Carnegie Mellon Automated Car Grabs the Gold at DARPA Challenge

An autonomous robotic SUV from Carnegie Mellon's Tartan Racing team won the US$2 million prize in the DARPA (Defense Advanced Research Projects Agency) Urban Challenge by successfully completing an urban obstacle course faster than 10 other finalists in the race, held Nov. 3. "Boss," as the robotized 2007 Chevy Tahoe is called, averaged 14 miles per hour over 55 miles.

The DARPA Urban Challenge is the third in a series of competitions DARPA, an organization that does military research for the U.S. Department of Defense (DoD), has sponsored to help foster the development of autonomous robotic ground vehicle technology. The DoD hopes such research will lead to developments that could help save lives on the battlefield by performing hazardous missions.

DARPA held its first Grand Challenge in March 2004 on a 142-mile desert course in California and Nevada. Fifteen autonomous ground vehicles attempted the course, but none finished, and the $1 million cash prize went unclaimed. In 2005, in DARPA's second challenge, four robot vehicles completed a 132-mile desert course in under 10 hours. A team from Stanford University took home that $2 million prize.

"The 2004 event was equivalent to the Wright brothers flight at Kitty Hawk, where their airplane didn't fly very far but showed that flight was possible. I believe that the significant progress after 2004 was due to the fact that the community now believed that it could be done," DARPA Director Tony Tether said.

"The DARPA challenges have been tremendous technology accelerators for the robotics, and in particular the mobile sensing, industry," he told TechNewsWorld. "Prior to the Grand Challenges, the field of mobile robotics was a loosely organized confederation of research projects occurring in academia and in the hobbyist communities. What Dr. Tether, the DARPA director who orchestrated the Grand Challenge series, has done is give these communities a mission to rally around."

The Details

After reviewing judges' scorecards overnight, DARPA officials concluded that Boss completed the course by following California driving laws as it navigated the course -- and that it operated in a safe and stable manner. Surprisingly, many of the robots made good decisions, said Tether. That meant speed became the determining factor, he added, and Boss was the fastest of the competitors. Boss completed the course about 20 minutes faster than the second-place finisher, Junior, from Stanford Racing, which brought home the $1 million second place prize.

Victor Tango's Odin for Blacksburg, Va., came in third to nab a $500,000 prize.

The course was set at George Air Force Base in Victorville, Calif., which is used by the U.S. military to train for urban operations. It has a network of roads on the site that simulate the type of terrain American forces operate in when deployed overseas. Vehicles that competed in the Urban Challenge were required to operate entirely autonomously, without human intervention, as they obeyed California traffic laws and performed maneuvers such as merging into moving traffic, navigating traffic circles, and avoiding obstacles, DARPA noted. The vehicles had to "think" like human drivers and continually make split-second decisions to avoid other moving vehicles and safely pass through intersections.

"The urban setting added considerable complexity to the conditions faced by the vehicles and was significantly more difficult than the fixed desert courses featured in the first two Grand Challenges," added Urban Challenge Program Manager Norman Whitaker. "Tartan Racing, Stanford Racing and Victor Tango all did a great job getting their vehicles to navigate the course quickly and safely despite the challenging conditions."

Robot Eyes

Continental Automotive Systems was one of the sponsors for Tartan Racing. The company provided sensors to help the vehicle identify objects.

"We got off to a rocky start because there was an issue with the GPS (global positioning system)," Dean McConnell, director of occupant safety and driver assistance systems for Continental Automotive Systems, told TechNewsWorld. A huge television monitor jammed GPS signals to Boss, but after DARPA shut down the big screen, the GPS came back to Boss. Stanford's bot started about 20 minutes ahead of Boss and completed the course first, but Boss made up the difference and completed the course more quickly.

Massive Efforts

Tartan Racing includes Carnegie Mellon faculty, staff and students from the School of Computer Science's Robotics Institute, as well as Carnegie Mellon's College of Engineering, the university noted, adding that Tartan Racing received major support from General Motors, Caterpillar and Continental. Engineers from those companies plus Intel (Nasdaq: INTC) worked on the project.

McConnell said the total team likely averaged 20 people who were 100 percent dedicated to the project at any one time during the last year, with some experts taking occasional time off from the project. In terms of the money, including all the materials and time that was donated, McConnell is sure the project cost somewhere in the millions of dollars, but he's not aware of the actual total.

"The prize of $2 million doesn't pay off the bet, so to speak, but it does allow this nucleus of people involved in the project to further their development and research in the technology areas and continue to refine the robotics for autonomous vehicle capabilities," he noted.

What's Next?

As of right now, future DARPA challenges are not yet scheduled.

"I know DARPA will have their normal challenges of funding and all the things that go along with government activities. I do think this was a positive step toward their goal of having a significant percentage of their fleet be autonomously guided vehicles in the future," McConnell said.

"We've proven we can do it. The next challenge is to refine it, get the costs out and ensure that the safety and robustness of the systems are there," he added. "We've proven our technology in terms of the radar sensors we've provided for the vehicle and the capabilities to fuse the [radar] data with the vehicle. It's a direct link to the future next steps of what we're working on."

Continental plans to accelerate the use of automated systems on consumer vehicles in the future for active safety systems. "Our goal is, as soon as possible, without a date defined, is to get to that vision zero, which is to avoid all the accidents if we can, and that's what we'll continue to work on," he said.