None of the systems listed are very complex. One or two lines, or a loop of track, with very few level-grade crossings. Most of the ones in the US are just airport shuttles. Sure, that's easy to do.
I wouldn't move the subway systems in New York City to UTO operation, though. Not a chance.
It's already done. The docklands light railway in London is more complicated than the metro systems of most cities and it's been driverless since the 1980's
http://en.wikipedia.org/wiki/Docklands_Light_Railway#Map.
On the other hand, it still has a staff member, 'Passsenger Service Agent', on every train to deal with customer service. They are also trained to drive the train at very low speed (I'm guessing <10mph) when the automation fails, which it does occasionally.
"Passenger Service Agents" are on some, not all trains. And on about a third of the trains they are on, they are manually operating the train at full speed (generally for "diagnostic" (read as sanctioned joyriding) purposes, I am informed by a friend at TFL).
What's the difference between an automated train failing to stop at the end of a line in a system with 200 stations to an automated train failing to stop at the end of a line in a system with two stations?
Of course in the rest of the system if can be incredibly complex to ensure safety, but that's not where the accident you're relying on to make your case occurred.
I wouldn't move the subway systems in New York City to UTO operation, though. Not a chance.