Gadfly

A Plane Lesson for Driverless Cars

Lighter-touch regulation will hold back progress in autonomous driving.
From
Andreas Arnold/Bloomberg

Something remarkable happened last year: For what may be the first time in the almost seven decades since a jet aircraft first carried commercial passengers, nobody died in a passenger-jet crash.

While recent events, and the two fatalities listed in an annual aviation safety update by the Dutch consultancy to70, showed the vulnerability of turboprop planes, airline passengers travel about 7 trillion kilometers a year virtually without incident on jets.

End of an Era

Deaths of airline passengers hit a multi-decade low in 2017

Data: Airfleets.net; graphic by Bloomberg Gadfly

Note: Statistics include both passenger and cargo fatalities. Four deaths recorded for 2017 were all on a cargo flight.

The airline safety record is one of the chief arguments in favor of a glorious future for self-driving cars. If aviation technology can all but eliminate accidents, surely driving -- with its far greater death toll -- is ripe for similar innovations. Tesla Inc. has even called its driver-assist technology Autopilot, a linguistic nod to aircraft control systems that German regulators seem to consider dangerously misleading.

If only it were so simple. While autonomous driving will one day become a standardized (and life-saving) technology, the road there is likely to be bumpier than its evangelists expect -- just as it was with aviation.

Take complexity. One reason aviation is unusually safe is that aircraft rarely get anywhere near each other. When a plane pierces the imaginary bubble that air-traffic controllers draw around their radar traces, a "loss of separation" incident occurs, typically requiring lengthy reports to regulators. 

That flight through empty space couldn't be more different from the difficult, dynamic environment traversed by cars, where balls rolling into the road, jaywalking animals and impatient fellow-drivers all present hazards that automated systems still struggle to comprehend.

Get Your Motor Running

The vast majority of U.S. passenger traffic is in passenger cars

Data: U.S. Department of Transportation; graphic by Bloomberg Gadfly

As aviophobes nervously awaiting takeoff are well aware, the majority of air accidents happen in the few minutes when planes are negotiating a complex environment. About 59 percent of fatal accidents over the past decade occurred during descent and landing, and 18 percent during takeoff and initial climb, with just 11 percent taking place during cruising flight, according to a 2016 Boeing Co. study. That's despite the fact that aircraft on a standard 90-minute trip spend about 57 percent of their time cruising.

Some of the most successful innovations in road safety are the ones that try to mimic that pattern by simplifying the road environment -- by grade-separating major highways, for instance, or keeping heavy trucks away from passenger cars. But a successful fully autonomous car won't have the luxury of choosing where it travels.

Another secret of aviation's record is that it's among the world's most rule-bound industries. Apart from loss-of-separation reporting, any time an airline or manufacturer detects the smallest defect in a mechanical part, regulators blast out airworthiness directives to ensure that maintenance engineers worldwide are on the alert.

There currently are some 15,000 such directives outstanding at the U.S. Federal Aviation Administration alone, covering everything from the inspection of wing cracks on Boeing 737s to the mechanisms to operate tailplanes on Airbus A330s.

More serious incidents, such as fatal jet crashes, result in wreckage being reconstructed in disused hangars and tests being conducted on metal fragments in the service of vast investigations like the three-year, 221-page report into Air France 447, the Rio de Janeiro-bound flight that crashed in the Atlantic Ocean in 2009.

All this goes to ensure that as much information as possible is shared among manufacturers, airlines and governments to steepen the safety learning curve.

In the U.S. at least, regulation of autonomous cars is moving in the opposite direction. Obama-era plans that would have allowed regulators to approve self-driving systems before they go on sale -- similar to the type certifications common with aircraft -- were dropped last year in favor of a lighter-touch approach.

Perhaps automakers' self-interest will spontaneously cause them to make safer cars under that regime than they would under a rule-bound system. But that's the opposite of the experience in the history of aviation.

Generally, the world makes transportation safer by learning the grim lessons of accidents. Treat that information as the proprietary data of businesses rather than a means to advance the common good, and the path to safer roads will become longer, not shorter.

This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
    Before it's here, it's on the Bloomberg Terminal.
    LEARN MORE