Abstract: A hierarchical path planning algorithm was proposed, considering the factors of terrain features, obstacles and vehicle motion characteristics, to address the problem of finding a safe, efficient and right of way for unmanned tracked vehicles in off-road terrain. Firstly, improving the node expansion mode of A * algorithm with the 3D reference path planning module, a smooth reference path was obtained based on the gradient free iterative smoothing algorithm and B-spline curve fitting. Then, using the adaptive sampling strategy, a candidate path set was generated in the multi-stage state space sampling module, and a greedy strategy was used to select the optimal path. The experimental results show that the proposed method can not only ensure feasibility, stability, and real-time performance in off-road environments, but also have advantages in stabilizing vehicle posture.