Researchers built robotic skin with a sense of touch

It's powered by Intel's brain-mimicking neuromorphic chip Loihi.


Using Intel’s neuromorphic chip, Loihi, researchers from the National University of Singapore (NUS) developed an artificial skin that allows robots to detect touch 1,000 times faster than the human sensory nervous system. The system can also identify the shape, texture and hardness of objects 10 times faster than the blink of an eye. The researchers believe this work could improve human-robot interaction, making things like caregiving robots and automated robotic surgery more feasible.

The NUS team first taught a robotic hand fitted with their artificial skin to read Braille. The robotic hand passed the tactile data to Loihi, which translated it. Loihi was over 92 percent accurate and used 20 times less power than a standard Von Neumann processor.

The researchers then combined tactile and visual data. They taught the robotic hand to classify containers using both the artificial skin and an event-based camera. They sent the data to a GPU and to Loihi and demonstrated that combining event-based vision and touch, via a spiking neural network, made the system 10 percent more accurate than it would have been with visual data alone.

They also found that Loihi processed the sensory data 21 percent faster than the best-performing GPU, while using 45 times less power. The researchers presented their findings at Robotics: Science and Systems this week.

Robotic hands have come a long way in the past decade, and grippers are becoming smarter and more delicate. While some robots are capable of tasks like sorting plastic and paper recycling by touch, the next hurdle in robotics will be combining a sense of touch and sight, which makes this NUS research encouraging.

“We’re excited by these results,” said Harold Soh, an assistant professor at the NUS School of Computing. “They show that a neuromorphic system is a promising piece of the puzzle for combining multiple sensors to improve robot perception. It’s a step towards building power-efficient and robust robots that can respond quickly and appropriately in unexpected situations.”