【摘要】 针对休眠期枣树自动剪枝过程中机械臂路径规划的问题，提出了基于改进RRT~*算法的机械臂路径规划方法。为了获得真实休眠期枣树的树形结构，搭建了一台视觉采集系统，对上下布置的2个深度相机进行位姿标定，同时提出了枣树三维点云预处理的算法流程，通过对背景噪声、地面噪声的去除得到单棵无噪声的枣树点云。结合真实枣树结构和机器人工具箱构建了剪枝过程中机械臂路径规划的仿真环境，试验结果表明：改进后的RRT~*算法能够进一步减少路径规划的轨迹长度和运行时间；对于剪枝点3到剪枝点4的平均路径为5 140 mm、运行时间为64.6 s,其优化后的路径长度为1 244 mm、运行时间为29 s;而对于剪枝点4到剪枝点6的平均路径为1 169 mm、运行时间为60.2 s,其优化后的平均路径长度为672 mm、运行时间为17 s;标准差和变异系数变化范围较小，验证了提出的方法具有可重复性。更多还原

【Abstract】 To solve the path planning for automatic pruning of dormant jujube tree, a method based on improved RRT~* algorithm was proposed.To obtain the real structure of jujube tree, visual acquisition system was built, and the pose calibration of the two depth cameras arranged up and down. For obtaining a single jujube tree, 3D point cloud preprocessing algorithm was proposed to remove background and ground noise.Combined with the real jujube tree spatial structure and Robotics Toolbox for Matlab, the path planning simulation environment for the manipulator arm was constructed. The test results showed that: the improved RRT~* algorithm could further reduce length and running time of path planning; For pruning point 3 to point 4, the average path was 5 140 mm and the running time 64.6 s. The optimized path length was 1 244 mm and the required time was 29 s.For pruning point 4 to point 6, the average path was 1 169 mm, and the required time was 60.2 s. Meanwhile, the average path was shortened to 672 mm, and the running time was reduced to 17 s after the improvement.The variation range of standard deviation and variable coefficient was small, which verified the reproducibility of the proposed method in the paper.更多还原