Abstract: Aiming at the application of unmanned ground vehicle (UGV), a local path planning algorithm based on multilayer Morphin search tree is proposed. By considering the nonholonomic constraints of vehicle, a multilayer Morphin search tree is constructed. Then control of dynamic behavior is learned by fuzzy Q-learning, as a basis for the evaluation of search tree to get a smooth and trackable trajectory with nonholonomic constraints of vehicle. This algorithm overcomes the inflexibility of Morphin algorithm in trajectory searching. The simulation results and the test on a real vehicle verify correctness and effectiveness of the algorithm.