In grid map environments, path planning algorithms based on Voronoi diagrams offer good globality and completeness. However, the resulting paths often suffer from excessive turning points, significant redundant paths, poor followability, and low planning efficiency in dynamic obstacle environments. To address these shortcomings, this paper proposes a path planning algorithm (BV-R-GDWA) that integrates an improved Voronoi skeleton diagram with the dynamic windowing approach (DWA). This algorithm first utilizes key point extraction and topology reconstruction techniques from the Voronoi skeleton diagram, combined with an obstacle inflation model and inflection point screening mechanism, to replan the initial path, resulting in a shorter and smoother globally optimized path within safety distance constraints. During the local planning phase, this paper innovatively designs a dynamic weighted global path guidance function, enabling the robot to adaptively adjust its tracking strategy based on the deviation between its current position and the global path. Experimental results show that in simple environments, compared with the Voronoi skeleton graph method, the proposed global path algorithm reduces planning time, path length, and the number of turning points by 26.3%, 12.9%, and 27.3%, respectively. In complex dynamic environments, the BV-R-GDWA algorithm can still maintain high planning efficiency and path quality, showing good robustness and adaptability. The main innovation of this paper is the proposed key point extraction and dynamic weight guidance mechanism, which achieves an effective balance between global path safety and local obstacle avoidance real-time performance. It has important theoretical significance and engineering application value for improving the navigation performance of mobile robots in complex scenarios.