LEAN back, let go of the steering wheel, ease your feet off the pedals and relax: your car is now in charge. The dream of a car that can drive itself has grown over the last decade as the necessary technologies have gradually proved their worth, but the idea has faced major legal hurdles.
Not for much longer. Politicians are now scrambling to make self-driving cars a reality. From Hawaii to Florida, and Oxford to Berlin, the race is on to get driverless cars onto our streets.
Promising improved safety, better fuel-efficiency and freedom from the boredom of long drives, autonomy has been coming piecemeal to our cars for some time ā and it has always had its critics. In 1994, on a UK motorway, Jaguar and Lucas Industries demonstrated the safety of adaptive cruise control and automatic lane keeping; both technologies are now commonplace on our roads. The media were not impressed, describing the idea of cars that drive themselves as āmadnessā.
Advertisement
But concerns about the safety of autonomous cars are misplaced in a world where 1.2 million people die every year in road accidents due to human error, says Paul Newman, a robotics engineer at the University of Oxford, whose team is .
āItās crazy to imagine that we are going to keep driving cars like we do now ā that in 10 to 20 years weāll still have to sit behind a wheel, concentrating hard, not falling asleep and not running over people,ā he says.
This notion now has powerful backers ā and barriers are beginning to fall. In that came into force on 1 March, the state of Nevada now allows driverless cars to ply the stateās road network provided they sport a special red licence plate, and the owners pay a $1 to $3 million insurance bond. Similar legislation is in California, Arizona, Florida, Hawaii and Oklahoma.
āDriverless cars can ply Nevadaās road network provided they sport a special red licence plateā
The phenomenon is not confined to the US either. In Germany, a led by Tinosch Ganjineh at the Free University of Berlin has permits to use the abandoned Templehof airport for autonomous tests. When necessary, team members get special permits to drive on Berlinās streets, and hope to drive on the autobahn soon. The Oxford team plan to approach the British government for similar permits.
The Berlin team are automating a VW Passat, patriotically named MadeInGermany, while Oxford is turning a BAE Systems WildCat military jeep into a self-driving machine. Nissan has just joined the Oxford project, so the Leaf all-electric car may end up driverless too.
Driverless cars first appeared in a meaningful way in the US Defence Advanced Research Project Agencyās āgrand challengesā. Cars competed to drive fastest around desert courses in 2004 and 2005, and in an urban setting in 2007.
Mike Montemerlo and Sebastian Thrun of Stanford University, California, whose car won the 2005 prize, lead Googleās . Their cars, based on the Toyota Prius and Audi TT, typify the approaches of the Oxford and Berlin teams. All the cars have laser rangefinders, radar and optical cameras to sense the vehicleās changing real-time environment with high accuracy. They know where the traffic lights and road signs are, and which moving objects are animals, people, bikes, motorbikes or trucks. Newmanās team are studying how algorithms can make sense of data streaming from a 3D laser rangefinder and quickly decide whether an object is a car or pedestrian, for example. His team is also looking at how a robotic visual system can build up a picture of its world and adapt to changing conditions, varying light levels or even seasons. The commercial sensors and software to make this happen are still some way off, though.
āThe Velodyne ā the 64 spinning lasers on top of most driverless cars ā give a quickly updated 360-degree, 3D view of the surroundings up to 40 metres away,ā says Newman. But cars of the future wonāt have unwieldy spinning lasers on them, he says.
Ganjineh agrees that driverless technology has to be refined. āThe size and price of these systems needs to come down. Today, half a trunk of equipment is needed for autonomous driving,ā he says.
Another challenge, says Newman, is getting the cars to recognise the precursors to risky events ā like sudden bright sun reflections on the road, truck spray, which may blind some sensors, or simply a burst tyre.
Googleās cars, meanwhile, tell each other about the roads they have travelled, such as exchanging data on how to negotiate awkward junctions, says Vinton Cerf, a Google technology evangelist. Ganjineh wants similar technology to broadcast GPS map changes car-to-car, when there are roadworks ahead, for example.
However, driverless cars will not need to communicate wirelessly with expensive roadside technology as they need to be āindependently smartā and aware of all risks around them at all times, says Newman.
āAutomation of cars is going to happen,ā he says. āComputing has caused devastating change and transport is going to be its next target.ā
If the goingās tough, the car gets cover
Driverless cars could reduce insurance costs, says Paul Newman of the University of Oxford, by allowing the car to add to its own insurance as road conditions change.
āOn a dark icy night, when it is riskier to drive, the car could go online and bid for extra insurance cover until conditions change,ā he says. āIf that proves too expensive, because conditions are tough for the autonomous system, the owner could take the wheel.ā
Meanwhile, clear standards for programmers and developers of such cars need to be drawn up, says Tinosch Ganjineh at the Free University of Berlin, Germany, as accident liability may fall more often on software or sensor-makers.
