Abstract: Based on the characteristics of urban distribution in the TSP problem, an ant colony algorithm is put forth for local clustering. First, urban distribution is treated by means of spherical clustering. Subsequently, the remaining cities are subjected to the treatment of linear clustering and isolated points clustering respectively. The ant colony algorithm for local clustering is characterized by a faster velocity of convergence and a better sought solution as well. Experimental results show that, by use of such an algorithm, the solution thus worked out is not only better but also the speed can be 3-13 times faster than by means of the basic ant colony algorithm.