As part of a broader project that designs machines capable of mimicking the adaptive qualities of humans and animals, scientists have uncovered on Wednesday a new tiny robot that is able to continue pursuing the given mission even after suffering damage.
Researchers hope to one day provide the world with first-responder robots which would not be stopped or put out of function by dangerous crisis situations.
Leading scientist Jean-Baptiste Mouret of the Pierre and Marie Curie University in Paris reported that the feat is trying to create robots with survival capabilities when faced with hostile environments, even if it’s the nuclear disaster kind.
Sending robots in recon mission inside a reactor hasn’t been possible before, as they would be damaged and become inactive. The article posted in the journal Nature was inspired by the adaptive qualities the natural world is presenting.
For example, scientists noticed that if a dog’s paw has been hurt, the dog will alter the way it walks in order to avoid pressure on the painful wound. Learning to become adaptive can only come from experience. When you know how a body moves or reacts, the options that can be explored are endless.
This ability is what Mouret’s team tried to clone in the six-legged 50-centimeter robot, which is guided and powered by an ingenious computer program. They attempted to teach the program to create a database with information on how the robot moves, and then learn how each movement can be used in a crisis.
These bits of knowledge are called “values,” or in other words, hints of intuition. In the case of the wounded dog, for instance, the animal is aware that in order to walk again, it has to shift its weight according to the levels of pain.
This is how Intelligent Trial and Error was then created, a learning algorithm comprising a collection of “values.” Researchers proceeded to conduct experiments to see if the robot would use compensatory behavior in order to continue its mission even when damaged.
It was impressive to see that, in the end, the prototype robot learned that it could walk even if it was missing two legs, relying on an intelligence which rules out a walking type if it becomes useless – if walking mostly on its hind legs seems to hinder the mission, the robot will switch to waking mostly on its front legs.
The most surprising feat was getting the damaged robot up and running in less than two minutes after it was crippled. Limping away without six functional legs became the new norm in no time.
Image Source: Pacific Standard Magazine