Cooperative coevolutionary algorithms for fuzzy vehicular routing problem: An analysis of efficiency vs. geographical distribution

Abstract

In this article, a cooperative coevolutionary genetic algorithm for the solution of fuzzy vehicular routing problem (FVRP) is presented. FVRP is a variant of VRP with time windows; based on additional use of fuzzy due-times for customers' preferences. The objectives of FVRP are the minimization of total number of vehicles in service, total travel distance, and the total waiting time over all vehicles; while maximizing the average grade of customer satisfaction. Benchmark problem instances classified by geographical distribution of customers are used in experimental studies and results of cooperative coevolutionary GAs are compared with those of panmictic GAs. It has been observed that cooperative coevolutionary approach generates significantly improved solutions for multi-cluster tightly grouped problem instances, while it also performs equally well for other problem classes.