Google’s self-driving cars have been in accidents “surprisingly often” -- 14 times, to be exact -- since the company began testing the vehicles on the open road in 2009. And the project’s director, Chris Urmson, takes this as an especially encouraging sign.
That’s because in every instance, Google’s automated vehicles were hit by other drivers, Urmson writes in a new blog post detailing the latest collision -- which was the first to result in injuries (Google staffers reported minor whiplash.)
Urmson said that such crashes are “a big motivator” in that they illuminate pervasive human fallibility and distractedness on the road. (Of the 14 total accidents, 11 were rear-endings.) Given that self-driving cars have never caused an accident, “We’ll take all this as a signal that we’re starting to compare favorably with human drivers,” Urmson wrote.
Until recently, it’s been hard to make such comparisons because national crash rates have been calculated with police-reported accidents -- even though there are millions of unreported fender benders every year, Urmson writes.
But now that Google’s cars are driving 10,000 miles per week -- which is what a typical American adult drives in a year -- it is ascertaining a more accurate picture of crash frequency, and how to avoid accidents.
“Our self-driving cars can pay attention to hundreds of objects at once, 360 degrees in all directions, and they never get tired, irritable or distracted,” he concludes.