This website uses cookies primarily for visitor analytics. Certain pages will ask you to fill in contact details to receive additional information. On these pages you have the option of having the site log your details for future visits. Indicating you want the site to remember your details will place a cookie on your device. To view our full cookie policy, please click here. You can also view it at any time by going to our Contact Us page.

Novel learning rule explains the development of 'sensorimotor intelligence'

28 October 2015

Using a novel learning rule, researchers demonstrate how robotic systems controlled by a neural network spontaneously develop self-organized behaviours.

Simulated humanoid robot in a crawling position (courtesy of IST Austria)

A robot is able to explore its physical possibilities and surroundings and subsequently develop different self-taught behaviours without any instructions. In a paper published in The Proceedings of the National Academy of Sciences, Professor Ralf Der from the Max Planck Institute and Georg Martius, a Fellow at the Institute for Science and Technology (IST Austria), demonstrate the emergence of this 'sensorimotor intelligence' in robots based on their proposed learning rule.

How brains or artificial neural networks develop autonomous, self-directed behaviour is a fundamental challenge for both neuroscience and robotics. Traditionally, the self-organized development of behaviour is explained by using concepts such as intrinsic motivation or curiosity. However, in their paper, Der and Martius argue that the emergence of such behaviour can be grounded directly in the synaptic plasticity of the nervous system.

To test their hypothesis, the authors use bioinspired robots consisting of a humanoid and a hexapod robot in physically realistic computer simulations. The robots receive sensory input from their bodies but are not given any form of instruction or task. What can then be observed is a rich spectrum of rhythmic behaviours of the robots as they explore various movements.

Solely because of the tight coupling of environment, body, and brain (in this case an artificial neural network), the robots can obtain feedback from their situation and adapt quickly. This, together with a simple, learned self-model, allows them to develop a form of sensorimotor intelligence.

Different scenarios show how they acquire the ability to crawl, walk on changing surfaces, or even cooperate with another robot. The authors explain this phenomenon with the proposed synaptic plasticity, a coupling mechanism that allows a simple neural network to generate constructive movements for almost any given body.

Potentially, this concept can also lead to a new understanding of the early stages of sensorimotor development in the natural world and even be used to elucidate some evolutionary saltations.

“It is commonly assumed that leaps in evolution require mutations in both the morphology and the nervous system, but the probability for both rare events to happen simultaneously is vanishingly low," says Georg Martius. "But if evolution was indeed in line with our rule, it would only require bodily mutations — a much more productive strategy. Imagine an animal just evolving from water to land: Learning how to live on land during its own life time would be very beneficial for its survival.”

Print this page | E-mail this page

Coda Systems