If you are ever trapped in a collapsed building, SCHAFT is the robot you’ll want by your side. The a bipedal robot, built by a Japanese team, just took the top score in the DARPA Robotics Challenge trials, where it scored 27 out of a possible 32 points.
IHMC Robotics (20 points), Tartan Rescue (18 points), MIT (16 points), RoboSimian (14 points), WRECS (11 points), TracLabs (11 points) and Trooper (9 points) filled out the next seven spots. DARPA will soon announce up to eight teams that will receive funding to proceed on to the finals next year.
SCHAFT proved itself to be the leader early on when, during day one of the competition, it completed every task except driving a car.
Those who went into watching the Robotics Challenge blind were likely confused. The robots moved slowly, often stopping altogether for minutes at a time. They fell. They fumbled. By human standards, they just weren’t very good at the eight tasks in which they competed.
But the 16 teams competing were working with some of the best robots in the world, and, historically, the fact that they could complete any of the tasks at all was impressive. DARPA designs its competitions to push robotics farther, faster, and in the past, that mentality has paid off. In 2004, it challenged teams to develop an autonomous car that could travel 150 miles through the Mojave Desert. No team made it to eight miles. In 2005, four teams successfully made it through a 132 miles course. Six teams completed a 60 mile urban course in 2007.
The winning teams will move on to the Robotics Challenge finals next year, where expectations will similarly be higher. Teams were rushed to build hardware and software since the competition’s announcement 14 months ago. They will now have another year to further develop their robots.
If the DARPA Robotics Challenge teaches us one thing, it’s that a robot is the sum of many parts. Valkyrie, the Iron Man-like robot entered by NASA’s Johnson Space Center, scored zero points. It’s not enough to be a piece of impressive hardware. Sensing systems and artificial intelligence are just as important.