Google's quest to build and deploy cars that drive themselves has been going well. On Monday, Chris Urmson, director of Google's self-driving car project, said the company's autonomous vehicles have logged more than 700,000 miles.
"With every passing mile we're growing more optimistic that we're heading toward an achievable goal -- a vehicle that operates fully without human intervention," Urmson said in a blog post.
Since its last progress report on the project in August 2012, Google has improved the software that controls its cars to enable the detection of hundreds of distinct objects at once. Urmson says Google's technology can track pedestrians, road signs, cyclists, and a variety of other objects and entities without human liabilities like fatigue or distraction.
The chaos of a city street is fairly predictable to a computer, said Urmson; Google has trained its software by driving thousands of miles on the streets of Mountain View, Calif., where the company is based. Google has built software models for a wide variety of scenarios, from cars stopping at red lights to cars ignoring stoplights, he said. Thousands of situations that would have stumped Google's cars two years ago -- such as the unexpected placement of orange construction cones in a road -- can now be navigated without human aid.
To teach its computers to drive, Google sends employees out to ride with its cars and document anomalous conditions. These scenarios are presented to engineers who then have to implement an appropriate response. Beyond creating algorithms to navigate through areas with road work, Google's cars now slow down when approaching large objects, like a truck parked on a road's shoulder.
Google has also trained its cars to handle railroad crossings, where last year there were 2,087 train-vehicle collisions, 251 fatalities, and 929 injuries, according to the Federal Railroad Administration. When its autonomous vehicles detect train tracks and crossing signs, Google's software waits to make sure the tracks are clear of other vehicles before driving across, to eliminate the chance of being caught behind another car and waiting there as a train approaches.
Cyclists have been accorded special status in Google's software. When an object identified as a cyclist uses a hand signal, Google's cars have been trained to slow down in anticipation of the cyclist's pending lane change. The car will continue to yield to cyclists even when it gets mixed signals from the cyclist, a Google video explains.
Such deference, while advisable for safety, may make Google self-driving cars the object of scorn among human drivers, who tend to prize speed over caution. Imagine how you would react if, during your commute, every other driver stuck to the speed limit and yielded to everything. It would feel as if the road had been taken over by the elderly. What's more, human drivers may not feel the need to drive politely around robot cars.
Then there's the question of how to deal with rage against the machine, something Google already confronts as it buses employees to and from work. In a town like San Francisco, where the relationship between cars and cyclists is often antagonistic, cyclists might be delighted to bring autonomous cars to heel with hand gestures. In a similar vein, imagine how easy it would be for vandals to blind autonomous car sensors using stickers, spray paint, or some other opaque substance.
Google may be able to teach its cars how to behave around people, but judging by the way people abuse computers online, by the tech-hostile climate in San Francisco, and by the social issues facing wearers of Google Glass, the company will have a much harder time teaching people how to behave around its cars.
Anthony Levandowsky, manager of the self-driving car project, said last year that Google was aiming to deploy its self-driving car technology in some form by 2018.
What do Uber, Bank of America, and Walgreens have to do with your mobile app strategy? Find out in the new Maximizing Mobility issue of InformationWeek Tech Digest.