Multiple times this week, I sat on the floor at the annual meeting of the Transportation Research Board during sessions packed beyond capacity where driverless cars were up for discussion. In one session, where Google's Chris Urmson showed images of how our streets appear through the eyes of machines, the curious spilled out of the doorway. In another — "Is transportation infrastructure ready for driverless cars?" — the crowd filled a sizable ballroom.
These people are not debating whether autonomous vehicles will come to be, but how it will happen and what it will mean. Some of the most interesting questions, after all, are not about the technology itself, but about the behavioral, legal, economic, environmental and policy implications around it. Compared to all of these other knotty issues still to be worked out, the fundamental engineering problem — can we really build these things? — seems like the easy one.
Here are some of the mind-benders occupying the people trying to plan for this future:
Will consumers want to buy driverless cars?
The potential benefits of driverless cars are many. The Big One: There are about 30,000 traffic fatalities in the U.S. each year, and research suggests that as many as 95 percent of them are attributable to human error (drunk driving, distraction, slow reaction times). Take the human out of the equation, and fatalities could plummet. Most engineers and policymakers are expecting that computers will be better drivers than we are.
Better (autonomous) drivers would also mean that driverless cars could reduce congestion. There would be no rubber-necking effect on highways. And an individual road could accommodate more cars, all of them moving through space more efficiently. If we shared autonomous cars among us, we could also dramatically cut down on the amount of scarce urban land needed for parking.
Now, here is the economic conundrum: Many of the biggest benefits of driverless cars will be shared by society, but the costs of the technology will be borne by individual consumers. That means that we have to find a way to better align the costs and benefits of driverless cars for the people who might buy them. That may mean offering subsidies. Or it may mean building into these cars clear value for individual consumers from the moment they drive off the car lot. It's easier to value, for instance, great fuel economy (which saves me money now) than great safety (which produces the absence of future accidents).
"We need to make sure the individual is really seeing a lot of benefit," says Peter F. Sweatman, the director of the University of Michigan Transportation Research Institute. "If it's a purely societal benefit, it ain't gonna happen. This will only happen by being highly attractive to consumers."
How will humans behave behind the (self-driving) wheel?
What if your autonomous car doesn't drive like you do? Say you're one of those people who drives reallllly slowly on the highway. How will you react when you're sitting in the front seat of a vehicle that drives faster than you're used to? Or vice versa. Will people be tempted to take control from these vehicles? Can we learn to trust them? If a vehicle requires you to suddenly take wheel for some reason, will you be able to quickly turn your attention away from whatever you were doing while your car was doing the driving for you?
Monali Shah, an official with HERE (a Nokia company working on mapping technologies for autonomous cars), says engineers are trying to create different driving profiles that you might select for your vehicle. "It has to feel natural," Shah said during one panel. "In order for people to adopt it, people have to be able to drive the way they would drive." That means learning a lot about how people would behave in driverless cars, as well as how to make driverless cars behave more like us.
A related question: How will people still driving old-school cars behave around autonomous vehicles when we still have a mix of the two on the road? It's plausible that some human drivers may behave more recklessly around autonomous cars — weaving or speeding around them — because they expect autonomous cars to correct for their behavior.
Who will be liable when a driverless car crashes?
The vast majority of collisions today are the fault of one driver or the other, or the two in some shared responsibility. One of the primary jobs of insurance companies is to apportion this liability. Few crashes are deemed the responsibility of the vehicle itself — or whoever manufactured it. But that will change. "If a vehicle drives itself," says James Anderson, a behavioral scientist at the RAND Corporation, "it’s going to be a lot harder to conventionally blame one driver or the other."
Liability will inevitably rise for automakers, which is one reason they might be reluctant to see a broad and speedy rollout of autonomous cars. Google has been looking to partner with automakers on its autonomous technology. If a car designed by multiple companies crashes, do we blame the people who made the hardware? Or the software? Or the mapping platform? Or maybe we blame another car that sent a faulty signal on the highway.
How will driverless cars change our travel and consumption patterns?
Here are just two competing theories: Autonomous cars will reduce car ownership, because we'll simply be able to order them when we need them, and they'll come to function as shared assets akin to public transit. Alternatively: Autonomous cars will increase car ownership because, as the utility of each vehicles rises (now you can send your 7-year-old to ballet alone!), people will want to own even more of them. These hypotheses are equally plausible and mutually exclusive.
So are a lot of other theories around how autonomous cars will change our travel behavior. If they make travel easier, perhaps autonomous cars will induce new trips that we aren't making today. You'll never have to say to yourself on a Friday night, "I think I'll stay home because I can't find parking/don't want to deal with traffic/don't want to drive home drunk." As a result, the number of trips and the number of miles we collectively travel could increase.
Or, maybe, autonomous cars will create new efficiencies, enabling better carpooling, less idling in traffic, and smarter route-planning. Computers won't waste gas getting lost or circling for parking spots. And, as a result, total miles traveled and greenhouse gases will decline.
One more! Autonomous cars will make it easier to live in a dense city without a car, rendering urban living more attractive. Or: Autonomous cars will turn your 20-mile commute into a productive joyride. As a result, they'll make it easier to live in the exurbs, spurring the next explosion of sprawl.
Is our infrastructure ready?
To function at their best, autonomous cars will communicate with each other ("you go ahead and change lanes first") as well as with the infrastructure around them ("I am ready for a green light"). That means that we have an awful lot of work to do to prepare our roads and communities for these things, even after we're done designing the cars themselves.
These infrastructure questions are both high- and low-tech. What kind of lighting do we need on city streets if we're trying to optimize for radar vision instead of human sight? Can a computer process a street sign that's covered in graffiti? Will automakers want to make autonomous cars if only a few places in the country are ready for them? And if we need to invest in radically changing our roadways — networking streetlights, installing sensors — how will we pay for that?
These questions about physical infrastructure don't even scratch the surface of the legal and policy infrastructure that must come before any widespread adoption of driverless cars.
Bonus question: Do we even want this future?
This is a pretty big question, worthy of its own discussion. So more on this later...