Singapore National University is trying to give robots the sense of touch. The National University of Singapore, a member of the Intel Neuromorphic Research Society, presented new findings that, together with Intel’s neuromorphic processor, they could bring event-based touch sensing to robots.. They’re trying to teach robots to detect objects’ hardness, shape, and touch. Bringing a sense of touch to robots will significantly improve the ability and functionality of robots compared to today’s visual systems. Furthermore, these studies highlight that neuromorphic processors outperform traditional methods in processing sensory data.

Human touch is sensitive enough to feel the difference between surfaces. National University of Singapore researchers are working to allow robots to sense 1,000 times faster than the human nervous system. They hope to achieve this by using artificial leather, which can determine the shape, texture and hardness of objects 10 times faster than the blink of an eye.. For example, robotic arms equipped with artificial leather can easily adapt to sudden changes in factory-made products.. The ability to sense and perceive the environment better, giving robots a sense of touch, could allow robots to be used in more critical tasks in the surgical field. needs a chip that can run it efficiently. In short, an artificial brain is needed to perform perception and learning.

Intel Loihi Neuromorphic Chip 21 Times Faster

The University of Singapore team has used Intel’s Loihi neuromorphic research chip to provide sensory data from artificial skin. plans to process. In their first experiment, the researchers used a robotic hand equipped with artificial skin to read Braille.. Passing tactile data to Loihi, they transformed microbumps felt by the hand into something meaningful.. Loihi achieved more than 92 percent accuracy in classifying Braille letters while using 20 times less power than a standard Von Neumann processor. They further enhanced their robotic sensing capabilities by combining them in a neural network.. To do this, they sent sensory inputs to the robot through artificial skin and used an event-based camera.

After capturing this sensory data, the team ran two experiments with the GPU and Intel’s Loihi neuromorphic research chip to compare its processing capabilities.. Intel Loihi processes sensory data 21 percent faster than the top-performing GPU, while requiring 45x less power.

“We are excited by these results.. Combining multiple sensors to improve robot perception of a neuromorphic system shows a promising piece of the puzzle. This is a big step towards creating power-efficient and reliable robots that can respond quickly and appropriately to unexpected situations,” said Harold Soh. .