Reverse Search Heuristic Sampling-Based Path Planning Algorithm With Path Smoothing for Autonomous Vehicles

Path planning is a critical component of autonomous driving, particularly in complex unstructured environments characterized by the absence of clearly defined drivable corridors and the presence of irregular obstacles. A widely adopted two-stage framework under such conditions involves generating an initial path followed by its optimization. In this paper, a bi-directional rapidly-exploring random tree (RRT) based path planning algorithm, namely reverse search heuristic RRT* (RH-RRT*), is proposed to generate an initial path composed of Reeds-Shepp (RS) curves. Unlike traditional bi-directional tree-based path planning algorithms that directly connect two trees, RH-RRT* employs a reverse search tree as heuristic information. This design achieves a faster convergence speed. The RS curves inherently suffer from curvature discontinuities, posing challenges for vehicle tracking control. Therefore, a fast optimization-based smoothing algorithm is applied to the initial path to ensure kinematic feasibility and smoothness. Extensive simulations and real-world experiments demonstrate that the proposed algorithms outperform existing methods in terms of planning efficiency, path smoothness and tracking performance.