A team out of MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) spent more than three years preparing for this past weekend, when it took part in the DARPA Robotics Challenge in Pomona, Calif.
During the competition, MIT’s robot, Atlas, finished inches away from winning a $2 million prize. The 6-foot-2-inch, 400 pound robot needed to perform eight different tasks—such as opening a door, turning on a power tool and drilling a hole in a wall, climbing stairs, and even driving a car, all within an hour. (See the video below.)
MIT’s team, comprised of a professor, a postdoc and fewer than a dozen graduate students, faced adversity in the competition when Atlas broke its right arm getting out of the car after the driving task. But the team was able to reprogram Atlas to do all of the remaining tasks left-handed.
Each team was given two chances to complete the tasks as quickly as possible. On its second run, Atlas ran through the first three tasks faster than any other robot, but on the drill task, the drill slipped out from the wall, ending MIT’s chances of first place.
“This is without a doubt the most ambitious project that any of us have ever undertaken,” said Professor Russ Tedrake, the CSAIL investigator who led the team at the Robotics Challenge. “From perception to motion-planning to manipulation, the breadth and depth of challenges have forced us to think creatively, program nimbly – and sleep sporadically.”
Contrary to some approaches which focus on moving a robot incrementally to its goal, MIT developers solved for motion in advance with a method called “whole-body motion-planning.” The MIT team developed algorithms for every one of the robot’s micro-actions.
“It’s no longer a question of whether we’ll ever live in a world filled with useful robots,” said Tedrake. “The real question is just how soon that time will come.”
The DARPA Robotics Challenge, which is sponsored by the Pentagon, is aimed at developing robots to perform tasks in disaster-relief situations.