Trains and buses are somewhat "bursty" because... they're trains and buses. Lots of people have to be riding to make it worth the trip. Robocars will work for the single-traveller-at-1AM case. But they also work for the 2-miles-from-the-metro-station-to-the-office case, so they'll make public transport more useful. Robocar owners can optimize rush-hour pricing to emphasize lots of valuable low-speed quick-turnaround trips to and from the dedicated lane in front of the train station. They'll sell a long trip through thick traffic in five outlying suburbs, but it will cost you.

Not every Uber arrives inside five minutes, so I think travelers will be similarly understanding of robocars. That means the fleet doesn't need a giant parking structure anywhere, they just need to keep a few units in every sector (either cruising or parked in surplus spots at the remote edge of store parking), with other units near enough to fill in when demand spikes in a given sector. Lots of existing parking lot owners will make a little more money when the algorithms determine they're optimally located for overnights. It's not like the supermarket needs all its parking spaces outside normal shopping hours. Robocars probably won't typically park on the street, because it's not secure, especially against taggers.

I do expect that robocruising (driving slowly in a parking-limited area while waiting for customers) will be decried as the greatest scourge to afflict red-blooded single-occupancy commuters since the bicycle. I don't expect anyone important will care.