As a system an elevator is working in a infinitely more controlled environment when compared to a driverless car please consider these problem areas: -Localisa

As a system an elevator is working in a infinitely more controlled environment when compared to a driverless car please consider these problem areas: -Localisation and perception in all matter of conditions e.g. terrain or wheather changes, interaction with traffic, pedestrian behaviour, unexpected or unclassified agents (road works, car stopped in the middle lane of a highway) - Vehicle dynamics sensing, vehicle diagnostics. Any bog standard car will have in excess of 30+ control units (from micro controllers to full on embedded computers). How do you keep tabs on all of them, which one's can you trust, if my LIDAR says I'm good but my RADAR false triggers on a low beam tunnel which one is accurate? Should I slam the brakes or should I drive through what could be a front collision? - Actuation, how smooth is smooth? I can be within the speed limit and stil scare the crap out of you driving above an "adequate" speed for the environment, you can brake with an ABS pump far better than a human can and avoid a collision at all cost, but the G forces involved would probably snap the neck of everyone in the car, what would be the acceptable outcome in that instance? There is a lot of complexity in all of this and realistically the only way you can trust the current levels of driverless is by limiting/controlling the environments they are allowed to be enabled in. Edit: paragraphs