Stanford Racing Team's unmanned Volkswagen, races through the Mojave Desert near Primm, Nev., in this Oct. 8, 2005, file photo, during the Defense Advanced Research Projects Agency (DARPA) 2005 Grand Challenge robot race sponsored by the Pentagon. (AP Photo/Damian Dovarganes)

Stanford Racing Team's Volkswagen Touareg unmanned vehicle won the DARPA 2005 Grand Challenge robot race, sponsored by the Pentagon. (Damian Dovarganes/Associated Press)

A Pentagon-sponsored robot race through model city streets that began its qualifying round on Oct. 26 is the next step in military technology — a test to see if the machines can handle the dangerous environment of an urban war zone — but it could hold great potential for civilians, as well.

The technology at play at the Defense Advanced Research Projects Agency (DARPA) Urban Challenge could have far broader implications on the cars we drive, said Alan Mackworth, a Canadian robotics expert. Or more accurately, the cars that could one day drive us.

"This competition is forcing development of what will be the next generation of smart cars," said Mackworth, a University of B.C. computer science professor and past president of the Association for the Advancement of Artificial Intelligence.

"We're talking about autonomous cars that are learning to talk to each other and drive in complex urban environments," he said. "As this kind of technology advances, you could see a situation where stop signs and traffic stops will no longer be necessary. The robots will just talk amongst themselves and figure it out."

Links

• Military robots
• British contestants get $25M to build military robots
• Singapore looking for a few good robots

Military funding contest

In the DARPA Urban Challenge, 35 teams are bringing their robot-driven vehicles to a 96.5 km test course modelled after a real city, complete with manned and unmanned vehicles, in Victorville, Calif.

It's the first of a wave of robotics competitions that are taking artificial intelligence to the streets. Both Singapore and the United Kingdom have announced similar contests to test robots in urban environments, with the final round of both competitions scheduled to run in August 2008.

All three contests are being funded for the purposes of advancing military technology. In the U.S., DARPA is holding the Urban Challenge to help make good on a congressional mandate that "one-third of the operational ground combat vehicles are unmanned" by 2015.

The 20 vehicles that make it into the final round of the competition on Nov. 3 will attempt to complete a simulated military supply mission in six hours or less without any input from their human creators.

They will have to do the kind of everyday driving most city dwellers are used to: merging into moving traffic, navigating traffic circles, negotiating busy intersections, and avoiding obstacles. And they'll have to obey California traffic laws.

It's that everyday practicality that ultimately spurred automakers' interest in the competition, said Mackworth.

"The major car manufacturers ignored the first DARPA competition three years ago, but now I think they see the benefit," he said. Manufacturers such as Volkswagon, General Motors and Ford all have vehicles in the race.

Rocky beginnings

The first DARPA road race — a 228 km trek through a desert course in Nevada — ended without one of the 15 vehicles finishing. In 2005, four teams completed a slightly shorter desert course within the 10-hour time limit, with a team from Stanford University taking the $2 million US top prize.

The Stanford team and its unmanned Volkswagon — dubbed Stanley — are back to compete in the urban race, as are teams from prominent U.S. schools Carnegie Mellon University, the Massachusetts Institute of Technology and the California Institute of Technology.

To navigate the course, the robots will rely on the latest hardware and software, including laser detection and ranging — or Lidar — technology to sense the immediate environment, and Global Positioning System and other navigation systems to keep on course.

The street winds on part of the course at the 2007 DARPA Urban Challenge. The course, in Victorville, Calif., is designed to mimic city driving conditions. (DARPA)

At least ten teams, including Stanford's, will be using a hybrid navigation system developed by Applanix Corp. of Richmond Hill, Ont., which uses internal gyroscopes and precise instruments to provide accurate information even when GPS is unavailable.

Robots taught to prioritize

"The software is also getting better," Mackworth said. "One of the hardest things to do with a robot is teach it how to handle all of the information from multiple sensors, whether it's GPS, or lasers, or infrared vision or motion detectors. It creates sensor fusion, and the question becomes, which one do you trust? It's the kind of question drivers face every day, but now machines are being taught to make those distinctions and prioritize."

The advances will likely make their way into consumer vehicles as driver-assist safety technologies, such as cruise control, and may eventually lead to robot-driven vehicles, Mackworth said. But for consumers to get to that stage of trust with robots, the artificial intelligence at work has to be proven safe, and the urban road race should be a good test of that, he said.

Developing that AI is a good part of Mackworth's work. He helped start his own robotics competition, the Robocup, which has gained international attention for its novel approach: pitting robots against each other in a game of soccer.

The Robocup, the DARPA race and competitions like them are all moving towards a common goal, Mackworth says: teaching robots to work together in dynamic environments.

"When we first started, it was dismissed as toys solving toy problems in a toy domain," he said. "But the essence of soccer is it's a multi-agent situation, with 10 teammates and 11 opponents, and in those conditions, you have to develop co-ordinated strategies and communication methods. Robotics depends on those things.

"It's funny, because it's exactly the same dynamic that happens in motor vehicle traffic," he said.

[an error occurred while processing this directive]