In the era of social distancing, humans tend to interact more with digital edges/devices and robotic avatar. Users mostly interact with cyber-physical systems through audiovisual devices, while tangible interfaces are limited to touch panels. Touch perception not only brings sense of physical contact, texture; but also conveys the intention and emotion of users. Here, we introduce a novel soft tangible device, called IoTouch, as a combination of soft robotics and deep learning, that can act as both robotic skin or haptic sensing interface. The technique aims to implementation of touch sensing with a vision-based approach. We showcase on barrel shape (ellipsoid with two flattened ends) of the IoTouch, where a dataset of tactile images was collected through an autonomous process, in order to train a network for recognition of physical contact states (contact location, contact depth). A Finite Element Method model of the barrel skin was also setup for data acquisition of skin's deformation. The proposed method can detect multiple contacts simultaneously on the entire area of the barrel skin. Obtained result promises a platform for creation of robotic avatars that can sense diverse states of physical touch, through which the users' experience and communication of haptics can be enhanced.

在远离社会的时代，人类倾向于更多地与数字边缘/设备和机器人化身互动。用户主要通过视听设备与网络物理系统进行交互，而有形界面仅限于触摸面板。触觉感知不仅带来身体接触感、质感；同时也传达了用户的意图和情感。在这里，我们介绍了一种称为 IoTouch 的新型软有形设备，它是软机器人和深度学习的结合，可以充当机器人皮肤或触觉传感界面。该技术旨在通过基于视觉的方法实现触摸感应。我们展示了 IoTouch 的桶形（具有两个扁平末端的椭圆体），其中通过自主过程收集了触觉图像数据集，为了训练识别物理接触状态（接触位置、接触深度）的网络。还建立了桶状蒙皮的有限元方法模型，用于获取蒙皮变形的数据。所提出的方法可以在桶状皮肤的整个区域上同时检测多个接触。获得的结果为创建机器人化身提供了一个平台，可以感知身体接触的不同状态，通过它可以增强用户的体验和触觉交流。