A Topological Map-Based Path Coordination Strategy for Autonomous Parking

This paper proposed a path coordination strategy for autonomous parking based on independently designed parking lot topological map. The strategy merges two types of paths at the three stages of path planning, to determinate mode switching timing between low-speed automated driving and automated parking. Firstly, based on the principle that parking spaces should be parallel or vertical to a corresponding path, a topological parking lot map is designed by using the point cloud data collected by LiDAR sensor. This map is consist of road node coordinates, adjacent matrix and parking space information. Secondly, the direction and lateral distance of the parking space to the last node of global path are used to decide parking type and direction at parking planning stage. Finally, the parking space node is used to connect global path and parking path at path coordination stage. After optimizing nodes and smoothing path utilizing a fourth-degree polar-polynomial function, those two types of paths can be merged without deviation. Experiments show that the proposed topological map-based path coordination strategy can effectively generate a feasible path to guide vehicle from the drop off zone to the desired parking space. The designed controller meets real-time requirements. At the same time, continuous curvature variation of path and steady speed can improve accuracy of path tracking.