The self driving car will be much more energy efficient, especially if they can communicate with each other to optimize traffic flow
Clearing up 1 mile of stop-and-go is not going to make up for the 9 miles from your house to the city streets on your 10-mile commute.
Personally, I'm also pretty sure the net effect of self-driving cars will also be to spread out the suburbs even more too. Most people imagine self-driving cars that are still owned by you and have single-owner use patterns, but the economics of that don't work; in nothing flat owners start renting their car out during their off times, AirBnB style, and it's a short hop from there to simply providing cars as a service. Somebody is going to get very rich being the somebody who wins on the network effects and creates "the" transportation-as-a-service company. It's going to be an enormous change, and I find it unlikely it's going to play to the city's advantages in terms of people actually moving there.
Notice that cars have markedly different highway vs. city mileage ratings. The old BMW I used to drive would get better than 26 (not great by today's standards) on an open freeway, but 17-18 in city traffic due to repeated acceleration and idling. Algorithmically optimized traffic flow would allow cars to spend more time in the highway range.
This is an over-generalization. Yes, city mileage is generally lower than highway, and yes, it is because of stop-and-go. But markedly different? My old car got 32 freeway, 29 city. My current car gets 27 freeway, 25 city. Cars with a large difference, like your BMW, suffer either from poor aerodynamics or engine configuration. (That is to say, the particular engine setup operates very inefficiently when accelerating from a stop) Now, the most damning evidence of all- the Prius gets 40 freeway, 44 city.
Automating drafting is an interesting suggestion, but anecdotal evidence from hypermilers suggests that even when drafting close behind a big-rig, you typically only net a few percent. In addition, to get any kind of measurable improvements from drafting behind a regular car, you'd basically have to ride their bumper. It seems unlikely we will permit computers to try this any time soon.
In the end, my point is not that improved traffic flow is worthless- it is a win in nearly every aspect- but that it is a small, incremental win. It is not going to reduce our energy consumption enough as to obviate our dependence on cheap energy, which ties back in to the earlier point this discussion spawned from:
Suburbs are built on the principle that energy is cheap. That is no longer the case.
We can drastically increase fuel efficiency by making our cars out of carbon fibre, and it doesn't have to be more expensive or less safe. Check out Amory Lovin's concept car for one example.
We believe lighter = less safe, despite evidence to the contrary (in reality it's smaller = less safe, lighter = safer). Perhaps computer controlled cars will allow us to stop feeling the need to fortify our cars more and more, making the road less safe for everyone.
Whoops, you're right, that is a silly mistake on my part.
And that's if you're driving a Mini Cooper.