1) Replacing taxi drivers.
2) Replacing commuters who don't want to drive.
Both of those problems exist on the scale of hundreds of thousands / millions of people. Solving them with good margins is a business model.
Good assistive systems (probably leading in the nearish term to a level of full autonomy under some conditions) has already started to arrive from a variety of auto makers--but Google doesn't seem to be interested in participating there. So there's a definite business model for assistive/autopilot systems and there will be, some day, for fully autonomous systems. However, it's unclear whether the latter, taken by itself, will be profitable in any reasonable commercial timeframe.
1) The human driver. 2) The fuel. 3) Maintenance.
Even at a minimum wage, what really breaks bus routes is the human driver. You could send a smaller (more fuel efficient for low loads) vehicle on less profitable routes, but that doesn't modulate the biggest cost at all.
Google, Ubur, Lyft, all seem to see a future where (to borrow from the older Total Recall movie) Johnny Cab is able to eliminate a lot of public transit and ride-sharing issues.
I believe that the technical challenge is within our grasp (it'll be difficult but possible).
It might, with sufficient volume, be cost effective and when factoring in the difference in accident rates surely will be.
The hardest sell will be the truly glacial speed of modifying laws to get these vehicles on the road without drivers.
I disagree with this. Yes, autonomous vehicles will be held to a very high standard--both in terms of statistical safety and in not doing "dumb things." A higher standard than humans almost certainly. But that is not an unreasonable bar to set.
When vehicles can indeed go door-to-door reliably across a wide spectrum of roads and road conditions, I don't really expect laws to be the limiting factor. But I expect that to be pretty far out.
If you can replace a driver you are paying ~$20,800 [$10/hr] every year, that is ~$100k of savings [I'm assuming the more complex car also costs ~$100k to service over 10 years, so it may be more than that if I'm being overly conservative].
https://www.washingtonpost.com/news/innovations/wp/2015/12/0...
Right now, its ~$100k to produce a self driving car. If they can get it down to $50k and maintenance costs of ~$5k/year ... it is much, much cheaper than any human driver.
I actually think that your human driver costs are quite conservative. And I expect that the incremental costs associated with autonomy--once the technology exists--will be quite modest.
IOW: people tend to acquire multiple vehicles out of necessity, not poor planning.
To me, at least, that seems like a use case where you'd need drones.