Study: Self-driving cars less likely to crash
A Virginia Tech analysis shows self-driving cars have lower accident rates than regular vehicles.
According to the newly released study, self-driving cars are involved in fewer crashes than vehicles which are steered by a human driver. The research project was conducted by the Virginia Tech Transportation Institute and commissioned by Google.
In the meta study, the researchers aggregated national crash data and information from real-world driving studies. This they then compared to data gathered by Google while test-driving its autonomous fleet of more than 50 cars.
After adjusting for severity and unreported incidents, the analysis concluded that cars driven by humans were involved in an average of 4.2 crashes per million miles covered, whereas autonomous vehicles were in 3.2 accidents per million miles.
The researchers, however, recommended not to jump to conclusions too early: “The limited exposure of the Self-Driving Car project to real-world driving increases statistical uncertainty in its crash rate”, they say. “That uncertainty will decrease as it receives more on-road, in-traffic testing."