无人车运动规划算法综述

A Review of the Motion Planning Problem of Autonomous Vehicle

摘要:

回顾无人车运动规划问题.无人车的运动受微分约束，且运行环境既包括结构化的道路也包括非结构化的野地.根据具有阿克曼转向性质的车辆模型所具有的微分平坦性质，可以简化无人车的轨迹生成问题.相比直接轨迹生成法，路径速度分解法更常用.回旋线、样条曲线、多项式螺旋线是使用较多的路径生成曲线.具有重要实用意义的两大类无人车运动规划算法分别是：以快速随机扩展树算法（RRT）为代表的基于采样的规划算法和以A*搜索算法为代表的基于搜索的规划算法.

Abstract:

The motion planning problem of unmanned autonomous vehicle (UAV) is reviewed. UAV operates in both structured road and unstructured field with differential constraints. The problem of trajectory generation can be simplified with the differential flatness of Ackerman steering vehicle. Compared to direct trajectory generation method, pathvelocity decomposition method is more popular. Clothoids, splines and polynomial spirals are used for path generation. The two major planning algorithms of great practical significance are rapidly random tree(RRT) in the name of samplingbased and A* in the name of searchbased methods.