In the six years that Google's self-driving cars have been on the road in California, they've been involved in only a handful of accidents. (And gotten just one traffic ticket that we know of.) These accidents have all been the fault of humans—either humans who plowed into Google's cars, or humans who were driving Google's cars in manual mode, overriding the car's autonomous driving system.
But two weeks ago, for the first time, one of Google's self-driving cars was at fault in an accident. It hit a municipal bus in Mountain View, according to new documents filed with the California DMV.
The accident happened on February 14, according to the documents, when a Google self-driving car pulled into a right-hand turn lane on El Camino Real in Mountain View. Moving to the rightmost side of the lane in order to turn, the car came up against sandbags on the road, blocking its way. The self-driving car veered left to avoid the sandbags, which put it in the path of an oncoming bus. The Google car's software told it that the bus would likely yield, but the bus kept going, and…crunch.
According to Reuters, "as the Google car in autonomous mode re-entered the center of the lane, it struck the side of the bus, causing damage to the left front fender, front wheel and a driver side sensor."
Luckily, nobody was hurt—the bus was moving only 15 miles an hour when the collision happened—and Google says it is working to fix the issue with a software update. Google sent over an advance copy of its February self-driving car report, in which it said:
"This is a classic example of the negotiation that’s a normal part of driving – we’re all trying to predict each other’s movements. In this case, we clearly bear some responsibility, because if our car hadn’t moved there wouldn’t have been a collision. That said, our test driver believed the bus was going to slow or stop to allow us to merge into the traffic, and that there would be sufficient space to do that."
In other words, the Google car didn't realize that bus drivers don't like to get out of the way of other cars on the road.
Last fall, Fusion's Alexis Madrigal reported on the ways in which Google is trying to accident-proof its self-driving cars. The company, now known as Alphabet, maintains a 60-acre test facility at Castle Air Base near Merced, California, where it sets up lifelike simulations of one-in-a-million road occurrences and tests the cars' ability to respond properly. The company hired Jaime Waydo, a former NASA engineer who built autonomous rovers, to put its self-driving cars through a range of odd tests, like: what if a person in a wheelchair suddenly pulled out in front of the car?
“No matter how much experience you gain, there are always going to be things out in the world that happen to you that you have never encountered before,” Dmitri Dolgov, the lead engineer on Google's self-driving car project, told Fusion.
Google's self-driving cars are getting better at recognizing and dealing with these fringe occurrences. And there's still no question that a world with self-driving cars would be dramatically safer than a world where humans are allowed to drive themselves.
But self-driving cars still aren't perfect. And as companies like Google and Tesla edge their self-driving technologies closer to the mass market, they'll have to be ready to answer questions about every accident caused by their software, no matter how small.