PC - Teaching robots how to fall with grace

News Bot

Your News Bitch
Console: Headset:
10-15-2015 04:01 PM

For a robot, a fall can be devastating (and costly). It sounds silly, but the damage from a robot tipping over and doing a face-plant can result in a broken components and/or motors. As robots move from safely wheeling around an office to walking in a number of environments, precautions need to be taken to make sure they’ll survive a hurdle as small as a fall.*Researchers Sehoon Ha and Karen Liu from Georgia Tech have developed an algorithm, giving robots the information they need to correct themselves and fall with grace.
“When you fall down, you try to dissipate energy,” Liu said in an interview with Technology Review. “And every time you make contact with the ground, some of that energy is dissipated.”*In the video below, the researchers demonstrate how the algorithm works. The robot on the bottom of the screen is not using the algorithm:
[embedded content]
This research mirrors what the folks over at Boston Dynamics, which is now part of Google, are doing with their BigDog and smaller, more domesticated Spot robot. The developers happily kick and send the robot over slippery surfaces, and the machine manages to correct itself.*But for all the correcting it can do in these scenarios, there’s always a need to consider how to prepare it for the inevitable fall. The developers say they found making the robot go limp, rather than tensing up, was the best option to protect its internal components.
Liu plans to expand the falling algorithm to make sure when a robot does fall, it doesn’t do so on a human. It’s a complex system of factors. “That’s why we have reflexes,” Liu says. “We are thinking of building something like a nervous system for robots.”
Image credit: Christopher Schmidt/Flickr
This entry passed through the Full-Text RSS service - if this is your content and you're reading it on someone else's site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.