A team of scientists from the Korea Advanced Institute of Technology (KAIST) has created an artificial sensory nervous system for robots that allows them to distinguish between safe and potentially dangerous touches, imitating the behavior of living organisms. This was reported by the publication Interesting Engineering.
This technology is built on a neuromorphic semiconductor — a new type of memristor that allows it to respond to external stimuli without the need for complex software or large computing power.
In animals and humans, familiar stimuli, such as the noise of an air conditioner or clothing on the body, are gradually ignored, while novel or dangerous signals, such as sudden touch or pain, trigger an immediate response. This mechanism is regulated through two processes: habituation (silencing the response to repeated safe stimuli) and sensitization (increasing the response to threatening stimuli). The KAIST team was able to recreate these mechanisms electronically.
The innovation lies in the special design of the new memristor. Unlike previous models that could only change conductivity in one direction, this device has an additional inner layer that allows it to change conductivity in the opposite direction. This allows the memristor to reduce sensitivity to repeated touches while at the same time quickly activating in the event of a painful signal.
To test the development, the researchers installed the system in an artificial hand. At first, it responded to every touch. Over time, if the touches were repeated without threat, the reaction decreased. When the same touches were combined with an electric current, the hand became very sensitive again.
This breakthrough demonstrates that artificial nervous systems can be compact, energy-efficient, and operate without complex software, paving the way for a new era of robotics.