A new set of abilities to learn to adapt to ‘injuries’ makes robots more effective in quickly recovering from damage and keep performing their task.
Until recently, the sturdiness or robots and their stubborn abilities to survive any environment, regardless of how harsh it would seem for a human, was overthrown by the fact that once they sustained damage, they would stop functioning.
They were not programmed, much like animals and human are, to adapt to damaging challenges. Now, thanks to a team of robotics specialists at the Pierre and Marie Curie University of Paris, robots won the capacity to self-heal or adapt by easily accessing simulated damaging life-experiences.
That is good news for the fields that would undoubtedly need increasingly autonomous and efficient robots. It could extend to natural disaster relief programs, dangerous factory-related tasks that could be undertook by the enhanced creations.
During the experiments for instance, one robot featuring no less than six legs had two of them damaged. Under normal conditions, it would fail to complete its task, not understanding what had happened. Now, it took the robot a minute to recover and adapt to a new pattern of action.
Jean-Baptiste Mouret and Antoine Cully, authors of the study and robotics specialists at the Piere and Marie Curie University developed the program that allows a robot to adapt to damage by learning, as humans and animals do, via trial and error.
Once the program succesfully learned, a robot should be able to simply adapt without the need for self-diagnose sensors or other installed contingency plans.
Both authors emphasized the added value that their results could bring to the world. Robots that can work in natural disaster relief, or be sent to toxic environments inaccessible to humans without human supervision are indeed a goldmine. They function for longer, barely require maintenance and are highly cost effective.
The pattern was initially observed in both humans and animals. Once we hit out leg for instance, we intuitively learned that limping surely eases the pain. However, this is knowledge obtained from trial-and-error experiences.
Referring to animals, Mouret stated:
“These intuitions allow them to intelligently select a few, different behaviors to try out and, after these tests, they choose one that works in spite of the injury. We made robots that can do the same.”
So, how does it work? Once a robot is selected to perform a certain task where damage might infer, a computer simulation maps out all the motions the robot is able to perform and inform it as to what has the highest chances of working in spite of damage sustained.
The map in fact substitutes the real-life trial-and-error scenarios. According to the scientists, it also enables the robot to rapidly adapt as it doesn’t need to have information on why it is unable to perform a task and analyze that reason. Instead it knows that performing the task is the main goal and chooses from optimal patterns of action.
The robots will be present in Pomona, California where they will compete in the Defense Advanced Research Projects Agency (DARPA) Robotics Challenge. Good luck!
Image Source: cnbc.com