Tuesday 20 March 2018

Look, no hands on the wheel

Robocars will take us out of the driver's seat and Germany is leading the race to build them, reports Jeevan Vasagar

Jeevan Vasagar

The journey begins conventionally enough, with the driver turning the steering wheel to ease the car out of the driveway. But then he flicks a switch and raises his hands in the air. The laptop in the passenger seat, connected to the car's electronics, is now in charge.

On a busy Saturday evening in suburban Berlin, the modified VW Passat drives itself down a tree-lined street crowded with pedestrians. Spotting a cyclist ahead, the car veers left to give her a wide berth, and then slows as a minibus filled with children pulls out. The car's electronic driver judges the minibus's speed, and decides on a surge of power to overtake.

Germany is at the forefront of a global race to develop the world's first driverless car. Devised by a team at the Free University of Berlin, the self-driving Passat is a highly advanced autonomous car, capable of navigating a safe path through unpredictable city traffic.

It "sees" the world through a spinning silver cylinder mounted on the roof, a laser scanner, which generates a million data points per second to give the car's computer a real-time map of its surroundings. Cameras and radar sensors provide a further wealth of detail, alongside an ultra-precise GPS based on aircraft navigation systems.

It will be many years before a Volkswagen this sophisticated is commercially available. But Germany's luxury car makers have begun introducing an array of autonomous features which enable some of their leading models to drive and steer themselves.

Rather than the sudden advent of robot vehicles, car makers believe autonomous driving will be introduced gradually. Daimler, which owns Mercedes Benz, predicts that at low speeds – such as in traffic jams or parking – cars will operate with full autonomy "in a matter of years".

At higher speeds, several manufacturers plan for highly automated driving within the structured environment of the motorway. The latest Mercedes S-Class, unveiled in Hamburg last month with a firework display and a performance by Alicia Keys, is the most significant advance yet in automated driving for the mass market.

"In critical situations, the driver assistance systems in the car react more quickly than humans ever could," says Professor Ralf Herrtwich, director of driver assistance and chassis systems at Daimler. "Unlike drivers, they are never inattentive."

The aim of greater automation, say the car makers, is not to take away the pleasure of the open road but to save motorists from some of the tasks that make driving a chore. Customers would be relieved to hand over to a computer "in specific situations where driving simply isn't enjoyable, for example, in a traffic jam on the daily commute to work," a BMW spokeswoman said.

Raul Rojas, a professor of artificial intelligence at the Free University of Berlin, who leads its autonomous car project, has a more ambitious vision.

He predicts that the public transport of the future will be fleets of robot taxis, cheaper and safer than the human kind, and capable of operating for 24 hours a day without fatigue. In a few years' time, Professor Rojas suggests, people will use their mobile phones to summon a robot taxi.

There are still technical barriers to the mass production of automated cars.

The key problem is vision. The laser scanners used by the Free University's car cannot pass through a solid barrier, so a pedestrian stepping out from between two parked buses would be invisible until the last moment.

An automated car "does not just need to watch out for other cars," Professor Herrtwich says. "It needs to check for pedestrians and cyclists, for lanes, stopping lines, traffic signs, and traffic lights. It needs to understand that a lane ends in a few hundred metres or that there is a stalled vehicle on the highway ahead of it."

This technology does not just need to be reliable, but affordable within the budget constraints of a typical mass production vehicle. The Free University's team is working on improving the computer's vision, but faces a huge challenge replicating the human ability to see.

"Almost a fifth of the brain is just for image processing," Professor Rojas says. "There's lots of things that we don't yet understand about how human vision works. Computer visual recognition is still very, very primitive."

There are legal barriers too, as the Vienna Convention of 1968, the global legal framework for driving, requires that motorists are in full control of their vehicle at all times. Another important question is that of legal liability, as a BMW spokeswoman points out: "Who is responsible for an automated driving system in case of an accident or failure?"

In the US, three states – Nevada, Florida and California – have all passed laws permitting the testing of driverless cars. This follows lobbying by Google, which is also developing technology for automated vehicles.

At present, the Vienna Convention means that all automated cars must have human "safety drivers". That means that when the Free University's car goes out on the road, a member of the research team sits in the driving seat and a colleague holds the laptop that directs the car.

The scientists here are convinced that the law will change once the technology has been proven. Before fully driverless cars become a reality, there is one more problem to solve. The relationship between automated vehicles and human road users is the greatest challenge of all.

After all, machines lack our social skills. Computer-driven cars cannot judge, for example, whether a pedestrian waving at them from the kerb is urging them to stop or go.

"That's the last frontier: predicting human behaviour," Professor Rojas says.

Irish Independent

Life Newsletter

Our digest of the week's juiciest lifestyle titbits.

Also in Life