A hierarchical RRT-DWA planner for autonomous parking in dynamic environments

Aiming to tackle the problems of low adaptability to dynamic environments and low planning efficiency of traditional automaticp arking path-planning algorithms, this paper proposes a hierarchical path-planning framework that integrates the improved rapidly-exploring random tree (RRT) algorithm and the dynamic window approach (DWA). Firstly, at the global planning level, a Gaussian-uniform mixeddistribution sampling strategy is adopted to optimize the growth direction of the random tree, and a dynamic step-size mechanism is incorporated to improve the algorithm expansion efficiency. Secondly, the artificial potential field (APF) method is introduced to optimize the RRT-generated path nodes, ensuring the geometric safety clearance for the vehicle chassis. Subsequently, at the local planning level, these optimized nodes serve as waypoints to guide the DWA. Dynamic obstacle-avoidance weight is introduced into the evaluation function of DWA, and this RRT-DWA collaborative framework effectively solves the problems of dynamic obstacle-avoidance and local stagnation. Finally, for the terminal parking maneuver, the Reeds-Shepp (RS) curve is used to smoothly adjust the vehicle pose to match the parking end-point. Finally, a joint simulation is carried out in Carsim/Simulink through the pure-pursuit control algorithm. The simulation experiments show that the maximum tracking error of the planned global path in parallel, perpendicular, and diagonal parking scenarios is within 0.35 m, and the distance from dynamic obstacles is greater than 2 m, which confirms that the planned path is rational.