基于点云采样权重估计的未知物体抓取位姿生成方法

针对机械臂在非结构环境中对未知物体抓取位姿生成困难及抓取稳定性差的问题,提出一种基于点云采样权重估计的抓取位姿生成方法.首先通过移动深度相机的方式拼接得到较完整的物体点云信息,并对物体的几何特性进行分析,有效避开物体不宜抓取的位置进行抓取位姿样本生成;然后结合几何约束条件实现抓取位姿搜索,并利用力封闭条件对样本稳定性进行评估;最后为了对实际的抓取位姿进行评价,根据其稳定性、夹取深度、夹取角度等设定抓取可行性指标,据此在工作空间输出最佳抓取位姿并完成指定的抓取任务.实验结果表明,采用所提方法能够高效生成大量且稳定的抓取位姿,并在仿真环境中有效实现机械臂对单个或多个随机摆放的未知物体的抓取任务.

Grasping pose generation method for unknown objects based on point cloud sampling weight estimation

To resolve the problem of generating grasping poses for unknown objects in an unstructured environment, a method based on point cloud sampling weight estimation is proposed to generate grasping pose. Firstly, the object point cloud information is spliced by moving camera, then the geometric characteristics of the object are utilized to improve grasp quality. In addition, the candidates pose can be evaluated by geometric constraints, while the force closure condition is used to quantify pose stability. Finally, in order to select the optimal pose to perform the grasping task, set reliability indicators based on pose stability, depth and angle. Experimental results show that the proposed method can efficiently generate a large number of stable grasping poses, and complete the grasping tasks of different unknown objects by the manipulator in the simulation environment effectively.