Robotic cars were never meant for city driving; the presence of pedestrians (and cyclists) makes it too dicey a proposition. The moment a robotic car even grazes a human, the whole program would be scrapped. The developers and investors themselves would never allow a city-based pilot to go forward for fear of that. Robotic city driving could only be feasible with major infrastructure changes that totally isolate cars from non-car traffic (ie, pedestrians but also bikes, scooters, motorcycles).

I think the real purpose of robotic cars is and always will be the following: to reduce highway congestion. Robotic cars in this environment present several advantages--chief among which is their ability to "chain" together, ie drive extremely close to each other (front to back but also side to side). This not only increases traffic capacity several fold (in effect eliminating traffic jams), but also decreases individual car fuel consumption by ~20% due to drafting. Safety and collision avoidance would be nearly perfect also, as every car would know the precise location, speed and heading of every other car in the "train."