Having been involved in the development of autonomous vehicles for several years now, I can confirm that replicating the driving capabilities of a human driver is a very hard problem. To me the aspiration of full Level 5 autonomy is akin to an effort to produce a humanoid as a versatile aid for a variety of tasks. Technically challenging and interesting as both cases are, the approach is questionable.

That being said, it makes more sense to me to evolve autonomous driving capability from Level 4 autonomy, starting with very, very constraint operating design domains, while building an infrastructure that facilitates autonomous driving, such as dedicated lanes and other technologies. Over time, cities may evolve in conjunction with the development of autonomous vehicles.

However, the automotive industry must not dominate the design of cities, like it happened in the 50ies. Rather it would be preferable to question personal mobility and the ways it can be delivered, including avoidance, walking, cycling and most importantly, shared forms of mobility such as public transport. More on that on https://medium.com/@peter.wurmsdobler/adaptive-frequency-public-minibus-service-for-towns-and-small-cities-75cfd36e274e