More than a dozen people will take their places this morning before two rows of computer monitors in a dingy garage at Carnegie Mellon University. For an hour, about the only sound to be heard from this group will be a flurry of mouse clicks.

	John Beale, Post-Gazette CMU graduate student Erick Tryzelaar works on route planning for the Grand Challenge robotic race. Click photo for larger image.

That is the sound of racing.

Those "click-clicks" are every bit as significant for the success of Carnegie Mellon's Red Team in next month's $1 million Grand Challenge race as the diesel bark of the team's well-worn, heavily modified Humvee, dubbed Sandstorm.

In the space of an hour, this group will attempt to plan every twist and turn Sandstorm might take in a 210-mile scramble across the back roads and open spaces of the Mojave Desert -- which side of the road to favor in a turn, which obstacles to go around and which to barge over.

Today's exercise, like yesterday's and those of last weekend, is a dry run for the March 13 race. And if the concentration displayed by the students and other volunteers seems particularly intense, it's because what they are doing is crucial for Sandstorm's success.

"If you had the perfect map and the perfect [route] plan, you'd win this race," said William "Red" Whittaker, the renowned roboticist who organized the team last spring and has since pushed its members to the edge of their endurance. "There's no sense that any of these things are perfect."

Not yet, at least. But Whittaker also is quick to point to the mapping and routing effort as one of the major accomplishments of the team to date and a major reason to hope Sandstorm will prevail.

The maps developed since last summer by a group coordinated by undergraduate Vanessa Hodge now occupy trillions of bytes of computer memory and, with some features resolved to one meter or less, are probably the most detailed ever compiled of the Mojave.

		MAP Possible race routes

Up to 20 robotic vehicles, ranging from a six-wheel, heavy-duty truck to an autonomous motorcycle, will compete for the cash prize offered by the Defense Advanced Research Projects Agency. The winner will be the first machine to cover the still-undisclosed route from somewhere outside Barstow, Calif., to somewhere in the vicinity of Las Vegas within 10 hours.

No robot has ever done anything like this. Never has an autonomous vehicle gone so far, so fast without any human intervention. It's quite possible, DARPA officials acknowledge, that no one will win the inaugural race.

But if any team is favored, it's the Red Team. Both Scientific American and Popular Science magazines this month cited it as the consensus favorite. Esquire magazine, though hardly a techie bible, likewise rated the Red Team first, with 7-1 odds.

And perhaps the sincerest compliment, the backhanded kind, came earlier this week when one team acknowledged on a tech Web site that it wouldn't be able to finish the race and simultaneously took a dig at Carnegie Mellon.

Sandstorm is now chewing up the desert at a test facility outside of Carson City, Nev., where the vehicle team relocated this month, frustrated by the persistent snow and ice cover that limited field testing at the old LTV site in Hazelwood during January.

But once Sandstorm moved out of the drafty Planetary Robotics building at Carnegie Mellon, the routing and mapping team expanded into that space, installing several top-end computers provided by team sponsor Intel and an array of high-powered workstations, including lightning-fast graphic cards straight off the ATI Technologies production line.

Speed is essential for route planning, explained Michael Clark, a former NASA engineer who now heads the Red Team's route group. No one will know the exact race route until 4:30 a.m. March 13, two hours before the start of the race. Each team will be handed a computer disc containing thousands of Global Positioning Satellite coordinates, or "waypoints," that will define the route. That gives the route planners no more than an hour, by Whittaker's reckoning, to put together the most detailed TripTik ever.

Human drivers might be able to cover the course without much more than the GPS waypoints. But the human mind has evolved over millions of years into a sophisticated mapping machine.

Electronic computers aren't nearly as good at mapping, and Sandstorm's array of laser rangefinders, radar and video cameras gives it only a fuzzy view of the world. If Sandstorm or any robot is to navigate 210 miles on its own, it will need much more knowledge of the route.

"Even great racers who pre-drive a course can't memorize every turn," Whittaker said. But the route planners will attempt to provide Sandstorm with every move in advance, leaving it to Sandstorm's on-board computers and sensors to detect unexpected obstacles and navigate around them.

"Mother Nature does a great job of presenting profound challenges to a machine of this type," he said. No matter how good the map, a rock slide, a washout or a sandstorm could easily force the robot to plot its own course.

The map team initially was faced with a gigantic task; the race theoretically could occur anywhere inside a 54,000-square-mile area. But federal and state environmental impact regulations required DARPA in late December to designate potential routes, greatly narrowing the area to be mapped in detail.

They began with U.S. Geological Survey aerial maps that are 10 or 15 years old and typically can resolve features no smaller than about 30 meters.

"If you think about what that means for driving a vehicle, that's a big gap," said Jeff Shufelt, of TerraSim, a Carnegie Mellon spin-off that produces digital maps. A 30-meter miscalculation easily could put the vehicle in a ditch instead of on a road.

So team members headed west and began driving areas that might be part of the race route, recording GPS coordinates of features. Space imagery, showing areas of vegetation, has been added to the mix, and, last weekend, an aerial photography firm flew over 100 miles of the potential route with a laser radar device capable of submeter resolution.

The resulting sandwich of aerial and space imagery, GPS coordinates and elevation maps is then used to produce a checkerboard-like "cost map," which identifies areas of greatest cost to Sandstorm in terms of risk.

"There is data that had never been combined in that way before," Shufelt said. "There's no map in the world quite like it."

Using the cost map, Clark said, an automated program can spit out a rough "lowest-cost" route within a matter of minutes. But it takes human eyes to go over the entire route and make adjustments to make sure Sandstorm plays it safe, but not so safe that it loses the race.

For instance, during one trial run, a route editor had Sandstorm navigate around an obstacle. A closer look, however, showed that the "obstacle" was a truck that had been caught in the aerial photography of a decade ago and is long gone.

In the first dry run, Clark said, eight people worked to edit the route, but came up 60 miles short at the end of the hour. In the next run through, 11 editors came up 40 miles short. And last Sunday, 12 editors came within 16 miles of completing the 210-mile planning exercise.

"Ten was my number," Clark said of his race-day plans for editors, "but I don't know if 10 is going to do it." He now hopes 12 editors and three alternates will be able to finish within the hour.

Originally, Whittaker had planned to let the route-planning group operate out of Pittsburgh on race day. But now the plan is to load the computers and workstations into a 56-foot trailer and drive them to California, fly the team out and set up the operation onsite in Barstow.

"It's a tight-knit group, those that have survived," said Clark, who began working on the team as a systems engineer for Sandstorm before Whittaker asked him to take over the route group a month ago.

"We tend to pull, like, two days a week with one hour [of sleep] in between," said Tugrul Galatali, who graduated from New York University last spring with a computer science degree and has been volunteering with the Red Team since October.

"I came in around 4 p.m. yesterday," Galatali said Wednesday evening, just before heading out for dinner. "I was up all night, then slept from 10 to 1." He wasn't sure when he'd leave. "It's pretty much consuming all my time."

Clark already is resigned to celebrating his 50th birthday tomorrow in the Planetary Robotics building and has spent more than a few nights sleeping outside in his van, running an extension cord from the building to power a space heater.

DARPA proposed the Grand Challenge as a way to spur innovation, and Whittaker maintains the mapping and route planning efforts are an example of what can result.

"I think the world has not yet begun to dream what is possible when you get maps down to the resolutions we have them," he said.