Hybrid Manipulator Path Planning with Probabilistic Guidance and Adaptive Retraction

Abstract

As manipulators are increasingly used in intelligent manufacturing, agricultural harvesting, and medical applications, efficient motion planning in complex environments has become essential. However, high-degree-of-freedom planning in continuous joint space is often limited by high search complexity, inefficient exploration, and poor path quality. This paper proposes a hybrid planner that combines probability-driven A*-guided sampling with global RRT exploration, together with adaptive retraction and trajectory post-processing. Three representative scenarios are constructed in MuJoCo for evaluation. In the complex scenario, compared with conventional A*, the proposed method reduces planning time by about 54%, explored nodes by about 90%, joint-space path length by about 47%, task-space path length by about 49%, and terminal error by about 46%. The results demonstrate that the proposed method improves both search efficiency and path quality, providing an effective solution for manipulator motion planning in complex environments.