改进RRT算法的无人驾驶车辆路径规划研究

Research on Driverless Vehicle Path Planning with Improved RRT Algorithm

针对基础快速扩展随机树(Rapidly-exploring Random Trees,RRT)应用于无人驾驶车辆路径规划时缺乏导向性,收敛速度慢,路径平滑性差及规划结果并非最优解等问题,提出了一种基于RRT的路径规划改进算法。首先,设计了启发式采样策略:提出基于权重分配的目标指向的局部扩展方式,解决了节点盲目扩展的问题,避免了因目标偏向而出现路径陷入局部最小值的情况,并通过设置转角阈值约束节点转角范围,同时采用变步长采样策略,提高了算法局部避障能力;其次,对已得路径进行后处理:提出了节点优化策略,并用B样条曲线进行路径拟合,实现了路径长度的优化并满足平滑性要求,路径末端与目标点采用Reeds-Shepp曲线连接,解决了车辆抵达目标点时的航向问题。最后利用Matlab软件,将改进算法与基础RRT及其衍生算法进行了对比分析,验证了所提算法的有效性和优越性。

Aiming at the problem which is lack of guidance,slow convergence,poor smoothness and the planning result is not the optimal solution in RRT applied to driverless vehicle path planning,an improved path planning algorithm based on RRT is proposed.Firstly,a heuristic sampling strategy is designed:The local expansion mode of target orientation based on weight redistribution is proposed to solve the problem of blind expansion of nodes,the path falling into the local minimum due to the target orientation is avoided,the steering angle range of the nodes is constrained by setting the steering angle threshold,and the variable step size sampling strategy is adopted to improve the local obstacle avoidance ability of the algorithm;Secondly,the obtained path is post processed:The strategy of node optimization is proposed,the path is fitted with B-spline curve to optimize the path length and meet the smoothness requirements,and the end of the path is connected with the target node by Reeds-Shepp curve,which solves the heading problem when the vehicle reaches the target point.Finally,the improved algorithm is compared with the basic RRT and its derivative algorithms by using MATLAB software to verify the effectiveness and superiority of the proposed algorithm.