How Google’s Driverless Cars Work

A Google self-driving car caused a crash on Valentine’s Day after traveling more than 1.3 million miles since 2009, Wired reported Friday. No one was hurt and the damage was minimal, but the crash shows companies developing the technology still have a way to go before autonomous vehicles are actually commonplace.

In mid-February, as it was driving on a three-lane road in Mountain View, California, a Google self-driving car trying to merge into the center lane hit a public bus. The car had expected the bus to yield.

In a statement issued Monday, Google said it “bears some responsibility” in last month’s crash, Reuters reported. In the same statement, Google said it has reviewed the crash and made some changes in its cars’ software to prevent similar incidents. Unlike human drivers who can react quickly to unexpected behavior or obstacles on the road, autonomous vehicles aren’t always able to do that.

This is just one of the few important wrinkles Google and other tech companies testing autonomous vehicles need to smooth out. Another flaw with Google’s cars is they require the roads they’re traveling to be mapped out precisely, otherwise the vehicle is unsure how to behave. These maps are extremely detailed digitizations of the physical world where things like even the height of every single curb is measured.

Before February’s incident, Google reported its driverless cars had been involved in only 17 accidents, all of which had been caused by human error.

What allows self-driving cars to be such “careful” drivers is the combination of GPS systems, cameras and various sensors that tell the cars where they’re going, how to get there and what’s around them at all times.

The autonomous vehicles use GPS to match their position with customized Google maps. This allows the cars to select a starting point and an end point, as well as to choose the best routes to take. Radars in the front and back of the cars keep track of other vehicles on the road. The LIDAR sensor rotates 360 degrees and detects the distance between the autonomous vehicles and surrounding objects. A video camera behind the front windshield is able to read road signs and traffic lights.