Abstract
Traveling salesman problems (TSPs) are well studied as combinatorial optimization problems. In this paper we propose a multiple heuristic search algorithm for solving the TSPs. Three kinds of TSP instances are used in our experiments: double concentric circle, fractal and random instances. The double concentric circle instances were set with tighter conditions compared with conventional experiments and the city numbers range from 48 to 14,400. Firstly, the algorithm was applied to the double concentric circle instances and it shows that the search ability of the proposed algorithm is robust with the increase of the city numbers in the double concentric circle instances. Furthermore, the algorithm was applied to the fractal and random instances. A considerable result was presented in fractal instances. The results of our experiment show that the proposed algorithm is extremely more effective for the double concentric circle and fractal instances which have the self-similarity than for the random instances.
