Curvature Continuous Path Planning with Reverse Searching for Efficient and Precise Autonomous Parking

In autonomous driving, vehicles usually need to complete parking tasks. The existing parking planning algorithms based on hybrid A∗ can ensure the smoothness of the planned path, but it is inefficient due to the consumption of time. In addition, these algorithms do not fully consider the particularity of the parking scene, that is, the unbalanced distribution of obstacles around the lane and parking slot. To solve the above problems, we propose a curvature continuous parking planning method with reverse searching based on hybrid A*, called RS-HA*, which detects the collision of dense obstacles around the parking slot to prevent useless expansion nodes and complete the pruning of the algorithm to improve the real-time performance of the algorithm. We call this processing OCP (Obstacle-Collision-Pruning). RS-HA∗ can improve the planning efficiency and ensure the precision of the goal pose without reducing the smoothness of the path. We verified RS- HA∗ performance through simulation environment and real scene. The experimental results show that compared to the existing methods, RS-HA∗ greatly improves the parking efficiency and accuracy.