Since my original post you released your Safety Report. Kudos to Waymo for putting ego to the side and letting objective information guide your path. There are few companies doing this.
It appears you are on the right path
- Not exaggerating your capabilities like most companies and providing the public a false sense of confidence
- Not using L2+/L3/Handover or Public Shadow Driving as they are unsafe and will never lead to an autonomous vehicle
- Switching to simulation for most AI, engineering and test
I have two remaining questions though
Earlier Waymo mentioned using remote control for vehicles where unexpected issues arose. That is not mentioned in the Safety Report. Did you abandon that approach for the others mentioned in the report? Those being various methods for direct human intervention and the vehicle finding a safe location to stop. While I could see remote control being a last option resort I would want to know if that is being balanced with it being an even less safe form of handover. That being because of the added delay as well as it being harder to pay attention when not in the vehicle. The other concern being if driving remotely does not involve being in a full motion simulator. In some conditions the lack of motion or “feel” cues could be problematic.
The other question I have is if Waymo is availing itself of aerospace level simulation. The reasons for that being they are far ahead of other industries. And the capabilities of simulation products in the AV space are currently not adequate to get to L4/5. The risks of having less than adequate simulation may result in a false level of confidence you don’t find until someone else does in the real world later. And using more public shadow driving that is necessary.