What people are missing here is thinking of the fully functioning automated car springing up fully formed over the course of 5 to 10 years, that’s absurd and of course it won’t happen but it’s also a straw man. Instead, it will be a gradual but steady evolution of incrementally more useful advancements.
First of all, the intelligent car is almost certainly not going to be implemented first in America, instead, America will probably be the last place in the developed world that gets intelligent cars and for many of the reasons pointed out above (if you think that’s unrealistic, the average person in rural china probably has a more advanced cell phone than the average american). It has a well established infrastructure, huge capital requirements for extensive population of existing cars and a population of highly trained drivers. Instead, wher it will probably happen is in a rapidly developing country, probably in South East Asia such as Malaysia for which the cost/benifit ratio skews much higher.
Automated cars are not going to be one giant monolithic technology, instead, it will arise from the intersection of a whole bunch of different technologies that are already coming to market.
Driver support systems:
These are systems which doesn’t aim to replace the driver, but to augment their driving ability. For example, Toyota is currently experimenting with a system where cameras detect the lane markers for the road and then adjust the force feedback on the steering wheel so it feels like you’re driving in a slightly concave bowl which allows you to stay in the centre of the lane easier. Intelligent Cruise Control is already present in Mercedes cars which can keep your car a certain distance away from the car in front of you. Each of these enhancements is incrementally useful for the driver and the immediate benifit is obvious but they also form an economic incentive to build in much of the neccesary infrastructure for truely automatic cars. Right now, they’re using vision to detect lane markers, but theres problems with lighting, worn markers, road construction and so forth, RFID based lane tags would work much better. But you don’t have to RFID an entire nation for this to be useful, you simply make cars that use RFID if it’s present and defaults back to vision or eventually to “dumb cars” on rural dirt roads. The most high volume roads will start adding RFID markers first and then gradually, cities will compete with each other to be “90% RFID instrumented” (which should only require instrumenting the most common 20% of roads which carry 90% of the traffic) and then “97% RFID instrumented” (which would require about half the roads to be instrumented) and then “complete RFID instrumentation”. Voila, you have your magic intelligent roadways which make the automated car problem a whole lot easier to solve.
Car mesh networks:
Car mesh networks allow cars to communicate with each other and allows for a whole bunch of innovative driver support systems. For example, cars could brake in unison and start up again in unison which would greatly increase the efficiency of traffic lights. You could also imagine hybrid-electric cars actually touching each other as they drive and sharing a common power train so that, if you had 10 cars driving in a train, only 5 would have their engines on at full efficiency instead of all 10 and less than peak efficiency. It also allows for real time monitoring of traffic and a whole bunch of other stuff.
Because the benifits of car mesh networks become so much more compelling when every car on the road has it, there’s a strong government incentive to force compliance by saying all new cars had to have at least rudimentary mesh capabilities built in and all old cars had to be retrofitted within the next 5 years. In America, that’s a daunting prospect and one that’s politically infeasible, but if you’re in Malaysia, in 5 years time, you’re going to be earning twice the income you are now, cars with new features have come onto the market and you can ship your old car to be sold in Cambodia, upgrading is a no-brainer.
Smart GPS:
GPS is already here but it has a lot of problems with canyons and signal loss and such. It’s going to be more accurate with the addition of supporting technologies that sprout up to augment it. Ground based GPS stations can improve your accuracy to a couple of feet and help eliminate a lot of urban canyon problems. Instrumented road signs means that you no longer have to rely on GPS to tell you “turn left in 300 feet” when it’s actually 250 feet, just stick a RFID beacon exactly 300 feet from the exit. Car mesh networks means that you can get bona-fide real time traffic information for your entire trip instead of the ersatz stuff we have now and all that academic research on intelligent path planning is going to become obsolete, simply track the GPS of every single taxi in a city and you automatically get the optimal route from any location to any other location.
Smart roads:
Instrumentation is the golden word, every single bit of matter is going to have a wireless tag attached to it which tells the world exactly what it is. Every road sign, every lane marker, even every tree alongside the road. Once you have instrumentation, automated driving becomes a whole lot simpler because the world is working with you to get you to the right location, not against you. Instrumentation is coming anyway and all that’s going to be neccesary is for your car to tap into that. City councils are tagging every tree, construction managers are talking about tagging every brick in new buildings to get real time maintenence information etc etc.
Automated cars are coming, it may seem hard to spot the trends from the outset because right now, nothing coming along really looks like an automated car. There never is going to be any “special, dedicated, high speed toll roads” for atuomated cars only because there’s never going to be a clear distinction between automated and non-automated cars. Instead, everything will steadily blur into a continuum until every car turns automated and we didn’t even notice. All of these different technologies are only going to combine and reinforce each other so that the economic calculus looks inevitable.
An intersection control system could just weave cars through each other.