Network of robotaxis has a wonderful quality of increasing utilization with density.
It's a self-healing system: the more people use it, the more efficiently we can use it. Efficiency, it this context, is number of people we can transport per hour on a given stretch of the road.
It's important, so i'll emphasize it: high utilization of roads is a good thing. Roads are a fixed resource that have already been paid for. High utilization means we're providing more transportation value for the same, fixed price.
Most of the time roads are empty. Case in point: it's 1 PM in a busy part of San Diego. I just looked out of the window and the road is empty.
Even if there were 10 more, single occupancy cars on that road it would not degrade transportation service for anyone. Plenty of capacity.
The bad part is congestion caused by going over capacity. That's the friday night scenario in my neighborhood or commute to/from work traffic during weekdays.
Robotaxis give us obvious tools to combat that.
When the need for transportation at a given time is so hight that it would cause bumper-to-bumper traffic, it implies that a lot of people are sharing their route with other people.
It is the case today but we have no practical way of arranging for those people to share the car.
With robotaxis, the fleet can see that 4 people on the same block want to go to roughly the same place so it can direct them to the same car, so that they can more efficiently utilize fixed resource of cars and roads.
What if they don't want to share a car with other people?
Make them with congestion pricing.
Make them pay $10 for a single person ride vs $3 for shared. If $10 doesn't work, then jack it up to $20, $50 or $100. There is a price that will make 90% choose shared ride during congestion.
Or if things are really that bad and the trip is non-essential, they'll choose to shift their trip. Maybe 5:30 PM is not the best time for trip to walmart.
We can certainly do better than what we do today during congestion: mostly empty, overly big cars.