First I want to make it very clear that I would like to see this tech available ASAP. However the use of public shadow driving for primary development will never result in anything close to L4. It is impossible to stumble and restumble on all the scenarios needed and the cost of trying is over $300B. In addition to that there are massive safety issues around handover and many more lives will be lost needlessly as the scenarios move toward accident scenarios. If the approach is untenable and harms people for no reason you will not win the public’s trust, especially when the first child or family is killed for no reason. The answer is to flip the public/simulation paradigm to 99.9% simulation and move to aerospace/DoD/FAA simulation technology, engineering and safety practices. Not the IT/gaming based simulation products currently being used or the Agile approach.
Please find more detail here
The Crash of the Autonomous Vehicle Industry