基于拓扑地图的自主泊车路径协调与优化策略

为了使全局路径与泊车路径无偏差对接，得到曲率连续的可行驶路径，为泊车模式切换提供精准位姿，提出基于拓扑地图的自主泊车路径协调与优化策略。首先，定义一种精简的停车场拓扑地图描述形式与道路拓扑设计原则，通过采集停车场内关键特征的定位数据建立停车场拓扑地图。其次，基于道路拓扑设计原则与泊车规划原则，设计“第1次平滑处理-路径协调-第2次平滑处理”的流程，对规划出的全局路径与泊车路径进行协调与优化。其中，“第1次平滑处理”利用对称极多项式，以车辆最小转向半径为约束，对全局路径的转折处平滑处理；“路径协调”通过节点替换的方式实现全局路径与泊车路径的无偏差对接；“第2次平滑处理”利用梯度下降的方法，分别优化前进段与倒车段的路径曲率，消除路径对接与均匀插值之后引起的曲率抖动。最后，在PreScan/Simulink联合仿真平台和实车试验平台上，验证路径协调与优化策略的有效性。研究结果表明：所提策略能够无偏差对接全局路径与泊车路径，并生成曲率平滑的可行驶路径，为横向控制提供了平滑变化的曲率值，有助于提高路径跟踪时转向的平稳性；在路径跟踪前为自主泊车控制系统提供了精准的泊车切换位姿，保证了低速自动驾驶与自动泊车之间的准确切换。

A Topological Map-based Path Coordination and Optimization Strategy for Autonomous Parking

In this study,a strategy was proposed for merging the global path and the parking path without any deviation to obtain a drivable path with continuous curvature. Moreover,to obtain an accurate pose for parking mode switch,an autonomous parking path coordination and optimization strategy based on the parking lot topological map was proposed. First,a compactness expression of parking lot topological map and the design policy of road topology were defined. In addition,aparking lot topological map was built by collecting the positioning data of key features in parking lot. Second,based on the design policy of road topology and the parking planning policy,the planned global path and parking path were coordinated and optimized under the designed "First Smoothing Operation-Path Coordination-Second Smoothing Operation"procedures. The "First Smoothing Operation"was used to smooth turning points of global path with the constraint of vehicle minimum turning radius by using a symmetrical polar-polynomial function. The "Path Coordination"was used to merge global path and parking path without deviation by using node replacement. The"Second Smoothing Operation"was used to eliminate curvature jitter caused by path merging and interpolation by optimizing the path curvature of the forward and reverse parts based on the gradient descent method. Finally,the effectiveness of path coordination and optimization strategy was verified using the PreScan/Simulink co-simulation platform and real vehicle platform. Experimental results indicate that the proposed strategy can connect the global path and the parking path without deviation,and generate a drivable path with smooth curvature. This path provides smooth curvature value for lateral control to improve the steering stability during path tracking. Simultaneously,the drivable path provides an accurate parking switch pose for the autonomous parking control system before path tracking,and the pose guarantees the accurate switch between the modes of low-speed automated driving and automated parking.