To address the issues of low search efficiency， poor path smoothness， and limited local obstacle avoidance in traditional A* algorithms within complex 3D environments， this paper proposed a hierarchical fusion algorithm combining A* and DWA. The algorithm first extended the A* algorithm into 3D space using a 26-neighbor node search strategy. It introduced a dynamic adjustment term into the cost evaluation function to achieve adaptive weighting， and employed the Douglas-Peucker algorithm combined with cubic B-spline curves to smooth paths. Second， it integrated a 3D-extended DWA algorithm to compensate for A*'s local obstacle avoidance limitations. A 3D dynamic window model was constructed via kinematic decoupling， and a cosine similarity measure was introduced to enhance the evaluation function， thereby improving real-time obstacle avoidance performance. Finally， a dynamic feedback mechanism was designed to enable adaptive correction of the global path， forming a closed-loop optimization system： “A* global planning -DWA local obstacle avoidance-Dynamic feedback.” Simulation results demonstrated that in both static and dynamic 3D environments， the A*-DWA hierarchical fusion algorithm significantly outperformed other comparison algorithms in path length， planning time， and path smoothness. Obstacle avoidance success rates exceeded 90% across multiple scenarios， validating the effectiveness of the A*-DWA hierarchical fusion algorithm.