To address the issues of the traditional A* algorithm, such as insufficient adaptability to complex environments due to its reliance on the orthogonal node layout of grid maps, as well as redundant node traversal, long search time, and poor path smoothness, a path planning and speed collaborative optimization method based on topological maps and an improved A* algorithm was proposed. Firstly, the obstacle point cloud map and OpenDrive map were fused to construct a topological map, and road center points and boundary points were extracted to generate a core point set, which solves the problem of redundant storage in grid maps. Secondly, the search neighborhood was optimized to adapt to non-orthogonal node scenarios, and a dynamic direction heuristic function with adaptive weights was introduced into the evaluation function to suppress invalid exploration that deviates from the target and improve search efficiency. Thirdly, the path smoothness was optimized using cubic spline curves with curvature constraints, and speed planning was conducted based on the curvature distribution of the optimized path. Finally, a preview tracking strategy was adopted to verify the effectiveness of the path and speed planning. Simulation results show that the path smoothness is significantly improved, verifying the collaborative optimization effect of the method in terms of efficiency, smoothness, and safety.