Robots Get Artificial Skin That Feels Like Human Touch
▼ Summary
– The biological nervous system processes sensory information using complex, noisy streams of activity spikes that are integrated across many neurons.
– Researchers have created an artificial robotic skin using spiking circuitry, inspired by how sensory neurons transmit and integrate signals.
– The artificial skin system is designed to integrate with energy-efficient, spiking neural network chips for potential AI-based robotic control.
– The system uses flexible polymer skin with pressure sensors, converting sensor inputs into electrical activity spikes to mimic neural signaling.
– It encodes information primarily through spike frequency (for pressure intensity) and uses other spike characteristics like shape and magnitude to identify the specific sensor source.
The development of a robotic skin capable of sensing touch like human skin represents a significant leap forward in robotics and prosthetics. This innovation hinges on mimicking the complex way our nervous system processes sensory information. Researchers have now engineered an artificial skin that uses spiking neural networks to translate pressure into electrical signals, much like biological neurons communicate through bursts of activity. This approach allows the system to pinpoint the location and intensity of touch while integrating seamlessly with energy-efficient AI hardware designed to process these specific signal types.
Our biological skin is a marvel of sensory engineering. It contains a dense network of specialized receptors that detect everything from a light brush to painful pressure. These signals travel through neurons to the spinal cord for initial, reflexive processing before potentially being relayed to the brain for conscious perception. The recent breakthrough involves creating a synthetic version of this sophisticated system for application on robotic hands.
A team based in China constructed a flexible polymer skin embedded with an array of pressure sensors. These sensors connect to the processing system through conductive polymers. The core innovation lies in the next stage: converting the raw sensor data into a stream of electrical spikes. In biological systems, neurons communicate through these brief pulses of activity, and the artificial system adopts a similar language.
Information within these spike trains can be encoded in several ways: the shape of an individual pulse, its magnitude, its duration, or the frequency at which spikes occur. The research team primarily uses spike frequency to convey the intensity of pressure detected by a sensor, mirroring the most common method found in nature. The other characteristics, pulse shape, magnitude, and duration, are cleverly combined to generate a unique identifier for each sensor, functioning like a neural barcode. This allows the system to not only know how much pressure is applied but also exactly where on the skin the contact occurred.
This design enables the artificial skin to perform localized processing. It can identify the site of an input and even detect potential damage, much like our nervous system alerts us to injury. While the current prototype is limited to sensing pressure, the underlying architecture that borrows principles from neurobiology paves the way for adding other sensations like temperature or texture in the future. The use of spiking circuitry is particularly promising because it is compatible with emerging, low-power neuromorphic computer chips, which are engineered to run artificial intelligence models using this very type of signal. This synergy could lead to more responsive and energy-autonomous robots or highly intuitive prosthetic limbs.
(Source: Ars Technica)
