New cities for self-driving cars
100 mph roundabouts when?
Electric autonomous vehicles (AV’s) are not just a safer way to drive. They’re an entirely different mode of transportation that happens to look like a car. AV’s will change how cities are organized.
Self-parking cars
One of the biggest perks of self-driving cars is their ability to park themselves. They can bring you directly to your destination. No parallel parking. No competing for a spot while the backseat yells suggestions. No squinting at parking rules. Just get out at the doorstep and let your car glide away.
The car could use a parking space nearby, but there’s an opportunity to completely decouple parking spaces from human spaces.
Parking takes up almost a quarter of the land in city centers. This prevents cities from achieving a gentle density by stretching everything out. Just look at what downtown Houston used to look like with parking:
Parking lots suck1. The sim city designers had to hide much of the parking because it would’ve ruined the game. With self-parking cars, humans don’t ever need to spend time in them again. But how will we store them?
Beautiful parking structures and solar battery lots
To get the full benefits of self-parking, we need to reclaim all of the land used for flat parking spaces and use a tall, dense parking structure for AV’s. Since people won’t need to use the structure, we can do away with amenities like lighting, fire sprinklers2, railings, ventilation, stairs, and high ceilings.
Since we’re not worried about the time it takes to park, these structures can have fewer roads and a single entry and exit point. Imagine a one-way parking structure where cars spiral all the way to the top and back down to exit. Taken together, structures for self-parking cars can be denser and taller than modern parking structures.
There’s just one problem: parking structures are ugly as sin. Beautifying them is tricky too. Parking structures are pretty cheap to build, so an expensive makeover can add a lot to the cost.
One option is to hang vines down the side of the structure supported with wires and irrigation. The mass of greenery will also keep the neighborhood cool on hot days. More elaborate schemes like hanging gardens, murals, or carved stone walls are possible. But keeping costs down is critical.
For long-term parking and excess rideshare cars, we need a cheaper lot. Outside the city, flat lots can store many more cars. Solar panels can protect the AV’s from weather and produce power to charge the batteries. This concentration of solar and batteries can benefit the local grid34.
Changing road layouts
Recently, I got so frustrated at a long red light that I redesigned the traffic system while I sat there. I wanted to flow through the city without stopping until I reached my destination.
Inspired by UPS drivers, my first solution was to abolish left turns entirely. By blocking traffic in certain directions and allowing cars turning right to merge smoothly, you can get everywhere without stopping. The problem is that it takes longer to get places.
On a second pass, I think it’s better to replace every intersection with a roundabout. The transition would be expensive but worthwhile5. They’re safer, with higher throughput, and offer full flexibility of movement. The center of large roundabouts can be used for art, parks, housing, or parking.
Self-driving cars can adapt to new traffic rules and use roundabouts at higher speeds without collisions6. Once AV’s are common, switching the whole city over to a new traffic system is as easy as pushing a software patch.
Georgism for cars
One benefit of putting an internet-connected computer in every car is automated billing. Dynamic congestion pricing can be implemented on every road with no hassle for users. This will fix traffic, ensuring a smooth flow throughout the city. Parking charges will do the same for parking lots, ensuring that parking space is used efficiently and encouraging more people to offer their car for ride sharing services.
Large cars take up more space on the road and more space in parking lots and should pay for the resources they use. Ideally this tax on car size will cause a shift to smaller cars, improving traffic flow and space efficiency78. The same principle means that busses and alternative forms of transport9 should be subsidized. EDIT: AV’s can also be used to clear up jams.
Designing a tax regime that encourages efficient use of shared resources is precisely the goal of Georgism. With cars that pay automatically, tax systems can be tailored to each city without bothering passengers. The revenues can be used for road maintenance and other infrastructure.
Bigger urban agglomerations
Because AV’s are much safer than human drivers, they can drive at much higher speeds. The flow of traffic on highways could reach 100 mph if every car is self-driving.
Because AV’s are more pleasant, they should raise Marchetti’s constant, the amount of time per day people are willing to commute10. High speeds, roundabouts, and the time savings from self parking mean that people can get much further in the same amount of time11.
Taken together, urban agglomerations can be much larger under a self-driving paradigm12. By expanding our cities slightly, we can fit far more people1314. This is extremely valuable because productivity increases with city size.
You might worry that with bigger, denser cities we’d be turning the Earth into a capitalist hellscape. But bigger cities mean more nature; by densifying, some built land can return to nature.
Maybe the city isn’t your thing. Fortunately, AV’s and urban agglomeration are a boon for people who prefer rural life. Agglomeration effects will draw people to the metropolitan area, lowering demand for land outside the city’s orbit. In case you want to visit, self-driving cars make the trip easier since they are more pleasant to ride in, travel faster, and can drive through the night.
City blocks to quiet towns
The story so far has revolved around making cities better for cars. Robots whizzing through roundabouts makes things less pleasant for people at street level. What are we going to do for them?
Walking next to cars is dangerous and cross walks interrupt the flow of traffic. Pedestrian bridges would pay for themselves by reducing injuries and slowdowns. With a network of bridges between city blocks, people can stroll anywhere in the city without having to dodge traffic.
This encourages walking by making it more pleasant15. No need to wait for a car to pick you up if the destination is only a few bridges away.
Without crosswalks, there’s little need for sidewalks next to the street. The pedestrian bridge can land in the center of a city bloc, allowing shops and houses to turn inwards, away from the street. These perimeter blocks would have internal spaces insulated from the bustle and noise of the street. Large enough blocks could have a mixture of parks, residences, and restaurants, forming a small town-like unit.
I think Barcelona’s superblocks are a good starting point.
Auto delivery
With city blocks turned inwards and pedestrians off the street, the rear of each building is free to interface with the cars. Automated delivery trucks can facilitate exchange of goods anywhere in the city16. In addition to long haul trucking, this is a significant step towards an internet of physical goods.
Reliance on delivery over direct shopping will change the composition of the city. Supermarkets, pharmacies, and retailers are glorified warehouses for goods and will be replaced by actual warehouses outside the city that can deliver goods faster and cheaper. Ghost kitchens and multi-brand restaurants will absorb fast food franchises.
Places that sell a desirable atmosphere will grow. Coffeeshops, bookstores, libraries, parks, and sit-down restaurants don’t diminish in value when delivery is omnipresent.
Curious new business models may arise. Parks could charge a fee to use a table for lunch, using the revenue to maintain the greenery. Retail stores use salespeople to encourage customers find a product, but if all the products move to a warehouse outside the city, better to send a rack of clothes and a fashion expert directly to your home. Or to a small try-on spot nearby. Grocery stores can do something similar, a small shop sampling of thousands of available items and a salesperson present to suggest meal plans and encourage you to bring an extra treat home.
This is an extension of the point I made about hyperlogistics; with less labor required for manufacturing, warehousing, delivery, and checkout, employees are freed to help customers enjoy their options.
Conclusion
Life would be markedly different in such a city. Commuting is as simple as being picked up at the edge of your town, reading in the car as it swings around the city, and hopping out at the door to your office. Walking from town to town, cars buzz under the bridges; they punctuate city life rather than dominate it.
Shopping is a thing of the past, replaced by direct deliveries and visits to sample shops. When you go to a new block, it’s for the people and the atmosphere. The insulated interiors are filled with cafe’s, restaurants, offices, libraries, and parks.
Towns diverge in character, with some celebrating particular cultures or catering to particular lifestyles. The physical insulation from other blocks and a desire to attract visitors will encourage experimentation.
Each change reinforces the value of AV’s. Parking consolidation makes the city denser and more walkable. Neighborhoods with pedestrian bridges eschew street access and turn inwards. Deliveries enter at the exterior and feed an interior that emphasizes presence over products.
The result is cozy blocks made of houses and pleasant businesses all fed by the rush of cars outside. Agglomeration makes everyone more productive and the density preserves the wild lands at the edge of the metropolis. With foresight, self-driving cars are an opportunity to build a better city around them.
Further reading
Devon Zuegel has a syllabus on cities.
The prophet of parking. A profile on Donald Shoup’s work on economics of parking.
Austin vernon on self driving cars.
Handmer on congestion and city design.
The forthcoming book The End of Driving is relevant.
I discuss other implications of self driving cars here.
Edit Feb. 26th onward: notes on the comments
After reading people’s comments, a few points of clarification:
The point of having all of these rules is to free up land use for whatever people want to do. We constrain the cars to free the people. I’m aware of how often top-down city planning fails. I hope to avoid that by using rules to guarantee mobility so that we can be agnostic about how the land is used or organized. Like Barnett’s skyscraper example in The Structure of Liberty, by structuring a building and how people can move, we can create more freedom and more space overall.
I envision the superblocks being quite large, perhaps half a kilometer on a side. Since there aren’t internal roads or parking, these blocks can effortlessly achieve high density while still feeling like a suburb. Parking, roads, and sidewalks can use half of the land in the city center, so a density of 12K/km^2 (a little denser than NYC) feels like 6K/km^2 (about the same as Garfield, NJ). So each superblock would house around 3000 people, similar in size to a pretty small town like Wales, NY.
Some were concerned that pedestrian bridges would make it infeasible to wander around. Or that all the up and down would be too cumbersome. On the contrary, I think it would make walking easier and more pleasant. Most of the time is spent walking through the quiet superblock-towns with brief excursions on soundproofed bridges. The superblocks have far more amenities than a typical city block due to the higher density. You only need to take 2 bridges to reach 13 superblocks with 40K people. It doesn’t seem like that much trouble to walk up a ramp every once in a while, but some small, 1 story bike elevators at each end should make walking even easier. Though see an alternative pedway system below.
Some were just generally concerned by the emphasis I put on cars in this piece. But this is about de-emphasizing cars. We’re condensing all the infrastructure cars normally require so there’s more room for people. And as people have pointed out, AV’s mean fewer people have to own cars in the first place.
No discussion of cities is complete without train advocates. I think trains, subways, busses and other forms of transit are great too. But no transit mode is better than all the others, trains included. All of them trade off in terms of speed, cost per passenger-mile, and space efficiency. Cars are fast, cheap, and space inefficient. Trains are slow, cheap, and potentially space efficient if you build a subway. Ideally, cities use a portfolio of approaches to transit. In this case, underground rail makes the most sense. Surface level rail would interfere with roads and and raised rail might be too expensive if it goes over pedways. Since this design has lots of short-hop transit options, longer routes running the perimeter or connecting nearby cities are more valuable.
After thinking about the pedestrian bridges more, I think it makes most sense to build them running along the road. Imagine a grid layout oriented along the cardinal directions. Every block could have a pedestrian lane running north-south raised over the two east-most lanes of the road. There would be east-west lanes over the southernmost two lanes laid down every few blocks to ensure connectivity. 2 car lanes is about 20 feet, enough to fit 2 bike lanes (~6 ft each) and 2 pedestrian lanes (3 ft each) comfortably. Pedestrians can walk down to the basin of a city block using a spiral ramp on the west face of the block (or perhaps the northwest corner). This utilizes space more efficiently and is more pleasant than any plan that puts pedestrians and cyclists on the same level as cars. It also requires far less up and down ramps per trip since pedestrians at ground level would have to be raised over the cars at every roundabout. Though the length of pedestrian track is much longer this way, the overall cost might be the same or lower since it requires 4x fewer ramps and doesn’t take land from the streets or the core of the city blocks. These pedways also have the benefit of shielding the buildings above from street noise.
A city that adopts pedestrian roads as outlined above can organically grow more pedestrian infrastructure. Adding a rim of walkways 1 story above the basin of the city block and adding skyscraper bridges between stories of the city blocks. A pedestrian and micromobility layer can grow on top of all the roads, to the point where the cars are essentially underground. No noise. Houses or hotels could have a balcony that connects directly to the pedway. Note that building over the roads is cheaper than building tunnels for cars or putting the cars over the pedestrians.
It probably makes most sense to put the bike lanes on the outer-most side of the pedway so that pedestrians can reach the stairs without having to cross the bike lanes. There can be bike racks near the stairs and on the outer side so cyclists can dismount and park on their right side without having to cross the other bike lane. Pedestrians will still have to cross bike paths at pedway intersections, but hopefully the traffic is low enough that this isn’t an issue. Pedway intersections probably need bike roundabouts and pedestrian crosswalks, I don’t think another layer of grade separation or bridges is necessary.
Another potential option is to abolish cars entirely, moving to micromobility on the raised pedways, turning streets into superblocks, and running subways underneath. Cars would be for longer trips between cities and bulk delivery.
There’s a natural progression to building the city as I’ve described. First start with congestion pricing to immediately improve traffic flow and generate revenue. Then add roundabouts which work best without congestion. Build pedways next to get the full benefits of traffic flow through the roundabouts. Then build the big parking structures, self-parking technology, and regulation hand in hand. At that point, everything is in place to remove sidewalks, let city blocks expand, and turn inwards creating cozy, walkable towns embedded in a bustling city.
While congestion pricing makes money and parking structures will be made with private money, the roundabouts, pedways, and subways require public funding. However, these developments will be largely self-financing through higher local land values. A city can best capture these gains by issuing bonds and collecting land value taxes.
Liu Jiakun’s West Village incorporates some of the elements I’ve discussed here. Though the ramps seem excessive.
Not to mention that despite the low speeds, 20% of accidents happen in parking lots, with 60K injuries per year.
What if there’s a fire? Well the first thing you want to do when a battery is burning is NOT put water on it. EDIT: I’m wrong. Water can cool down the battery and halt thermal runaway. The amount of water you need to put out a battery fire is substantial, such that fire sprinklers won’t do much good.
I wonder if all those self-driving computers sitting idle would be useful for something. Running AI inference while they are idle? AI training run in a parking structure?
One gripe I have with at-home vehicle charging at night is that it uses electricity when it’s expensive. Most charging should happen during the day.
If you really wanted to spend money, every intersection could be grade separated. Then you could go any direction without stopping! A small amount of grade separation might help flow in a city full of roundabouts. Since the cars are going at high speed, it’s better to keep one direction level and dig the other direction underneath so they maintain traction better.
Smoother driving and better use of regenerative braking (slow, consistent braking) combined with less congestion and roundabouts means that electric AV’s in this city will use less energy per mile than human drivers.
See footnote 8. In theory, cars of a certain size category should share the cost for the road widening they cause.
Combining dynamic tolls with route planning makes for interesting policy. Essentially, you want cars to pay their shapely value and design tolls that encourage cars to minimize their shapely value.
For example rail, micromobility and eVTOL’s.
There’s some interesting economics here. The more comfortable travel becomes, the more people are willing to travel. The more people travel, the more productivity due to agglomeration. The more productivity, the less people want to travel since their time is more valuable.
A numerical example: if the car spends 20% of it’s travel distance in roundabouts going 25 mph and 80% of its travel distance going 45 mph then the average speed is 41 mph in the city. In half an hour on a grid system, the car can cover a square diamond shape with a diagonal length of about 41 miles (20.5 miles to the north, south, east, and west from center, respectively). The area is 840 square miles, larger than almost every major city and about the size of Tokyo.
With self-driving cars, these agglomerations are far more accessible to children, the elderly, and people with disabilities.
Moving to smaller cars and trucks in the city would mean roads have a smaller footprint. Consider, for example the value of reducing the size of fire trucks and road widths:
Micromobility vehicles or gondola systems might also find use here.
Drones and delivery bots will also come in handy, serving households in the interior without street access.
Brilliant ideas! I am very bullish about the future of self-driving cars, and I'm sure at least some of these ideas will get implemented in the next few decades.
Nice work, Sam. I did some similar writing about what life would be like without gas stations: https://powermetal.substack.com/i/154912880/goodbye-gas-stations-hello-homes-and-gardens