Ant Colony Optimization for Solving Electric Vehicle Traveling Salesman Problems

Abstract

Electric Vehicles (EVs) have become very popular nowadays due to their affordability, lower operating cost and maintenance cost, and environmental benefits. In this paper, we study an electric vehicle variant of the traveling salesman problem (EV-TSP) by using a modified ant colony optimization (ACO) method. Unlike the traditional traveling salesman problem (TSP), the electric vehicle variant of the TSP is more complex due to the limited battery range and the requirement of recharging along the trip. To tackle these difficulties, our proposed algorithm incorporates the battery level constraint as a heuristic information for choosing the next visiting node. We use a sigmoid function to capture the probability of node selection in addition to the pheromone value and visibility of ant. The type of nodes is also considered during the node selection. When the remaining battery level of the vehicle is low, the charging station nodes will have higher probability of being selected compared to the customer nodes, and vice versa. The algorithm was implemented in C++ and tested by using a number of benchmark instances from TSPLIB. The problem instances were modified by adding vehicle range and charging stations with their associated locations. Computational results were reported to show the effectiveness of the algorithm.