Evolutionary Algorithm Based on Partition Crossover (EAPX) for the Vehicle Routing Problem

Abstract

Problems associated with seeking the lowest cost vehicle routes to deliver demand to customers are called Vehicle Routing Problems (VRPs). Over the last decades, increasing research efforts are being dedicated to handle the VRPs. Most of the solution approaches have been metaheuristics, such as the evolutionary algorithms (EAs). This paper proposes a new EA (EAPX) based on Partition Crossover (PX), a recombination operator proposed by Whitley et al., which demonstrated very good performance in solving the Traveling Salesman Problem (TSP). PX strength lies in the characteristic of tunneling between local optima: if the parents are both local optima, with a high probability PX will generate two local optima offspring. Experimentations show that EAPX is competitive with the existing solution approaches.