Cars moving take up more space that stationary parked cars. What you've stumbled on is actually a concern for transit planners - more traffic because self driving cars are circling without any passengers.
Well, I think this is probably a case where a bit of number crunching will save the day - considering a world in which driverless cars are the norm and fully embraced by the city, the coordination system can likely determine based on historical data and also current queue request pressure how many cars to launch into traffic at any given time to satisfy demand for a summon in less than 5 minutes or whatever target they want. When not summoned or being sent to meet demand, the cars return to an enclosure outside the city for maintenance.
Admittedly, there is a lot of planning that needs to happen for even a prototype of this to happen, as well as some major cultural shifts, but when it does happen, I think it will run pretty smoothly. The biggest issue in my mind isn't the logistics as much as the cultural attitude that will need to happen - right now everyone seems to assume ownership of driverless cars, but the more reasonable approach seems to be summoned fleets instead. I can envision something like the city owning the control infrastructure and leasing slots to competing companies, or even just competing companies and leaving the city controller out of the picture.
Admittedly, there is a lot of planning that needs to happen for even a prototype of this to happen, as well as some major cultural shifts, but when it does happen, I think it will run pretty smoothly. The biggest issue in my mind isn't the logistics as much as the cultural attitude that will need to happen - right now everyone seems to assume ownership of driverless cars, but the more reasonable approach seems to be summoned fleets instead. I can envision something like the city owning the control infrastructure and leasing slots to competing companies, or even just competing companies and leaving the city controller out of the picture.