It's been a little over two weeks since Carnegie Mellon's "Red Team" vehicles competed in the DARPA Grand Challenge and took second and third place. Team leader William "Red" Whittaker, a research professor at Carnegie Mellon's Robotics Institute, is already, without pause, rattling off a list of future objectives. In the immediate future is fence maintenance, lunar polar explorations, subterranean mapping, and driverless motor sports—all to be accomplished by robots. The results of the Grand Challenge have not deterred Whittaker's enthusiasm for Carnegie Mellon robotics.

"Airplanes and computers are the past, they're the technology of the last century," said Whittaker. "Robots are this century."

Carnegie Mellon's Red Team competed quite successfully in the DARPA Grand Challenge, a fiercely competitive 130-mile autonomous vehicle race across the Mojave Desert. Red Team's two vehicles, H1ghlander and Sandstorm, beat all of their competitors except Stanford University. Now that it is over, we must ask, what was it like, what happened, and, most importantly, what is going to happen next?

First, what exactly was the Grand Challenge about? A few years ago, the United States Congress asked the Department of Defense to develop a fleet of autonomous ground vehicles by the year 2015. The Defense Advanced Research Projects Agency (DARPA) spearheaded the task and came up with a competition to stimulate research by non-military agencies. The DARPA Grand Challenge was a 131.6-mile race through the Mojave Desert that started and finished in Primm, Nevada, near Las Vegas. The race had to be completed in less than 10 hours. The vehicles in the race were required to be completely autonomous; that is to say, they could not have drivers and could not be controlled remotely. The prize was $2 million. This follows last year's disappointing $1 million challenge from Barstow, California to Primm, in which there were no winners.

The DARPA challenge is a feat for the world, but it is especially meaningful for the Carnegie Mellon community. DARPA Director Tony Tether was not exaggerating when he said, "These vehicles haven't just achieved world records, they've made history." When the first Grand Challenge was announced, most people in the industry did not think that the race was possible. At the time, it was a feat for a robotic vehicle to autonomously travel 200 meters over such terrain, let alone 130 miles.

The first year, a number of factors including a GPS error and a "divide by zero" computer glitch doomed the Red Team vehicle to running off onto the side of the road. The Grand Challenge was repeated this past October 8, with a larger prize and larger ambitions. This time, the Red Team entered two vehicles, both highly modified Humvees (Humvee and Hummer, to be specific) donated by General Motors. Carnegie Mellon's team made corrections and improvements over the past year and was ready to go for the 2005 Challenge.
    
The team spent the last 12 days before the race in the desert, making the final preparations for the big event. A trailer donated by NASCAR champion Chip Ganassi served as its center of operations. The team started each day at around 2:30 a.m., practicing the schedule they would follow on race day. On site was a NOVA film crew, who were making a documentary about the team and their epic challenge. As one team member, CIT first-year Matthew Delaney, described it, the experience was "completely exhilarating."

4:30 a.m., Race Day: Red Team Too, the Red Team division in charge of H1ghlander, received a CD with the route data. That was the first that anyone in the competition had seen of the actual racecourse, because H1ghlander was the first vehicle set to race at the Grand Challenge because of preliminary runs. Over the next two hours before the start of the race, the rout planning team planned out the specifics of how vehicles would navigate the course. By intelligently choosing specific way points and speeds to reach each location checkpoint, the Red Team hoped to gain an advantage over other, less well-directed teams.

6:30 a.m., the race began: Red Team's H1ghlander vehicle left the starting shoot in the pole position. The remaining vehicles started on five-minute intervals. H1ghlander was scheduled to finish the course in six hours and 19 minutes and in test runs it never missed its mark. Watching in the crowd that day were many people interested in seeing the future of robotics, including Carnegie Mellon's President Jared Cohon and Apple's co-founder Steve Wozniak. Everyone on the team was confident, though apprehensive.

Within the first few hours of the race, the team knew that something had gone wrong. H1ghlander, the team's newer vehicle was to complete the course quickly and increase its lead throughout the run. Sandstorm, which started in position three, was to run a slower, more reliable race. However, from the times being reported back to them, the team could see that H1ghlander was not extending its lead. Instead, it was running in lockstep with Sandstorm. Stanford's Stanly vehicle was between the two, keeping a slightly faster pace. What had happened? Had the Red Team loaded the same route onto both vehicles? Was H1ghlander critically injured in some way, causing its computers to put it into "limp mode"? The answer: they didn't and still don't know.

A little after 1:40 p.m., Race Day: Stanly became the first vehicle to cross the finish line with a time of six hours and 53 minutes. No other vehicle would match its pace. Carnegie Mellon's H1ghlander and Sandstorm came in shortly after, Sandstorm in second place, 11 minutes behind Stanly and right on its own schedule, and H1ghlander in third, 10 minutes behind Sandstorm and 55 minutes behind its schedule. In truth, both vehicles performed phenomenally. Entering this race were 195 teams that ranged in form from large companies and universities to garage-built-hopefuls. Of the 23 finalists who made it to the final competition, only four vehicles finished the course in the required time, two of which were from Carnegie Mellon. In fourth place was Gray Insurance Company's GrayBot. Oshkosh's TerraMax also finished, though outside of the time limit. Red Team celebrated its victory, though not without some concern for what may have happened to H1ghlander.

Data retrieved from both vehicles as well as a race day video has provided some evidence. When a human driver drove H1ghlander off the course by, its engine delivered almost no power in response to a light depressing of the pedal. Even on level terrain, H1ghlander never reached its pre-planned high speeds. Its underachievement became proportionally worse with larger slopes, and in some instances it actually stopped and restarted on the steepest terrain. Further analysis showed that the engine was delivering below-normal torque at low RPMs. The team is still investigating the root cause, with both vehicles now securely parked in the Planetary Robotics Building on campus. No matter what is discovered, most people would agree that both Humvees performed spectacularly.

What comes next? The only absolute is that Whittaker will not be standing still. Whittaker compares his passion for robotics to the passion that professional racecar drivers have for racing. Should the current members of the Red Team remain committed to Whittaker's goals, his post-race schedule should be no less ambitious than what it looked like a few months ago.

Consider the fence maintenance project: using some type of autonomous vehicle, Whittaker hopes to automate the process of clearing weeds from many miles of electrical fencing. Or the convoy project, for which he hopes to get several autonomous vehicles to reliably follow a human-driven vehicle. Or try Lunar Polar explorations, or subterranean mapping, or, if he can find a group of people as passionate as he is, driverless motor sports. These compose just a few of the items on Whittaker's immediate to-do list. The important point is that Whittaker's robotics teammates are examining the Red Team failures and using what they learned to move on to new projects. There will be greater things to come.

The Grand Challenge should be considered, by all involved, an enormous success, and certainly not the end of the Red Team. Even lacking the $2 million prize, Carnegie Mellon and the members of the team have gained something more important: enormous respect from within the industry. They will surely be earning it again with their future projects. 

 talkback to the pulse