基于KNN-RRT 的机械臂运动路径规划算法

Motion Path Planning Algorithm of Manipu-Lator Based on KNN-RRT

针对机械臂路径规划过程中节点生成容易陷入局部最小值、算法收敛速度慢等问题,以目标引力函数渐进最优快速扩展随机树(P-RRT)为基础,提出一种基于KNN快速查找的自适应步长的改进RRT算法(KNN-RRT)。首先,在目标引力的基础上引入AdaGrad方法来调整自适应步长系数,降低随机点采样陷入局部最小值的问题;其次,利用KDTree来存储节点并利k邻近快速搜索查找相邻节点,提高算法的效率,并结合三次B样条曲线优化搜索路径的质量;最后,基于KNN-RRT算法在不同障碍物环境下进行实验,实验结果表明该算法在路径搜索时间、路径质量等方面有显著提升,提高算法的稳定性。

In order to solve the problems such as local minima and slow convergence rate in the process of path planning,an improved adaptive step size RRT algorithm(KNN-RRT)based on the asymptotically optimal fast extended random tree(P-RRT)of target gravity function is proposed.Firstly,AdaGrad method is introduced to adjust the adaptive step coefficient on the basis of target gravity to reduce the problem of random point sampling falling into the local minimum.Secondly,KDTree is used to store nodes and quickly search for adjacent nodes with k proximity to improve the efficiency of the algorithm.The quality of the search path is optimized with cubic B-spline curves.Finally,experiments based on KNN-RRT algorithm are carried out under different obstacle environments,and the experiments show that the path search time and path quality of the algorithm are significantly improved,and the stability of the algorithm is improved.