Abstract: To address the issues of low path planning efficiency and insufficient safety in underground coal mines caused by complex environments and numerous dynamic obstacles, this study proposes a hierarchical collaborative path planning framework. In global planning, to address the shortcomings of the traditional Rapidly Expanding Random Tree（RRT） algorithm−namely, its high randomness and tortuous paths−this study introduces an artificial potential field strategy and designs a potential field-guided RRT algorithm（APF-RRT）. By using the potential field to guide random sampling directions, the algorithm reduces randomness and incorporates a dynamic step size adjustment mechanism, thereby generating smoother global paths while improving planning efficiency. In local planning, to address the poor adaptability of the Dynamic Window Approach（DWA） in complex environments due to its fixed weights, it employs a fuzzy-controlled adaptive dynamic weight adjustment strategy to enhance the safety of dynamic obstacle avoidance. Finally, an underground scenario is simulated and validated in MATLAB. The results demonstrate that the robot strictly follows the global path and, upon encountering dynamic obstacles, can promptly avoid them via the shortest detour path. This approach ensures both the guidance accuracy of the global path and the flexibility and efficiency of local obstacle avoidance, effectively enhancing the autonomous planning capabilities and operational safety of underground robots.