A robot car built by MIT faculty and students was one of just six autonomous vehicles to successfully complete the DARPA Urban Challenge competition this weekend. However, when the results were announced Sunday afternoon, Team MIT had missed out on the big prize money, which went to the first three finishers: Tartan Racing (led by Carnegie Mellon University), the Stanford Racing Team, and VictorTango from Virginia Tech.
The challenge, presented by the Defense Advanced Research Agency, was designed to advance the state of the art for autonomous vehicles for use by the military. After semi-final qualifying rounds in late October, 11 teams were chosen for the finals, which took place on Saturday at the abandoned George Air Force Base in Victorville, CA. Their task was to navigate a mock-city environment that contained busy four-way intersections and other obstacles designed to simulate conditions that vehicles might encounter in a military supply mission. The event offered $3.5 million in prize money, with $2 million going to the winner, $1 million to the second-place finisher, and $500,000 to the third-place team.
MIT’s robot vehicle, dubbed Talos, was one of just six of the 11 finalists to complete the three challenges given the finalists and finish the course in under six hours. We checked in throughout the day Saturday to watch the competition via DARPA’s live webcast. By noon EDT, all 11 teams were successfully on the course. But almost immediately, some ran into trouble. Intelligent Vehicle Systems, an entry from Ford, Honeywell, and Delphi, got stuck right out of gate and had to be manually backed up and restarted. Similarly, Team Oshkosh Truck, which looked like a giant yellow Tonka truck, also encountered early difficulties.
Around this time came what might have been the world’s first autonomous vehicle traffic jam—seven vehicles piled up at a four-way intersection. A little over an hour and a half into the challenge, Oshkosh (oh gosh), Annieway (from Germany and the only team led by a woman), and ISV were being removed from the course. A half-hour after that, the Knight Rider bot (University of Central Florida) drove into a house, unfortunately a disqualifying error.
But given all the challenges of an urban environment, with stop signs, intersections, moving traffic, and various additional obstacles, it was impressive that even six robots finished the course. Indeed, with many of the vehicles at times traveling very quickly while obeying all traffic rules, it was sometimes difficult to believe they were completely autonomous. In addition to the four already mentioned, Team Cornell and the Ben Franklin Racing Team (University of Pennsylvania and Lehigh University) completed all the tasks.
MIT’s robot, with by far the most sensors of any robo-vehicle in the race, according to commentators, might have been done in by its own sophistication. The car carried 40 computing cores cooled by a rooftop air conditioning unit, all of which was powered by a 6-kilowatt generator. It seemed to get stuck repeatedly. And indeed, webcast commentators began asking such questions as, ‘Does MIT’s car think too much?’
“They’ve got enough going on in there to operate two or three cars,” said commentator Jamie Hyneman, co-host of the hit Discovery Channel show Mythbusters. “You don’t hear the car, you hear the generator and the air conditioner.” However, he added, “once they get those bugs worked out, watch out, it’s quite a vehicle.”
At 4:44 pm, Stanford’s “Junior,” became the first team to finish all three missions. One minute behind it was Tartan Racing (the winner was not just selected by time, but based on its overall performance throughout the challenge, much like an equestrian competition). When Talos finally completed the course, the commentator enthused, “MIT, in their first-ever DARPA challenge, finishes fifth across the line.”
We reached MIT aeronautics and astronautics professor Jonathan How, one of the team leaders, for comment late Sunday. “I congratulate CMU, Stanford, and Virginia Tech on excellent performances,” How wrote in an e-mail. “While we trailed them by some time, we are very pleased that we finished the full 55 miles and came in fourth place. I think this was a great accomplishment for a first time team, and think that the students and postdocs did an outstanding job. I am very proud of what the team was able to accomplish in the past year and a half.”
“By the way,” How continued, “I thought the race was going to be fun, but I didn’t realize that it was going to be that much fun. It truly was amazing to watch these robots driving around at speed with all that traffic on the road. What a blast.”
