How Google's autonomous car navigates city streets

Google's self-driving car has been on the road for five years now, at various levels of autonomy. From the ease and relative serenity of California's arrow-straight highways, the car drove hundreds of thousands of miles with a greater level of concentration and mastery than the wandering attention spans of humans could accomplish. In 2012, Google shifted from the freeways to the cities, navigating a far more convoluted set of challenges: the slow-speed chaos that comes with any city, any suburb, any place with people and cars in it.

Google has driven around its Mountain View, Calif., home for thousands of miles, dodging cyclists and delivery trucks and the riff-raff of techy suburbia. Eric Jaffe of The Atlantic's Cities blog was one of the first journalists invited to sit in the backseat of Google's self-driving car, while software lead Dmitri Dolgov and lead test driver Brian Torcellini rode up front. During the drive, they did "no more to guide the vehicle than I'm doing from the back seat," reported Jaffe.

Most of Google's cars have been Toyota Priuses, but Google employs about two dozen Lexus RX 450h crossovers, large enough to handle all the equipment to make the autonomy work. A "bucket" on the roof spins 10 times a second as it emits laser beams to map its environment in 3-D. Four-direction radar bounces forth 150 meters "to perceive things a human driver never could." And there's a big red button -- "'Every robot has a big red button,' says Dolgov" -- as a kill switch.

Nobody at Google has reportedly ever had to use the kill switch.

Google cars have been involved in two crashes. The first crash, says Google, was

By Blake Z. Rong