Abstract
In Ant Colony Optimization (ACO), it is important to diversify the search so as not to fall into local optima. In this paper, the edges that are not used in the global optimal tour but used only in the local optima are defined as local optima's edges in the traveling salesman problems. We clarify the difference between the local optima's edges and the edges in the global optimum from the viewpoint of the change of the amount of the pheromone. In addition, multiple colonies are generated to search for various solutions in our proposed method. Each colony prohibits ants from using the local optima's edges. We compared the search performance between our proposed method and the conventional method, and showed the effectiveness of our proposed method.
