Robot Cars to Do Battle in Desert Race

Stefan Lovgren in Fontana, California
for National Geographic News
October 6, 2005
When 15 competitors lined up in Nevada last year for the U.S. Defense Department's first million-dollar robot race, hopes were high. The challenge: to drive a vehicle without a human driver or remote control some 150 miles (241 kilometers) through the Mojave Desert.

But those hopes quickly went up in a cloud of dust as most robots barely managed to get off the starting line. The best performer, a modified Humvee built by engineers at Pennsylvania's Carnegie Mellon University, traveled 7 miles (11 kilometers) before breaking down.

To robot devotees, however, it was a minor hiccup.

No surprise, then, that 43 teams showed up to try out for this year's race, dubbed the Grand Challenge. For the past week, teams ranging from garage enthusiasts to well-funded university engineers have been fine-tuning their machines at qualifying rounds here at the California Speedway in Fontana, California. (Watch the robots in action in our exclusive video.)

Twenty-three finalists were announced Thursday for Saturday's Grand Challenge. The 175-mile (282-kilometer) course starts and finishes in Primm, Nevada.

The race promises to be even tougher than last year's run. But 18 months is an eternity in the robotics world, and the technology has vastly improved.

Organizers believe several teams have a real shot of finishing the race in less than ten hours to earn the grand prize of two million U.S. dollars.

"When the first team out of the chute—Mojavaton, a small team out of Colorado—made it successfully around the 2.2-mile [3.5-kilometer] qualification course, I knew right there and then that we had something special," said Ron Kurjanowicz, the chief of staff for the Pentagon's Defense Advanced Research Projects Agency (DARPA), which is sponsoring the race.

Unknown Course

The aim of the Grand Challenge, Defense Department officials say, is to spur development of autonomous ground vehicles that can operate in dangerous environments, such as war zones, keeping soldiers out of harm's way.

A U.S. Congress mandate requires that one-third of military ground vehicles drive themselves by 2015, but the technology to meet that mandate does not yet exist.

So the government looked to enterprising teams to develop the technology for driverless vehicles, sweetening its offer with the two-million-dollar purse.

None of the 23 teams knows what lies ahead for this year's race. DARPA won't reveal the exact route until two hours before the start of the race on Saturday.

But the obstacles on the Fontana qualification course—including a steel-enforced tunnel that wipes out a vehicle's global positioning system—are made to resemble the rugged, real-life conditions that the vehicles will have to navigate.

The vehicles use sensors such as lasers, cameras, and radar to help them avoid obstacles such as rocks and cliffs. The computer's brain has to figure out how to resolve unexpected conflicts, like a boulder sitting in the middle of the road.

"Think about all the decisions that you and I have to make when we drive from our house to the store," Kurjanowicz said. "These vehicles have to do the same thing, without a driver."

Among the top contenders in Saturday's race is TerraMax, a massive truck originally built by the Wisconsin-based Oshkosh Truck Corporation for the U.S. Marine Corps.

In last year's race, TerraMax managed to go only 1.2 miles (2 kilometers). Team leader Gary Schmiedel expects to do much better this year. He pointed to the new all-wheel steering feature on the truck as an important addition.

"We can move this large, 15-ton [13.5-metric ton] payload vehicle in a turn that's equivalent to that of a Humvee," he said.


The resources of teams like TerraMax or Carnegie Mellon University, which has two vehicles in the race this year, are a far cry from those of some of the other competitors, including inventors, electricians, and even a high school team.

One entry, from a Southern California team of engineers, racers, and hot-rodders, is called It Came From the Garage. It has a beer keg stuck on the back and an on-off switch that says "brain."

"Most of the schools and organizations we're up against are just accessorizing conventional vehicles," said team leader Chris "C.J." Pedersen, a former actor. "Our [vehicle] is a custom-built, 21st-century hot rod … complete with hood scoop and exhaust coming off the side."

Anthony Levandowski, a robotics builder from Berkeley, California, is back with Ghostrider, the only motorcycle robot in the qualifications. Studded with sensors and computers, it toppled over after 3 feet (1 meter) in last year's race.

Levandowski, who had to postpone his graduate studies when he couldn't find a faculty advisor who believed it would be possible to build the motorcycle robot, says his vehicle has some distinct advantages.

"We're smaller and go a lot more places," he said while tinkering with the robot before another trial run. "We're also a lot less expensive. This bike costs as much as a tire or a wheel of some of these other guys' machines."

Smart Money

Neither Ghostrider nor It Came From the Garage made the final cut at this week's qualifying races.

However, another crowd-pleaser, Cajunbot—or the Ragin' Cajun—a converted all-terrain vehicle developed by a team from the University of Louisiana in Lafayette, did.

The smart money in Saturday's race may be on Stanley, a converted Volkswagen Touareg made by a team at California's Stanford University. It was the only vehicle that didn't hit an obstacle in the trial runs.

Even if none of the vehicles finishes the race this year, DARPA's Kurjanowicz said, the event has succeeded in galvanizing robotics developers and pushing the creation of new technologies.

"The beauty of the Grand Challenge is that it doesn't tell people how to solve the problem," he said. "The community has come up with its own elegant solutions."

Free E-Mail News Updates
Sign up for our Inside National Geographic newsletter. Every two weeks we'll send you our top stories and pictures (see sample).


© 1996-2008 National Geographic Society. All rights reserved.