基于改进RRT算法的无人车路径规划

Path Planning of Unmanned Vehicle Based on Improved RRT Algorithm

针对无人车在复杂环境中进行全局路径规划时存在的盲目搜索、节点冗余、路径不光滑及不安全等问题,提出一种基于快速扩展随机树(RRT,rapidly-exploring random tree)的综合改进路径规划算法;首先引入目标动态概率采样策略和人工势场引导随机树扩展机制;其次根据汽车运动学模型,对规划的路径进行转角约束和碰撞检测,保证路径的安全性;然后引入Reeds-Sheep曲线用于直接与目标位姿进行连接,避免多余的位姿调整;最后对路径进行剪枝和平滑处理,得到一条更短更光滑的路径;在实验部分,针对不同仿真环境,以规划时间、路径长度和节点数目作为评价指标,对比了RRT算法、RRT*算法和文章算法的路径规划效果;实验结果显示,文章算法相比于RRT算法和RRT*算法,节点数目分别减少了58.94%和85.22%,规划时间分别缩短了61.20%和79.23%,且路径长度相比于RRT算法缩短了17.26%,并和RRT*算法规划的最优路径长度相近。

aimed at the problems of blind search, node redundancy, unsmooth and insecure path planning of unmanned vehicles in complex environment, a comprehensive improved path planning algorithm based on rapid exploring random tree(RRT) is proposed. Firstly, the target dynamic probability sampling strategy and the artificial potential field guided random tree expansion mechanism are introduced. Secondly, according to the vehicle kinematics model, the planned path is subject to corner constraints and collision detection to ensure the safety of the path. Then the Reeds-Sheep curve is introduced to directly connect with the pose of the target to avoid the redundant pose adjustment. Finally, the path is pruned and smoothed to get a shorter and smoother path. In the experiment, the path planning effects of RRT algorithm, RRT* algorithm and the algorithm are compared by the evaluation indexes of the planning time, path length and node number in the different simulation environments. The experimental results show that compared with the RRT algorithm and RRT* algorithm, the node number and path length of the algorithm are reduced by 58.94% and 85.22% respectively, the path length of the algorithm by 61.20% and 79.238 5% respectively, compared with the RRT algorithm, the path length is shortened by 17.6%,which is close to the optimal path length by the RRT* algorithm.