A team of roboticists from the German Aerospace Center’s Institute of Robotics and Mechatronics has discovered that combining traditional force-torque sensors with machine learning algorithms offers robots a new method for sensing touch.
In their study, published in Science Robotics, the researchers explored a novel approach to endow robots with touch sensitivity, without the need for artificial skin.
For living organisms, touch is a two-way interaction: you perceive the texture, temperature, and other characteristics of an object when you touch it, but you also feel when something touches you. In this research, the team focused on mimicking this latter type of touch in robots by integrating internal force-torque sensors with machine learning.
The team recognized that the sensation of being touched is largely driven by torque, like the tension felt in the wrist when pressure is applied to the fingers. To replicate this, they equipped a robot arm with ultra-sensitive force-torque sensors in its joints, enabling it to detect pressure applied from different angles simultaneously.
Next, they trained the robot with a machine learning algorithm to analyze and interpret the varying types of tension. This allowed the robot to distinguish different touch scenarios and identify specific locations where it was being touched. The method eliminated the need for artificial skin while still enabling the robot to sense detailed touch inputs.
In one experiment, the AI system became so refined that the robot could recognize which of the numbers painted on its arm was being pressed. In another, it identified numbers drawn on its arm by a person’s fingertip.
This breakthrough could lead to new ways for robots to interact, particularly in industrial settings where they often work side-by-side with humans.
We have helped 20+ companies in industries like Finance, Transportation, Health, Tourism, Events, Education, Sports.
