Based on the characteristic of autonomous underwater vehicle path planning, the method of path planning was analyzed by genetic algorithm. Firstly, by means of grid, plan space was modeled into two markers. Best path was searched by genetic algorithm. Method of giving birth to initial groups was improved. Sufficiency function of path planning was given. Chamfer operator in genetic algorithm was imported to speed up the convergence. It guarantees the request of planning time. Through simulation experiment, the important function of chamfer operator was proved in path planning. The method can produce an almost best path for Autonomous underwater vehicle in short time.