A Motion Planning Method Addressing Arbitrary Lots for Autonomous Parking Vehicle

Considering the particularity of the parking scene, an improved hybrid A* algorithm is proposed in this paper. In order to apply to a variety of regular lots, and satisfy requirements of comfort and meet the vehicle nonholonomic constraints, the cost rules employ driving direction and steering angle, and the heuristic rules combine RS length and Euclidean distance of eight neighborhood expansion. The improvements of collision detection and velocity planning method have also been introduced. According to the regular conditions, 5 parking scenarios of 0/30/45/60/90° lots are designed to verify the improved hybrid A*. And the algorithm is compared with A* algorithm. The results show that the improved algorithm could generate the vehicle executable parking trajectory, and can adapt to the regular parking scenarios. By 1000 cyclic planning experiments, the time consumption is less than 120ms, which means the improved algorithm has good real-time performance. And through 369 tests at different starting points, the success rate is no less than 97% compared to A*, which illustrates the strong generality of the improved A*.