The primary objective for self-driving cars, at least the ones pushed by proponents of the technology, isn’t the convenience of the driver or passenger. That is just one of the goals or even a by-product of what these autonomous vehicles bring to the table. After all, one can only sit back and relax and take one’s hands off the wheel and eyes of the road when you’re 200% sure the self-driving car can safely drive itself through any road condition or obstacle. It turns out, however, that our current autonomous driving technologies can only avoid 34% of crashes that human drivers caused.
To be clear, the study conducted by Insurance Institute for Highway Safety’s Highway Loss Data Institute (IIHS-HLDI) doesn’t exactly examine self-driving technologies directly. Instead, they looked at and categorized the causes of more than 5,000 police-reported car crashes due to driver-related factors. Comparing that with what current self-driving technologies are capable of yielded some rather disheartening conclusions.
These self-driving cars won’t have problems with distractions, impeded visibility, and even impairment due to alcohol or drugs, of course. These perception and incapacitation problems, however, only accounted for 24% and 10% of those 5,000 plus crashes, respectively. And while autonomous vehicles can easily avoid these problems, that also requires them to be 100% functional and error-free.
More problematic, however, are the other causes of those crashes. 40% of those were due to decision-making errors, particularly those that involved speeding or illegal turns, deliberate actions by the drivers that led to unfortunate consequences. And then there are accidents related to drivers erroneously predicting what a pedestrian or another vehicle would do as well as more mechanical performance problems like overcompensation or incorrect evasion.
Based on accidents involving self-driving cars, systems are still unable to sufficiently address those errors. The IIHS-HLDI concluded that it won’t be enough for autonomous vehicles to have perfect or even near-perfect perception. They also need to be better drivers and better decision-makers than humans if we’re to ever live with fully self-driving vehicles.