While our brain makes sense of the external world using 'chaotic' electrical bursts, much like a fireworks display, rather than the regular structure of computers, a China-based research team developed a neuromorphic electronic skin called ‘NRE-skin’ for robots, capable of sensing pain and injury, inspired by these principles of the human nervous system.
The system was designed not to directly copy the working principles of biological neural networks, but rather to be inspired by these mechanisms, and was specifically configured to work with energy-efficient AI hardware.
Pressure and Location Detection in the Same System
The developed artificial skin was built on a flexible polymer layer designed to cover a robot hand. Pressure sensors embedded within this layer were connected to the system via conductive polymers. Although researchers have currently limited sensing to pressure only, they created a multi-layered structure capable of simultaneously evaluating the intensity, location, and potential damage of a stimulus, similar to the nervous system in human skin.
Analog pressure data from the sensors is converted into "spike" signals, consisting of short electrical pulses similar to those in humans, in the system's next layer. These signals can carry information in four different ways: pulse shape, amplitude, duration, and frequency. As in biological systems, the intensity of the pressure in this artificial skin is encoded by the spike frequency. Other features are used like a barcode to identify which sensor the signal originated from.
The system not only detects pressure but also monitors whether the sensors are operational. Each sensor sends a "I am here" signal at regular intervals. The interruption of this signal indicates a malfunction or damage in the respective sensor.
Robots Will Be Able to Give Reflex Responses
The second layer, where spike signals are processed, accumulates incoming data and compares it to a predefined pain threshold. When this threshold is exceeded, the system automatically generates a "pain signal." This allows reflex-like responses to be triggered without the intervention of higher-level control units. Researchers successfully demonstrated that a robot arm covered with this artificial skin automatically retracts when exposed to damaging levels of pressure. Experiments also showed a robotic face changing its expressions based on the pressure applied to the robot arm.
To facilitate practical use of the system, the research team adopted a modular skin design. The electronic skin consists of segments held together by magnetic locks. These segments automatically establish the necessary electrical connections and each emits a unique identification code. Thus, when damage is detected, the faulty part can be easily removed and replaced with a new one.
Researchers call this system "neuromorphic robotic e-skin" (NRE-skin). However, the neuromorphic approach used here is based on a simpler definition inspired by the nervous system's operating logic, rather than being a direct model of it. For example, while the biological nervous system uses a direct positional map on the body, NRE-skin encodes positional information into the characteristics of spike signals. In this respect, although the system is inspired by biology, it is not a direct copy of biological processes.
In its current state, NRE-skin can only detect pressure. However, real human skin can simultaneously process temperature, cold, chemical irritation, and many other different stimuli. While adding such perceptions is theoretically possible, it would require parallel processing layers to prevent signals from interfering with each other.
0 Comments: