# X2-VelociRoACH Smashes Speed Record for Tiny Legged Robots

“In this work we explore increasing the maximum attainable speed of a legged robotic platform by pushing its stride frequency to an extreme value.” Need we say more? Okay, how about this: you’re about to see one of the fastest legged robots ever built.

Don’t blink, because if you blink, you’ll miss UC Berkeley’s X2-VelociRoACH zipping past. I’m not exaggerating even a little bit, just watch:

It only weighs 54 grams, but the X2-VelociRoACH is running here at 4.9 meters per second (17.6 km/h, or 11 mph), sustainably. This is by far the fastest legged robot of its size, although as far as legged robots go, we’re obligated to point out that it’s slower than Boston Dynamics’ enormous, complicated, and expensive hydraulic quadrupeds.

Making VelociRoACH go faster was, in some sense, relatively straightforward. Legged animals have two different ways of increasing their speed: either they increase their stride frequency (taking more steps over a given distance), or they increase their stride length (taking longer steps). Most animals, real cockroaches included, have a maximum stride frequency imposed on them by their musculoskeletal systems, and so they tend to prefer to run at a consistent stride frequency while adjusting their speed by varying stride length. VelociRoACH doesn’t work that way, because its legs are coupled directly to its motors, so stride length can’t be dynamically varied. The only option to make the current design go faster is to just start cranking the motors up to increase stride frequency and see what happens, which is what the researchers did.

The X2-VelociRoACH maxed out at a stride frequency of about 45 Hz, meaning that the robot was completing 45 stride cycles every second. The upper limit on the stride frequency is due to structural failure of the robot itself. In other words, it can go a bit faster, except that when it does, pieces of it start to fly off. The researchers did their best to robustify key components, employing 3D printed plastic and even carbon fiber reinforcement when necessary. The legs are fiberglass instead of rubber, and the material for flexible joints (previously PET film) is now ripstop nylon, after both Kevlar and a Spectra fiber composite couldn’t handle the speed of the robot. At this point, it’s the structural composite making up the body of the robot itself that fails, becoming delaminated near flexing joints.

What’s particularly interesting is that up until this point of structural failure, the X2-VelociRoACH exhibits an increase in speed that’s in linear proportion to stride frequency. In other words, if it weren’t for the robot starting to explode, there’s no reason to think that increasing the stride frequency even more wouldn’t lead to corresponding increases in speed: the researchers say that if they can find new processes or materials to replace the structural composite, “more dynamic performance from this robot is achievable.” Scary.

These tests were all carried out with an aerodynamic stabilizer (which had no effect on speed) to keep the robot runing mostly straight, and no dedicated steering system was implimented, as this was purely an investigation of top speed. We’re told, though, that adding a tail for either inertial or aerodynamic steering might be the way to go.

“Running Beyond the Bio-Inspired Regime,” by Duncan W. Haldane and Ronald S. Fearing from UC Berkeley, will be presented this week at ICRA 2015 in Seattle.

### Automaton

IEEE Spectrum’s award-winning robotics blog, featuring news, articles, and videos on robots, humanoids, drones, automation, artificial intelligence, and more.

Editor
Erico Guizzo
New York City
Senior Writer
Evan Ackerman
Washington, D.C.