To　improve　the　computation　performance　of　Ant　Colony　System　for　solving　large　scale　Traveling　Salesman　Problems　(TSP),　a　new　algorithm　is　presented　in　this　paper.　It　adopts　a　set　of　multistage　strategies　to　look　for　an　optimal　solution.　The　analysis　for　computation　complexity　and　experimental　results　for　TSP　problems　demonstrate　that　the　proposed　algorithm　can　greatly　improve　the　speed　of　convergence　while　keeping　more　optimal　solutions.