SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Photonics West 2019 | Call for Papers

2018 SPIE Optics + Photonics | Register Today



Print PageEmail Page

Sensing & Measurement

CMU robot car first in DARPA Urban Challenge

Autonomous vehicle technology took a big step closer to reality as six robotic cars completed a 60-mile course on city streets.
5 November 2007, SPIE Newsroom. DOI: 10.1117/2.2200711.0001

Video: QuickTime | Windows Media Player

The Tartan Racing Team from Carnegie-Mellon University (CMU) took home the highest honors and $2 million as the winner of the 2007 DARPA Urban Challenge in Victorville, California on November 3.

The winning vehicle was "Boss," named for Charles Kettering, founder of General Motors Research Labs. Stanford University took second place, with their car named "Junior" after the founder of the university, and Virginia Tech took third, with "Victor Tango."

The race of autonomous vehicles took place under sunny skies in the high desert of southern California in front of enthusiastic crowds there to revel in the low-speed excitement. The course was laid out on the winding streets of abandoned residential neighborhoods at the former George Air Force Base, which was closed in 1992.

Stanford's car crosses the finish line first, as DARPA director Tony Tether waves the checkered flag. Stanford finished second in elapsed time on the course.

Six vehicles completed the 60-mile course, reflecting noteworthy progress from the first two DARPA Challenges, held in 2004 and 2005. No vehicles completed the first, and four finished the second. Both were run on courses through the desert. This competition involved navigating successfully through city traffic, consisting of both human and robot drivers, while obeying traffic laws. Winners were determined by the shortest total time after adjustments by judges for infractions, driving etiquette, or other factors.

"This technology is at least -- at least -- an order of magnitude more complex than what was done in the previous challenges," said Raj Rajkumar, co-director of the General Motors-CMU Collaborative Research Lab and a member of the Tartan Racing Team.

University of Central Florida's 1996 Subaru Outback leaves the starting area at sunrise.

Spectators brave the fence line to take photos as the massive driverless truck from Team OshKosh Truck thunders past.

Unlike most auto racing events, this one was not characterized by speed, but by technology and computer processing power. As one race team leader put it, "this is not a car race, it's an artificial intelligence race." Many of the cars in the National Qualification Event (NQE) leading up to the final day featured as many as 10 on-board computers just to handle the "data cloud" generated by sophisticated laser sensors and other telemetry the vehicles carried. From the NQE field of 35, there were 11 teams selected for the Urban Challenge final.

The competition required each vehicle to perform three simulated missions, with six or seven sub-missions in each. Each mission consisted of instructions to visit a series of specific GPS points successfully. At the end of each mission, the vehicles returned to the starting area to receive their next assignments.

DARPA director Tony Tether said that before the race, nobody was sure what would happen when all the robot vehicles were out on the course. Despite the months of testing by each team, and the week of qualification events, the final race represented the first time that robots had interacted with other robots in traffic.

"We weren't certain that we'd get more than an hour into it before they all destroyed themselves," he said. But after the first two robots encountered each other on the road, going in opposite directions at 30 km/h, organizers breathed easier. When watching the robot cars perform, "I forgot after awhile that there was nobody in there," Tether said.

The artificial intelligence required to perform successfully in the race meant that the computers had to "plan in the wild for unexpected behaviors, and still make the right kind of decisions," said Norm Whitaker, program manager for the Urban Challenge. Whitaker was optimistic not only about the prospects for technology advancement, but by the interest in science and technology that this event will stimulate.

"We're excited to see the turnout we've gotten, and we need more of the same. We're learning what gets people excited," Whitaker said.

While DARPA's professed goal in sponsoring the Urban Challenge is to improve battlefield safety by making one third of all military vehicles autonomous by 2015, as mandated by Congress, there were other motivations mentioned by teams sponsored by automakers. Heavily involved companies included General Motors, Ford, Volkswagen, and heavy equipment manufacturer Caterpillar. Their presentations to media tended to stress the progress already made with such technologies as intelligent cruise control, in which a vehicle can sense another slower one in front of it and adjust speed accordingly. Participating in this race offers an opportunity to advance such technologies toward a day when cars will be able to take their owners to work while the occupant might be already working, free of the need to pay attention to traffic.

Besides $2 million for the winner, $1 million was the second place prize, and $500,000 for third. Eleven of the participants were "Level A" teams who qualified for funding up to $1 million from DARPA by passing numerous qualifications and benchmarks in preparation for the event. Those who didn't were "Level B," but on race day there was no distinction between the two categories and all competed together. Of the 11 finalists, seven were from Level A and four from Level B.

Those in the crowd who hoped for a robot "demolition derby" got only a tiny taste of it when Team MIT and Cornell had a fender-bender on the course. Cornell's vehicle was stopped and MIT attempted to pass. Just as the two cars were beside each other, Cornell began to pull forward slowly. MIT was beginning to pull back into the lane ahead, but instead, bumped front fenders with the other car. Some tense moments ensued before the cars were separated carefully. Key sensors are located on the front corners of most of the vehicles, where the cars collided, but the damage was insignificant, and both continued in the race.

Everyone involved in the Urban Challenge said the results have changed the landscape of robotics and transportation forever. CMU professor William "Red" Whittaker, team leader of Carnegie-Mellon's Tartan Racing Team, said the most important thing is the awareness generated by an event like this.

"Robots sometimes stun the world," he said. "This is a phenomenal thing for the future of robotics automation and what it can do."

CMU's Tartan Racing Team after receiving the $2 million check for first place.

Watch the video: QuickTime | Windows Media Player

DARPA Urban Challenge website