Global trajectory planning for Autonomous Valet Parking with segmentation, rewind and relaxation

A significant amount of time and fuel energy can be wasted in the process of parking with so many vehicles in this world, while autonomous valet parking is capable of liberating drivers for parking with high efficiency. In recent years, various proposals for autonomous valet parking emerged. However, this problem still cannot be promptly handled with reliable trajectories for vehicles to follow. Therefore, this paper proposed a novel global trajectory planning method which generates reliable trajectories with given start configuration and vacant parking space. To be detailed, global path obtained by an improved A* algorithm, which is a rough solution, provides waypoints for the trajectory planning as the reference. Moreover, this planning of trajectory is transformed into an optimal control problem for the solving of trajectory. Considering the probability of failure in the solving process, segmented planning is utilized to reduce the size of constraints, as well as rewind and relaxation applied to improve the convergence. The performance of this global trajectory planning algorithm is evaluated through the simulation in a designed parking lot scenario.