An Improved Hybrid Particle Swarm Optimization for Travel Salesman Problem

Abstract

Travel salesman problem (TSP) is a typical NP-complete problem, which is very hard to be optimized by traditional methods. In this paper, an improved hybrid particle swarm optimization (IHPSO) is proposed to solve the travel salesman problem (TSP). In IHPSO, there are four novel strategies proposed for improving its comprehensive performance. Firstly, a probability initialization is used to add prior knowledge into the initialization, so that much computing resources can save during the evolution of the algorithm. Furthermore, in order to improve the algorithm’s convergence accuracy and population diversity, two kinds of crossover are proposed to make better use of Gbest and Pbest. Lastly, a directional mutation is applied to overcome the randomness of the traditional mutation operator. Comparison results among IHPSO and other EAs, including PSO, genetic algorithm (GA), Tabu Search (TB), and simulated annealing algorithm (SA) on the standard TSPLIB format manifest that IHPSO is more superior to other methods, especially on large scale TSP.