In 1935, two years after his death, Fritz Malcher’s 91-page manifesto was published by Harvard University Press. The Steadyflow Traffic System summed up the late engineer’s ideas for resolving a dirty, dangerous problem: cars and humans trying to share space in the Depression-era American city. Malcher envisioned threading the city with wide boulevards, linked by U-turn ramps and roundabouts, on which a driver would never need to stop.
This would be good for pedestrians, too. “Imagine a city,” Malcher wrote,
... where the street system permits vehicles to move without obstructions, traffic lights or officers with automatic regulation of speed and capacity; where pedestrians can walk continuously through the whole city areas—no matter whether this be in the outskirts or in the center—without any fear and danger of vehicular traffic. … Such a city ideal we can make come true.
Malcher and a number of like-minded contemporaries were half-right: “Steadyflow” roads did come to dominate the American city, whether by billions of dollars spent on elevated and sunken highways, or smaller changes like stoplight synchronization and pedestrian overpasses. Pedestrian fatalities did decline but only because as cars conquered more and more space, people stopped walking.
Today, Malcher’s “city ideal” also sounds a lot like the visions of the American future promised by the pioneers of the autonomous vehicle industry. “Eventually, we’ll be able to turn parking lots back into parks,” Lyft co-founder John Zimmer wrote in September. “We’ll be able to shrink streets, expand sidewalks, and make room for more pedestrians.” Tesla CEO Elon Musk has predicted less congestion and big safety increases. Researchers at the Massachusetts Institute of Technology’s Sensible City lab imagine autonomous cars could shoot through “smart" intersections without stopping.
But those utopian visions ignore the lesson of Malcher: What’s good for cars has rarely been good for people. “My whole career, people have been saying: We wish we could have known the social costs of driving, we would have done this differently,” says Costa Samaras, an assistant professor and civil engineer at Carnegie Mellon University in Pittsburgh, where self-driving Uber cars hit the streets earlier this year. “Policymakers have to think about this now, because the decisions they make affect the landscape for a century.”
Last week, Musk proclaimed that a self-driven Tesla will make a cross-country trip by the end of next year. Most companies plan to have AVs rolling off the line within five years. Goldman Sachs predicts North American auto sales could be almost 60 percent autonomous by 2030, divided between “limited self-driving” cars, which may require driver control during difficult conditions (like encountering highway maintenance) and “full self-driving” cars, which can drive alone in all situations. Carmakers and suppliers say full autonomy is possible within five years, at least in contained areas like corporate headquarters and university campuses.
Overall, this is a good thing. First, autonomous vehicles will be much safer. If three existing automobile AI technologies—forward collision warning, lane departure warning, and blind-spot monitoring—were deployed across all U.S. cars, they would prevent or reduce the severity of more than 1 million accidents every year, including more than 10,000 fatal crashes, according to research out of Carnegie Mellon. And that’s just using the technology we have now. Second, AVs will use space more efficiently than regular cars, accelerating and tailing their peers more efficiently, parking more tightly—and shrinking traffic’s footprint. Third, they will expand the driving population, giving better transportation access to people who can’t drive because of factors like physical disabilities, advanced age, or youth.
But that also means more traffic: If nondrivers, seniors, and people with medical conditions could access automated mobility, Samaras’ research shows, U.S. vehicle miles traveled could increase 14 percent. That would add 295 billion miles of driving each year.
As AVs bring mobility to new populations and change the way we use space, it’s inevitable that urban environments will be transformed, too. But exactly how cities will change depends entirely on one thing: who owns all of these self-driving cars.
There are three options for ownership when it comes to autonomous vehicles. We could continue with our current system, in which people own private cars. We could begin using shared fleets, owned by companies like Google, municipal cab companies, or cities themselves, that operate a bit like taxis, picking up one person at a time. Or—and this is the method preferred by many urban planners—we could turn to shared fleets that also offer shared rides, like Uber Pool, in which you take the backseat with some strangers headed in the same direction.
The Rocky Mountain Institute, a sustainability think tank in Boulder, Colorado, argues that AVs will quickly challenge the private ownership model. In a report released in September, RMI calculates that self-driving cars will make automated taxi service in cities as cheap, per mile, as personal vehicle ownership. Jon Walker, a manager at RMI and co-author of the report, anticipates that autonomous vehicles’ superior use of road space—optimal acceleration and spacing, for example—will unleash a wave of urban transformation. Even if the number of cars on the road doubled, he argues, traffic would still move faster.
Sharing the backseat with strangers could be a crucial factor in keeping traffic from exploding. One OECD study found that shared, autonomous cars in Lisbon—in combination with a good public transit system—could cause peak-hour traffic to fall by two-thirds.
Large numbers of streets could be decommissioned and reused as promenades, parks, and sites for housing. Most downtown parking could also become obsolete. The average car is parked 95 percent of the time, and parking spots are required, at great cost, in housing, retail, and office construction. San Francisco, to take a city not famous for car use, has 250,000 free, on-street parking spaces. Given what land is worth in San Francisco, that’s an unfathomable subsidy for private car ownership and an enormous waste of space.
But let’s say that we can’t get everyone on board with a citywide Uber Pool model. Even if riders use cars one at a time, the shared fleet model has some pretty nice perks. Car-buyers tend to like SUVs in part because they can handle off-road travel or full-family excursions—even if those make up only a small fraction of trips. (Automakers also have better profit margins on bigger vehicles.) But drivers who pick a car by the trip tend to be satisfied with smaller vehicles, as the success of car-sharing programs demonstrates. This is known as “right-sizing” vehicles, and it can cut energy-per-mile by 20 to 40 percent, according to Don Mackenzie, a professor of engineering at the University of Washington. But the real advantage for public space is in behavioral economics. Ownership of an AV will be a big fixed cost, with driving a very small marginal cost, leading people to take more trips. Mobility-as-a-service, depending on the pricing scheme, might have almost no up-front cost but keep taxi-style fares for individual trips. Since consumers sweat costs at the margin, the latter model is likely to push mileage down.
Finally, shared fleets can smooth out congestion by optimizing traffic patterns. “Cars will be routed for a higher-level objective,” says Karl Iagnemma, the founder of Singapore’s automated taxi company Nutonomy. Imagine if Google or Waze sent you on what would be the best route for society, instead of just for you.
Of course, the socially optimized system has rarely been the American way. What if everyone still wants a private car? And wants that car to drive in a way that’s personally, not societally, optimal? It’s true that private ownership can co-exist with shared cars. One such model is Elon Musk’s Tesla Network idea, in which private Tesla owners can lease their cars to a shared fleet to make money when they’re not using them.
Still, many of the promised benefits of our hypothetical AV future depend on sharing. Take parking. “All the space of downtown parking really gets to be opened up in an automated world,” says Samaras. The change is more fundamental than parking garages packing automated cars in like sardines. “You’re decoupling the economics of downtown parking from the location. The constraint is no longer: ‘I need to park closest to where I need to be,’ but ‘I’m going to park where the cost is minimal to me.’” If your car will drive itself, at virtually no cost to you, it can drop you off at the office and then spend your workday in the cheapest place possible. That is the scenario that unfolds in a video Tesla released last week.
That might mean two extra trips through downtown as your car heads out to a cheap parking spot on the edge of town. Or, if your stay is short enough, it might mean your car simply circulates empty through downtown while you get your prescription filled or have a drink. One of government’s key functions would be to create incentives against unmanned vehicle travel for private cars to curb rising traffic.
The most significant source of congestion, though, lies in a lesson 100 years in the making: When driving gets easier, people drive more. Traffic engineers call this “induced demand,” and it explains why freeway-widening projects never solve traffic jams. We’ve tried for a century to build our way out of congestion by adding more lanes. What’s different when we add capacity by changing the vehicles?
Ken Laberteaux, a scientist at the Toyota Research Institute of North America, has predicted the early stages of automation will increase automobile vehicle miles traveled, lengthen commute distances, and accelerate existing trends towards suburbanization of homes and jobs. The initial results of a study TRI is running, he said, suggest that long-distance driving is getting easier before urban driving does.
Pittsburgh Mayor Bill Peduto, whose city hosts Uber’s first self-driving taxis, says, “There are some opportunities for cities on the front end of it to help to steer the industry.” The city recently received a $10.9 million federal grant to invest in stoplights with cameras and sensors that could respond to traffic patterns, buses, or waiting pedestrians automatically.
Already there is talk of giving autonomous vehicles special zones or lanes so that they may function optimally. Madrona Venture Group has proposed converting a lane of Interstate 5 between Seattle and Vancouver, for example, into an autonomous lane. Pittsburgh has considered giving Uber access to busways in exchange for helping run its paratransit program, ACCESS. In order to keep pedestrian fatalities down or the air clean, some cities may ban manually driven vehicles from the city center, suggested Kara Kockelman, a professor of engineering at the University of Texas in Austin. Drivers of older cars—who, in the future as today, tend to be poorer—may be excluded from certain parts of the city (as they already are in Paris) or shunted into traffic jams while AVs race by in another lane.
Transit planner Jarrett Walker, who recently designed the bus network in Houston, believes that vehicle occupancy is the only metric that can really make our use of limited urban space more efficient. Autonomous taxis can only offer marked improvements to city life, he suggested, if door-to-door driving solo is more expensive than driving together on fixed routes.
It is essentially the same problem American cities have had for a century: Driving is too cheap to account for its costs—the deaths, the pollution, the sprawl, the gargantuan investment in roads, and all the wasted time. Driverless cars will solve old problems and create new ones. But why would a nation of driverless car owners be any more likely to confront them than we were?