Modified Ant Colony Optimization Algorithm for Multiple-vehicle Traveling Salesman Problems

Yindee Oonsrikaw,Arit Thammano

doi:10.2991/ijndc.1970.1.7.4

Abstract

In this work, we extended the original Traveling Salesman Problem (TSP) to cover not only the case of multiple vehicles but also to constrain the minimum and maximum numbers of cities each vehicle can visit. Our algorithm is a modified Ant Colony Optimization (ACO) algorithm which has the ability to avoid local optima; our algorithm can be applied to transportation problem that covers either a single vehicle or multiple vehicles. To the original ACO, we added a new reproduction method, a new pheromone updating strategy, and four improved local search strategies. We tested our algorithm on several standard datasets in the TSP library. Its single-vehicle performance was compared to that of ant system (AS) and elitist ant system (EAS) algorithms. Its multiple-vehicle performance was evaluated against that of ant colony system variants reported in the literature. The experiments show that our proposed ACO’s single-vehicle performance was superior to that of AS and EAS on every tested dataset and its multiple-vehicle performance was excellent.

Highlights

Business and industry sectors are giving increasingly higher priority to transportation and distribution of goods because the oil price which determines a transportation cost is ever increasing
We measured the performance of our algorithm as well as other major algorithms on Traveling Salesman Problem (TSP) data from TSP Library (TSPLIB) [19]
For the single- vehicle TSP, the results of our algorithm were compared with those achieved by the ant system algorithm and elitist ant system (EAS)

Summary

Introduction

Business and industry sectors are giving increasingly higher priority to transportation and distribution of goods because the oil price which determines a transportation cost is ever increasing. Meta-heuristic algorithms have been used successfully by manufacturers to reduce transport costs and enhance the business. Algorithms of this type generally find a satisfactory solution to problems, which are inherently nondeterministic polynomial-time hard (NP-hard), in an acceptably short time [1,2]. Meta-heuristic algorithms have been applied to Traveling Salesman Problem (TSP), a classic NP-hard problem that tries to find the shortest tour that a salesman can take to visit all of his customer sites in a single tour, with starting and end points at the same city.

Objectives

Results

Conclusion