When we pick up an object, we can adjust our grip if that object proves to be a slippery customer and we can use a combination of senses to determine if an object needs a delicate touch. ETH researchers have developed a sensor that could help robot grippers to do the same.
The prototype comprises a silicone skin with colored microbeads and a camera pointing up. When the skin comes into contact with an object, the pattern of microbeads changes and this is picked up by the camera.
The researchers set up experiments to precisely control contact with the sensor, including varying the size of the object, its location on the sensor skin and the force applied. Machine learning was then called upon to help match different types and levels of contact with registered changes in microbead patterns, and train the sensor system to accurately register the vertical and lateral forces applied using the experimental dataset.
"Conventional sensors register the applied force at only a single point," said doctoral student Carlo Sferrazza. "By contrast, our robotic skin lets us distinguish between several forces acting on the sensor surface, and calculate them with high degrees of resolution and accuracy. We can even determine the direction from which a force is acting."
The team says that such a system should be fairly inexpensive to produce, and sees potential in giving robot grippers the ability to adapt grip force based on the object being handled, for example, and make adjustments on the fly when an object seems to be slipping away. The system could also be used by athletes or fitness enthusiasts to measure the level of force being applied to pedals or running shoes. There could be a future for the tech in virtual reality setups, too, to provide gamers with tactile feedback.
For the moment, the prototype measures just 5 x 5 cm (1.96 x 1.96 in) and is 1.7 cm (0.66 in) thick. The researchers are working on extending the system to cover a larger surface area, monitored by several cameras, and are also aiming to reduce the thickness of the sensor.
A paper on the project has been published in IEEE Access. The video below has more.
Source: ETH Zurich