You’ll note that even the autonomous cars aren’t quite ready to be given the keys to… themselves… before being allowed to head out on the open road sans driver overlord behind the wheel. If anything, Google’s autonomous driving experiments can be thought of as a kind of cruise control—they aren’t quite at the I, Robot level of automation where one can simply tell a car a destination and be whisked into a high-speed, automatic-travel highway.
Nevertheless, the technology employed by the vehicles does allow them to create a 360-degree view of their surroundings for quickly mapping out any changes that need to be made to a given course. Lasers mounted at various points around the vehicle keep track of road markers as they pass by, what’s directly in front of the vehicle for a fairly long range, and what’s surrounding the vehicle at any particular moment.  Computers quickly analyze the collected data and spit information back out to the car’s driving system, which makes corrections to the car’s trajectory as appropriate.
According to a 2007 article about Junior, an improved version of Stanford University’s Stanley car that won the DARPA Grand Challenge in 2005, the autonomous vehicle can apparently process the various information it receives within 300 milliseconds—an ideal time frame in case someone, say, cuts off the self-navigating vehicle or simply tries to merge into a lane in front of it, amongst other nail-biting scenarios.
