Alex Pell – Times Online November 4, 2007
Last week the sci-fi fantasy of a self-driving car came one step closer to reality. On a former US airbase in Victorville, California, some 30 vehicles – most looking less like Knight Rider’s Kitt than like an average family runabout – gathered to put their driverless skills to the test over a 60-mile urban course, complete with junctions, speed limits and traffic.
From the outside they may appear nothing special, but these cars are competing in an engineering race to develop the first “autonomous vehicle” that could one day be the prototype for the cars sitting on all our driveways. The competition is sponsored by the US government’s Defence Advanced Research Projects Agency (Darpa). Known as the Grand Challenge, it also has potential military spin-offs (a war could one day be fought with robot vehicles).
Darpa was set up to sponsor scientific research after the Russians beat the Americans into space with the Sputnik satellite in 1957. Its projects have already led to the creation of various futuristic robots, superdrugs and, of course, the internet.
With the US military looking to have one-third of its ground-based combat vehicles unmanned by 2015, Darpa wants to rapidly develop the first self-driving car – and is offering big money to the first team of engineers to achieve it. The winner of this year’s contest – the third one in the six years since the Grand Challenge was launched in 2001 – will receive a $2m (£960,000) cash prize.
A team from Stanford University in California won the last event, held in an empty stretch of the Nevada desert back in 2005, using a heavily adapted Volkswagen Touareg dubbed Stanley.
The Stanford team returned for this year’s event, hoping to retain its crown, but the challenge has moved up a gear and is now designed to reflect the real-life challenges of modern urban warfare.
The cars don’t just have to follow a set course, they need to be able to manoeuvre their way around the urban jungle, recalculating their route if a way has been blocked by an unexpected obstacle. The vehicle to complete the course in the fastest time, and without major incident, is the winner.
The Stanford team has an all new car, this time adapted from a Volkswagen Passat saloon and called Junior. It has a roof-rack-mounted device that includes 64 lasers capable of reading everything around the car, up to a range of 65 yards, 15 times a second. The computer in the car’s boot can then adjust its route (preprogrammed using global positioning satellite technology) to navigate around obstacles.
The car’s “brain” contains image-recognition software that enables it to recognise traffic lights and stop, slip around awkwardly parked cars and pull an emergency stop if a pedestrian steps out in its path.
The course, only announced a few hours before the final, was set amid houses and various other buildings. The competing vehicles had to obey all Californian traffic laws, such as signal turns, stick to speed limits and merge into traffic.
The traffic in this instance consisted of other robot vehicles together with cars piloted by the competition judges, who were also keeping nervous fingers on kill switches, designed to instantly stop any robot car going on the rampage.
Tony Tether, director of Darpa, concluded his introductory briefing for the competing teams with a stark message written in huge letters on a board behind him: “Don’t hit anyone.”
Nevertheless, neither this warning nor the judges’ kill switches could prevent Axion Racing team’s driverless Jeep, known as Spirit, from slamming into a judge’s Ford Taurus. The judges were glad of the roll cages in their vehicles as some of the more maverick robot models went awry.
The Golem Group’s car decided at one point to make a break for it and accelerated rapidly towards the barriers at the edge of the circuit before mounting the kerb and blowing a tyre. “The course is dynamic, but fair,” says Tether. Tell that to the Sting Racing team from Georgia Tech, Atlanta, which exited the competition early after ploughing into the concrete barriers at a narrow section of the course, only to find this had later been widened to accommodate the girth of TerraMax, the Oshkosh team’s oversized robot truck.
A winner had not been established as InGear went to press – Junior was a hot favourite, but facing stiff competition from the Tartan Racing team of Carnegie Mellon University, Pittsburgh, which set the early pace by completing the first tranche of parking and navigation tasks without incident. Its adapted Chevy Tahoe SUV was dubbed Boss – which, rumour has it, stood for “Bugger off Stanford”. And who says engineers have no sense of humour?
Diva, the control freak, does it all by herself
Diva may look like a shy, unassuming Volkswagen Passat from the outside, but step inside and you’ll find she’s an outrageous control freak. Don’t even think about getting your grubby hands on her steering wheel. Diva does the driving and you are just along for the ride.
She is the latest in robotic vehicles created by engineers at Stanford University, winner of the previous Darpa (Defence Advanced Research Projects Agency) Grand Challenge for self-driving cars in 2005, and a few weeks ago I was invited to take a “test ride”.
Diva is a prototype for the actual car, dubbed Junior, which last week competed in the latest Grand Challenge. I’m in the passenger seat and Sven Strohband, one of the team’s engineers, is sitting in the driver’s seat, although this is something of a misnomer – the only control is a big red engine-kill button if Diva decides to veer off into a balustrade.
Strohband flicks three switches below the dashboard onto automatic pilot and leans back as Diva bursts into action. His hands are by his side as the steering wheel turns 45 degrees and the pedals depress as if touched by invisible shoes, and we are suddenly pulling a curve at 30mph.
I’m looking outside for a beardie bloke with a remote control lurking in the bushes, but Diva is doing this all by herself. She is being guided by a computer in the boot and multiple sensors around the bodywork. The course, a large ellipse in a car park outside Google’s Californian headquarters, has been programmed in advance, but after that she is on her own.
Junior is even more sophisticated, with a battery of roof-mounted lasers that scan for obstacles several times a second. In 2005 the course was in the Nevada desert; this time the cars had to negotiate an urban driving environment. Junior’s “brain” can now receive images, which look like a real-time video game, and image-recognition software enables it to recognise and obey traffic lights and negotiate obstacles.
Later I watched as Junior, driverless and self-reliant, zipped around the car park, dodging traffic cones and 4x4s parked in its path. Only once did it fail to make the grade – when it decided it could not squeeze through a gap and, paralysed by fear, refused to move back or forward until a handler gave it a metaphorical kick.
Back in Diva, I can’t help expecting her to start talking, a sexier version of David Hasselhoff’s Kitt. Maybe for the next challenge.
Last updated 06/11/2007