Suche innerhalb des Archivs / Search the Archive All words Any words

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[] NYT 09.10.03: Road Trip For Robots,

New York Times
October 9, 2003

Road Trip For Robots

By Ashlee Vance

ST. LOUIS--TWO sturdy men edged through the Saturday crowd at a diner here, 
sat down and began planning what they hope will someday be the vehicle of 
choice for the United States Army. The partners, Warren Williams and Bill 
Zimmerly, call themselves Team Phantasm. By day, Mr. Williams works 
calibrating electrical instruments; on evenings, weekends and holidays, he 
creates robots for the BattleBots competitions, in which machines fight 
each other to destruction.

Mr. Zimmerly is a semiretired programmer and network administrator who 
likes to write software; he has a particular passion for the Forth 
programming code. Mr. Williams, talkative and energetic, is in many ways 
the opposite of his reserved technophile friend, but the two men share a 
common goal. They are trying to create a robotic vehicle for the Pentagon 
that can drive itself for hundreds of miles across rugged terrain.

While no large military contractor has met the challenge and some people 
may say the task is impossible, the Defense Department has placed a $1 
million bet that Team Phantasm, or one of the dozens of other teams like 
it, has the creativity and imagination to pull off the feat.

To that end, teams from universities and small companies, as well as 
hobbyists, will compete next March in the Grand Challenge, a contest 
sponsored by the Defense Advanced Research Projects Agency, or Darpa, an 
arm of the Defense Department. Participants will race vehicles they have 
designed through the Mojave Desert to Las Vegas in a quest for the 
seven-figure bounty.

The winning vehicle ­ if there is one ­ will be the first to cross the 
finish line after traversing roughly 200 miles of desert in less than 10 
hours without a driver or remote control.

Theoretically, a robot cavalry has the potential to save soldiers' lives by 
fighting battles, performing reconnaissance and rescuing wounded troops. 
But the technology that would enable a machine to guide itself at a decent 
speed across obstacle-laden terrain has yet to emerge. Over the years, 
Darpa has handed out millions of dollars to military contractors for new 
designs for robotic land vehicles, but it has received only a modest return 
on its investment.

Nonetheless, a Congressional mandate calls for one-third of ground combat 
vehicles to operate unassisted by 2015. Darpa sees the Grand Challenge 
contest as the quickest way to get new robots going.

"We are asking these teams to take a quantum leap toward the next 
generation of technology that will be needed in 2015," said Col. Jose 
Negron, the project manager for the race. "There are solutions out there in 
the community and nation that people weren't offering because they don't 
deal with the military complex. So we are inviting little mom-and-pop folks 
out there to help spur advancement and take us where we need to be."

Team Phantasm may be the smallest of the 74 teams taking part in the 
contest, but its members are optimistic. "The biggest ace in the hole that 
we have is the fact that we are two stubborn, independent, unreasonable 
people," Mr. Zimmerly said. "We won't follow what the textbooks teach, and 
we don't have any boss telling us what to do."

All of the teams face the same hurdles: x their vehicles must see, steer, 
accelerate, brake and navigate without outside assistance. The machines 
will receive external guidance only from a series of Global Positioning 
System coordinates that will be revealed by Darpa two hours before the race 
on March 13. Each team must enter the data into computers on its vehicle, 
then turn the vehicle over to Darpa.

The vehicles will start about 20 yards apart, each governed by an emergency 
on-off system. They will have to guide themselves from one G.P.S. point to 
the next, passing over or around ditches, water, rocks, barbed wire and 
other vehicles. The only information about the course that Darpa has 
provided is that obstructions on the route can be overcome or dodged by a 
commercial 4-by-4 truck.

Many of the contestants agree that the race poses two main problems. First, 
robots are not well suited to recognizing "negative terrain" like a ditch 
or cliff. To avoid a robotic demise at the bottom of a canyon, many teams 
are loading detailed maps ­ often relying on satellite imagery ­ of the 
area between Los Angeles and Las Vegas into the onboard computers. They are 
also equipping the machines with radar, laser radar and a host of other 
sensors. Ideally, the detection tools would be able to distinguish an 
insurmountable boulder from an insignificant bush, and a shallow ditch from 
certain destruction.

The second main area of concern is rooted in the restrictions imposed by 
Darpa, which has posted its rules at At 
certain points, the robots will be required to pass through 10-foot 
corridors at a specific speed. The machines must average 20 miles an hour 
over all, while avoiding houses and other buildings.

"Getting within the 10-foot corridors will be the toughest part," Mr. 
Williams said. "You can only plan so much, and then it gets down to chaos 
theory. We're going to map the course down to one square meter, have every 
conceivable route checked, and then everything just has to go right."

The two teammates hope that the imagination demanded by their low-budget 
approach will help them overcome some of the obstacles. Using gear donated 
by Kawasaki Motors, they decided to build a custom all-terrain vehicle 
outfitted with a dome made of bulletproof material that can open up like 
flower petals if the robot rolls over. The machine can then right itself 
and take off again. The teammates have designed an efficient battery to 
power the robot.

Their $50,000 approach is in stark contrast to the one taken by the 
Carnegie Mellon University team, considered one of the favorites to win.

The Carnegie Mellon crew, known as the Red Team, has millions of dollars 
and a long list of corporate sponsors at its disposal. More than 30 faculty 
members and students have joined in outfitting a Hummer with the latest 
autonomous technology. The $1 million prize matters little to the Red Team; 
the school's reputation as a robotics powerhouse is on the line.

William Whittaker, the a professor at Carnegie Mellon's Robotics Institute, 
is a pioneer in building unmanned robots. His past work, some of it for 
Darpa, has included building robots to investigate hazardous waste sites 
like the Three Mile Island nuclear reactor. He is currently building 
systems for planetary exploration.

Some of the smaller teams argue that Carnegie Mellon's experience and 
financing run contrary to the innovative spirit of the Grand Challenge. But 
Carnegie Mellon has laid out one of the clearest plans for completing the 
race, while discussing strategies other teams have yet to fathom.

"You have to be able to assess your situation in the race and make 
strategic determinations," said Mr. Whittaker, whose nickname, Red, 
inspired the team's name. "If you are not in first place, then you have to 
decide when it's best to trail other vehicles and when you are going to 
make your move."

Carnegie Mellon has run several trials and is using an array of on-board 
supercomputer-class computers to make tactical calculations.

Approaching the challenge on a smaller scale is the team from the 
California Institute of Technology, which has a budget of $400,000 and a 
1996 Chevrolet Tahoe ready for the race. Team Caltech has enlisted 
volunteer help and financing from Northrop Grumman, which has won defense 
contracts in the past for its robotics work.

"The undergraduates at Caltech are fearless in being willing to try new 
things, which in an area like this is a plus," said Alex Fax, who is part 
of the technical staff of Northrop Grumman's navigation systems division. 
"We have to design things that can be manufactured and ultimately 
profitable. The Grand Challenge vehicles are designed to win a race, so 
there is a different motivation."

Darpa, noted for being an incubator for new technologies, including the 
Internet, is counting on the competition format to tap into that 
motivation. By dangling the $1 million prize (and the potential for 
bragging rights), the government has managed to extract a vast investment 
from mechanics, engineers and scientists.

"Darpa has been funding contractors for a while, and they want to speed 
things up," said David van Gogh, the project manager for Team Caltech. 
"This contest lets people take risks, and plays on human nature."

Despite the zeal of the competitors, the consensus seems to be that no one 
will win the race this year. "I don't want to get your hopes up too high," 
Anthony Levandowski, a graduate student at the University of California at 
Berkeley and leader of the Robotic Infantry team, told his group. "There is 
a very large probability that we will not go too far."

Each team must submit a technical paper this month that will determine 
whether it is eligible. So far only a handful of teams have passed this 
test, but others are busy revising their proposals to meet Darpa's safety 
and feasible-use requirements.

The teams, from as far away as Alaska (Team Arctic Tortoise), generally 
wish they had more time and more money. Some participants hope to start a 
business as a result of their work, while others hope that a victory will 
establish their credibility as robotics experts. Darpa plans to hold the 
competition roughly every 18 months until someone wins, the funds dry up or 
everyone gives up. The agency retains right of first access to the 
technology developed by the teams, but the groups are to retain ownership 
of their intellectual property.

Win or lose, the teams describe the event as worthwhile and tremendous fun. 
"Six months from now we will all be celebrating in Las Vegas," Mr. 
Whittaker said. "I don't know whether there will be a winner, but there 
aren't going to be any losers."

Liste verlassen: 
Mail an infowar -
 de-request -!
- infopeace -
 de mit "unsubscribe" im Text.