and how the problem structure influences the effectiveness of different configurations of the genetic algorithm. Computational results are presented for different classes of vehicle routing problems, varying in their coverage with time windows, time window size, distribution and number of customers. The results are compared with a simple, but effective local search approach for multi-objective combinatorial optimization problems.