The way I see truck automation going is that the drivers will still be there, but the skill of the job will be dramatically reduced, at least for long haul loads. I'm thinking it will be a lower paying job for drivers that "babysit" automated driving trucks. I don't believe this is optimal. A distracted driver is more dangerous, regardless whether they have a computer helping them drive. But, the incentives are there. I would much rather it looks like the airline industry, where computers have come a long way in assisting pilots, but the job still has stringent requirements of the pilot.
I think the first step will be long-haul trucking across long, technically easy, high-demand routes. Think of a cross country route taking something from San Francisco to Chicago. A driver gets you out of San Francisco, through the valley, and through the Sierras. Once it reaches Reno, the driver exits the truck, then the truck self-drives across the Nevada and Utah deserts until you reach Salt Lake City. A new driver gets you across the Wasatch but then I-80 is easy again all the way from Wyoming to Chicago so it goes back to self-driving. A driver takes over again to navigate Chicago's busy roads. You've cut the driver out of 80%, and no driver is going more than 300 miles in one shot.
One nitpick, the driver will more likely simply drop the trailer off, and pick up a new trailer to go back back home with.
The self driving truck will be valuable enough that you wouldn't want to waste any costly mileage by having it being actually driven.
Another likely scenario, you'll want to design big drop off hubs that are optimized for self driving trucks to successfully park and drop off/pick up trailers.
