Years of Testing Needed to Prove Safety of Autonomous Vehicles

Self-driving cars may be safer than those driven by humans, but it will take years to prove it, according to the Santa Monica, Calif.-based Rand Corp.

It’s a simple matter of statistics, says Rand senior scientist Nidhi Kaira. She points out the traffic fatality rate in the U.S. is about one death per 100 million miles driven. To prove with 95% confidence that autonomous vehicles are at least as safe, she says, they would need to accumulate some 275 million miles of travel without a fatality.

Kaira figures it would take a fleet of 100 automatic cars operating nonstop for 12 years to cover that many miles. But 10,000 self-driving cars could do the same in about six weeks.

So far, there simply aren’t enough driverless cars in service to come close to that total anytime soon. Google Inc., which operates the world’s largest autonomous fleet, says it has accumulated a mere 1.3 million miles of driverless testing over the past several years.

Kaira cautions that aiming for perfection—or standards that demand it for robotic vehicles—could slow or even prevent autonomous-driving technologies from being deployed at all.

She opines that regulators might instead consider a “not quite there” starting point. Such rules would allow carmakers to deploy self-driving systems that are approximately as safe, but not necessarily safer, than the average human vehicle operator. Doing so would help accelerate the development and testing to determine just how safe self-driving cars could become.