The first day of the DARPA Robotics Challenge Finals was spectacular. We’re going to have lots more for you tonight and tomorrow, but we wanted to get you this video right away. A bunch of robots fell during runs today, and in every case, humans rushed in with a gantry and hoisted the robot back up again. This is what the rules allow, but the spirit of the competition is really looking for robots that can operate independently in disaster areas without human assistance. We’re likely to see more attempts at robots getting themselves up tomorrow (as it’s the last competition day and there isn’t as much to lose), but during today’s run, CMU’s CHIMP robot showed everybody what a resilient disaster robot should be able to do. And it was amazing.
Is that the coolest thing ever, or what? The atmosphere in the stands was electric, with the entire audience transfixed on CHIMP’s run, and exploding in cheers with each success. And there were a lot of successes. The driving task was flawless, but on egress, one of CHIMP’s limbs got stuck on the side of the vehicle. The operators may not have realized this, and CHIMP kept trying to get out, and eventually, it toppled back into the vehicle. But, no matter: the robot realized something was wrong, paused to collect itself, and repositioned to give itself a clear path out of the vehicle. And it worked.
The big fall happened after CHIMP had opened the door and was about to go through the doorway. Apparently it wasn’t in a stable position as it moved its leg tracks, falling down and landing sideways, against the doorframe. It’s something that the CMU team had really ever thought about, to say nothing of prepared for (we asked them while all this was going on). It’s worth mentioning that there appeared to be no discussion of running in with a gantry: the team had faith that CHIMP would be able to get up on its own.
However, because of CHIMP’s weird angle against the doorframe, the robot’s autonomous get-up behavior couldn’t work. The remote operators stepped in and manually moved CHIMP’s limbs to get it to flop over on its belly, and at that point, its autonomous behavior could take over, and just like that, CHIMP was up and running and no worse for wear.
CHIMP also had some issues with basic shapes, as it started to cut a triangle and then changed its mind to cut a square instead. The target shape was a circle, but CHIMP managed to get all of the circle inside the square, so it earned a point for the task.
And lastly, on the stair climbing task, CHIMP got down on all fours to drive straight up the stairs, but slipped back down twice. Time was running out (just five minutes left), but finally, the robot made it to the top, and it was the perfect success to cap off an awesome run.
Besides just being awesome, we want to make the point that this is exactly how these robots are supposed to work, especially looking ahead towards operations in a real disaster area. Bad things happen. Robots fall down. Problems occur that you may be utterly unprepared for. But your robot should be able to just suck it up and deal with unexpected adversity, and this is exactly what CHIMP did.
If you want to get a sense of what the DARPA Robotics Challenge Finals competition is like at its absolute best, here’s CHIMP’s full run at 20x. We’re working on getting a copy of this that has audio (DARPA hasn’t provided that yet), but in the mean time, this is a very cool 3 minutes.
Evan Ackerman is a senior editor at IEEE Spectrum. Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.
Erico Guizzo is the Director of Digital Innovation at IEEE Spectrum, and cofounder of the IEEE Robots Guide, an award-winning interactive site about robotics. He oversees the operation, integration, and new feature development for all digital properties and platforms, including the Spectrum website, newsletters, CMS, editorial workflow systems, and analytics and AI tools. An IEEE Member, he is an electrical engineer by training and has a master’s degree in science writing from MIT.