Path Planning Algorithm of Integrating Improved RRT-Connect and APF

An optimized path planning algorithm integrating improved bidirectional rapidly-exploring random tree (RRT–Connect) and artificial potential field (APF) was proposed to enhance the real-time performance and safety of autonomous vehicles. First, a dynamic step size strategy was adopted to adaptively adjust the expansion step size according to the distance between nodes and obstacles, significantly improving path search efficiency. Second, the characteristics of APF were incorporated, where its attractive potential component was utilized to bias sampling towards the target direction for accelerated convergence, while its repulsive potential component was employed to achieve obstacle avoidance for enhanced path safety. Furthermore, a bidirectional pruning strategy combined with cubic B-spline curve optimization was introduced to effectively shorten the path length and improve trajectory smoothness. Particularly, the APF repulsive function was modified by adding a target distance component to address the goal-unreachable issue while ensuring stable arrival at the target position in dynamic environments. To validate the algorithm’s effectiveness, a simulation platform was established based on the robot operating system (ROS), and tests were conducted in various complex obstacle scenarios. The experimental results demonstrate that compared with the benchmark RRT and RRT–Connect algorithms, the proposed integrated optimization algorithm achieves approximately 30% and 12% reduction in path node quantity, 30% and 13% shortening in path length, and 50% and 3% decrease in search time respectively through the improvement of dynamic step strategy and sampling function. The path smoothness is further enhanced and the length is additionally reduced after being processed by the combined optimization of bidirectional pruning strategy and cubic B-spline curve. The modified repulsive potential function not only effectively solves the goal-unreachable problem but also improves the real-time obstacle avoidance capability of the algorithm in dynamic complex environments.