So smart roads? A road that provides relative location and speed of each vehicle for computational efficiency and avoidance. This seems like the most reasonable autonomous driving scheme. Enter destination. Network provides minimal path for vehicles not to smack into each other as the distance between cars is maintained for safety and efficiency. The only safety drawback I can think off is non-vehicle hazards in the roadway.
I realize there are governmental infrastructure requirements, which would stop such an endeavor. It also means less money to be made so less incentive for commercial capital investment.
I have a long car commute, and this thought experiment occasionally occupies the drive. What other considerations am I missing?
You still have the sensor challenge either way. Decision-making could be centralized, but that likely requires <i>everything</i> to follow it - a mix of 50% controlled and 50% human-driven cars would probably be a mess.<p>I think you sort-of get this, in industrial environments with self-driving transports (both small ones in warehouses and truck-sized ones on mining roads etc). There it's relatively easy to ensure everything is known to central control, access control helps with non-vehicle hazards, ... For public roads, this level of integration seems hard to achieve anytime soon, and would also be a larger trust issue that everything actually supplies accurate data and follows commands. Big standardization effort.<p>I think there's also been a few public prototypes of self-parking parking garages, but there similar things apply: only for a limited type of vehicle with the necessary support, humans kept out of the building, ...