Fuda van Diggelen, Nicolas Cambier, Eliseo Ferrante, and Guszti Eiben published an article on robot learning in Nature Communications. Their work is inspired by the fact that many animals can walk within minutes after birth. They propose a novel approach to learn locomotion skills in robots with various morphologies, similar to how motor commands from the brain initiate locomotion patterns in the spinal cord. The new method achieves good locomotion within minutes without the need of a detailed model of the robot or environment, and is particularly useful for learning multiple skills in parallel on any arbitrary robot design. The paper has been selected as editor’s highlight in Nature Communications.
To read the open access article, click here.
For a digestible blogpost, please click here.