We are 5 or 10 years or more away from a car you can start up, set the destination, say go and start to read the paper. Just finding the right reflective paint and painting the lines on the lanes and edge of the roads may take that long.
I built one of the first if not the first networks that would carry a J-1939 packet for farm machinery on 2 MHz 68HC11 with 6.5 K of memory while working for one of the original SAE J1939 / ISO 11898 committee members the late Marvin L Stone PhD.
Running a single process takes lots of bandwidth on the bus. If automotive programmers continue to write code as they have in the past there will be hundreds of millions of of lines of code running all those devices with constant ACK’s and NACs. Proving the safety of the hundreds of thousands of logic braces the the code must take to choose what to hit in a Kobayashi Maru situation may be impossible. Every day divers are forced to choose to take the lesser of a group of bad alternatives on very short notice. The next problem is which insurance covers them.
We can probably get the interstate highways to the point they can carry automatons vehicles in 2 or 3 years if the payout is good enough. Sate highways will take from 2 to 10 years for most of them and some will never be ready for trucks due to the “not in my back yard” folks.
And finally can automatons vehicles share the road with human driven vehicles as the number of human driven vehicles gets smaller and smaller or do the irrational human divers get run off the road for being in the wrong place at the wrong time because all the correct positions are occupied by automatons vehicles that run closer and closer together and a human piloted vehicle can’t negotiate new positions in traffic as fast a most of the cars around him.