Connect with us

News

Artificial cerebellum can learn to control a robot’s movement

Published

on

Researchers at the University of Granada in Spain, working in the Human Brain Project, have designed an artificial neural network that mimics the structure of the cerebellum, one of the evolutionarily older parts of the brain, which plays an important role in motor coordination.

When linked to a robotic arm, their system learned to perform precise movements and interact with humans in different circumstances, surpassing the performance of previous AI-based robotic steering systems.

Researcher Ignacio Abadía said it is the most biologically realistic and detailed model of the cerebellum to date that is capable of work in real-time, and replicates not only aspects of the structure, but also the adaptability and capacity to learn.

“Our bio-inspired robotic controller responds to technological challenges on last generation collaborative robots by taking advantage of the millions of years that allowed evolution to find the most efficient biological solutions. This is at the essence of research fields such as neuromorphic engineering, computational neuroscience and neurorobotics ”

By taking inspiration from the brain in this way, the scientists were able to solve one of the common technological challenges in robotics: Their cerebellar spiking neural network enables the robot to deal with so-called latency, or time delays, which is a central real-world problem for computational systems in robotics, especially during wireless or remote steering.

Research leader professor Eduardo Ros said the work could also help to control new bio-inspired robots, which are equipped with elastic and flexible components that replicate the muscles and tendons of the human body.

Such ‘co-bots’ are safer for human interaction, but their flexibility makes it difficult to use classical control techniques.

“As next steps are integrating our cerebellar model with spinal cord models in collaboration with EPFL (Auke Jan Ijspeert and Alice Bruel) in order to better understand how spinal cord and cerebellum complement each other in tasks requiring accurate motion control.”

The work of the Applied Computational Neuroscience research group has been published in the journal Science Robotics.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Health tech alerts



Sign up for HT World updates

Copyright © 2021 Aspect Publishing Ltd.