Robots taught to “feel” objects thanks to artificial skin

Robotics does not stand still. Engineers taught robots to understand human language and walk independently. But there is still a problem of tactile feedback, when a smart mechanism can not only move an object, but also feel it. Scientists have created a new sensor that works on the principle of human skin, distinguishing between fragile, soft, hard and other properties of objects.

Created by researchers from the Swiss Higher Technical School of Zurich, the sensor consists of elastic silicone skin with colored plastic microspheres and a conventional camera mounted on its underside. Thanks to machine vision, the robot “sees” which part of the skin comes into contact with the object. A fisheye lens captures the size and depth of a dent on silicone skin, calculating the required force to hold the object, depending on its characteristics.

Silicone skin is able to distinguish several types of effects on the surface of the sensor, accordingly calculating the degree of effort to hold with high accuracy. Also, the robot will be able to determine the specific direction in which you need to put more or less effort. After the robot has thus studied the properties of various objects, machine learning allows it to more accurately identify the pattern on the skin with its size, weight and density.

Currently, the research team is working on new sensors that will teach the robot to recognize more complex objects, and are also considering the possibility of miniaturizing the sensor.