The Washington PostDemocracy Dies in Darkness

Here’s what it would take for self-driving cars to catch on

Placeholder while article actions load

Some day in the near future, cars will drive themselves. Traffic jams and deadly accidents will become obsolete. Morning commutes will evolve into less-stressful affairs, as riders can sit back with their coffee and let computers handle the trip.

That's the dream of many a transportation visionary, at least. And it's not totally implausible. Google is developing "autonomous vehicles" that are improving rapidly each year — Sergey Brin thinks they could hit the road within five years. Nissan and Volvo are planning to add a few self-driving features to their cars by 2020 to help minimize accidents. A future of completely self-driving vehicles doesn't seem too far off.

And yet, as a new report (pdf) from the Eno Center for Transportation details, there are all sorts of obstacles that still need to be overcome before self-driving cars ever take over our highways.

The costs remain high and the technology has encountered some unexpected sticking points. What's more, state and federal regulations, as well as fights over liability and data privacy, could impede widespread adoption of self-driving cars.

"Self-driving cars have the potential to monumentally transform transport as we know it," explained report co-author Daniel Fagnant — and bring billions of dollars worth of benefits. But getting to that point won't be easy.

The case for self-driving cars

It's easy to rattle off various benefits of a world packed with self-driving vehicles. Safety is the biggest. Right now, more than 30,000 people die each year in the United States from automobile crashes. And roughly 40 percent of fatal accidents are caused by alcohol, distraction, drugs or fatigue. Letting robots take the wheel would save lives.

Or take congestion. Cars driven by robots could travel closer together at steadier speeds. They wouldn't bunch up in traffic jams caused by a ripple of brake lights. More cars could squeeze onto the road and move more quickly. The savings in time and fuel would be enormous.

The authors of the Eno report, Daniel Fagnant and Kara Kockelman of the University of Texas, tried to tally up some of these benefits:

If just 10 percent of the vehicles on the road were self-driving cars, the authors estimate, the country could save more than $37 billion a year — fewer deaths, less fuel, more free time. If we reached a point where self-driving vehicles constituted 90 percent of the cars on the road, the benefits would rise to some $447.1 billion a year.

Now, as Fagnant pointed out in a presentation Wednesday, these are "ballpark, rough estimates ... prognostications, really." So don't get too fixated on the numbers. They're meant to be illustrative, not definitive.

The authors also didn't try to quantify any of the costs of self-driving vehicles. What's the price tag on these gadgets? How does that compare to the benefits?

And what about unforeseen consequences? Researchers can't predict how, exactly, self-driving cars might reshape society. Maybe the vehicles will induce even more travel and congestion will get worse. Or maybe they'll lead to a fresh wave of suburban sprawl and increase air pollution. It's impossible to know at this point. Still, the upside is tantalizing.

What's standing in the way of self-driving vehicles?

Even if the benefits are massive, though, self-driving vehicles have a lot of hoops to leap through. First, the technology is still very pricey. By one estimate, the first wave of autonomous vehicles could cost over $100,000 — five times the cost of the average new vehicle. Even that might be an underestimate: The 3-D sensors alone on Google's autonomous car cost about $70,000.

Those costs would presumably come down over time, but no one knows how rapidly. And that's a big hurdle. One survey by JD Power and Associates found that only 20 percent of Americans would "definitely" or "probably" buy a car with self-driving capabilities even if the price dropped to $30,000. (The figure rose to 37 percent when price wasn't mentioned at all.)

What's more, while the technology is making major strides, it's far from perfect. Google's self-driving cars have now traveled more than 435,000 miles in California, but the cars have yet to be fully tested in urban environments.

Most likely, self-driving technology will emerge gradually, piece by piece. Automakers are already installing features like adaptive cruise control or systems that warn drivers when their cars stray out of their lanes. Volvo is developing technology to allow cars to interact with each other over wireless spectrum and alert each other when they inch too close.

But even intermediate features pose unique challenges. As Will Knight reports in MIT Technology Review, early autonomous cars will likely require their human drivers to take the wheel during especially complicated situations. But as BMW is discovering, it's difficult to get people to drift in and out of attention while driving — and the process of switching back and forth between robot and human could well make these cars less safe, at first.

Policy problems with driverless cars

That's just the technology. But as the Eno report details, there are also plenty of legal and policy obstacles.

For starters, self-driving cars will almost certainly have to meet more rigorous standards than regular cars. "The first accident that's caused by a computer malfunction will freak everyone out far beyond the thousands of car accidents caused by humans," explained ENO Director Joshua Schank. That makes regulations and litigation much trickier.

So far, only California and Nevada have passed laws allowing licensing of self-driving vehicles — and standards in those two states are different. If states aren't consistent in what sorts of licensing and safety standard they require, it will be extremely difficult for manufacturers to figure out how to comply with the rules. That's why Fagnant argues that the federal government should get involved here, to set uniform standards.

But even that's not as easy as it sounds. "You still need to understand what the safety requirements actually are before you can determine licensing," said Mary Lynn Tischer of the Federal Highway Administration. No one's even begun to develop a framework for what that might look like.

Lawsuits raise another challenge. Even if self-driving vehicles are safer, they'll likely never be crash-free. What happens if a deer or pedestrian suddenly jumps out in front of a car? Who would be at fault if the self-driving vehicle hits something? The manufacturer? The person behind the wheel? Fagnant notes that self-driving cars are likely to be held to a higher liability standard than regular cars, though if the standard is "grossly higher," that could make the vehicles prohibitively expensive.

There are also all sorts of privacy issues raised by autonomous vehicles. Most of the benefits from self-driving cars come from the vehicles being able to communicate and share data with each other. Similarly, crash data will almost certainly be stored for use by manufacturers and to sort out liability (California's laws require this). But how much data will be stored? And how will it be shared?

Driverless cars could also cause serious disruptions in the labor market — particularly if they start displacing the nation's 240,000 taxi drivers or its 1.6 million truck drivers. "Any time you're looking to replace a human job with an automated job," Fagnant says, "there's going to be opposition."

The Eno report also notes that that there's still a ton of research yet to be done about the prospect of self-driving cars. How secure will the systems be against cybertattacks? What sort of timeline are we talking about for adoption? What sorts of changes might need to be made to our roads or transit systems?

"As long as these and other crucial questions go unanswered," the report concludes, "the nation will be hampered in its ability to successfully plan for and introduce [autonomous vehicles] into the transportation system."

Further reading:

— The weird, counterintuitive science of traffic jams

— Will driverless cars solve energy problems — or create brand new ones?