Whether you're Beyonce or Kelsey Grammer, everyone looks like a hot mess when they fall. However, researchers at the Georgia Institute of Technology may have found a way to make sure robots can at least look good while doing it.
Georgia Tech PhD graduate Sehoon Ha and Professor Karen Liu have come up with a new algorithm that communicates to a robot how to react during a variety of falls. This research, titled "Multiple Contact Planning for Minimizing Damage of Humanoid Falls," was initially presented earlier this month at the IEEE/RSJ International Conference on Intelligent Robots and Systems in Hamburg, Germany.
The researchers used a BioloidGP humanoid to test out the algorithm and show how it helped the robot fall more gracefully after a slight nudge, heavier push and a really big shove, which you can see for yourself in the video below. In each case, the algorithm helped the robot land on its hands, knees or roll over onto its back instead of just falling flat on its face.
Aside from saving robots the embarrassment of making an ugly fall, if robots have a cleaner stumble, it could help minimize damage to them, which could not only help save money but could also make them safer as robots become increasingly used in health care or domestic situations around the elderly, children, the injured or pets.
"We believe robots can learn how to fall safely," Ha said in a press release. "Our work unified existing research about how to teach robots to fall by giving them a tool to automatically determine the total number of contacts (how many hands shoved it, for example), the order of contacts, and the position and timing of those contacts. All of that impacts the potential of a fall and changes the robot's response."
This new research builds upon Liu's previous work studying how cats change their bodies when they fall, which revealed the importance of the angle of landing in a fall and that robots had the capability to achieve a softer landing.
"Our new planning algorithm takes into account the hardware constraints and the capabilities of the robot, and suggests a sequence of contacts so the robot gradually can slow itself down," Liu said in a press release.
Now if only there was a way to teach humans to fall more gracefully. But then again, if that were the case, what would we watch on YouTube?