Self-driving cars on the roads won't be able to prevent all crashes if they're designed to process information like humans and defer to occupant preferences, the IIHS found in a study released Thursday.
Only about one-third of crashes now would be easily avoided by future autonomous vehicles, the study found. To prevent more crashes, those systems would need to be designed to prioritize safety over speed and convenience.
“Building self-driving cars that drive as well as people do is a big challenge in itself,” the study's lead author and IIHS researcher, Alexandra Mueller, wrote. “But they’d actually need to be better than that to deliver on the promises we’ve all heard.”
The study analyzed more than 5,000 crashes that involved emergency medical services, and required at least one vehicle towed away. Driver error was the final fault in 94% of those crashes. Researchers broke down the chain of events leading up to the crash into five categories that included driver inattention or not recognizing road hazards, misjudgment of other cars' behaviors, poor driving or vehicle control, bad decision-making or unlawful driving, and incapacitation or intoxication.
Of those categories, researchers discovered that autonomous vehicle systems would only prevent crashes related to driver inattention or incapacitation, which accounted for 34% of the crashes studied. The other 66% of those crashes may not be prevented if autonomous systems were designed to prioritize the preferences of its occupants, rather than safety.
The study underscores the difficulties self-driving systems and programmers will face developing driver-less cars and their systems. In extensive testing, Google's self-driving engineers discovered that while its cars were safer drivers than humans, those cars could be still at odds with how to interact with unpredictable human drivers. Google's self-driving cars were hit by human-driven cars multiple times.
Researchers at the IIHS modeled their study assuming an all driver-less fleet, and noted that driver-less cars sharing the roadways with human-driven cars would constrain some advantages that autonomous vehicles would have.
Prioritizing safety in future driver-less cars may have disadvantages as well. Researchers noted that some complex logic by drivers to willingly break laws may be necessary to complete journeys, such as driving around double yellow lines to move around parked vehicles, for example. Changing road conditions could also challenge autonomous-vehicle logic.
Researchers wrote that driver-less cars could result in fewer crashes, but only when programmed with a safety-first preference.
“Our analysis shows that it will be crucial for designers to prioritize safety over rider preferences if autonomous vehicles are to live up to their promise to be safer than human drivers,” Mueller said in a statement.