Differential Evolution Algorithm for Generalized Multi-Depot Vehicle Routing Problem with Pickup and Delivery Requests

Abstract

This paper presents a Differential Evolution (DE) algorithm for solving generalized multi-depot vehicle routing problem with pickup and delivery requests (GVRP-MDPDR). The GVRP-MDPDR does not require the restricted assumptions of CVRP, VRPSPD, etc. and it contains nearly all characteristics of real world vehicle routing problems. The solution is represented as a multidimensional vector where each dimension is filled with random number and a population of vectors is evolved via the mechanism of differential evolution. A decoding scheme (SD1) is applied to decode the vector into priority of requests and construct the routes of vehicles under the restricted constraints. Five groups of test problem instances, A, B, C, D, and E, with differences geographical data and number of requests are used to evaluate the performance of the algorithm. Each group of instance composes of three different location scenarios of requests: clustered (c), randomly distributed (r), and half-random-half-clustered (rc). The computational results demonstrated that DE algorithm is very competitive when compared to the results obtained by using Particle Swarm Optimization (PSO).