The model:
Take the Renault Twizy (the car above). It's a two-seat tandem though primarily single-seat vehicle (see here). It's an electric commuter car with a top speed of about 80 km/h.
The Twizy can exist as a hire car, which is collected and dropped-off at or very near shopping centres, where it automatically recharges. It also has a 10 kw range-extender so there's no concern with battery range.
It only operates in driverless mode to redistribute itself to where the demand is. When people actively use it, they use it like a traditional automatic car, though the Level-2 driverless option will be provided (which already exists in some cars today).
-Level-2 driverless operation means you can have your car drive itself, though the driver must keep their hands in continuous contact with the steering wheel for security.
So, after say a 5 minute walk to the local shops, you collect your Twizy car and actively drive to the stop-point closest to your destination, like a bus stop.
The Twizy cars only redistribute themselves along designated routes, just like a bus. We can have painted lines on the road that they follow, for precise road positioning.
The Twizy cars can also have blue pulsing (not flashing) LED lights in the top four corners of the vehicle, to warn everyone that the car is in driverless mode, in case they have any fear that the car might fail to respond to them. Add to this, the cars could have a small horn embedded in each corner of the vehicle, so when an approaching pedestrian is detected the car can sound a small "beep, beep" out of the appropriate horn, making the cars virtually unmissable to a daydreaming pedestrian.
Further to this, they can run close to the centreline of the road, like a bus or truck does. The effect is that the cars further become virtually unmissable and, due to their narrowness, inherit considerable extra room for collision-avoidance in a worse case scenario.
Hence, just like with a train or bus, people can choose to not mindlessly walk in front of the car while in driverless mode (though it would stop anyway).
Service:
This service would replace buses for a lot of peak demand and most off-peak demand. It would also cut aggressively into private car demand.
It would be cheap to use and remove parking costs. You would need to tolerate a 5 minute walk to collect your Twizy (a pleasure for many) though it would otherwise be convenient, especially where parking is otherwise a challenge.
Safety:
When in driverless mode, these highly visible and very light cars, operating on designated routes at speeds usually not exceeding 50 km/h, would be perfectly safe compared to other forms of travel.
For example, buses in Wellington, New Zealand (where I live) drive insanely close to all surrounding traffic on diabolically narrow roads, and the buses are a constant hazard to other traffic which must continuously accommodate them. Yet there are virtually no accidents involving physical harm...
The driverless Twizy system that I have suggested is far safer than the current Wellington bus system, for both pedestrians and traffic. They have much more room on the road, can stop rapidly, turn instantly (to avoid), are highly visible and weight a mere quarter of an average car.
Expandability:
My model is a cautious yet highly practical first-step for implementing a driverless system, as a public transport service. As the technology matures it would expand to larger vehicles that can eventually move to a complete robo-taxi system, over time.
We don't need a "courageous revolution". We just need a comfortable starting point.
Public advantage:
It's a great boon for poorer people. Many will not need to own a car at all with this system. It reduces costs overall, quietens the streets, and gives back a huge amount of parking space.
In my view, the New Zealand government should work with Waymo or GM to develop a system like this, and maybe even work with busing companies as well.
The safety concerns are benign and the economic and social advantages are great.
Extended article (the transport end-game):
Public Transport and driverless cars: How they will play.
A device like this would use the same technology already established on the most basic smartphones today. It would probably cost no more than about $20 a unit for every car. That's peanuts relative to the material advantages of the device.
This is how it would work: A driverless car, once within 300 meters of your car, would signal to the device in your car to signal in return your cars coordinates - as it would. The electromagnetic frequency for communications would be set low so it can easily penetrate common objects, including very bad weather.
The result, is that the driverless car can "see" long range very accurately, no matter the weather, and can even see around corners and in-between cars. Hence, it simply expands the driverless cars map as though it's linking to a drone that 'sees all' from above.
It's important to note that computers and sensors are brilliant for measuring and reacting to objects that are close. Far better than humans in fact. The weakness (comparative weakness, that is) for computer sensors is seeing long distance, and especially in bad weather. However, long distance vision naturally does not need to be overly precise to the end of safety, so a system like this could be installed easily with perfectly good reliability.
Installing this device in all cars could also be valuable for safety in general. Imagine having an app on your phone that warns you that there's a car driving towards you recklessly, on the wrong side of the road, and you can't even physically see it coming yet. The technology on your mobile can essentially do this right now.
Addition 2:
Major car manufacturers will be doing their best to suppress the implementation of robo-taxi's.
