WHY SELF-DRIVING CARS COULD BE A DREAM COME TRUE FOR CAR THIEVES

Vehicles that stop for pedestrians could be a gift to thieves

By Corin Faife | VICE News | February 22, 2017

One moment you're cruising to a halt at a red light; the next, being pulled roughly out of your vehicle by an armed man in a mask. It's every driver's nightmare and the crime that helped Grand Theft Auto become a multi-billion dollar franchise: carjacking.

There are endless depictions of carjacking in popular culture, perhaps tied to the role that the automobile occupies in the American psyche. Certain places like Los Angeles, New Jersey or Detroit are notorious for it, with some incidents linked to organized crime rings who will steal to order and arrange international shipments for their discerning (if unscrupulous) foreign customers.

But the vehicles we drive are changing, as is the way we drive them. As a world of widespread autonomous vehicles moves from the realm of science fiction into near-future possibility, it's important to consider how vehicle crime will change, and how the benefits from the elimination of certain categories of crime linked to vehicles—drunk driving, for example—could be tempered by the growth of other types.

When parked, advances in security have made modern cars far more difficult to steal than used to be the case. But if a thief can stop a car with the driver in it, say by bumping it deliberately with another car, creating a distraction or blocking the road with some kind of physical barrier, immobilizing technology is made irrelevant by having access to the ignition keys and an unlocked vehicle.

This creates a problem for self-driving cars, which are explicitly designed to be risk averse when it comes to interactions with pedestrians. In a paper in the Journal of Planning Education and Research, Adam Millard-Ball, an assistant professor at the University of California, Santa Cruz, argued that pedestrians would learn to become more aggressive when crossing roads in front of autonomous vehicles in the knowledge that the vehicle would be forced to stop. Although Millard-Ball wasn't addressing malicious intent, it's not hard to imagine criminal gangs using the same principles to "herd" an autonomous vehicle into a position where it could be commandeered, or the driver harmed. (Of course, the possibility of stealing a car in this situation also depends on the steering system: a vehicle with a manual override would be simple enough to take, but a fully autonomous vehicle with pre-programmed destination, say an airport-to-city shuttle taxi, would be difficult to make off with.)

Mary "Missy" Cummings, Director of the Humans and Autonomy Lab at Duke University, says that the optimism surrounding driverless cars has discouraged people from talking about the darker aspects of the behavior they could enable.

"I've been saying for a long time [that] I think that there's a lot more criminal behavior that will happen in these cars than people understand," Cummings said in a phone call. "Interiors of cars will be gutted, hubcaps will be destroyed, headlights will be taken out of these cars. Without a human driver in the car, we're going to see some really bad behavior."

Cummings says that she's more cynical about this kind of behavior happening in America than in, say, Europe, citing the sad tale of Hitchbot as an example of what can happen when people choose to enact violence on anthropomorphic robots. But in the case of carjacking it's not just the robotic cars, but their occupants, that could be at greatest risk.

"There's a section of society, like it or not, that's going to be predatory towards these vehicles," she said. "Hacking of cars is real. It's not hard at all to GPS spoof any vehicle that relies on GPS—that's true of drones and driverless cars—and so at a minimum we have a real vulnerability that if you're in a driverless car you could be kidnapped, electronic locks could be engaged externally, and if there's no steering wheel, good luck!"

Nonetheless, Cummings is clear that the possibility of police being able to remotely control driverless cars is not an appropriate response. Instead, closer liaisons will need to be made between the police and a new kind of dispatch center run by Tesla, Uber or other driverless car manufacturers, which will allow for quick reaction when a problem occurs—whether mechanical or criminal—that necessitates the intervention of an engineer, either physically or remotely.

Although autonomous vehicle systems look set to revolutionize urban transport, designing for these kind of edge cases will be a crucial part of the transition from human-controlled to self-driving transport networks.