GM has been testing it’s autonomous Bolt EV, but I’m having difficulties finding any data like the number of miles driven, and the rate of accidents compared to usual drivers, or other AV systems in testing. Any sources out there?
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0184952 This has a decent amount of info. Might have what you’re looking for. Although it is only through early-to-mid 2017 I believe. “Overall, accident frequencies computed for all manufacturers showed that conventional vehicles drive one order of magnitude more miles compared to AVs before encountering an accident, with a mean mileage before a crash for conventional vehicles of about 500,000 miles, compared to 42,017 miles for AVs. Detection and disengagement issues were also analyzed, indicating that the AT technology suffers from the same “deficit” human drivers have in its limitation for detecting and reacting to rear-end type of collisions.“
There’s a lot of good data presented in that, but I think they make a false comparison of conventional vehicles accidents per mile to AVs. They are comparing all miles driven everywhere for conventional cars. Versus where AVs are being tested and driven and California only reporting, which appears to be mostly highly challenging city environments. Additionally, they are reporting events with no damage and minor damage. What percent of vehicle incidents nationwide go unreported, because there is no damage or only minor damage and people don’t want to get the police or insurance involved?
Unreported crashes appear to be estimated at anywhere from 30% to 80% based on different studies. https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812183
Yes, certainly a good point. A car crash that results in a minor ding or scratch probably means a driver isn't gonna bother getting the police or even insurance involved. But by law companies with autonomous vehicles on the road in California have to report every single minor accident. Unfortunately, since these two things are both true, it's hard to get a real accurate comparison until one of them changes. Most likely the reporting requirement would change. But since other states are now crafting more relaxed laws for testing, we may start to get a better idea of how often significant accidents actually occur and in different regions. But the data in the report is a better compilation than I had seen elsewhere.
