作者：Sameer Pai Tao Chen Megha Tippur Edward Adelson Abhishek Gupta Pulkit Agrawal
我们研究了在视觉感知被禁用、物体形状事先未知、物体可以自由移动（比如从抽屉里抓物体）的情况下的物体检索问题。成功的解决方案需要定位自由对象，识别特定的对象实例，然后仅使用触摸反馈来抓取已识别的对象。与视觉不同，在视觉中，相机可以观察整个场景，触摸传感器是局部的，只观察场景中与操纵器接触的部分。此外，通过触摸传感器收集信息需要在触摸表面上施加力，这可能会干扰场景本身。因此，用触摸进行推理需要随着时间的推移仔细探索和整合信息，这是我们面临的一个挑战。我们提出了一种系统，能够使用灵巧手上指尖触摸传感器的稀疏触觉反馈来定位、识别和抓取新物体，而无需任何视觉反馈。视频可用athttps://taochenshh.gith
We study the problem of object retrieval in scenarios where visual sensing isabsent, object shapes are unknown beforehand and objects can move freely, likegrabbing objects out of a drawer. Successful solutions require localizing freeobjects, identifying specific object instances, and then grasping theidentified objects, only using touch feedback. Unlike vision, where cameras canobserve the entire scene, touch sensors are local and only observe parts of thescene that are in contact with the manipulator. Moreover, information gatheringvia touch sensors necessitates applying forces on the touched surface which maydisturb the scene itself. Reasoning with touch, therefore, requires carefulexploration and integration of information over time — a challenge we tackle.We present a system capable of using sparse tactile feedback from fingertiptouch sensors on a dexterous hand to localize, identify and grasp novel objectswithout any visual feedback. Videos are available athttps://taochenshh.github.io/projects/tactofind.
论文链接：http://arxiv.org/pdf/2303.13482v1
更多计算机论文：http://cspaper.cn/