Pittsburgh, PA
August 13, 2022
    News           Sports           Lifestyle           Classifieds           About Us
Health & Science
Place an Ad
Running Calendar
Travel Getaways
Headlines by E-mail
Home >  Health & Science >  Science Printer-friendly versionE-mail this story
L.A. to Las Vegas off-road race to feature robotic vehicles

Monday, April 07, 2003

By Byron Spice, Post-Gazette Science Editor

William "Red" Whittaker isn't a gambling man, but he's convinced a million dollars is waiting for him in Las Vegas.

That's because the wagering mecca will serve as the finish line in March for a $1 million, winner-take-all race through 250 miles of desert. Neither Whittaker nor any other member of his "Red Team" from Carnegie Mellon University will be at the wheel of the winning vehicle, but then again, no one else will be either.

That's the whole point of the Grand Challenge race -- to eliminate the need for human drivers. Sponsored by the Defense Advanced Research Projects Agency, the race will pit robotic vehicles against each other in what could turn into a wild scramble from Los Angeles to Las Vegas.

Or, as one wag has suggested, it could end with a pileup of wayward robots 300 yards from the starting line.

What lies behind the contest is the Defense Department's desire to automate many of its ground vehicles, both fighting vehicles and cargo carriers, to reduce risk to life and limb by replacing soldiers on the battlefield when possible.

So far in the Iraq war, for instance, nearly 10 percent of the U.S. and British soldiers killed have died in vehicle accidents.

Unlike the fighting machines on Comedy Central's "BattleBots" or even NASA's robotic Mars explorers, the robots in the Grand Challenge race will not be remotely controlled by people. Each machine will be on its own as it traverses the largely off-road course, negotiating hills and water hazards, avoiding fence posts, gulches and other obstacles, and deciding when to pass or challenge competitors.

"Go" and "Stop" will be the only commands the robots will receive.

On Tuesday, the first day the Defense Advanced Research Projects Agency began accepting entry applications, Whittaker's Red Team became the first to enter the race and, as of week's end, apparently remained the only entrant.

"There's a lot of gamesmanship at this point," explained Air Force Col. Jose Negron, who is in charge of the race for DARPA. Teams have until mid-October to enter the race, he noted, and many may simply be waiting before announcing their intentions to potential competitors.

Is it impossible?

Negron is confident that at least three to five strong teams will enter and that as many as 20 vehicles of all shapes and sizes may start the race.

But one factor that may keep the ranks thin is the widespread belief that the task is impossible, at least for now.

"I would be surprised if somebody won it this [first] year," said Ronald Arkin, director of the Mobile Robot Laboratory at the Georgia Institute of Technology. Robotics technology has made great strides in the last decade, he acknowledged, but designing an autonomous vehicle capable of traveling a long course over varied terrain at high speed remains a daunting task. "To do it in one year? I don't think anybody can do it in that time frame."

A forum on the Grand Challenge's Web site includes several postings from roboticists and other gearheads that are decidedly skeptical, if not downright derisive, about the chances. On Thursday, for instance, one MIT student offered "generous odds" to anyone who would take his bet that no robot can win the contest this year under the existing rules.

Whittaker, the longtime head of the Field Robotics Center at Carnegie Mellon's Robotics Institute, not only believes the race is winnable, but also that the most competitive teams will cross the finish line in substantially less than the 10 hour maximum set by DARPA.

"I actually believe this race will change the view of what's possible," Whittaker said. "This will inspire a generation of technology and people. We can't pass on it."

A robot that can autonomously race off-road from Los Angeles to Las Vegas is a feat akin to Charles Lindbergh flying solo across the Atlantic in 1927, Whittaker said.

It's a comparison that DARPA officials also are eager to make.

The Grand Challenge and its $1 million prize, Negron said, is in the tradition of technology-oriented prizes of the past, including the $25,000 prize that Lindbergh claimed for his feat which would be worth about $240,000 in today's money.

In Lindbergh's case, sponsors hoped to spur development of commercial aviation. With the Grand Challenge, he said, DARPA Director Anthony Tether hopes to accelerate development of technology for autonomous ground vehicles.

Last year, DARPA contracted the with National Robotics Engineering Consortium in Lawrenceville to develop a six-wheeled Unmanned Ground Combat Vehicle, a multipurpose, Humvee-size machine that can keep operating even if it flips upside down.

Robotic aircraft, such as the Predator that gained fame for surveillance in the skies over Afghanistan, already represent a mature, deployed technology. But flying unimpeded in a vast sky is far easier than negotiating obstacles on the ground.

Attracting new ideas

Negron said DARPA was convinced that good ideas existed for overcoming some of the problems plaguing vehicles that drive themselves. But officials also suspect that they aren't hearing all those ideas because some people are unable or unwilling to run the bureaucratic paperwork gauntlet necessary to secure a DARPA contract.

"Who's out there in their garages, their bedrooms, in their labs, working on this?" Negron said. "We want to know." The race might appeal to some people who simply want to show what they can do, without all the red tape.

That said, it's not a race likely to be won on the cheap. Whittaker estimates it will cost about $5 million to win the $1 million prize. He is now lining up partners and sponsors. An alliance with an automotive company, such as General Motors or Toyota, is essential, and similar relationships may well be necessary with groups that develop sensors, software and other necessary components.

Though Whittaker would love to put his hands on some of the components from, say, a GM Hummer, the actual racer probably will be unlike most manned vehicles.

"Computer driving won't even approximate human capability," Whittaker explained, so the machine will have to be designed with unprecedented degrees of "forgiveness."

Its suspension, for instance, will have to be unusually steady for an off-road vehicle, so that the sensors and onboard computers receive a smooth ride. Generous ground clearance will be needed to avoid road hazards that sensors and computers fail to recognize. The machine must be highly maneuverable and powerful -- all-wheel-steer in addition to all-wheel-drive -- to get out of tight spots and quagmires no human driver would be stupid enough to get into.

DARPA promises that the race route -- to remain a secret until two hours before the race -- could be negotiated within the time limit by a human driver in a four-wheel-drive truck. It will be a combination of paved and unpaved roads, dry lake beds, hills, rocky terrain and creeks.

Negron said even he didn't know yet how difficult the course would be. DARPA has laid out three to five possible routes, ranging from easy to hard, and will be obtaining all necessary local, state and federal permits for two or three of those routes, just to be sure no one guesses what the course will be.

Whittaker said he anticipated that some hill climbs would force the racers to creep at 10 miles an hour. Consequently, to stay within the 10 hour time limit over the 250-mile course, the racers will have to go wide open in the flats.

"You will have to hit 100 to be competitive," he said.

Carnegie Mellon experienced

Few organizations in the world have more experience with autonomous driving than the Robotics Institute. Carnegie Mellon has designed cars that can steer themselves and control their own speed on highways. Two Carnegie Mellon researchers drove coast-to-coast in a self-steering minivan in 1995's "Hands Off Across America" tour. Since the early 1990s, researchers have worked on autonomous off-road vehicles.

Autonomous vehicles can handle some situations, such as flat terrain, reasonably well, said Chuck Thorpe, an autonomous driving researcher and director of the Robotics Institute. On a salt flat, "all you would need is a brick to drop on the gas pedal."

At their best, off-road robots are able to cover 90 percent of rugged terrain autonomously, Thorpe said. For the Grand Challenge, however, a racer will have to be 100 percent effective over the same terrain to even finish the race.

"This is certainly more difficult than anything that's been accomplished," he added.

Georgia Tech's Arkin worries that DARPA may be inflating expectations of what robots can do while setting them up for failure in a high visibility competition.

"I'd hate to see a bunch of Comedy Channel competitors out there turning the field into a joke," he said.

That characterization, he hastened to add, does not apply to Whittaker.

"Red's probably the best guy to do it," Arkin said, noting Whittaker's strength is his ability to integrate many components. "If anybody can do it, he can do it."

Whittaker first became famous for designing robots to climb inside the crippled Three Mile Island reactor in 1984. One of his walking robots, Dante II, explored inside an active volcano in 1994, and his Nomad drove autonomously for more than 100 miles across Chile's Atacama Desert in 1997. Whittaker now heads a research team developing technology for an autonomous robotic Mars explorer, beginning initial testing this month in the Atacama.

"I would put my money on Red," Thorpe agreed. "Red really hates to lose."

Byron Spice can be reached at bspice@post-gazette.com or 412-263-1578.

Search | Contact Us |  Site Map | Terms of Use |  Privacy Policy |  Advertise | Help |  Corrections