How Autonomous Driving Will Change Our Cities

Traditional Car Ownership on the Decline

Does your commute involve a car? Odds are, yes. 74% of Americans who work in metropolitan areas commute by car. The rest of the time, our cars sit idle, waiting for us to need them about 95% of the time, taking up a great deal of space. A significant amount of real estate in our cities is dedicated to cars that are not in use, between the surface lots, parking garages, and curb-side spots.

Surely, this valuable real-estate could be put to better use, such as housing, employment opportunities, or common space. But, what to do with all of these cars each day that might be needed at a moment’s notice? Do we need them at all? Enter, the autonomous car.

Clockwise from top-left: Tesla Model S, Google/Waymo self-driving car, the Mercedes F 015 Concept, and an Uber-LIDAR-retrofitted Ford Fusion

What is Autonomy?

The Society of Automotive Engineers defines six levels of autonomy when it comes to the various levels enabled by ADAS (Advanced Driver-Assistance Systems):

  • L0 — No Automation: steering, brakes, throttle, all controlled by you, the human
  • L1 — Function-Specific Automation: adaptive-cruise-control, for example
  • L2 — Combined Function Automation: this could include adaptive cruise control plus lane-keeping technology
  • L3 — Limited Self-Driving Automation: under certain environmental and traffic conditions, no human input necessary
  • L4 — Human-Mostly-Expendable: the car can handle pretty much all aspects of driving within its sphere of comfort, whether that be a city environment, or anywhere without inclement weather, etc.
  • L5 — Human-Fully-Expendable: the car can handle any and all driving tasks without the need for human intervention, nullifying the need for a steering wheel or pedals

Right now, the best autonomy we have access to sits around levels 2 and 3, where many new luxury cars can handle highway driving and shifts in traffic patterns, but rely on human intervention when leaving the highway or dealing with severe weather.

Current Teslas have all the parts needed to attain the higher levels of autonomy from a hardware perspective, and the only areas that need to be hammered out to get fully there are software and policy-based. Adaptive cruise control and lane-keeping are both very common features in new cars these days.

When level-5 autonomy is achieved, many assumptions around what it means to use a car can go out the window. Even car interiors are ripe for innovation, as true autonomy can allow cars to become a mobile office and entertainment hub, with seating all pointing inward around a central “coffee table”.

A good example of what this might look like can be seen in the Mercedes-Benz F 015 Concept car as demonstrated by The Verge in this video here (longer video from Mercedes at the end of this post):

To be completely clear, from a technical perspective, this is a matter of *when* we get there, not *if* we do. This is not science fiction.

Own a Car

With an autonomous car, imagine driving or being driven to work by getting out right at the entrance to your office building. You did not need to look for parking, and because your car is autonomous, you spent your commute getting a few emails handled, bills paid, and video-chatted with some friends.

Now that you are at work, instead of having your car sit idle for the next 8 hours, it can go park itself in a purpose-built garage off-site where the real-estate is cheaper. While here, it can charge itself to be ready for you at the day’s end.

Remember that the car is autonomous. It can spend the day driving around the city, giving people rides from Point A to Point B, delivering Amazon packages, and so on. It can become part of the “smart fabric” of the city, an additional node for productivity on already existing large logistics networks of the likes of Uber and Amazon.

Your car knows when you are done with work, so it will never leave you waiting to be picked up. Meeting get done early? Still working on a PowerPoint? Your car knows what you are up to, and can adapt its own agenda accordingly.

This daytime activity for your car means it can effectively become a revenue stream that helps offset its own maintenance, insurance, and energy costs.

This sounds pretty good, right?

What if you don’t even need to own a car in the first place to live this life?

Don’t Own a Car

Imagine the previous scenario, but instead of having to own the expensive, fancy autonomous car of the future, what if you were on the other side of that transaction, merely using cars owned by others to get around?

They could be owned by individuals (as in the previous example), private fleets, or as city-provided “public transit”. Ultimately, it does not matter to you as the user where the cars come from.

Think of this as the hyper-efficient “next turn of the crank” for the likes of Uber and Lyft for urban transportation. Maybe I still have my car on Uber’s platform to make money for me, but instead of being the driver I can be at home watching football or doing my day job while my car is out “in the wild” working for Uber and their users on my behalf.

As a user of an autonomous ride-share, the key difference I see has to do with my ride now navigating in a manner guaranteed to be the most efficient, all the likely while running up a more affordable tab, as it turns out robots do not expect to be compensated for their efforts.

Here is where things get really interesting. By not owning a car, I don’t have to commit to a specific type or class of car in my life. I can have my default on-demand ride of a four-door car for shuttling the family around, but can just as easily call for a truck for a trip to Home Depot, or an off-road-capable SUV for a trip to the mountains, or a svelte convertible for date night.

By not owning a car, this sort of platform can actually make me feel like I “own” many different cars!

Changes to Cities

So, what does the concept of autonomous motoring mean for us in aggregate, in our cities?

Imagine a city where every car is fully autonomous. What does this get you?

For example, congestion automatically drops dramatically, especially on highways, where following distances between cars can drop to near zero. Today, we only maintain following distances while driving to safeguard against the unpredictability of other motorists, as well as to factor in our own poor reaction time.

Autonomous cars behave more predictably, have faster “reaction times”, and can even sync data across each other in a sort of P2P (peer-to-peer) mesh network to introduce additional efficiency.

This is something that Cadillac and others are already experimenting with, though of course for this to actually work, we can’t have a General Motors P2P standard that fails to communicate with and understand a competing P2P standard from the German automakers.

Especially at high speeds, the other key benefit of this “mesh network caravanning” (where cars cluster together with very tight following distances as though they were part of a single train) comes through in the form of greater energy efficiency: drafting. Drafting is commonly-used in cycling and motorsport (NASCAR, etc.) to increase speeds without losing efficiency and to enable easier passing of competitors.

Drafting at work

Not to oversimplify, but drafting (also referred to as “slipstreaming”) is made possible by a quickly moving entity (vehicle, cyclist, etc.) creating a pocket behind it where the air is thinner. In this close-following thinner air, any other vehicles in that space do not face as much wind resistance as they otherwise would and can therefore achieve greater speeds with the same effort, or maintain the same speed with less effort. This is commonly used to then slingshot ahead of the lead vehicle.

Drafting also benefits the lead vehicle too, as drag coming off the rear of the lead vehicle is minimized if there is another vehicle occupying that space. As such, a large caravan of vehicles drafting off each other not only occupies less space on the roadway, but also achieves greater fuel efficiency.

While P2P-drafting cars will certainly ease commutes, once in the city the primary benefit comes in the form of there being no need for parking in high-value real estate. Cars can stay active all day shuttling people and goods around. If demand dips and some cars need to become fallow resources, they can go park themselves in a giant parking depot outside city limits where they can charge up, get a wash, undergo maintenance, and so on.

Now we get the street-side parking, surface lots, and in-city garages back. These areas can become common spaces, housing, offices, and so many other things that boost productivity and happiness in greater ways than just being holding space for our vehicles in a way that tries to optimize for having them as close to our offices as possible.

Beyond obviating the need for convenient parking at the expense of what else that space could be, do you even need signs and stoplights?

Imagine an intersection like this, but without the obvious risk and room for human error since all vehicles are autonomous and communicating flawlessly in real-time:

  • Think of what this does for emergency response in crowded cities: cars are able to all move out of the way of ambulances in unison like a large, metallic school of fish.
  • Airport pickups and drop-offs become significantly more effective, as vehicles can arrange themselves in the most space-efficient fashion and home in on the exact GPS location of where the person they are picking up is located.
  • Parking garages can become more tightly packed as there will be no need for space between cars since people will not be getting in and out of cars in these autonomous garages. So long, door dings!

While this is going on in the cities, what happens to the suburbs? Before automobiles were commonplace, many people lived in crowded urban environments since the only way to have everything you needed accessible was to have everything be within walking distance. With the advent of the automobile, the mid-twentieth century saw a massive surge of population radially outward from these urban cores, launching suburbanization.

With cars, the range of everyday people extended dramatically, and the draw of cheaper real estate and a still-reasonable commute time was irresistible. That is, until commute times began swelling faster than our highways could grow to accommodate this growth. As such, this trend then swung the pendulum back the other way, with a large amount of people now opting out of extensive commutes and expensive cars, and into more expensive real estate that allows them to better centralize. That said, there is still a not-insignificant part of our population that undergoes “extreme commutes” (NYT), and many others are still spending too much time getting to and from work in a manner that keeps them from being able to double-count that time as productivity.

With autonomous cars available at a reasonably affordable level (someday), one can see how this might push things back in the direction of suburbanization, and perhaps relieve the pressure of overcrowding in our cities. So much of America’s landmass is utterly devoid of people, it certainly would not hurt us to spread out a bit more!

How do we get there? (Partner up!)

When it comes to making this vision of our urban future a reality, there are quite a few stars to align, including huge areas of overlap between technology, business, and policy.

In general, there seems little doubt that the technology involved will not be the bottleneck to realizing this safer and more efficient future. In fact, Tesla has claimed that every car they produce is capable of full autonomy down the line, and all that needs to catch up is software that can be pushed as an OTA (over-the-air) update overnight, and regulatory policy (which will definitely take more time).

In terms of business partnerships, there are already alliances forming around each other’s strengths. For example, there is a tripartite bond between General Motors, Lyft, and Waymo (of the Alphabet/Google family) which makes a great deal of sense. Lyft already has a ride-sharing network, and Waymo has been working on the R&D around autonomous cars for such a long time that they arguably began this conversation in the first place. But, why General Motors?

The cars in use need to come from somewhere, and does it seem like a good use of effort for Google or Lyft to enter the space of building their own cars? Lyft does not need expertise in dealing with airbags, and partnering with an organization that has been working in R&D and supply chain relationships in the automotive industry for the past ~100 years. (Sure, Tesla is doing everything in-house, but this is why they are considered a special “unicorn” of a company, particularly with the distinction of being the only new American automaker to surface in many years that has stuck around against the deeply rooted incumbents.)

Today, when you summon an Uber or a Lyft, the car that comes to you is owned (or leased) by the driver. How does this work if we are removing the driver from the equation? As previously mentioned, perhaps I personally own the car and am making it available on platforms as a sort of modern GetAround. Companies like Uber and Lyft do not necessarily want to be in the business of maintaining their own fleets. After all, their distributed model has allowed them to be so successful without owning single car, just as Airbnb is the largest “hotel chain” in the world without owning a single room for rent.

Without relying on cars owned by individuals, General Motors could then provide “seed fleets” in urban areas and take financial kick-backs from the owners of the sharing network. This would save the ride-share startups from having to worry about fleet maintenance, and this helps General Motors hedge their risk of collapsing into obsolescence due to declining car ownership caused in part by these very same sharing networks.

With enough funding, cities could also provide fleets as a sort of more expensive public transit, but the pricing model would need to be such that it does not yield overcrowding of that resource while existing infrastructure such as subways gather cobwebs.

These fleets of autonomous cars do have to coexist with all sorts of existing modes of transport as a complement, not a replacement, at least in the near term.

Devil’s Advocate

So far, this all paints the picture of a hyper-efficient urban utopia, and if that sounds suspicious to you, you would be right.

One of the assumptions I stated earlier was around flipping a switch and imagining what we could do with cities if they were full of only autonomous cars. Of course, this is not going to happen instantaneously, without a period of awkward transition.

The transition will necessarily be gradual, while the technology improves, becomes more affordable, and regulations adapt to allow this change to happen.

I like to compare this to what it was like for automobiles when they first came on the scene and threatened the dominance of horses a hundred years ago. You begin with a mixture of horses and cars trying to share the road, and the result was chaos and many horrifying crashes.

Similarly, mixing human-driven with machine-driven cars will yield more problems in the short-run while we work toward full autonomy. For example, when you pull up to a stop sign, how do you communicate with the other driver to ensure that you are both doing what you are supposed to do (since trusting that everyone understands proper right-of-way etiquette is simply too much to ask for)? This commonly involves some sort of hand gesture, a shared glance — a common understanding.

As humans, we are able to pass and try to understand vague signals like this. Autonomous cars will likely struggle with reading our intent (and therefore err on the side of caution), while we cannot read their “intent”, since there is no intent — rather a complex set of training data, decision trees, and so forth. Two autonomous cars could share exact data at an intersection and make decisions instantly. So, two autonomous cars are definitely more effective at an intersection compared to two humans, with the coordination between one human and one autonomous car being the most inefficient combination.

This sort of situation could be mitigated by having crowded urban centers require vehicles be in “autonomous mode” within city limits, but this of course assumes everyone has access to a vehicle with such a capability. Requiring everyone to own an autonomous car is not affordable or reasonable on any near-term time frame, but access to autonomous rides could indeed be commoditized.

Beyond reading each other’s “minds”, there is also the colossal discussion around safety and liability. The Trolley Problem is a classic ethics thought experiment that goes as follows:

There is a runaway trolley barreling down the railway tracks. Ahead, on the tracks, there are five people tied up and unable to move. The trolley is headed straight for them. You are standing some distance off in the train yard, next to a lever. If you pull this lever, the trolley will switch to a different set of tracks. However, you notice that there is one person on the side track. You have two options:
1. Do nothing, and the trolley kills the five people on the main track.
2. Pull the lever, diverting the trolley onto the side track where it will kill one person.
Which is the most ethical choice?

You can consider the trolley problem in the context of autonomous cars, and what they mean for pedestrians. For example, if someone hops out in front of your autonomous car, should it do everything in its power to protect the pedestrian, or if swerving means careening off a bridge, should the car instead press on and ultimately prioritize protecting its own occupants?

How does this calculation change if there are many pedestrians instead of one? What if there is a child involved? Or the elderly? Or a family pet? Do you react differently for wild animals rather than those pets that have microchips, if you can detect that?

How morally weird and horrifying does this get if those are preferences that you, the owner of the car, can set and fine-tune? If something happens, are you liable? Is the manufacturer of the car liable? Is the software developer who let you change that setting liable? What about the engineering manager who told the software developer to write the code to enable those settings?

Data coming off the car should help resolve these questions, but it would be naïve to think this would hold across 100% of all possible cases. Tesla has claimed that they are absolved from guilt in a case where an owner died when using one of their cars irresponsibly, citing logged data on the car in their case. However, the NTSB believes that Tesla is still partly to blame, which I personally feel sets a dangerous precedent. If something terrible happens due to an accident in a non-autonomous car, is Toyota still “partly to blame” since they produced a vehicle that allowed someone to achieve speeds that caused the incident in the first place?

Naturally, if you could take a step back and say, “well if this car is autonomous and can react so quickly, shouldn’t that keep it from getting into these sort of situations in the first place?” Ideally, yes, but at the massive scale of human activity across a global population, even the most improbable events become almost guaranteed.

There are additional factors to keep in mind, such as how 2nd and 3rd order consequences for the economy are effectively impossible to predict, for better or worse outcomes. Cities also rely heavily on fees and fines associated with parking infractions for their revenue, so for policy to get on board, questions around these sort of issues need to start seeing answers.


There will always be the need for human-driven vehicles, but on what scale? As the payoff of this infrastructure disruption carries less weight for rural areas, farming vehicles and related tasks likely will not be going autonomous anytime soon.

That said though — on a long enough time scale, human-driven cars will likely go the way of the horse:

  • a specific tool for a specific task
  • and a medium for recreation in a controlled environment (dressage for horses, track day for cars)

I know I cannot speak for everyone, but I am absolutely ecstatic for this future to arrive. BCG claims that by 2025, autonomous cars will be here with some level of sophistication and will cost somewhere around $10,000 more than an otherwise identical vehicle. Especially in a distributed, decentralized model of “ownership”, this absolutely gets things within the realm of accessibility for the sufficiently-sized urban populace.

I’ve had the thought of writing this post in my head for the last few months, and only recently got around to putting pen to paper on this topic. This is partially due to the fact I spend so much of my life driving my “dumb” car manually through commute traffic, weaving in and out of people who are all burning days of their lives by doing the exact same thing.

Additional resources

Thanks for reading!

Blake <@b_t_walsh>

This story is published in The Startup, Medium’s leading publication for entrepreneurs and startups.

Join +12,417 people who get the top stories here.