The range of actions that robots are able to take is constantly growing and expanding – the latest discovery is an algorithm that allows a robot to learn how to keep moving, even if one or several of its limbs are non-functional. This could be an absolutely life-saving ability whenever robots are used to enter disaster areas where humans can’t go.
Jean-Baptiste Mouret of the Pierre and Marie Curie University in Paris explains: “The idea is to have robots that can survive in hostile environments such as a Fukushima-type nuclear disaster”. The results of their work was published in the journal Nature.
By closely studying how animals learn to use three legs instead, if the fourth leg is injured, scientists could determine how to create the logical ways for robots to reason in a similar situation. By evaluating how its body moves and try out the remaining options, the test robots have learned in under 2 minutes to adjust to the lost functionalities and find a new way to use their legs or arms and continue on with their assigned mission.
The scientists taught the robot to use predetermined “values” for its different body parts, making it similar to using intuition to decide what to try out next if its damaged. It wouldn’t be useful for a dog to use its nose to compensate for an injury in the paw, for example, so by assigning similar values to equally useful parts of the robots, it knows that if one leg goes, the other three or five could be valid alternatives.
Mourets colleague, Antoine Cully, talks about their new, exciting progression in the robotics field of science: “What’s surprising is how quickly it can learn a new way to walk. It’s amazing to watch a robot go from crippled and flailing around to efficiently limping away in about two minutes.” The algorithm is called Intelligent Trial and Error.
Besides from watching one of their test robots learning to use its other legs when two of them had been broken, they’ve also observed another one of their robots that used a flexible arm to drop a ball into a bin, and after having several of its joints broken, it could quickly learn to keep performing its task again after evaluating its different possibilities from there.
Image: JD Hancock