Artificial nervous system senses light and learns to catch like humans - New Scientist News

sekirta.blogspot.com

An artist's impression of neurons — An artist’s impression of neurons

KTSDESIGN/SCIENCE PHOTO LIBRARY

A simple artificial nervous system is able to mimic the way humans respond to light and learn to perform basic tasks. The principle could be used to create more useful robots and prostheses.

Humans, when confronted by external stimuli such as heat or light, can react rapidly and automatically – think about how your hand withdraws from a hot surface, or how your leg flicks up when tapped on the knee. These are unconscious responses. But conscious responses, such as catching a ball, must be honed by repeated stimulation.

Researchers at three universities in South Korea have developed an artificial system capable of simulating a conscious response to external stimuli. It consists of a photodiode – which converts light into an electrical signal, a transistor acting as a mechanical synapse, an artificial neuron circuit, which acts as the system’s brain, and a robotic hand.

When the photodiode detects light, it sends an electrical signal through the transistor that the light is on. That signal is carried to the artificial neuron circuit. There, the message is received, and that circuit then learns how to respond to the signal, sending a command to a robotic hand it controls.

At the same time as the light is turned on, starting the whole process off at the photodiode, a ball is dropped from above the hand. The idea is for the contraption to learn how to cup the hand quickly enough to catch the ball.

The process is similar to the way our eye transmits electrical signals via synapses to our brain, which then translates those signals, decides on a course of action and sends a command to muscles to move – all within a fraction of a second.

In the early stages of the experiment, the brain of the system was slow to translate the light signal into a decision to cup the hand. Before “learning” how to react, the system took 2.56 seconds to do this. After it had been exposed repeatedly to the light signal and allowed time to process what to do, this decreased to 0.23 seconds. The researchers say the artificial neural system is imitating something like a conscious biological response.

The system isn’t the first to try to mimic the biological response humans have to external stimuli. A paper in 2018 detailed attempts to recreate sensory neurons within skin, while a 2019 paper focused on the development of artificial synapses. One of the authors of that paper even used an artificial nervous system to control cockroach limb movement.

One of the goals of this type of research is to help people with neurological conditions regain control of organs and limbs that they can’t control as quickly as before.

“The operation of the device shows great promise, especially in human assistance tasks, or in training robotic systems based on human movement,” says Jonathan Aitken at the University of Sheffield, UK.

Aitken believes the system could be combined with wearables tracking how people move to create robots trained to behave in a similar fashion. It could, for example, allow robots to do manual tasks that require responding to external circumstances.

Journal reference: Science Advances, DOI: 10.1126/sciadv.abe3996