Google's self-driving car makes strides

MOUNTAIN VIEW, Calif. — What's it like riding in one of Google's driverless cars? Picture yourself at the intersection of Liberation Highway and Panic Road, in a white Lexus SUV with a spinning laser on its roof.

Will the car sense all of its changing surroundings? What if someone runs a red light? Should we panic or relax?

"We are all about safety," says Brian Torcellini, lead test driver for the search engine's five-year-old self-driving car. "We realize that everything we do is about upholding the reputation of this technology."

On Tuesday, Google trotted out key technologists working on company co-founder Sergey Brin's pet project and provided media with rides in the vehicle's latest iteration. But this road trip to the future is still zipping up the on-ramp.

Google has mapped thousands of miles of this Silicon Valley suburb and done preliminary mapping in unnamed cities in states such as Florida and Texas. Yet, it remains in test mode with a dozen pricey machines loaded with lasers, cameras, radar and GPS.

"The maps tell the car what to expect, and its (on-board) sensors compare that to the real world," says Chris Urmson, the project's director, noting that the worst accident to date involved a Google car getting rear-ended. "Some 90% of accidents involve human error. What if we can make this go away?"

Google engineers showed digital maps that indicated how its self-driving cars managed to navigate construction cones and brake to avoid a collision. One video showed how a driverless car waited until a stream of bicyclists passed before making a right turn.

"You and I might turn and see a few cyclists coming, but we can't see back a hundred yards," says Urmson.

As exciting as the promise of safely texting while driving may be, hurdles remain high, says Susan Shaheen, co-director of the Transportation Sustainability Research Center at the University of California-Berkeley.

"Among the big questions are, who is responsible if there's an accident? It will cause us to rethink how we insure our transportation networks, much like (ride-sharing service) Uber is already doing," says Shaheen, one of three automotive and urban planning experts on hand.

Other big issues include the high cost of the technology, its sensitivity to hackers who might reprogram a car's route, and serious weather.

"We can do rain and fog reasonably well," says Dmitri Dolgov, software lead on the project. "But we haven't given snow much thought just yet."

The upsides make the work worth pursuing, says Larry Burns, professor of engineering at the University of Michigan who is a consultant on the Google effort.

"This is not an evolution of the car, it's a revolution," says Burns, noting not only the 33,000 people who die each year in car accidents but also the growing cultural trend toward seeing vehicles as shared transportation as opposed to owned hardware.