Urban Challenge: Robots Rule the Road

The rules for the Urban Challenge read straight out of a driver's-ed manual: Right-of-way belongs to the first car to arrive at a four way stop. Maintain a distance of five to 10 meters between you and the vehicle moving in front of you. Execute a u-turn on a one-lane street.

The difference is that this competition pits robotically driven vehicles, completely autonomous of human interaction, in parking, passing, stopping, and acceleration drills that any human driver could expect to perform on a city street. The race will be staged on a 60-mile course at an undisclosed location. If one of the almost twenty teams can design a robot vehicle to handle the wiles of urban driving better than any other, it's worth a $2 million first prize.

In the same spirit of the university's successful completion of the 132-mile Desert Grand Challenge in October 2005, Carnegie Mellon will enter the Urban Challenge with an interdisciplinary team, Tartan Racing, fielding two Chevy Tahoes provided by General Motors and retrofit to be driven by computers. In addition to GM, Caterpillar and Google are also partnering with Carnegie Mellon's Tartan Racing team.

The strategy of the Defense Advanced Research Projects Agency (DARPA), which sponsors this and earlier Grand Challenge events, is to excite enough interest among leading universities and private competitors to achieve a breakthrough in autonomous vehicle performance. Such innovation could lead to robotic vehicles being assigned to dangerous supply missions through hazardous terrain instead of putting the lives of military personnel at risk.

Yet no solo-driving robot is squealing its tires out of the pit stop just yet. In order for a team to compete in the Urban Challenge, it isn't enough to create a vehicle that knows better than to tailgate. To simulate advanced navigation skills, DARPA requires that a robotic vehicle pass another moving car and recognize an opportunity to pull out in front of moving traffic once it sees there is at least a 10 second window. Violation of any and all traffic laws, such as failing to come to a complete stop at an intersection, will accrue a traffic penalty costing the vehicle's team valuable points. And preprogramming a route into the vehicle isn't possible, since details of a course that might be complicated by a traffic circle are revealed only five minutes before race time. As human drivers absorb MapQuest directions while they are driving, so too must the autonomous robotic vehicles interpret and apply the route data provided by DARPA on the fly.
Members of the Tartan Racing team include several veterans from the desert Grand Challenges, including team leader, William "Red" Whittaker, the Fredkin Research Professor at Carnegie Mellon's Robotics Institute. He is joined by Raj Rajkumar, Carnegie Mellon professor of electrical and computer engineering and computer science, and co-director of the GM-Carnegie Mellon Collaborative Research Laboratory; Chris Urmson, Tartan Racing's technology leader and graduate student in the Robotics Institute; Ed Schlesinger, head of the Department of Electrical and Computer Engineering Department; Anthony Stentz, research professor of robotics; Martial Hebert, professor of robotics; and John Dolan, Senior Systems Scientist in the Robotics Institute.

Teams who qualify in a trial round will compete in the Urban Challenge November 3, 2007.


—John Worlton, September 2006


For information on Carnegie Mellon's entry into the Urban Challenge, please visit the Tartan Racing web site.