When traditional rapidly-exploring random tree (RRT) algorithm is used in adaptive path planning of unmanned vehicle,unmanned ship,the branches growing randomness causes lack of goal orientation.To overcome the problem,a hybrid algorithm combining improved RRT with the dynamic window was proposed.Considering that the large amounts of redundant points in the random tree growth process make the path twist and turn,an algorithm for filtering redundant nodes was studied based on the selection strategy of subtarget punctuation.In order to meet the safety of unmanned vehicle and unmanned ship,a safety threshold distance model was established to avoid collision with obstacles.Finally,the proposed method was simulated.The results show that the improved RRT algorithm can be used to generate effective feasible path in complex dynamic environment; compared with traditional RRT algorithm,the search efficiency increases 43%,the planning path is smooth.It can achieve a on-line real-time path planning in dynamic environment for unmanned vehicles and unmanned ship.