Fresh out of college, Brian Torcellini dreamed about getting a job writing about surfing.
Instead, he wound up in a dusty, dimly lit garage near Google’s Silicon Valley headquarters six years ago, getting ready to ride a wave of technology that thrust him into an occupational oxymoron. He became a driver in a driverless car.
Torcellini, 31, now leads a crew of test, or “safety,” drivers who are legally required to ride in Google’s fleet of 48 robot cars that the Internet company’s engineers are programming to navigate the roads without human assistance.
“A lot of people go to work and sit in a cubicle,” Torcellini says. “Our cube just happens to move around the roads. And if we are successful, we are going to put ourselves out of a job.”
The driverless cars already have logged more than 2 million miles in six years of sometimes tedious testing on private tracks, highways and city streets located mostly near Google’s Mountain View, California, headquarters.
The vehicles have traveled more than half that distance in automated mode, with one test driver in place to take control of the car if the technology fails or a potentially dangerous situation arises. Meanwhile, another driver sits in the front passenger seat typing notes about problems that need to be fixed and traffic scenarios that need to be studied.
If the technology advances as Google envisions, the only people sitting in driverless cars by 2020 will be passengers looking for an easier way to get around.
Everything will be controlled through a combination of sensors, lasers, software and intricate maps, eliminating the need for the vehicles to be equipped with steering wheels or brake pedals.
“When you go scuba diving and take a moment to really think about it, you realize you are doing something that isn’t supposed to be humanly possible: you are breathing underwater,” Torcellini says. “It’s the same kind of feeling you get in one of these cars. It’s not supposed to be humanly possible.”
Before they are entrusted with the cars, Google’s test drivers must complete three-week training courses. The drivers are taught to take control of the robot car whenever there is any moment of doubt or danger.
Google employs “dozens” of test drivers but won’t reveal the precise number. It’s likely around 100 because California law requires two test drivers per vehicle, and Google’s fleet currently consists of 25 pod-like cars and 23 Lexuses.
A few of those self-driving cars Google also recently began cruising around Austin, Texas.
Besides having clean driving records, Google’s test drivers say the job requires a combination of good judgment, patience and fearlessness. The self-driving cars were in 16 accidents from May 2010 through August, but they are becoming more frequent as the vehicles spend more time on public roads. Half of the collisions have happened since February — a stretch when the self-driving cars were traveling an average of about 10,000 miles per week on public streets in autonomous mode. There have been no major injuries reported so far.
The self-driving technology hasn’t been to blame for any of the accidents, according to Google, though it says one collision was caused by an employee who was steering a robot car while running a personal errand. In all but three of the accidents, Google’s self-driving cars have been rear-ended, a trend that the company believes has to do with the large number of motorists who are texting, talking on the phone or otherwise doing something besides paying attention to the roads and their surroundings.
“There are tons of situations where we see people who just aren’t very good at driving out there,” Torcellini says. “It’s up to us to teach the (robot) cars to be better than those drivers, and even better than the best drivers, too.”