Robots, start your engines
Mar 11th 2004
From The Economist print edition

Innovation: Could a robot race funded by a military-research organisation help to advance the development of autonomous fighting vehicles?

AMONG government organisations, America's Defence Advanced Research Projects Agency (DARPA) has always been somewhat unusual. As the research arm of the Department of Defence, it is akin to a high-stakes venture capitalist, gambling large sums of money (its estimated 2004 budget is $3 billion) on risky technologies that will probably fail, but could pay off in a big way. It has had some stupendous successes, such as the internet, the Saturn rocket and micro-electro-mechanical systems (tiny machines that work at the scale of a human cell). There have also been some resounding duds, such as the Total Information Awareness project, a Big Brotherish plan to spot terrorists by combing through databases of personal information, which was swiftly abandoned.

But what is arguably DARPA's most outlandish scheme yet will start rolling on March 13th, when a gaggle of strange-looking vehicles will line up in Barstow, California to make a wild run across 250 miles of scrub and desert. They will be heading, fittingly, towards Las Vegas, where gambling and high stakes are as common as hotel-buffet specials. DARPA calls this race the Grand Challenge. What makes the race so grand, and so much of a challenge? The lack of humans.

Websites
DARPA provides information on the Grand Challenge. Amongst the participants are Team Phantasm and a team led by William "Red" Whittaker at Carnegie Mellon University.

Not a single person is permitted to be at the wheel (or remote control) of any of the racing contraptions. Each machine in the contest must reach the finish line unaided, within ten hours, or die trying. The first vehicle to cross the finish line within the allotted time will win its human creators a cool $1m—respectable winnings even in the luminescent, desert-bound gambling capital of the world.

The Grand Challenge is creative even by DARPA's standards. Normally, the agency takes its multibillion-dollar annual budget and allocates it to defence contractors, academic institutions and individual scientists who bring ideas to the agency or answer a request for proposals. As a result, those who usually get to play with DARPA money constitute a relatively small club. But that changed in 2000 when DARPA's congressional paymasters decided that, in addition to the normal procurement process for developing technological breakthroughs, it might make sense to devise special prizes to enable DARPA to reach out to a range of researchers wider than the usual suspects.

DARPA embraced the idea and quickly began to explore technologies that might be prize-worthy. After much discussion, Anthony Tether, DARPA's director, decided in 2002 that the first DARPA prize should challenge researchers to solve the stubborn problem of building land vehicles that navigate and drive themselves.

There were good reasons to do this. For one thing, Congress has decreed that by 2015, one-third of America's combat ground-vehicles must be autonomous. Furthermore, engineers and scientists have spent years, and millions of dollars, grappling with this problem without success. Driving through rough, unpredictable terrain requires machines with sensors and computational capabilities that no one has been able to devise. So DARPA's staff worked out the format for the competition, the prize money was put on the table and the Grand Challenge was born. “By having this race and awarding a prize of this size, we will bring out the people from the garages around the country and stimulate interest in this important field,” Dr Tether said when he announced the prize.

That was the idea, at least: to fire up the enthusiasts, geeks and robot evangelists huddled away in the American hinterlands, and give them a shot at applying their collective genius to problems normally left to corporate giants such as Lockheed, General Dynamics and Boeing. But DARPA was far from certain that the citizen inventors were actually out there. “We would have been happy to field four or five vehicles in the race,” says Jan Walker, a DARPA spokeswoman.

Enthusiasm, however, was not a problem. By early 2003, 106 teams had emerged from the ranks of university students, robot clubs and off-road-vehicle enthusiasts. Of these, 86 presented scientific papers containing the detailed technical specifications DARPA required, and 19 were accepted to participate in a four-day “qualification inspection and demonstration” (QID) to be held five days before the race. Another 26 entrants were judged “possibly acceptable”. To quell grumbling among these teams, DARPA offered each one a visit before making a final decision. This offer was accepted by 19 teams, six of which made the cut. However, says Ms Walker, any or all of the teams could be wiped out in the QID.

Teams of every stripe have been formed, from Alaska to Lafayette, Louisiana. One group from Palos Verdes, California, consists mostly of high-school students, but is not to be underestimated, because many of the students' parents work in the aerospace industry. Another group, Team Phantasm, is based in St Louis and consists of an inveterate tinkerer and a semi-retired computer programmer who have big plans but shallow pockets. Despite meagre resources, the duo have devised a creative flower-petal design that will right their vehicle if it flips over during the race, in much the same way that an insect gets back on its feet when turned over.

Insiders suspect that three teams—from the California Institute of Technology, Carnegie Mellon University and Ohio State University (funded largely by Oshkosh Truck Corporation, a truck-maker)—have the best chance of robotic victory. But the leader of the Carnegie Mellon team, William “Red” Whittaker, a leading field-robotics expert, insists he is taking nothing for granted. “I think it would be presumptuous and pretentious to say that anyone would have better than a 50-50 chance to win,” he says. There are too many variables, and “it's hard to appreciate what these races can do to the machinery.”

Even if the delicate digital hardware running these machines keeps going, and even if they are the smartest roving robots this side of Mars, navigating through the scrub, dust and rock of the Mojave desert represents a complex challenge. The vehicles will use digital cameras, radar and sonar, married with sophisticated home-grown software, to pick and ping their way through the desert. Two hours before the race begins, DARPA will provide the racers with a CD-ROM containing the locations of up to 1,000 waypoints to help teams map the course. But there will still be plenty of challenges. The Mojave is filled with cactus plants, boulders, ravines and ruts. “The vehicles can be repaired and refuelled,” Dr Whittaker says with a grin, “as long as they do it themselves.”

Who will win? Probably no one this time around, although the agency is more optimistic than it was a year ago that one of the teams might manage to claim the prize, according to Ms Walker. But even without an outright winner, there may be rewards of other kinds for those who compete. There are plenty of people in industry and the military who want to solve the autonomous-vehicle problem, says Ms Walker, and they will be watching the race closely.

Besides, says Dr Whittaker, the race is not really about the money. It is about doing something that's never been done before. “In the end, the best technology will win out, but this is really about the triumph of the human spirit,” he says. Which is somewhat ironic, when you consider that the race is for robots only.

Drivers want
Mar 11th 2004
From The Economist print edition

The Vehicles Dynamics Lab at Berkeley, headed by Karl Hedrick, has information on Automated Highway Systems. See also FROG Navigation Systems.

Motoring: It is already possible to build driverless cars, trucks and buses. But practical problems and safety concerns mean they may never be allowed on the roads

THE teams competing in DARPA's Grand Challenge have it easy. The driverless vehicles racing off-road in the Mojave desert merely have to avoid boulders, dunes and the occasional cactus. That is nothing compared with the hazards of the open road. Put those same autonomous vehicles on Interstate 15—the busy road that links Los Angeles and Las Vegas—and they would also have to contend with bleary-eyed weekenders, huge trucks and octogenarians puttering along in mobile homes. Even so, engineers and scientists at a handful of academic and industrial research centres are valiantly grappling with the problem of designing autonomous passenger vehicles, buses and trucks. They imagine a future in which convoys of cars would communicate with each other and with roadside sensors to navigate congested freeways, ensure smooth traffic flow and virtually eliminate accidents.

The first fleet of autonomous cars emerged from the Automated Highway System (AHS) programme funded by America's transport department in the 1990s. After spending $14m, AHS researchers showed off their handiwork in 1997. They implanted 92,000 guidance magnets along a closed seven-mile stretch of motorway near San Diego, California and then released more than a dozen vehicles on to the road. But while the cars performed flawlessly, the programme was cancelled in 1998.

The development of autonomous vehicles has not skidded to a halt, however. In some respects, the technology is quietly sneaking into cars already. Carmakers are starting to incorporate radar-based sensors, micro-cameras and intelligent cruise-control systems—all of which were once limited to research vehicles—into their luxury models. As such equipment becomes more widespread, it could make retrofitting cars for autonomous operation far cheaper and easier.

The technology has also migrated into “group” transit systems. These are small-scale automated transports designed as affordable alternatives to light-rail networks, since they run on cheap roads rather than costly rails. FROG Navigation Systems of Utrecht in the Netherlands, for example, makes an autonomous control system that can be integrated into just about any vehicle. FROG—an acronym for “free ranging on grid”—uses magnets embedded in the road as reference points. Front-mounted radar scans for obstacles, triggering the brakes if needed. It is elegant and simple, and it works: FROG's ParkShuttle whisks passengers from car parks at Amsterdam's Schiphol airport to the terminal building. Even more impressive is the CyberCab, a driverless taxi that can carry riders to designated locations at the press of a button.

Autonomous vehicle technology is also attracting interest from the trucking industry which is especially intrigued by the idea of “platooning”. Commercial trucks often travel along the same regular routes. As a cost-cutting measure, platooning would allow haulage firms to arrange trucks in tight convoys and eliminate drivers in all but the lead vehicle. Braking, gear-shifting and steering commands in the lead truck would be beamed via a wireless connection—called an electronic tow-bar—to the trailing vehicles. The result would be a serpent-like conga-line rumbling down the highway. Chauffeur, a European joint-venture funded by DaimlerChrysler, Iveco, Bosch and others, has demonstrated the technique on Iveco's proving grounds in Balocco, Italy.

But truck platoons are not going to appear on public roads any time soon, reckons Karl Hedrick, an engineer at the University of California at Berkeley. He spent several years developing AHS technologies in the mid-1990s and points out that it is a wary public—rather than inadequate technology—that is ultimately hindering the widespread introduction of autonomous vehicles. Chauffeur's test runs went off without a hitch. But who wants to share the road with an army of 40-tonne driverless lorries?

A more realistic approach for the convoys, suggests Dr Hedrick, would be to designate special platooning lanes, separated by fixed barriers. Then, as drivers become comfortable with the technology and see that it is safe, they might be willing to consider platooning in their own cars. But there is still the problem of managing the formation and dispersal of platoons as cars join and leave, notes Dr Hedrick. If one car needs to refuel, it seems silly for the entire platoon to stop.

Platoons might be safer than individual cars and allow roads to be used more efficiently, but they require drivers to give up their autonomy. For many drivers, that may prove too high a price. In America, in particular, the open road is associated with freedom. A dozen cars coupled into a half-mile-long motorcade, in contrast, sounds an awful lot like a train.