Abstract :
In this paper, the sub-ant-colony scheme is revisited to improve the solution quality of the ant colony system. In the scheme, an original ant colony is divided into several sub-antcolonies, and each sub-ant-colony individually works for its local optimal solution and the best solution among them is output as the final one. The number of ants in a total ant colony is the same as the sum of the ant numbers in individual sub-ant-colonies. Experiments with the TSP data are also conducted to show the approach in general performs better than the original ant colony system. The trade-off between local search and diversity for the scheme is discussed as well.
