One road, four wheels, no hands

The roads of Britain and other industrial countries may seem congested to bursting point; in fact, they are not. Because of the need to leave a safe stopping distance between moving cars, even the M25 at its busiest is more than half empty. If the gaps between cars could somehow be turned to use, road capacity could be doubled or tripled at a stroke - without building a single extra mile of road.

Until recently, such an idea has been no more than a pipe dream. But the technologies needed to allow cars to be driven in convoy, bumper-to-bumper, down the motorways at 70mph - ranging from video cameras to image recognition software, from satellite-based navigation to brute computing power - are advancing at a prodigious rate. What was formerly the stuff of science fiction is fast becoming commercial fact. At a conference in Paris last month, 2,500 specialists in Intelligent Transportation Systems (ITS), as the new area of research is known, met to compare notes.

According to a study carried out for ITS America, a not-for-profit organisation set up to link the efforts of US businesses, universities and the federal government, money is being pumped into intelligent highways in Europe, Japan and the US. The study says the Japanese government is spending most, dishing out some $1.9bn between 1985 and 1992, to complement the billions that Japanese firms have poured into satellite-driven systems which tell drivers in Tokyo, at a glance at a small TV screen, whether t o turn left or right.

But the other two continents are not far behind: the European Union, with its incurable appetite for alphabet soup, has an ambitious project called Prometheus, or Programme for a European Traffic System with the Highest Efficiency and Unprecedented Safety. And the US progamme aims to deliver, within 10 years, a working prototype of an intelligent system in which a number of cars can be driven on public roads without human intervention.

A visit to Carnegie-Mellon University (CMU) in Pittsburgh, Pennsylvania, where some of the most advanced research is being done, suggests this goal may be closer than expected. Ten years ago, the state of the art in autonomous navigation was a gadget called the Stanford Cart, which would move a few jerky inches and then take a 15-minute break to look around and plot its next step. With generous funding from the US Department of Defense, which would like to send driverless tanks into battle against the Iraqis and other malefactors, the CMU researchers have developed a full-sized vehicle, built on the platform of a military ambulance, which can find its way both on roads and off at normal driving speeds. Known as Navlab 2, the vehicle carries power generators, sensors, computers and - in case the software goes awry, and a human hand is needed on the steering wheel - graduate students.

In its early manifestations, the vehicle was the butt of jokes on the campus, with professors from other faculties claiming that it was responsible for running over scores of squirrels on the local golf course. The researchers responsible for it, Dean Pomerleau and Chuck Thorpe, deny this. Proof of their willingness to back their work can be seen in the fact that their latest self-driving vehicle is a smart new Honda Accord, Mr Pomerleau's own car. Thanks to the shrinking size of computers, it can be controlled by a powerful notebook computer on the back seat; and the camera on its windscreen is little more prominent than the extra mirror used to tow a caravan. The Honda has cruised around the Pennsylvania highways without incident. Testing it requiredno permission from the authorities, but attracted curious glances from traffic cops.

Perhaps surprisingly, given researchers' tendency to become over-enthusiastic about their creations, the scientists at CMU are realistic about the prospects of putting their work into practice. The next few years are to be spent working on the specifications of America's system, rather than actually designing and building it. "It'll be at least 2000 before we see cars steering themselves," says Mr Pomerleau.

One reason is that the US government will have to make difficult choices before it can hope to see cars driving themselves around the highways. The most important is whether the intelligence should be only inside the cars, or partly installed in the roadnetwork.

The arguments for confining work to cars are powerful. If magnets or beacons were to be used in the roads, the computational job of navigating a car might be simpler, but the system would be prohibitively expensive to install, since it would require roadworks to every mile of the country's road network. A system that depended on technology in the roads might also provide a useful source of material for sci-fi nightmares, unless it were designed so that a saboteur would need to do more than merely wave an electromagnet in order to lure cars off a mountain road on a foggy night. Most important, however, is the problem of redundancy: with technology advancing quickly, any road-based system would be doomed to obsolescence before it was complete - and countries that followed the US lead would find themselves in possession of cheaper and better road systems.

Car-based systems, however, have drawbacks, too. One obvious one is that even in mass production, it would cost several thousand pounds to equip a car with the navigational and computing gadgets needed, although the cost would be greatly reduced in time.A more subtle problem is that making decisions entirely on the basis of what a car's sensors tell it about the outside world is greatly more difficult, and will prove a prodigious challenge to researchers for years to come.

The programme is certain to yield benefits on the way. The work already done makes it possible to imagine collision-avoidance systems that could detect when a car is drifting towards the central barrier of a highway or getting too close to a car in front, and either warn the driver to change course or take control to avert an accident. Such systems, though, will gain public acceptance only when they are significantly more reliable than people - for accidents caused by faults in the software would generate acres of hostile publicity even though we shrug our shoulders at motorway pile-ups caused by human error.

When the system becomes a reality, it is certain to raise questions of public policy. Since people would be able to do something useful in their cars as they commuted, ITS would change patterns of land use and work as much as the building of highways in the US did in the first place. But it also raises a more frivolous question: will people want to give up driving, and spend their time reading in the back seat? For many people, cars are more than just a cheap, safe way of getting from A to B; a car is astatus symbol, emotional punchbag and game of skill all rolled into one.