Google: Self-driving cars in 3-5 years.

Google: Self-driving cars in 3-5 years.

 The self-driving car could be available to consumers in 3-5 years, the head of Google’s autonomous driving project says. That’s the most optimistic timeframe yet. Other projections have been for 2020 and beyond, which still beats “probably not in our lifetime.” The timeline came from Anthony Levandowski, Google’s product manager for autonomous driving, speaking at a Society for Automotive Engineers conference in Washington last week. “I can’t tell you you’ll be able to have a Google car in your garage next year,” Levandowski said. But he added, “We expect to release the technology in the next five years. In what form it gets released is still to be determined.” Releasing technology is different than announcing a self-driving car going on sale in 3-5 years. The real challenge could be getting the self-driving car approved for use of public highways everywhere, not just the handful of states that allow self-driving cars for test purposes.

 It’s a free country and if you want to mount a soap box and speak out in a public park, you can do it today. If you want to sell a motor vehicle, you have to meet an array of fuel-economy, stability, and crash safety tests, and those take time. Years sometimes. Self-driving cars would have to prove they can drive themselves, deal with jaywalking pedestrians, stop or at least slow when a crossing vehicle runs a red light, and deal with software glitches. Performance metrics for self-driving cars don’t exist yet and the Department of Transportation, which doesn’t offer bonuses for working at internet speed, may be more deliberate and methodical than Google.
“How will the government come up with a performance standard?”

Google 300,000 miles, Lexus self-driving carDan Smith, senior associate director for vehicle safety at the National Highway Transportation Safety Administration ,also spoke at the Jan. 31 SAE panel, Safer & Cleaner Transportation in a Digital Age. According to a report by Bloomberg News, Smith noted, “It gets to be a massive challenge to figure out how will the government come up with a performance standard that is objective and testable for so many different scenarios where failure could possibly occur. Part of that has to do with if we should be looking at the underlying electronics.”

When it comes to regulating auto safety, NHTSA is more tortoise than hare, partly because government moves methodically, partly because it may not be up-to-speed on technology. Last year, the National Academy of Sciences issued a report that took issue with NHTSA’s tech savvy, saying it needed to be more knowledgeable about current car electronics and take the lead in setting car electronics standards proactively. For what it’s worth, the guy Google sent to the forum  founded a company before joining Google and has two degrees in industrial engineering and operations research. NHTSA’s man on the scene has a law degree and two political science degrees; his employer does not offer stock incentives. The National Academies report was looking back at how NHTSA lacked expertise to deal effectively with unintended acceleration issues. It said NHTSA needs an outside board of experts it can call on to assist NHTSA’s engineers.

How to insure the self-driving car may also be an issue. Google’s Levandowski and others have said self-driving cars should be safer and cut accidents, reducing insurance rates. Insurers commonly offer discounts for high-tech safety gear such as stability control. But sometimes they balk. Anti-lock braking systems (ABS), which reduces skidding, might actually cause more accidents if it leads drivers to think they can drive at speed on snowy roads and ABS will stop them in time. So it’s not clear if the insurance on self-driving cars would be lower or higher.
Is the 3-5 year timeframe realistic?

When Levandowski described scenarios where Google would be ready to release the autonomous-driving technology in three years or five years (reports say it was described both ways), that’s not the same as dealer-ready cars. Google creates software technology. Ford, Toyota and Audi build cars. To create a car from the ground up, a new model (not an all-new-for-2013 refresh) takes 3-7 years. (The Chevrolet Corvette that was the hit of the recent Detroit auto show will be nine years between models.) The Department of Transportation will have to come up with tests to make sure self-driving cars don’t crash; the toughest part will be testing that shows the software doesn’t crash. Suppliers will have to bring down the cost of autonomous driving sensors. Currently the sensors and hardware (pictured below) cost more than the car they’re mounted on. All that suggests 2020 would be a more reasonable time to think about a completely self-driving vehicle.

In the meantime, look for assisted-driving cars that self-drive (loosely defined) under certain limited conditions. On limited access highways, they’ll maintain a safe following distance and keep pace with traffic ahead and they’ll center themselves in the driving lane, as long as the lane markings are clearly defined. If a driver were to nod off, the car would probably be able to drive safely for miles and miles. Most likely the car would need a drowsy driver alert (a camera watching for flickering eyelids or a movement sensor tracking minor corrections to the steering wheel), since the act of not driving while behind the wheel may be a little too restful.