In a robotics breakthrough reminiscent of a science fiction movie, computer scientists have developed a couple of robots capable of rapidly compensating for injuries, according to a recently published study.
The experiments, which were published in the journal Nature, led study co-author Jean-Baptiste Mouret of the Pierre and Marie Curie University in Paris to tell Live Science that he and his colleagues “were surprised” by the “extent of damage” which the robots were able to adapt to.
One thing we were surprised by was the extent of damage to which the robots could quickly adapt to (…) We subjected these robots to all sorts of abuse, and they always found a way to keep working.
Experimenting with a six-legged robot, the researchers created a six-legged creature capable of adapting to damaged, broken, or missing legs in less than two minutes. According to the researchers, this feat was accomplished by tapping into experiences from simulated lives.
A robotic arm employed by the researchers during their delve into self-healing robotics was able to learn how to successfully place objects in the correct position with several broken motors or joints.
As for the six-legged robot, it proved capable of walking even after up to two of its legs were damaged. A notion which, when combined with other recent advances in the fields of robotics and artificial intelligence, have lead some to question the safety precautions being taken. After all, who wants to live in a Terminator style reality? Surely, not SpaceX founder Elon Musk who donated $10 million to artificial intelligence safety or Stephen Hawking who previously warned of such a doomsday scenario.
In order for the six-legged walker to continue limping along with its damaged limbs, it employs an on-board camera used to detect that its movements have either encumbered or that it’s being prevented from walking in a straight line. Then, instead of diagnosing the problem, the six-legged bot simply attempts new patterns of motion until it finds one which enables it to restore an acceptable level of performance. The method serves as what Mouret refers to as a “shortcut” in the sense that it reduces the time it takes for the robot to adapt; this based on the 2006 breakthrough by roboticist Josh Bongard at the University of Vermont in Burlington who, along with his colleagues, created a six-legged robot capable of self-diagnosing injuries and slowly adapting after calculating new motion patterns which enabled it to resume operations.
