University of Southern California’s researchers have designed a robotic leg that learns to walk by itself through trial & error, pretty much like an animal. USC Viterbi School of Engineering’s research team feel that they are the first to develop a leg inspired by animal-like tendons. The robotic leg can gain its balance after being tripped so that it doesn’t fall. The robot was never programmed to recover its footing.
Professor Francisco J. Valero-Cuevas along with others have come up with a bio-inspired algorithm that has the ability to learn a novel task of walking on its own post five minutes of disorganized play. It can also take up various other tasks without the requirement of additional programming.
Their paper was released in Nature Machine Intelligence, introduces new possibilities for comprehending the movement & disability of human, developing responsive prosthetics, and also robots that plays with complicated & transforming surrounding such as search & rescue and exploration of space.
Francisco, who is the lead author says that it currently, it requires around months or years of training for a robot to get involved in the world. However, they want to accomplish fast learning & adaptations as witnessed in nature.
Paper’s lead author Marjaninejad explained that this advancement is similar to natural learning which takes place in babies. The robot was permitted to comprehend its surroundings in a free play process.
Leg’s random movements permits the robot to develop an internal map of its limb & interactions with the surroundings. Authors of the paper suggest that robots learn about things by actions sans any parallel computer stimulations to assist learning.
Additionally, this is specifically essential due to the reason that programmers can assume & code for several situations. However, not for all the situations, therefore robots that are pre-programmed are prone to failure. But if the novel robots take some learning from appropriate experience, they will ultimately discover a solution. Once it is found can be used & adapted as required. The solution needn’t be ideal, but it will be implemented if it is applicable for the situation.