首页|考虑点云骨架特征的未知物体六自由度抓取

考虑点云骨架特征的未知物体六自由度抓取

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对机械臂在非结构化环境下对未知物体生成稳定抓取位姿困难的问题,文章提出一种基于点云骨架特征的未知物体六自由度抓取方法.首先,通过深度相机获取包含物体的场景单视角点云,并在物体表面随机采样得到初始采样点,设计考虑L1中值骨架提取的迭代移动采样算法,保证抓取点最终均匀排布在物体的骨架上;然后,利用骨架点的分布信息和骨架点周围点云的几何信息生成候选抓取位姿,根据抓取器与物体之间的位置关系,使用启发式评价函数评估抓取位姿,从而保证位姿的最优化采样;最后,对不同形状的物体进行仿真实验和实际抓取试验.测试结果表明,文章所提方法可以对常见物体生成稳定的抓取位姿,并能有效拓展到更多形状复杂的未知物体.
Six-degree-of-freedom grasping of unknown objects considering point cloud skeleton features
Aiming at the problem that it is difficult for the manipulator to generate stable grasping posi-tion and posture for unknown objects in an unstructured environment,a six-degree-of-freedom grasp-ing method for unknown objects based on point cloud skeleton features was proposed.Firstly,the single-view point cloud of the scene containing the object was obtained by the depth camera,and the initial sampling points were obtained by random sampling on the object surface.An iterative moving sampling algorithm considering L1 median skeleton extraction was designed to ensure that the cap-tured points were finally uniformly distributed on the object skeleton.Then,the distribution informa-tion of the skeleton points and the geometric information of the point cloud around the skeleton points were used to generate candidate grasping positions and postures.According to the position relation-ship between the grab and the object,the heuristic evaluation function was used to evaluate the grasp-ing positions and postures,so as to ensure the optimal sampling of the positions and postures.Final-ly,the simulation and actual grasping experiments of objects with different shapes were carried out.The results show that the proposed method can generate stable grasping position and posture and can be effectively extended to more complex unknown objects.

unstructured environmentsix-degree-of-freedom graspingpoint cloud skeleton featuresgeometric informationheuristic evaluation function

吴航、谢远龙、王书亭、魏棋斌、熊体凡

展开 >

华中科技大学机械科学与工程学院,湖北武汉 430074

非结构化环境 六自由度抓取 点云骨架特征 几何信息 启发式评价函数

国家重点研发计划资助项目国家自然科学基金资助项目国家自然科学基金区域创新发展联合基金资助项目

2020YFB170830052105019U21A20151

2024

10.3969/j.issn.1003-5060.2024.07.002

合肥工业大学学报(自然科学版)
合肥工业大学

合肥工业大学学报(自然科学版)

CSTPCD北大核心
影响因子：0.608
ISSN：1003-5060
年,卷(期)：2024.47(7)