Robot cars rally for desert race

By day, Seth Cabe is a manufacturing engineer for a mannequin maker. By night, he's working on what could become the battlefield vehicle of the future.

Cabe, leader of Team Loghiq, is one of a number of engineers, researchers and robot aficionados who have signed up for the DARPA Grand Challenge, a contest designed to generate ideas that ideally will lead to the development of self-driving combat vehicles.

Put simply, the Defense Advanced Research Project Agency (DARPA) will give $1 million to the team whose robotic car drives itself the fastest from Los Angeles to Las Vegas, on an off-road course. The race, which must be won within 10 hours, will take place on March 13 next year.

The hard part is that the 250-mile course--which won't be revealed until two hours before the start--will require the computerized vehicles to drive through or around sand, mud, boulders, ditches, barbed wire, mountains and at least one overpass where onboard global positioning systems (GPS) for navigation won't work. (Except to send commands for an emergency stop and restart, the teams can't interfere with the driving.)

So though the vehicles are expected to travel at an average of 25 miles per hour, and to hit 50mph or more at times, they have to be capable of swerving or stopping suddenly.

"The major challenge, other than fundraising, is real-time obstacle detection and avoidance. It simply doesn't exist right now as an off-the-shelf item," said Team Loghiq's Cabe.

Like other contestants, Cabe harbors doubts that any machine can meet DARPA's challenge. However, he's taking part anyway. "I'm an engineer. I can't stay away from projects like this," he said.

The competition grew out of a U.S. Defense Department mandate to increase the use of technology in the field. The agency wants one-third of all combat vehicles to be able to operate unmanned by 2015, according to Don Shipley, a spokesman for the DARPA Grand Challenge.

Autonomous flying vehicles, such as the drones used in the Gulf War and in combat in Yemen, are already being used by the U.S. military.

Congress authorized the contest format a few years ago. "The idea is to attract fresh thinking on the subject, to get beyond the Lockheeds and the Grummans," Shipley said, referring to military contractors Lockheed Martin and Northrop Grumman.

DARPA will hold future contests involving different types of vehicles and technological challenges, he added.

Although the government agency came up with initial rules, it changed these after getting feedback from the 34 registered participants. Earlier, the rules stated that the vehicles couldn't be human-controlled machines. One participant asked if that meant it would be OK to get a chimp to drive. Now, the rules state no living thing can control a contestant vehicle.

Some vehicles are being designed specially for the race. Others are retrofits of existing vehicles. For example, the Arctic Tortoise, sponsored by the Geophysical Institute of the University of Alaska, is a 1992 Jeep Cherokee with sensors, beefy shock absorbers and a bevy of computers.

Mr. Magoo calling
Perhaps the chief underlying technical challenge lies in avoiding desert debris, according to contestants.

Commercially available GPS navigation systems--typically used to control tractors or to guide surveying equipment--are accurate to within a meter. That means race vehicles that get knocked off track should be able to easily and rapidly find their course again.

GPS systems, however, can't pick out rocks or ditches--which can ground a car--and sometimes fail to work in dust storms. To get around this, contestants are loading up their cars with sensors, cameras and accelerometers. These feed data about objects in the car's field of vision (and the car's position relative to those items) to onboard computers for supplemental navigation.

Some contestants will also rig their vehicles with radar and sonar to pinpoint location.

"If all of the sensors agree, we'll go full power," said Ivar Schoenmeyr, captain of the CyberRider team, whose day job is chief scientist at a pump company in Southern California. "But if all of the sensors for some reason are disagreeing, we'll go with the most reliable system (and) slow way, way down."

Calculations and decisions have to be made rapidly, however, and the room for error is huge. A vehicle moving at 45 miles per hour is covering about 60 feet per second.

"You only have about four seconds to stop or slow down," said Richard Ruhkick, an engineering technician at the University of Alaska at Fairbanks, who leads the Arctic Tortoise team. If the vehicle's computer can't absorb changes in data quickly enough, it could mean a trip into a gully.

Most teams are still in the preliminary planning and testing stages. The Arctic Tortoise team members, for instance, recently completed tests where they bashed about with the Cherokee on the University of Alaska's Poker Flat Rocket Research Range (the only university-owned rocket range, according to Ruhkick). Now, they are stripping the Jeep's interior to install computers.

Invite Michael Kanellos into your in-box Senior department editor Michael Kanellos scrutinizes the hardware industry in a weekly column that ranges from chips to servers and other critical business systems. Enterprise Hardware every Wednesday.

The vehicles entered in the Grand Challenge vary greatly, as do the design philosophies and budgets behind them. One of the acknowledged favorites is Sandstorm, out of Carnegie Mellon University. William "Red" Whittaker, the Fredkin Professor of Robotics there, is overseeing the project. (Whittaker devised robots that helped clean up after the accident at Philadelphia's Three Mile Island power plant.)

Another team, from CalTech, has several students and a full-time project manager working on its vehicle, along with professors. Its corporate sponsors include IBM and Northrop Grumman.

Some contestants have estimated that the budgets for the Caltech and CMU projects could run as high as $5 million or $6 million. Most other competitors will get by on a budget closer to $100,000, backed up with donations and lots of free labor.

Still, budget isn't a barrier to creativity. For example, Team Loghiq is building an invertible vehicle that can operate even if it flips upside down. The vehicle, still under construction, will consist of a relatively flat platform that is mounted on custom-made, five-foot-diameter wheels. It will largely be crafted out of carbon fiber, titanium and Kevlar.

"We should have the lightest vehicle there," said Cabe, who also races solar-powered, human-controlled vehicles.

The Loghiq vehicle is also expected to feature "tank" steering, which means its wheels are able to alternately lock and spin. Most other contestants are using robot-controlled steering wheels and movable front wheels.

Taiwan's Via Technologies is donating mini-ITX motherboards and microprocessors that will serve as onboard computers to the Loghiq team.

Start your engines
The next big milestone will come Oct. 14, when the participants must submit technical papers detailing their work to DARPA. A panel will then determine whether the plans conform to the contest rules and are technically feasible. It will also dispense advice, said Shipley. Many of the 34 registered contestants will likely drop out by this stage.

Most teams are then hoping to have vehicles in operation by the beginning of 2004. All will then begin to gather in Southern California in March for a DARPA symposium and a technical inspection. Two hours before the race, the route will be given to teams for downloading onto the vehicles' onboard computers, which will then create digital maps.

And, while participants said the million-dollar prize money would be nice, it's generally not the motivating factor.

"What drew me to the contest? The prospect to expand my horizons and do something that had never been done before," said Schoenmeyr. "Mount Everest has been climbed, the North Pole discovered..."