A few weeks ago, I slipped in the shower, and after a cartoonish sequence of body contortions, I caught myself while landing. Robots aren’t as lucky. As June’s DARPA Robotics Challenge taught us, when machines fall, they tumble terribly and without the ability to brace themselves.
But those days might be over thanks to a computer program created by scientists at the Georgia Institute of Technology in Atlanta.
A new algorithm from allows robots to fall with style. Video by Georgia Institute of Technology
As IEEE Spectrum describes, the new algorithm uses “techniques adapted from judo,” allowing a robot to learn how to position its appendages while tumbling.
So rather than the full impact being felt by a single part of the robot, the robot can displace the kinetic energy created during the fall over multiple parts of its body. By learning how to tumble, robots reduced impact intensity to the head by 30 to 90 percent.