Searching for the City in the Self-Driving Car
Why people who care about cities should be making friends with automotive engineers.
Popular media is filled with speculations about how self-driving cars will change urban life. Too many of these accounts begin from the wrong premise. They assume that the self-driving car is a fully formed thing.
What if, instead, we treat autonomous vehicles as an emerging technology that people in cities and neighborhoods help to create? What happens when we conceive of autonomous driving as both an act of engineering and of building relationships? These questions were inspired by a recent gathering I attended in Washington, DC where leaders in autonomous vehicles from General Motors, Nissan, and Toyota shared their hopes and frustrations.
Co-creation is not as weird as it might at first seem. Imagine any street or highway or even rail station or airport. Transportation is an experience that people create together. To engage in co-creation with autonomous vehicles, however, first requires finding the city in the self-driving car. Here’s what I mean:
I’ve spent years listening to transportation engineers, city planners, and real estate developers. Those of us who work at the intersection of transportation and land use learn to speak engineer, planner, and developer because, from our perspective, transportation and the development of land are so inextricably linked that each is always present in the other.
To transportation professionals, the city is an abstraction, hidden in technical terms like functional classification or access management. At its most abstract, transportation requires a trip, which in turn requires an origin and a destination. Land uses — houses, offices, shopping centers, and churches, etc. — generate trips, which are catalogued and tallied in fat trip generation manuals and their companions, parking generation manuals.
Transportation is a type of land use, of course: sidewalks, roads, and rails all sit on land. Transportation is also implied in the question, “can we get there from here?” Land use professionals, however, usually invert this question to “how many people can get here from there,” where “there” is a circle and “here” is a dot at its center.
In addition, transportation can be found in the design of the specific places where one type of land use connects to another. Think doors and gates. Driveways and alleys. Bus stops and rail stations. Think also garbage cans and dumpsters. Loading docks and delivery entrances. Mail boxes and that space between the storm door and the front door where UPS leaves packages.
Land use must be in self-driving cars, but where? My two big takeaways from the event in Washington:
“Operational design domain” is the term automotive engineers use for land use, neighborhoods, and the city.
When it comes to operational design domains — understanding them and interacting with them — the self-driving car is not yet close to being fully formed. There are still many choices to be made.
Two additional modifiers are important: “limited” and “low speed.” Through defining and separating out “limited” operational design domains, vehicles can leap to full autonomy in specific places without being expected to drive perfectly everywhere and all the time.
The definition of a limited domain is why automated driving already exists for high-speed expressways (interstates or motorways). Without the need to stop at intersections or anticipate turning vehicles, the same simplified and standardized environment that makes driving safely at high speeds possible for humans, it turns out, also works pretty well for robots.
Typical urban driving, on the other hand, still flummoxes self-driving cars. Ranked by degree of difficulty for autonomous vehicles are office parks, residential neighborhoods, and finally, downtown-style shopping, entertainment, and business districts. What these places have in common is an increasing complexity, which human drivers navigate safely by driving slowly.
This insight leads into a very different way to limit an operational design domain: define it as “low speed.” And, again, robots are not that different from humans. Low speeds provide more time to react to surroundings and less risk of harm if something goes awry. In addition, operating at low speeds makes remote control feasible when necessary. The automotive industry’s interest in low-speed domains coincides with a growing consensus among urban planners that streets with speeds as low as even 30 miles per hour are inhospitable to anything other than the movement of motor vehicles.
Defining operational design domains — defining the city — is an unavoidable requirement of autonomous driving. The National Society of Professional Engineers, in the context of California’s regulations for autonomous vehicles (PDF), warns that such definitions should not be “overly broad.” Definitions should, instead, “reflect the myriad of sub-domains a vehicle may be required to enter and exit.”
In this spirit, I offer a definition of the Kilmarock operational design domain, also known as my neighborhood. Although speed may be a number to a traffic engineer, to the people in my neighborhood, “low speed” is a co-created experience that moves to the rhythms of the day, the week, and the seasons.
I live in a residential neighborhood in what today’s urban planners call an inner-ring suburb. It’s a couple of miles from a railroad suburb founded in the 1880s, which was later served by streetcars. The streetcars, however, never came out quite this far. Although motor vehicles quickly became assumed, my neighborhood originated during an era when residential streets accommodated cars, but refused to be in awe of their size and speed.
The neighborhood nestles into the side of a line of hills, with a creek in the valley and a busy avenue along the ridge. Most of the residences date to the 1920s, 1930s, and 1940s. Set back from the street and surrounded by lawns and trees, they look like substantial, single-family houses. Some of them are. Many of the “houses,” however, are actually small apartment buildings in disguise. What looks like one house may contain four, five, or six small apartments.
The residents of each house or apartment usually have at least one car. Driveways range from grass with two tracks of gravel to concrete wide enough for two cars. Most of the apartment buildings, however, have little or no off-street parking. People park their cars on the street.
People also walk. They walk uphill to the avenue to catch the buses that connect to the region. They walk downhill to enjoy the creek park and to cross the wooden bridge that leads to the elementary school. They walk to the small university at one end of the neighborhood and to the stores and restaurants at the other.
The street network favors the pedestrian. A grid of short blocks and frequent intersections offers direct connections and varied routes. The occasional curving street cuts a diagonal along the slopes, creating intersections with acute angles. The narrow streets come in two varieties: streets with two travel lanes and one parking lane and streets with two parking lanes and one travel lane.
The sidewalks, on the other hand, appear to follow a design convention lost to time. Within three blocks of my home there are streets where asphalt meets grass; streets with a narrow sidewalk on only one side; and streets with curbs on both sides but no sidewalks. No streets have sidewalks on both sides. Many of the walkways that lead away from the apartment buildings connect directly to the street. These slim walkways become, in effect, pedestrian driveways.
Even where there is a sidewalk, to pass someone requires that one party step off the sidewalk into the street. Because of the parked cars, however, this maneuver often leads to someone angling around the vehicles into the middle of the street.
The haphazard sidewalk network encourages a walking culture that treats taking the sidewalk as recommended, but not required. Pedestrians follow their own maps for where to follow the sidewalk, where to walk to the side of the street, and where to walk down the middle of the street. When passing, they determine who walks where through body language and the conventions embedded in the lattice of social relationships.
The word “pedestrian” is itself a transportation abstraction. Some pedestrians are strolling. Some are running for the bus. Some are dressed for exercise. Some are dressed for church. A mother leads a walking school bus on weekday mornings. Neighbors stand chatting, and toddlers practice the art of one foot in front of the other.
Many pedestrians aren’t even human. My neighborhood is a haven for people walking dogs and dogs walking people. Over time we learn what to expect of anti-social huskies, overly protective border collies, and deaf Chihuahuas. Whether I’m walking or driving, I go on high alert whenever I see the older woman who struggles with her labradoodle, as goofy as he is large.
My neighbors manage the use of the right-of-way through reciprocity. People driving wait for people walking to cross a street, while people walking move aside to allow people driving to pass.
Near the creek park, the streets become even more ambiguous. Without any markings, the asphalt path that runs along the creek merges for a block into a street without curbs, before gently splitting to head back into the woods. Has the street entered the park or is the path skirting the street? Which perspective is correct?
This block of street/path is also flat and straight and without driveways or parked cars. These conditions would normally invite a car to speed up, but people seem to know — or learn — better.
This lengthy description still simplifies transportation in my little neighborhood. I’ve left out the people bicycling and learning to ride. The wheeled poly carts and blue bins waiting for the garbage and recycling trucks, whose workers leave them tipped and askew in their wake. The streetlights doing battle with encroaching winter darkness. The sheets of draining water every time it storms and the piles of autumn leaves that wait for weeks to be suctioned off the asphalt.
Nor does this account include extreme events such as the October when an eight-point buck in rut roamed the streets in search of a mate. Or the Sunday afternoon when two boys imagined an Olympic luge track in the sloping, curving streets and used their skateboards to zoom down, feet first and on their backs.
Technically, the speed limit is 25 mph, but cars rarely go anywhere near that fast. Even the drivers of the increasingly present delivery vans ease into their next stop. Very low speeds prevail because it is difficult for a human to drive even 25 mph in a place like my neighborhood. The parked cars, ironically, do much of the work of taming motor vehicles.
At times, especially during the morning dash to work, it can still feel like drivers are going too fast. Movement is slow enough, however, that people walking, bicycling, and driving can catch each others’ eyes. Through wordless expressions, we glare, apologize, and forgive. This is the safety that happens through being witnessed, and feeling like you might be witnessed.
A precise balance reigns. The apartments produce a bit more traffic than if the neighborhood contained only single-family houses. The density of dwellings, however, creates the critical masses of people walking and parked cars that complicate the streets just enough to slow, but not halt, movement. With fewer parked cars, people would drive faster and the sidewalk-optional walking culture would start to unravel. More than safety is at stake. I want to enjoy being on the street — relax in the fresh air, walk my dog, and get to know my neighbors.
Every day, as they have done for decades, the people in my neighborhood collectively create a network of shared streets. Four transportation elements — narrow streets, parked cars, stop signs, and “no parking” signs — make it work.
The most important ingredient, however, is the Golden Rule. Is autonomous driving about teaching a self-driving car to do onto others as you would have them do onto you?
For a vehicle to autonomously interact with people, does it need to believe that it is capable of being a person walking? A child playing basketball? A man a bit drunk on a Friday night? At the gathering in Washington, I heard hints that vehicle designers may indeed be experimenting with emulating empathy.
Or, should my neighborhood be “fixed?” Should my neighborhood’s streets — and the behavior of its people — be defined, simplified, and standardized? Perhaps. But fixed by whom, for what purpose, and for whose benefit?
And whose neighborhood streets get fixed first? What happens to the neighborhoods whose streets get fixed last? This raises a third way to limit an operational design domain: by geography. Autonomous vehicles could be confined to certain districts or neighborhoods, slicing and dicing urban space in very new ways.
These are difficult questions, to be sure, but they come with the larger opportunity to create different types of streets. Which operational design domains should be low speed? We could expand the number of places where people and transportation happily co-exist at low speeds. Or, we could create places where people are content to live, shop, and dine on streets where motorized vehicles whip by as fast as the technology can bear.
These choices will shape how people experience life in cities, suburbs, and small towns across the world. They will also influence how autonomous driving gets deployed. I propose this first principle: operational design domains belong both to automotive engineers and to the people who inhabit them.