Researchers at the University of Michigan Transportation Research Institute (UMTRI) have performed a preliminary analysis of the cumulative on-road safety record of self-driving vehicles for three of the ten companies that are currently approved for such vehicle testing in California (Google, Delphi, and Audi).
The analysis compared the safety record of these vehicles with the safety record of all conventional vehicles in the US for 2013 (adjusted for underreporting of crashes that do not involve a fatality). The study, by Brandon Schoettle and Michael Sivak, made four main findings:
The current best estimate is that self-driving vehicles have a higher crash rate per million miles traveled than conventional vehicles, and similar patterns were evident for injuries per million miles traveled and for injuries per crash.
Confidence intervals for self-driving and conventional vehicles overlap (95% confidence intervals). Therefore, the two concluded, they cannot rule out, with a reasonable level of confidence, the possibility that the actual rates for self-driving vehicles are lower than for conventional vehicles.
Self-driving vehicles were not at fault in any crashes in which they were involved.
The overall severity of crash-related injuries involving self-driving vehicles has been lower than for conventional vehicles.
There are caveats with these findings, the authors noted. First, accumulated driving distance is relatively low (about 1.2 million miles, compared with about 3 trillion annual miles in the US by conventional vehicles). Second, self-driving vehicles have been thus far driven only in limited (and generally less demanding) conditions (e.g., avoiding snowy areas). Therefore, their exposure has not yet been representative of the exposure for conventional vehicles.
The team drew its data from the contents of California crash reports combined with other public reporting of self-driving vehicle crashes. The lack of public data regarding the driving experience of most self-driving vehicle test companies and their self-driving vehicle fleets limited the analysis to the three of the ten currently approved testing companies in the report.
Brandon Schoettle, Michael Sivak (2015) “A Preliminary Analysis of Real-World Crashes Involving Self-Driving Vehicles” UMTRI-2015-34/p>