Robot hand “feels” pain and ignores harmless touch with a new sensory system

In the middle of the co-development of artificial intelligence and robotic progress, the development of technologies that allow robots to perceive and respond effectively to their environment as humans has become a crucial task. In this context, Korean researchers draw attention to the implementation of an artificial sensory nervous system which imitates the sensory nervous system of living organisms without the need for separate software or circuits. This revolutionary technology should be applied in fields such as in ultra-small robots and robotic prostheses, where intelligent and energy efficient responses to external stimuli are essential.

KAUST announced on July 15 that a joint research team led by the chair teacher endowed Shinhyun Choi from the School of Electrical Engineering in Kaist and Professor Jongwon Lee of the Semi-Conductive Department of Convergence of the National Chungnam University has developed an artificial new generation neuromorphic nervous system. This system imitates the functions of the sensory nervous system of a living organism and allows a new type of robotic system that can effectively respond to external stimuli.

The study is published in the journal Nature communications.

In nature, animals – including humans – safe or familiar stimuli and react significantly to safe stimuli or significantly to significant or dangerous stimulation. This selective response helps prevent unnecessary energy consumption while maintaining a rapid consciousness of critical signals. For example, the sound of an air conditioner or the feeling of clothes against the skin familiarized and is ignored.

However, if someone calls your name or a sharp object affects your skin, a concentration and a quick response occur. These behaviors are regulated by the “accustomed” and “awareness” functions in the sensory nervous system. Attempts have been constantly made to apply these sensory nervous system functions of living organisms in order to create robots that effectively respond to external environments such as humans.

However, the implementation of complex neural characteristics such as habitation and awareness in robots has been faced with difficulties of miniaturization and energy efficiency due to the need for distinct software or complex circuits. In particular, there have been attempts to use Memristors, a neuromorphic semiconductor.

A Memristor is a new generation electrical device which has been widely used as an artificial synapse due to its ability to store analog value in the form of device resistance. However, existing Memristors had limits to imitate the complex characteristics of the nervous system, because they only allowed simple monotonic changes in conductivity.

To overcome these limits, the research team has developed a new Memristor capable of reproducing complex neuronal response models such as habitation and awareness in a single device. By introducing an additional layer inside the Memristor which modifies the conductivity in opposite directions, the device can imitate more realistic the dynamic synaptic behaviors of a real nervous system – for example, decrease its response to repeated security stimuli but quickly find a sensitivity when a danger signal is detected.

Using this new Memristor, the research team has built an artificial sensory nervous system capable of recognizing touch and pain, and applied it to a robotic hand to test its performance. When safe tactile stimuli have been applied several times, the robot hand, which initially reacted with sensitivity to unknown tactile stimuli, gradually showed habitual characteristics by ignoring stimuli.

Later, when stimuli were applied with an electric shock, he recognized it as a danger signal and showed awareness characteristics by reacting again with sensitivity. Thanks to this, it has been proven experimentally that robots can effectively react to stimuli such as humans without separate software or processors, checking the possibility of developing neuro-inspired robots.

Vo-on Park, a researcher at KAIST, said: “By imitating the human sensory nervous system with new generation semiconductors, we have opened the possibility of implementing a new concept of robots which are smarter and more economical in terms of energy to respond to external environments.

“This technology should be used in various fields of fusion of new generation semiconductors and robotics, such as ultra-small robots, military robots and medical robots such as robotic prostheses.”