An improved ant colony algorithm for USV(unmanned surface vehicle) path planning is presented. A local tabu list and a global tabu list with different effect duration are set up to catalog the ants passed grids.Various deadlock processing strategies are introduced upon barrier deadlock and self deadlock situation in order to reduce the number of invalid ants and to improve the diversity of solutions. An adaptive heuristic function is designed by adopting the Euclidean distance between the ant and the destination to avoid the initial blindness and later singleness of ant path searching. The current worst path would be superseded by the historical best path when appropriate to retain the previous effort. Simulation results under different maps show that the improved algorithm considerably increases the number of effecive ants during the searhing process and the probability to find the optimal path, as well as the search speed. The improved algorithm performs even better in larger and more complex grid maps.