Self-Adaptive Ant Colony System for the Traveling Salesman Problem

In the ant colony system (ACS) algorithm, ants build tours mainly depending on the pheromone information on edges. The parameter settings of pheromone updating in ACS have direct effect on the performance of the algorithm. However, it is a difficult task to choose the proper pheromone decay parameters alpha and rho for ACS. This paper presents a novel version of ACS algorithm for obtaining self-adaptive parameters control in pheromone updating rules. The proposed adaptive ACS (AACS) algorithm employs Average Tour Similarity (ATS) as an indicator of the optimization state in the ACS. Instead of using fixed values of alpha and rho, the values of alpha and rho are adaptively adjusted according to the normalized value of ATS. The AACS algorithm has been applied to optimize several benchmark TSP instances. The solution quality and the convergence rate are favorably compared with the ACS using fixed values of alpha and rho. Experimental results confirm that our proposed method is effective and outperforms the conventional ACS.