A group of researchers from the National University of Singapore (NUS), has developed an artificial skin that gives robots the ability to detect a touch a thousand times faster than the human sensory nervous system.
The system also makes it possible for machines to identify the shape, texture and hardness of an object, ten times faster than the human eye.
Those responsible believe that this project can raise the level of human-machine interaction, opening doors, for example, the creation of caregiver robots (elderly, babies, animals) and the development of remote surgery technologies.
In a first phase of testing, the NUS team equipped a robotic hand with this artificial skin to read Braille.
The device transmitted the collected data to the neurometric chip Loihi, developed by Intel, which later translated it.
The chip reached 92% accuracy.
After this test, the researchers combined visual and tactile data to teach the robot to classify containers, based on what it felt, through artificial skin, and what it saw, through the camera.
The combination of these two types of data allowed the machine to be 10% more accurate than in cases where it relied exclusively on visual data.
Another promising sign was the fact that Loihi was able to process sensory data 21% faster than the best GPU on the market while consuming 45 times less energy.
The conclusions of this study were recently presented at Robotics: Science and Systems, a specialized conference.
"We are very excited about these results.
They show us that a neuromorphic system is a promising part of the puzzle needed to combine the multiple sensors that can improve a robot's perception.
A step that allows us to approach robust robots capable of respond quickly and appropriately to unexpected situations, "said Harold Soh, assistant professor at the Singaporean university's computer school.
