The "study" is you point at is not a study. It's an article which points to a simulation ran with perfect conditions for a self driving car demonstrating how they perform in _perfect_ conditions (key on that last part, it's like statistics but more blatantly fabricated).
The main point of that article is safety, which I agree with. Come to the SF Bay area and follow one of the Google cars around. Interesting that they don't do any better going from San Jose to San Francisco as anyone else, and in many cases do worse. The self driving cars would help if the issue was drivers being boneheads and having accidents. Reality is quite different, which is that having nearly 9 million people in a very small area all sharing the same 3 Freeways will have massive traffic jams. 10% autonomous cars won't get you off an exit ramp with thousands of cars waiting to exit any faster. That is what causes long backups on Freeways most of the time.
Turning cars into mass transit.. that would help congestion.