I've often wondered why consumer-level self-driving cars are being pushed as a thing when the logical step would be automating service vehicles first - low-risk vehicles that cover the most miles, which can go much slower if working 24/7. Instead of trying to replicate complex pedestrian routes at standard speed, why not handle established delivery routes instead, and optimise over time for new routes?

It's interesting to see how the numerous players in the field, including startups across the globe, are trying to handle this problem, but no one seems to be close. It's a slow race, where everyone is noisy in order to not seem they've been overtaken.