You could, you know, just Google it: https://waymo.com/safety/impact/. TLDR ~90% reduction of serious injury compared with human drivers.
Now, before you say this peer-reviewed paper is corporate propaganda, all self-driving companies are required by law to disclose accidents they are involved in, whether liable or not, in CA. You could access each raw accident report published by the CA DMV periodically and come up with your own statistics.
Is it from NHTSA with a NHTSA or an auto industry group diving deep into the software and testing its edge cases? No?
Yes, it is corporate propaganda. "Peer reviewed" doesn't mean anything if it is SPONSORED BY THE COMPANY. There is peer reviewed studies to kingdom come that are industry sponsored and have plagued our society for at least a century.
I AM NOT THE PERSON that should be doing this research. A FEDERAL AGENCY TASKED WITH HIGHWAY SAFETY should be doing it, barring that, the auto insurance industry groups should be doing it. Not a corporate sponsored shill paper.
Now, before you say this peer-reviewed paper is corporate propaganda, all self-driving companies are required by law to disclose accidents they are involved in, whether liable or not, in CA. You could access each raw accident report published by the CA DMV periodically and come up with your own statistics.