ORO GRANDE, Calif.--This is the experiment of all robot experiments.
In 30 minutes, the Defense Advanced Research Projects Agency (DARPA) will unleash 11 driverless cars on a 10-plus mile course here at the former George Air Force Base, as part of its $3.5 million robot race, the Urban Grand Challenge. It will be the first time DARPA has tested fully autonomous cars (with "no animals or midgets inside") driving on a course with other robots, as part of a test of the technology's capabilities as well as its safety, according to Norman Whitaker, DARPA Urban Challenge program manager.
"This is truly the first time we've taken robots and watched them interact with other robots," Whittaker said here Saturday before the start of the race. "They have not interacted so far."
As part of the competition, the robot cars must complete several driving missions within six hours at the closed air force base, which the government currently uses for military operations training in urban environments. The base is much like an environment where the government hopes to deploy autonomous cars by 2015 to complete missions like checking fence borders or clearing airport paths at night, Whittaker said.
On Saturday, the cars will be tested on driving skills much like they were 15-year-olds taking their California driver's exam for the first time, he said. They'll be faced with navigating four-way intersections, merging in traffic, and driving on the highway. About 100 officials from DARPA will be out on the field in safety boxes with ticket books, compiling data on how the cars perform and whether they're following basic traffic rules.
The robots may go as fast as 40 miles per hour on the highway, but watching the driverless cars move even at 10 miles per hour, "it catches your attention," Whitaker said. If there are difficulties on the course, each robot car has a so-called e-stop system installed that lets DARPA stop the car at any time.
In 2006, DARPA chose nine "track A" teams to receive $1 million in grant money to support their development efforts. Of those teams, only seven made it to the finals here Saturday. The seven track A finalists include Carnegie Mellon University, Cornell University, Honeywell Aerospace Advanced Technology, the Massachusetts Institute of Technology, Oshkosh Truck, Stanford University, and Virginia Polytechnic Institute.
The four teams with $1 million from the government that didn't make the cut: Autonomous Solutions from Utah; the California Institute of Technology, The Golem Group from Santa Monica, Calif., and Raytheon from Tucson, Ariz.
DARPA plans to name a winner (or not) on Sunday after compiling all of the data it collects from the race. That's no small task, given the information coming from 100 officials and data from cameras following the vehicles, Whitaker said.
"We'll take an accounting of all the data this evening and weight the evidence," he said.
Even if a vehicle finishes all of the missions within six hours, it's conceivable that the team wouldn't win if the robot violates traffic rules.
"We're looking for the best system, not the fastest system," he said.