News

net.wars: Drive by wire

The day in 1978 when I first turned on my CB radio, I discovered that all that time the people in the cars around me had been having conversations I knew nothing about. Suddenly my car seemed like a pre-Annie Sullivan Helen Keller

Judging by yesterday's seminar on self-driving cars, something similar is about to happen, but on a much larger scale. Automate driving and then make each vehicle part of the Internet of Things and suddenly the world of motoring is up-ended.

The clearest example came from Jeroen Ploeg, who is part of a Dutch national project on Cooperative Advanced Cruise Control. Like everyone here, Ploeg is grappling with issues that recur across all the world's densely populated zones: congestion, pollution, and safety. How can you increase capacity without building more roads (expensive) while decreasing pollution (expensive, unpleasant, and unhealthy) and increasing safety (deaths from road accidents have decreased in the UK for the last few years but are still nearly 2,000 a year)? Decreasing space between cars isn't safe for humans, who also lack the precision necessary to keep a tightly packed line of cars moving evenly. What Ploeg explains, and then demonstrates on a ride in a modified Prius through the Nottingham lunchtime streets, is that given the ability to communicate the cars can collaborate to keep a precise distance that solves all three problems. When he turns on the cooperative bit so that our car talks to its fellow in front of us, the advance warnings significantly smooth our acceleration and braking.

"It has a big potential to increase throughput," he says, noting that packing safely closer together can cut down trucks' fuel requirements by up to 10 percent from the reduction in headwinds.

But other than that, "There isn't a business case for it," he says sadly. No: because we don't buy cars collaboratively, we buy them individually according to personal values like top speed, acceleration, fuel efficiency, comfort, sporty redness, or fantasy

To robot vehicle researchers, the question isn't if self-driving cars will take over – the various necessary bits of technology are too close to ready – but when and how people will accept the inevitable. There are some obvious problems. Human factors, for one. As cars become more skilled – already, they help humans park, keep in lanes, and keep a consistent speed – humans forget the techniques they've learned. Gradually, says Natasha Merat, co-director at the Institute for Transport Studies at the University of Leeds, they stop paying attention. In critical situations, her research shows, they react more slowly; in urban situations more automated means they're more likely to watch DVDs until or unless they hear an alarm sound. (Curiously, her research shows that on motorways they continue to pay more attention; speed scares, apparently.) So partial automation may be more dangerous than full automation despite seeming like a good first step.

The more fascinating thing is what happens when vehicles start to communicate. Paul Newman, head of the Mobile Robotics Unit at Oxford proposes that your vehicle should learn your routes; one day, he imagines, a little light comes on indicating that it's ready to handle the drive itself. Newman wants to reclaim his time ("It's ridiculous to think that we're condemned to a future of congestion, accidents, and time-wasting"), but since GPS is too limited to guide an automated car – it doesn't work well inside cities, it's not fine-grained enough for parking lots – there's talk of guide boxes. Newman would rather take cues from the existing infrastructure the way humans do. But give vehicles the ability to communicate and share information – maps, pictures, and sensor data. "I don't need a funky French car bubble car. I want today's car with cameras and a 3G connection."

It's later, over lunch, that I realize what he's really proposing. Say all of Britain's roads are traversed once an hour by some vehicle or other. If each picks up infrastructure, geographical, and map data and shares it…you have the vehicle equivalent of Wikipedia to compete with Google's Street View.

Two topics are largely skipped at this event, both critical: fuel and security. John Miles, from Arup argued that it's a misconception that a large percentage of today's road traffic could be moved to rail. But is it safe to assume we'll find enough fuel to run all those extra vehicles either? Traffic increased in the UK by 85 percent since 1980; another 25 percent increase is expected in just the next 20 years.

But security is the crucial one because it must be built into V2V from the beginning. Otherwise, we're talking the apocryphal old joke about cars crashing unpredictably, like Windows.

It's easy to resist this particular future even without wondering whether people will accept statistics showing robot cars are safer if a child is killed by one: I don't even like cars that bossily remind me to wear a seatbelt. But, as several people said yesterday, I am the wrong age. The "iPod generation" don't identify cars so closely with independence, and they don't like looking up from their phones. The 30-year-old of 2032 who knows how to back into a tight parking space may be as rare as a 30-year-old today who can multiply three-digit numbers in his head. Me, I'll wave from the train.