Honestly, it seems like a more workable solution would be to improve the roadways with built-in beacons to indicate where a road is, and where it isn't. The cars would work only on the improved infrastructure, but they would work. I suppose there is a danger of nefarious actors spoofing beacons to make you drive off a bridge, but it seems far-fetched.
To me that removes the "self" in self-driving car. A car that stays on the road because there are magic beacons is like bowling with thise giant bumpers in the gutters. It'd be more like cars with uber cruise control.
That would be a train, essentially. Nothing wrong with that - but some of the stakeholders would need to eat crow. That could prove problematic, as some actors have doubled down
on the "full autonomy real soon now™".
it could be more easily done for limited access roads and lanes. like HOV and Express lanes. They are already heavily marked.
Still in the end it requires a combination of changes to how we mark roadways, indicate construction, detours and the like. Think of it like the ADA but for cars.
I would think of it less as an "accommodation" and more as an evolution. The way that roads are marked right now is a direct consequence of the fact that the markings need to be consumed by humans moving at high speed. When highways were for oxen, horses, and pedestrians, they didn't need guardrails, explicit breakdown lanes, and exit ramps with highly reflective signs. In the same way, when the majority of the piloting entities using the road have other input options besides visible light, to support them we can/should take advantage of those other channels.
