An AI-controlled limb able to teach itself to walk could lead to a new generation of adaptable prosthetics and robots.
Developed by a team from the University of South Carolina, the limb features animal-inspired tendons and bio-inspired algorithms. By moving the limb randomly (motor babbling), the robot learns the way it interacts with the environment, similar to the way babies learn. The robot can learn a new walking task within five minutes, adapting to changing terrain automatically.
The team believes this ability to adapt could have significant applications in responsive prosthetics that self-improve as well as space-faring robots able to adapt to the terrain of a distant planet.