To enhance the safety and efficiency of path planning for autonomous vehicles under sudden traffic incident disturbances, this study proposes an improved Rapidly-exploring Random Tree (RRT) algorithm to address the limitations of the traditional RRT, including strong search randomness, slow convergence, and suboptimal path quality. The proposed method introduces a dynamic sampling space optimization strategy to avoid invalid regions and integrates an A*-based heuristic cost evaluation to guide the random tree toward the target in a prioritized manner. Finally, simulation experiments are conducted to validate the algorithm’s performance. The results show that, compared with the conventional RRT algorithm, the proposed approach reduces the number of sampled points by approximately 74%, shortens the path length by about 13%, and decreases the search time by around 20%. These improvements demonstrate the algorithm’s significant advantages in enhancing both path planning efficiency and path quality, providing a safer and more efficient path solution for autonomous vehicles operating in complex and disturbed environments.