Everyone is wrong about long haul trucks, frankly. It will be easier to get fully autonomous cars than trucks. They're much longer and wider. They're at least twenty times heavier. They require much more room to maneuver, to start, and to stop. If they're involved in a crash, they cause a lot more damage.
Not to mention that there's nothing really predictable about a long route. Traffic, weather, construction, and every other variable is more likely to change over a longer route as well. It would be far better to focus on making autonomous cars first.
All those extra difficulties of long haul relative to cars are big challenges to humans because the scale is so far off from the baseline of our embodied intelligence, but not much of a difference to machines.
A human driver eyeballing a difficult curve will easily be half a meter or more off over the length the trailer. But for a machine it it does not really matter wether it plots a path for a Twizzy through the LIDAR point cloud or for a semi-trailer, as long as the model is accurate. Humans have all their relevant sensors at a single point, awkwardly mapped to vehicle dimensions with mirrors and guesswork, therefore human driving gets worse with increasing vehicle size. A driving machine however can take input from all over its body, its "skilll does not deteriorate with increase in vehicle size (arguably it might even improve because a bigger vehicle will have more computing and a wider range of sensors at the same fraction of total mass and cost). The bigger the rig the easier it becomes for robots to compete with humans.
It's not that trucks will have it first, per se. It's that trucks spend a much larger fraction of their time on freeways, so they can get a much bigger benefit from a system that only works there.
Specifically on freeways, the difference in difficulty is pretty minor. The lanes are wide and you don't need to accelerate very rapidly.
> It would be far better to focus on making autonomous cars first.
Mu. The same system's going to be on both types of vehicle.
Not to mention that there's nothing really predictable about a long route. Traffic, weather, construction, and every other variable is more likely to change over a longer route as well. It would be far better to focus on making autonomous cars first.