This website requires JavaScript.

Tac2Structure: Object Surface Reconstruction Only through Multi Times Touch

J. LuZ. WanY. Zhang
Sep 2022
摘要
Inspired by the ability of humans to perceive the surface texture ofunfamiliar objects without relying on vision, the sense of tactile can play acrucial role in the process of robots exploring the environment, especially insome scenes where vision is difficult to apply or occlusion is inevitable toexist. Existing tactile surface reconstruction methods rely on external sensorsor have strong prior assumptions, which will limit their application scenariosand make the operation more complex. This paper presents a surfacereconstruction algorithm that uses only a new vision-based tactile sensor wherethe surface structure of an unfamiliar object is reconstructed by multipletactile measurements. Compared with existing algorithms, the proposed algorithmdoesn't rely on external devices and focuses on improving the reconstructionaccuracy of the large-scale object surface. Aiming at the difficulty that thereconstruction accuracy is easily affected by the pressure of sampling, wepropose a correction algorithm to adapt it. Multi-frame tactile imprintsgenerated from many times contact can accurately reconstruct global objectsurface by jointly using the point cloud registration algorithm, loop-closuredetection algorithm based on deep learning, and pose graph optimizationalgorithm. Experiments verify the proposed algorithm can achievemillimeter-level accuracy in reconstructing the surface of interactive objectsand provide accurate tactile information for the robot to perceive thesurrounding environment.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答