It is absolute nonsense. It may happen in the next 10-15 years in controlled conditions, but you're dreaming if you think cars will be driverless by 2020 while mixed in with cars that still have drivers.
It is not going to happen. It will cost an absolute fortune to implement, just for people who buy cars for their own exclusive use. For a public taxi system, you'd have to introduce a way to ensure the cars were clean, safe and secure - which would probably mean every car going back to a depot to be serviced. That would be even more expensive, so how can you say it would be cheaper than a cab driver that can work as many hours as s/he wants?
But this has all been said before and you still go on about it, to the point that it's getting tiresome.
Why not just have driverless helicopters as an even quicker way of getting around, or personal planes? I mean, once you've developed the technology then those things are just as possible, right?
We have the technology for a car to read road signs and stick to lanes, as well as spotting fairly obvious obstructions, but just go into London and look around any street (with tourists and cyclists) and imagine how a computer could cope with all of that (clue; you'd be stationary most of the time or jumping like a kangaroo as the computer went mental). However smart a computer will be in 10 or 20 years time, it won't ever be able to manage dealing with such variable conditions - so we'll have to take all cars and bikes off roads, and stop pedestrians being able to come in harms way, to have any chance.
In other words, a system like the pods at Heathrow but on a vastly increased scale - and stupidly expensive when rolled out beyond those controlled routes. And of course, at risk of all sorts of problems when a vehicle breaks down, or someone decides to vandalise the system.
For all that nonsense, you may as well just build loads of tram routes. That would also cost loads, but would be far more sensible (and cheaper than your idea).