文章目录
概述
an object-centric, Actionable, Articulation-Aware Vision Language Model(A3VLM)
A3VLM focuses on the articulation structure and action affordances of objects.(结合了两者 关节结构和可操作性)
robot-agnostic representation
learning an object-centeric representation then translated into low-level manipulation actions
the first VLM capable of accurately and consistently locating and understanding the articulation structure for robot manipulation.
Related work总结
Manipulation of Articulated Objects
1、A common practice for articulated object manipulation is to determine its articulation structure first and then manipulate it with predefined action primitives.估计结构–>使用primitive操作
2、FlowBot3D [10], predict the perpoint articulation movement of every point on the object’s 3D point cloud.预测点云