通用抓取位姿
2024-08-09 11:25:08 0 举报AI智能生成
通用物体抓取思路拆解
作者其他创作
大纲/内容
夹爪深度
夹爪旋转
采样点选取
force analytic mode
c有效抓取位姿阈值（手动设定）
公式
Single Object Graspness
reconstruct the scene using object 3D models and correspond-ing 6D poses
碰撞检测
GraspNet-1Billion: A Large-Scale Benchmark for General Object Grasping
Ckij-collision label
问题1：抓取位姿与背景物或者临近物体干涉
associate the scene point cloud with the projected object point
通过对象的6D Pose将对象点投影到场景中
问题2：RGBD观测到的是单视野局部点云
归一化表示
外框
Scene-Level Graspness
graspness measurepoint-wise graspness scores & view-wise graspness scores
F - 高维特征
Two sub-function
point level过滤掉大部分不可抓取点；View-wise只需要计算剩下的点
功能
ResUNet14extraction of both global andlocal point features
backbone network
使用MLP建模to generate point-wise graspable landscape.
select points with graspness score larger than δp
adopt farthest point sampling (FPS) to maximize distances amongsampled points
Graspable Farthest Point Sampling
使用MLP建模to the sampled seed points and output M × V vectors for view-wise graspable landscapesand M × C residual features for grasp generation.
Fibonacci lattices生成V个观测视野
Graspable Probabilistic View Selection
cascaded graspness model作用：提取每个点的局部特征向量局部特征向量：
Crop-and-refine从点云中裁剪出潜在的抓取区域，然后对这些区域进行细化处理，以精确估计抓取位姿和提高抓取检测的准确性。通过这种方式，可以从复杂的场景中有效地识别和定位可抓取的物体。
Cylinder-Grouping from Seed Points
PointNetfor grasp generation
Grasp Generation from Candidates
minimum friction coefficient
Grasp Score Representation
Loss Function说明
Loss Function
grasp operation model
GSNet architecture
0 条评论
回复 删除
下一页
