Back in April this year, Google announced that its cars had driven 700,000 autonomous and accident-free miles. While these cars could probably perform well in an ideal driving environment, the reality is that they have yet to prove they can handle some common real-world driving scenarios.
According to Chris Urmson, director of the Google car team, the car hasn't yet been tested in bad weather conditions like heavy rain and snow, nor has it tackled big, open parking lots or multilevel garages.
While the driverless vehicle can detect and respond to stop signs that aren’t on its map, it wouldn’t know to obey a traffic stoplight that appeared overnight. Also, in case of an unmapped four-way stop, the car might fall back to slow, extra cautious driving to avoid making a mistake.
Among other unsolved problems, the car can’t differentiate between road obstacles like a rock and a crumpled piece of paper, and will try to drive around them. Also, it currently detects pedestrians as moving, column-shaped blurs of pixels, which means that it is impossible for the vehicle to spot a police officer managing traffic.
The car's video camera, which detects the color of a traffic light, gets blinded when the sun is directly behind a light, plus the vehicle also can’t detect potholes or spot an uncovered manhole if it isn’t coned off.
Urmson says these and other issues remain unsolved because engineers haven’t yet gotten to them, but he expects solutions are “going to happen more quickly than many people think”.