- Your car, and where it is
- The road you are on
- How to get to your destination
- Road signs
- The Other cars
- Any hazards that may come your way
- Radio/Cell Phone/iPod ect
But first, a word. Cars in general are not the best long term solution for transport. Today they consume huge quantities of non-renewable fossil fuels, and are the primary source of air pollution in most American cities. Renewable fuels and electricity would help with these, depending on their actual source, and assuming they can cover transfer losses. What I'm getting at is that just from a basic energy perspective, using two tons of metal to transport a 150 lb average person is just not efficient from an energy point of view, but I'm glossing over these issues today.
So as for the actual components of a potentially motorized car, it's not difficult for computers to drive in complex environments--the Mars landers, for instance, work mostly autonomously over very rough terrain. But no system has been developed yet that can work with active traffic. The predictive power simply isn't there yet, and I'm very sure no one will want to be the first. Still, I'm weary of Clarke's law--I'll say that I think that a self driving car in traffic very possible, but not likely to be practical for a long time.
What is more likely, and indeed is already happening, is that computers and sensors are being used to supplement weak spots in one's perception. Lexus cars, among several others, now have an option to parallel park themselves-see this YouTube link. Since the car "knows" exactly where its boundaries are relative to that of the other cars, this is fairly easy to automate. Other cars will now flash a small light in your rearview mirror's and beep at you if there is a car in your blind spot, or when you're backing out of a spot with poor visibility.