Advertisement

Robots are learning to fall with grace

"A fall can potentially cause detrimental damage to the robot and enormous cost to repair," said researcher Sehoon Ha.

By Brooks Hays

ATLANTA, Oct. 14 (UPI) -- Robots aren't yet perfect. As they learn new skills, they're likely to be even less perfect. For this reason and others, it's important for robots to learn to fall and fail gracefully.

At Georgia Tech, engineers are working to do just that -- teaching robots to fall.

Advertisement

One of the physiological skills that robots have had trouble picking up is improvisation. When a person begins to lose their balance, their innate reflexes kick in, allowing them to quickly re-calibrate their equilibrium and adapt -- often preventing a slip or fall.

Even when humans fall, most are able to adopt postures to protect themselves and dampen the landing. Robots traditionally lack this ability, too.

RELATED MIT's egg-clutching robot has soft but steady hands

But a new algorithm, designed by Ph.D. graduate Sehoon Ha and Professor Karen Liu, is helping to change that, empowering robots to react to their tumbles and prevent damage.

"A fall can potentially cause detrimental damage to the robot and enormous cost to repair," Ha, a 2015 graduate of Disney Research Pittsburgh, in Pennsylvania, said in a press release. "We believe robots can learn how to fall safely."

Earlier this month, Ha and Liu presented their work at the IEEE/RSJ International Conference on Intelligent Robots and Systems, held this year in Hamburg, Germany.

Advertisement
RELATED MIT engineers build, test bartending robots that work together

"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," Ha added. "All of that impacts the potential of a fall and changes the robot's response."

The researchers' algorithm was informed by observing the falling skills of cats, sequencing and transcribing the subtle movements that make up a fall and a reaction to a fall.

"From previous work, we knew a robot had the computational know-how to achieve a softer landing, but it didn't have the hardware to move quickly enough like a cat," Liu said. "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."

RELATED Water insects inspire tiny jumping robot

Latest Headlines

Advertisement
Advertisement

Follow Us

Advertisement