AI robotic leg teaches itself to walk

Researchers from the University of Southern California have developed a robotic leg that learns to walk through trial and error, much like an animal.

AI robotic leg teaches itself to walk

The team, from the USC Viterbi School of Engineering, believe they are the first in the world to create a leg driven by animal-like tendons, which can even recover its footing after being tripped so it doesn’t fall.

The algorithm operating the leg can learn a new walking task after just five minutes of ‘play‘ – making random movements that build up a map of the limb and its interactions with its environment.

“Nowadays, it takes the equivalent of months or years of training for a robot to be ready to interact with the world, but we want to achieve the quick learning and adaptations seen in nature,” said Professor Francisco J Valero-Cuevas.

Researchers hope that the technology could have several possible uses, including aiding our understanding of human movement and disability, and creating robots that can navigate difficult and rapidly changing environments (for tasks like disaster recovery).

“If you let these robots learn from relevant experience, then they will eventually find a solution that, once found, will be put to use and adapted as needed,” said Marjaninejad.