Abstract:

Based on the bi-directional asymptotically optimal RRT*-connect algorithm, the motion planning of the high-dimensional manipulator is analyzed, so as to make the search path in the planning process shorter and more efficient. Firstly, the target bias strategy is introduced into the sampling process, and the sampling area is constrained to ensure that each sample can be searched towards the target direction, therefore the optimal searching path can be obtained. Secondly, based on the above operation, the gradient descent method is adopted to optimize the searched path, smoothing the whole path and removing the large angle turn. Then, the improved RRT*-connect algorithm is simulated and analyzed by Matlab. It is demonstrated that the algorithm can guarantee the probabilistic completeness and asymptotic optimality of searching under various complex environment, and the shorter searching path is obtained in shorter time. Finally, a simulation experiment is conducted on the ROS platform with UR5 manipulator to verify the practicability and effectiveness of the algorithm.

Key words: manipulator, path planning, improved RRT*-connect, gradient descent method

摘要：

基于双向渐进最优的RRT*-connect算法，对高维的机械臂运动规划进行分析，从而使规划过程中的搜索路径更短，效率更高。将目标偏向策略引入采样过程，同时对采样点区域进行约束，保证每次采样都能朝着目标方向搜索，使得搜索路径更优。在此基础上，采用梯度下降法优化搜索出的路径，将整个路径做平滑处理，去除大角度转弯。利用Matlab对改进后的RRT*-connect算法进行仿真对比分析，从而证明该算法在各种复杂环境下都能保证搜索的概率完备性以及渐进最优性，并且搜索路径更短，用时更少。在ROS平台使用UR5机械臂进行仿真实验，验证该算法的实用性与有效性。

关键词: 机械臂, 路径规划, 改进RRT*-connect, 梯度下降法