The driverless car may soon leave the pages of science fiction novels and speed on to our roads, but is this a piece of technology too far, asks Paddy Comyn?
It is Dublin, the year is 2025. The Metro has just been finished and the M50 has just been reopened, again, after it was widened, again, to five lanes each way. The daily commute for the average motorist is slightly different than it was back in 2007.
The family saloon we drive is now powered by hydrogen. You can now refuel your car at home from a fuel cell station in your garage that extracts hydrogen from natural gas. You can step inside your car, set in your destination, read the headlines and curl up and fall asleep. The car will do the rest.
The idea that we could, one day, sit behind the wheel of autonomous vehicles is not a new one. The driverless car dates back to 1977, when the Tsukubu Mechanical Engineering Lab in Japan developed the world's first intelligent vehicle, based on machine-vision technology. The car achieved speeds of up to 20mph by tracking white street markers.
Later, in Europe, Ernst Dickmanns and his team at Bundeswehr Universität München created a Mercedes-Benz robot van that achieved 100km/h on streets without traffic.
The €800 million EU project Prometheus aided this development and, in 1995, Dickmanns's team created a Mercedes-Benz S-Class that took a 1,000-mile trip from Munich, in southern Germany, to Copenhagen, Denmark, and back, using saccadic computer vision and transputers to react in real time. The robot achieved speeds exceeding 170 km/h on the German autobahn and was, crucially, able to pass other cars in traffic situations.
Stateside in the same year, the CMU Navlab project achieved a 98.2 per cent autonomous journey on its 3,000-mile "No hands across America" trip.
Already, standard production cars are becoming more autonomous. Intelligent or "adaptive" cruise control uses radar to maintain a fixed distance from the car in front, rather than a fixed speed as with a normal cruise control system. Acceleration or braking is applied as required, and the distance is monitored using sensors at the front of the car.
Honda recently unveiled its Collision Mitigation Brake System in its new CR-V, which has recently gone on sale here.
It uses a radar to scan the road ahead and if the system detects an obstacle that is likely to cause a crash, it warns the driver. If the situation becomes more critical, the system prepares the car for sudden braking by moving the brake pads nearer to the brake discs and it will eventually apply the brakes itself where necessary. Similar systems are also employed by Lexus and Mercedes-Benz.
So how far has this technology got? In the US, the Defence Advanced Research Projects Agency (Darpa) - the research arm of America's ground forces - has an interest in creating autonomous military vehicles. In 2005, Darpa funded the Grand Challenge, which was a prize competition for driverless cars. The winner of this race was a Volkswagen Touareg from Stanford's artificial-intelligence lab.
"Stanley" finished the 210km course in a little under seven hours, claiming the $2 million prize money. The project's lead engineer, Sven Strohband, sees a limited future for the autonomous vehicle. "I don't believe the scenario will exist where we'll see fully autonomous vehicles, but we are likely to see increased progress in the avenue of driver assistance," he told The Irish Times from his Silicon Valley headquarters. Strohband is also part of a team of venture capitalists that evaluates investments in the areas of software, systems and materials.
Darpa will hold another challenge this year, but this time the Urban Challenge will present the teams from Stanford, close rivals Carnegie Mellon University (CMU) in Pittsburgh, as well as the other international participants, with an even greater challenge. This time the vehicles will be required to complete a 60-mile course through a mock urban environment in less than six hours. And this involves avoiding other participants, merging traffic and obeying a set of traffic laws.
While the 2005 desert race was a continuous run where objects needed to be avoided, this urban event requires the robotic cars to stop and go. "The cars use a combination of radar sensors, GPS positioning technology, laser range-finders and cameras to create a map, plan a route and execute it, processing its surroundings 20 to 30 times per second, while travelling at up to 56 km/h," Strohband explains.
He sees the future of this technology being very much in terms of "smarter cars that work on dumb infrastructure". Rather than governments having to rip up roads to suit the technology, it would, it seems, be better to allow the cars to adapt to the existing environment. "It might be difficult to get a car to negotiate a city environment, but easy for it be autonomous on a motorway."
Strohband's colleague David Stavens, who is in charge of writing the software for many of Stanley's systems, sees social issues are causing more of a delay. "The technology that allows autonomous cars will be there but it might take longer for people to trust it with their families," he says.
Stavens also thinks it won't be long before the robots' performance will better that of a human driver. "At the moment, we can produce a system that has the processing ability but not the capability of a human, but this will change with time."
There would, naturally be the worry that such autonomous cars would be prone to mischievous hackers or indeed viruses or errors. It would not be much fun to leave the house one morning to find that you car has driven itself to the hands of a car thief.
Imagine the Red Cow roundabout being jammed with cars that are all relying on radar to make a decision. And what would O'Connell Street be like late at night as drunken pedestrians, whose radars have long failed, do battle with driverless taxis. . . although a cab without the bad music and dreadful soliloquies might be a blessing.
Perhaps technology is not such a bad thing after all.