Self-driving cars twice as likely to be in a crash

Autonomous vehicles are more likely to be involved in a car accident than one piloted by a human, according to new research.

The reason, says researchers, is self-driving cars are twice as likely to be involved in an accident because they’re far too law abiding and people keep crashing into them.

So far, all incidents involving autonomous cars have been minor, non-fault accidents, but the growing number of human drivers crashing into the self-driving cars has prompted programmers at companies like Google to begin to
question whether autonomous cars of the future should adopt a more aggressive driving style and be prepared to break the law if it avoids an accident, however minor.

Speaking to newswire, Bloomberg, Raj Rajkumar from General Motors-Carnegie Mellon Autonomous Driving Collaborative Research La, Pittsburgh said: “We have basically decided to stick to the speed limit. But when you go out and drive the speed limit on the highway, pretty much everybody on the road is just zipping past you. And I would be one of those people.”

Rajkumar explained to the newswire how GM’s autonomous cars struggle to “trust” other road users when merging onto highways, especially when those drivers are speeding.

This has forced the operator to intervene, to speed up to squeeze into traffic - something the current software forbids, but changing the behaviour of self-driving cars is proving controversial, even among those who develop them.

“We end up being cautious,” Rajkumar said. “We don’t want to get into an accident because that would be front-page news. People expect more of autonomous cars.”

Analysis carried out by the University of Michigan’s Transportation Research Institute found that the most common cause of accidents involving autonomous vehicles were other drivers hitting the self-driving vehicles from behind.

In almost all cases, the university found inattentiveness or aggression from the human driver was the cause of the crash, with the autonomous vehicle completely blameless for the accident.

Google, one of the leaders in the development of autonomous driving, has recorded 17 minor crashes in over three million kilometres of real-world driving.

It, like GM, admits that its autonomous cars are currently overly cautious following an incident where a Google car was warned by Police for driving 24mph in a 35mph zone.

“We err on the conservative side,” said Dmitri Dolgov, an engineer from Google, speaking to Bloomberg, adding that its self-driving cars drive like a “cautious student driver or a grandma.”

It admits that in the future its vehicles will be more “aggressive” like humans so they “can naturally fit into the traffic flow, and other people understand what we’re doing and why we’re doing it”, but Dolgov rules out making driverless cars that would break the law.

Google also says it will continue to make its self-driving adopt more human behaviour on the road.

Already, its vehicles inch forward in a queue to turn right, rather than leave a big gap - which isn’t logical for the computer. These types of behaviour should, in the future, make it easier for humans to predict what a robocar might be about to do because, in Google’s engineer’s words: “Driving is a social game.”