Just months after awarding $2 million for a sport utility vehicle that drove itself over more than 100 miles of open road, the Pentagon on Monday unveiled a bigger, richer challenge for self-driving vehicles that can negotiate city traffic.
Veterans of the Defense Advance Research Projects Agency's earlier "Grand Challenges" said the technologies developed for the next contest will clearly benefit the U.S. military, which has set the goal of automating a third of its ground vehicles by 2015. But they said the innovations could have an even bigger impact on driving in America.
"It might fundamentally alter the way we use our highways and save trillions of dollars," said Sebastian Thrun, the Stanford University computer-science professor whose team won the Grand Challenge race last October.
Thrun and his closest competitor in that race, Carnegie Mellon University robotics professor Red Whittaker, both told MSNBC.com they would take part in the newly announced DARPA Urban Challenge competition, which is due to reach its climax on Nov. 3, 2007.
"We're about the future of robotics," Whittaker said. "Anytime there is anything going on in robotics anywhere, we're in."
This time around, autonomous vehicles would run a simulated military supply mission in a mock urban area. To succeed, the vehicles would have to complete a 60-mile course safely in less than six hours, obeying traffic laws and avoiding obstacles while they merge with moving traffic, negotiate intersections and even pull into and back out of parking spaces.
"Grand Challenge 2005 proved that autonomous ground vehicles can travel significant distances and reach their destination, just as you or I would drive from one city to the next," DARPA Director Tony Tether said in Monday's announcement (PDF file). "After the success of this event, we believe the robotics community is ready to tackle vehicle operation inside city limits."
The top prize would be $2 million once again, but DARPA would also offer a second prize of $500,000 and a third prize of $250,000. DARPA also will make funding available for contenders before the finals, through two tracks:
- Teams could submit detailed proposals for up to $1 million in technology development funds, with the government obtaining limited licensing rights to the resulting technologies. The selected teams would proceed to a semifinal known as the National Qualification Event.
- Teams could participate in a series of qualifying tests, just as competitors did in the 2004 and 2005 DARPA Grand Challenges. The teams selected for the National Qualification Event would get $50,000, and the teams that are successful at that event will get $100,000 and a spot in the November 2007 finals.
An initial conference for would-be entrants is scheduled for May 20 in Reston, Va. Details on the conference, as well as applications and rules, were posted Monday to the DARPA Grand Challenge Web site.
The first DARPA Grand Challenge was conducted in 2004 in the Mojave Desert, on both sides of the California-Nevada state line. In that $1 million contest, no competitor got farther than 7.4 miles along a rugged 142-mile course.
The second Grand Challenge, which was run along a similar desert route, was much more successful. Five vehicles finished the 132-mile drive, with Stanford's modified SUV (nicknamed "Stanley") narrowly beating out two Carnegie Mellon entrants.
Thrun and Whittaker both said the third challenge would be significantly tougher, but significantly more useful as well. Although the vehicles entered in the previous Grand Challenge negotiated a cleared road through the desert, they would be incapable of driving in a real-world environment, the two professors said.
"Any vision that we have for self-driving cars must address driving in traffic," Thrun said.
Whittaker said his team would probably have to upgrade their vehicles to understand rules of the road, recognize lane-divider lines or even road signs, operate in "urban canyons" where GPS signals can't reach, and read the intentions of other vehicles on the road. But he voiced confidence that his "Red Team" could build upon the foundation laid during the earlier challenges.
"It's a little bit like building the pyramids," he said. "You always have to get the foundation right."
Noting that an earlier DARPA-supported innovation, the Internet, revolutionized communications, Thrun suggested that autonomously driven passenger vehicles could revolutionize transportation:
- The development of safe, self-driving vehicles could drastically reduce the average annual traffic death toll of 42,000 Americans.
- The carrying capacity of the nation's highways could be dramatically raised from the current level of 8 percent, expressed in terms of the area actually occupied by vehicles.
- Commuters could relax or work during their average drive time of an hour a day.
- Children could conceivably have themselves driven to school or soccer practice, with a safety-conscious robot behind the wheel instead of a harried human.
Whittaker agreed that, in the long run, civilians could benefit from Urban Challenge technologies at least as much as the military.
"That's an inevitable future in the automotive industry," he said. "It could be that the automotive consumer will be the one who's the big winner."