Abstract: To improve the efficiency of path planning for unmanned vehicles and the safety during path tracking, an improved optimization algorithm was proposed that combines the improved bidirectional rapidly-exploring random tree (RRT-Connect) and artificial potential field (APF). A dynamic step strategy was employed to select appropriate step size for expanding the random tree based on distances between nodes and obstacles, and improving path search efficiency. The APF attraction component was used to guide the random tree towards the target direction for biased sampling, and accelerating algorithmic sampling efficiency. Meanwhile, the APF repulsion component was used to guide the random tree away from obstacles to optimize path obstacle avoidance performance.The length and smoothness of the planned path was further optimize by implementing a bidirectional pruning strategy and introducing cubic B-spline curves.By optimizing the APF repulsion function to increase the repulsive component of the distance between the unmanned vehicle and the target point, the vehicle could avoid dynamic obstacles and stably stop at the target point while tracking the path.Based on the robot operating system (ROS), simulation tests were conducted in different obstacle environments to verify the effectiveness of the fusion improvement optimization algorithm. The results show that compared to the RRT algorithm and RRT-Connect algorithm, the dynamic step size strategy and sampling function optimization reduces the number of path nodes by about 30% and 12%, shortens the path by about 30% and 13%, and shortens the search time by about 50% and 3% for the fusion improvement optimization algorithm, respectively.The path processed through bidirectional pruning and cubic B-spline smoothing further enhance smoothness and reduce length.Optimizing the repulsive function can effectively solve the problem of unreachable target points and improve the obstacle avoidance ability of local algorithms in dynamic environments.