The divided highway could be seen at the test case for autonomous cruise control. They key difference between regular cruise control and that is the steering wheel input, along with needed use of the brakes, without causing the system to shut off (or revert to manual mode) in the case of regular cruise control. But maybe that would be a good design feature.
I think cars need a lot of time using the divided highway as a test bed and then making that aspect of driving commercial available before we even think to solve the problem of driving in city traffic. Next, could come deserted country roads, maybe between checkpoints allowing someone to turn on such a system, and then it would turn off as soon as you approach a town, forcing you to manually drive through that.
In other words, I see a hybrid system for many years, before a fully autonomous vehicle is even possible.
I think the challenge is that while humans makemany kinds of mistakes, all the alternatives are even worse. Reading obscure warning signs or recognizing the significance of something out of the ordinary is a challenge for automated systems. Rolling a monorail train down a dedicated right-of-way is very different from navigating down a crowded road and recognizing that the cyclist on your right is about to take a spill.
@DrQuine, that's definitely an interesting notion. Insurance companies do have a big sway in this -- one way or another.
Two years ago, Bob Joop Goos, chairman of the International Organisation for Road Accident Prevention, was quoted by saying:
"More than 90 percent of road accidents are caused by human error. We, therefore, have to focus on people in our traffic safety programmes."
If that still holds true, getting human drivers out of the equation by asking them to rely on semi-autonomous, or full-autonomous cars, does make sense, and insurance companies shold like it, provided that autonomous cars work well withouth causing accidents.
The tipping point for autonomous cars will occur when automobile insurance offers a discount for travel in an autonomous car. If statistical analysis demonstrates the autonomous car is safer in all conditions than a car with a human driver, the technology has succeeded. I predict there may be some intermediate steps where the autonomous car is rated better than the worst drivers but not as good as the best drivers. This may result in some interesting insurance policies.
Furthermore, I rechecked my numbers. For 2011, NHTSA reported 21,253 deaths and 1.97 million injuries while the Wiki cites 32,367 deaths for the same year. The difference is that the Wiki includes pedestrian deaths. The Wiki number translates to about 2700 deaths per month while there were about 3000 9/11 deaths. So I overstated it a bit in my initial observation but I think the 164,000 monthly injuries makes up for it.
Conclusion that the days of human driven vehicles are very limited remains unchanged.
I think once Level 4 cars come to market at a mass market price then the phasing out of human driven vehicles on public roads will be extremely quick. How could it be otherwise? At present the monthly casualty count on US roads alone is equivalent to the 9/11 attack. Add to that the economic benefits of greatly increased roads capacities and an end to congestion, one can only conclude that the days of human driven vehicles are quite limited.