Google shows off the street smarts of its self-driving cars
When Google's self-driving cars were first spotted navigating California highways in 2010, that was the easy part. For the past year, Google has focused on a far more treacherous type of road: The city streets.
Google says its self-driving cars have logged thousands of miles on the roads of Mountain View, California since August 2012, learning to drive in the city.
There's just no comparison between highway and city driving when it comes to potential hazards a car might encounter. Railroad crossings, jaywalking pedestrians, double-parked vehicles, construction zones, and cyclists all contribute to potential hazards on city streets.
Google says it still has a long way to go in teaching its cars about city street hazards. But in a blog post and accompanying YouTube video on Monday the search giant showed the self-driving car making some impressive feats of urban navigation.
Google showed its self-driving car identifying hand signals from cyclists and anticipating the cyclists' next moves, such as a left turn in front of the car.
If the autonomous car is trying to make a right turn at an intersection, it can detect pedestrians and cyclists and wait before making a turn. The car's software can even detect cyclists charging to make the light from way behind and wait for the bicycle to fly past before moving.
Beyond cyclists, Google showed its self-driving cars navigating lane changes in construction zones by following orange pylons laid out on the road. It could also detect vehicles parked on the side of the road and then give those stationary vehicles a wide berth.
Finally, Google also showed the autonomous car safely dealing with a railroad crossing. At the crossing, the car would wait until the cars in front of it were clear of the tracks before proceeding.
Google says its cars have logged almost 700,000 autonomous miles since its program began. Google says that "with every passing mile" the company is more optimistic about a future where "a vehicle that operates fully without human intervention."