To solve Multiple Depot Vehicle Routing Problem (MDVRP) directly, ant’s transfer policy and method to construct valid solution were designed. When an ant began transfer, a temporary depot was prearranged to the ant, then the permanent depot was rooted when a depot was met with. After paths were constructed, several ants’ paths were selected randomly with the restriction of vehicles satisfied, and iterative depots were deleted and absent depots were inserted based on the regulars of “least decrease”，“least increase” and “nearest insertion”.Thus a solution of MDVRP was constructed. In order to improved algorithm’s performance, K neighbor was used to confine ants’ transfer objects, and ants’ paths and solutions were optimized by 2 Opt,and pheromone update policy was designed. The algorithm was tested on two standard data sets, and the results showed that the algorithm is an efficient one to solve MDVRP.
