Hi, it's called V2V in the automotive space. The US Dept. of Transportation has a thing called RITA (https://www.itscosts.its.dot.gov/ITS/benecost.nsf/SummID/SC2...).

As an aside, Federal Transportation guys don't bother invoking terrorism, they just go with "yeah these are two ton things moving around and we are the rule makers."

I think for safety you might get benefit out of cars communicating intent to neighbors. Pervasive tracking and reporting would be unnecessary - but that is the logic, of course various governments could require all sorts of stuff in the name of anything. So yea, good point.

Beyond uses of local comms, N Non-local coordination could improve long-range routing, and even support automated allocation of infrastructure (like the way some roads—especially expensive-to-build-and-maintain ones like tunnels—reconfigure lanes for different directions of travel at different times of day to match traffic patterns, but dynamically and in more places.)

I think non-local comms will end up going through more pedestrian modes.

Road usage configuration is something that external systems will already force self-driving cars to obey, a lot of bridges are 3 or 5 lanes with a middle that switches directions, or have some partitioned third lane group that switches similarly. Broadcasting that information over the web/whatever may help car co-ordination, especially it may help with route planning and traffic analysis, but these systems should be rather strongly isolated from the car itself and be offered from a source open to the general public.

There is a lot of interesting stuff here, especially when you get into traffic balancing, but none of it would need to pass through a car relay network or anything of the sort - there may be economy in doing so, but that sort of a propriety (or open yay) caching of state wouldn't impact or impose requirements on car-to-car communications.