DARPA Selects Boston Dynamics' Humanoid for Robotics Challenge (UPDATED)


UPDATE 13 April 2012, 12:02 p.m. A notice posted on the U.S. Federal Business Opportunities website confirms that DARPA has selected Boston Dynamics as a "sole source" to develop and build the humanoid robots that the software teams will use in the DARPA Robotics Challenge. "Of the few existing humanoid robots," the notice says, "[Boston Dynamics] was deemed to be the sole viable supplier for providing the necessary robotic platform capability within the specified timeframe." Boston Dynamics will build 8 identical humanoids, which will be based on PETMAN (see below an exclusive photo we obtained) and Atlas, robots that the firm build for the U.S. Army and DARPA, respectively. The result is expected to be "a one of a kind humanoid robot with state of the art capability."

Just as we were getting ready to put together a big long post speculating on which humanoid platform DARPA would select for their newest Robotics Challenge, a video posted yesterday to the official DARPA YouTube channel has made all of that completely unnecessary: to the surprise of nobody, it's going to be a derivative of Boston Dynamics' PETMAN/ATLAS humanoid robot.

We're probably not looking at the final form of the DARPA robot in the case of the platforms shown in the above video. We know that in its final form, ATLAS (we're not sure that the robot that Boston Dynamics is developing for the DARPA Challenge is actually going to be called ATLAS but we'll stick with that for now) "is expected to have two arms, two legs, a torso and a head, and will be physically capable of performing all of the tasks required for the disaster response scenarios scheduled in the Challenge." These tasks include driving cars, walking across rubble, moving heavy objects, opening and closing doors and valves, using tools, and climbing ladders. Based solely on the video above, it looks like Boston Dynamics is making good progress but still has a lot of work to do.

One of the primary differences between the final platform and the existing platform is going to be the addition of a head, which will likely contain the sensors allowing for the hybrid autonomous and teleoperated control that DARPA seems to have in mind for the Challenge (stereo cameras and LIDAR). Otherwise, what you'll expect to see is a 150 kg humanoid (hard to tell if PETMAN and ATLAS are in that ballpark or not) with a pair of two or three fingered hands, and we have no idea whether Boston Dynamics will be doing those in-house or not, especially considering how many excellent grippers are already available and all the money DARPA has spent on manipulation programs.

Boston Dynamics is very well known for developing (you guessed it) dynamic robots: machines that, rather than using static techniques to control their motion, move dynamically, relying on advanced control software and high performance hydraulic actuators. PETMAN and ATLAS are no exception. The trade-off is that hydraulics generally demands a fairly beastly engine. You'll notice that the robots in the video are both tethered to off-board hydraulic power, but luckily, DARPA doesn't seem to have a problem with this. From our interview with DARPA program manager Dr. Gill Pratt:

"We’re not going to disallow tethers in order to power the robot, but the tether will not be able to go all the way back to the operator. So what the utility vehicle can help with is, you can put the power supply for the robot on the flatbed of the vehicle and then the tether can go to the robot, and the vehicle is now a movable base that the robot can operate a certain distance from."

So, we're starting to get a pretty good picture of what the Government Furnished Equipment (GFE) humanoid robot is going to look like come October of this year, but there are also two other things to consider: for one, there's nothing that says teams have to use the GFE, and two, there's nothing that says teams have to use a humanoid platform at all, and indeed, DARPA has been very specific that the Challenge is "decidedly not exclusive to humanoid systems." There are lots of reasons why humanoids are a good idea (as Dr. Gill Pratt outlines in our interview), but that in no way means that humanoids are necessarily the best idea. Of course, completing tasks involving equipment that's been purpose-built for humans (like climbing a ladder) would be quite difficult for non-human robots unless some, uh, creative solutions are implemented.

And what about other humanoid robots? Well, there are some possible contenders, including ASH/SAFFiR (funded by the U.S. Navy) as well as Meka and UT Austin's hyper agile biped. Neither of these robots are nearly as far along as ATLAS presumably is, but on the other hand, ASH was deliberately designed to be able to do pretty much all of the stuff mentioned in DARPA's Challenge, as it's supposed to end up fighting fires in ships. And then of course there are hybrids, like Robospidernaut. Oh, if only!

The Challenge will last 15 months from October 1, 2012 to December 31, 2013, at which point there will be a competition. A second phase will happen from January 1, 2014 to December 31, 2014, ending with another competition where teams will try to improve their performances. We'll be keeping you updated!




IEEE Spectrum’s award-winning robotics blog, featuring news, articles, and videos on robots, humanoids, drones, automation, artificial intelligence, and more.
Contact us:  e.guizzo@ieee.org

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

Newsletter Sign Up

Sign up for the Automaton newsletter and get biweekly updates about robotics, automation, and AI, all delivered directly to your inbox.