Path Planning for Autonomous Vehicles Based on the Normal Distribution Transform

Traditional path planning methods for autonomous vehicles are performed on grid maps in a 2D plane. However, in real-world environments, which are complex and diverse, directly reducing the environment to a 2D plane can lead to incorrect estimation of traversability in certain areas, especially in scenes with significant height differences. In this paper, we propose a solution for autonomous vehicle path planning in 3D environments. Firstly, we simplify the original point cloud map using the normal distribution transformation and extract passable areas by considering real-world traversability constraints, further simplifying the map representation. Secondly, an improved A* algorithm is proposed, which incorporates an adaptive dynamic coefficient to significantly enhance the efficiency and quality of path planning in 3D environments. Experimental results validate that the proposed method provides an effective and efficient solution for autonomous vehicle path planning in 3D environments. In the parking lot scenario, the number of map units was reduced by 98.9%, and the path planning time and the number of search nodes, given the start and goal points, were reduced respectively by 90.7% and 72.4%.