Aiming at the problem that DWA algorithm is easy to falls into the local optimal solution, which cannot reach the target point smoothly, or even collide with obstacles, an improved dynamic path planning method for unmanned patrol vehicles based on DWA algorithm is proposed. Based on the original evaluation function, a gap evaluation subfunction for the judgment of unknown obstacles is added to predict the motion trend of dynamic obstacles in advance, and a new driving path is planned to avoid obstacles, to improve the obstacle avoidance ability of unmanned patrol vehicles. Moreover, the DWA algorithm is based on the node of the global path as the moving direction, so that the local path can fit the global path. Avoid situations where you cannot reach the target point. The results show that compared with the traditional DWA algorithm, the improved DWA algorithm proposed in this paper can predict the motion trend of dynamic obstacles in advance, plan a new driving path to avoid dynamic obstacles and solve the problem that the DWA algorithm reaches the target point due to the existence of locally optimal solutions.