The Study of Neighborhood Structure of Tabu Search Algorithm for Traveling Salesman Problem

Abstract

Tabu search (TS) algorithm is a powerful local search method. It has been successfully used in many discrete optimization problems, such as TSP, JSP, and QAP, etc. Neighborhood structure and size are key factors for a local search algorithm to get good performance. If hill climbing strategy is used, the bigger the size of a neighborhood is, the better its performance is in the cost of more computing time. Using the basic inversion and inserting move for TSP problem, this paper constructs a kind of linked neighborhood structure which uses the information get from previous move. Experiments were taken on some of the TSP instances from TSPLIB to compare the performance of different neighborhood structures. The simulation results show that the linked neighborhood structure has better performance.