Danny Crichton

Cross-posted on Medium

I’ve been told twice in the past week that red lights (everyone’s favorite traffic signal!) are going to disappear with the advent of autonomous cars. The first time was from a VC friend of mine, after which I got into a fairly extended argument about why red lights and traffic are still going to be with us for a very long time (i.e. forever).

[The second time came from TechCrunch], which interviewed Jeffery Owens, the CTO of Delphi, one of the largest auto suppliers in the world. In the video’s intro, Owens says that (slightly edited) “Ultimately, if every car was talking to each other, you wouldn’t need stop signs or stop lights at all. That would be kind of an end state and traffic flow would be incredibly smooth. No traffic jams. You wouldn’t need roundabouts, you wouldn’t need lines.”

I can expect venture capitalists to hype technology, but I found it more than a little disturbing that the CTO of one of the largest auto suppliers would continue to purvey this false concept. Traffic is here to stay, and so are the red lights and other traffic signals that we love to hate. That said, capacity can definitely increase, even while traffic remains. The distinction between the two is critical to understanding the future of transit.

A Pedestrian Problem

Mathematically, the easiest way to prove that this notion of no red lights is false is to just give a counterexample. In this future world of autonomous cars, people are still going to exist — especially in cities — and those people are going to continue to walk on sidewalks. One of the reasons that traffic management is so complicated is that it isn’t just designed for cars — it has to be designed for pedestrians as well.

An easy way to see the start of a traffic jam is to just look at cars that need to make a right turn in a city. Those cars have to intercept a crosswalk, which means they need to navigate around pedestrians who are crossing simultaneously. Anyone who has walked around Midtown Manhattan or SoMa in SF knows that it can take almost a full cycle of a light before a single car can get through that stream of people.

There is really no way around this situation. If you increase red light cycle times, pedestrians get more aggressive in crossing without a walk signal, increasing the risk of casualties and further blocking traffic. That’s why cycle times are often just 30 or 60 seconds, even though flow could be improved with longer lengths. If you add more than two cycles (left and right turning signals and the like), you will still have moments when there is no flow and pedestrians are crossing.

When you also add in the very real liability issues that autonomous cars are going to grapple with, cars will be very hesitant to push through a crowd of people, unlike say, NYC taxi drivers. So we can imagine it might even be harder in the future for cars to try to cross intersections than today.

So long as people walk and have to cross streets, there will be at least some traffic.

Flow and Capacity

But let’s step back. What people get excited about with autonomous cars is the ability to increase the capacity of existing road infrastructure.

While these concepts are simple to intuit, they are often quite complex in reality. “Road capacity” is the quantity of vehicles and people who can drive through a road, while “flow” is how objects on a road (cars, pedestrians, etc.) interact and ultimately influence capacity.

To me, both metrics can get significantly better under an autonomous regime. Flow will be improved for a host of reasons: accidents will be minimized (not eliminated — random events are still going to happen), cars will not block intersections, etc. If cars could interact with each other, then they could help each other navigate a street better. A package courier truck could let other cars know it is going to stop at an open spot up ahead, and other cars can be prepared to maneuver around it.

Better flow will increase capacity, but other benefits of autonomous cars will directly influence capacity. Cars could get closer to together both between lanes and between cars, which will directly increase capacity.

In addition, the advent of ride sharing means that more people will be comfortable with autonomous car pooling. Streets today are filled with cars with single individuals in them, but in the future, we can imagine a system of almost autonomous mini-buses based around technologies like UberPool which makes it super simple to pick up and drop off passengers. Indeed, much of the on-street parking infrastructure in cities today will likely go to just these sorts of loading and unloading zones.

The Problem with the Fundamental Law of Traffic

If all cars added a passenger or two, and more cars are able to drive on a road, how can it be that traffic doesn’t disappear?

For one thing, roads are so congested and overcapacity today, that even major improvements in efficiency may not be enough to get them flowing. Think of critical chokepoints like bridges and tunnels. The Midtown Tunnel is only going to have two lanes of traffic moving through it one way, and there isn’t much to be done on a technological front that will get that to flow at 5pm on a Friday.

Additionally, in a city grid network, there is no way to actually create constant flow. Every horizontal road intersects with every vertical one. Even if you believe that cars will interweave at intersections to avoid the red light cycles (which they won’t because you still have pedestrians!), those cars will have to leave enough space for cars to pass through, decreasing capacity.

But the real culprit will be much more annoying, which is induced demand. As roads increase their capacity to move passengers, more passengers will move from alternative transit options like subways onto the roads. In addition, people who might not have traveled at all will start to consider going to that extra restaurant or showing at the movie theater. Indeed, people are so rational about their transit options that economists have produced a [fundamental law of traffic congestion] with significant empirical data to back it up. Road capacity and road demand often have a linear relationship. A 10% increase in capacity often leads directly to a 10% increase in demand for roads.

As economists will tell you, the only way around that is to create a market for transit. If it cost $40 to cross the Lincoln Tunnel at rush hour, we can imagine many people thinking twice before heading out onto the road. Indeed, [such models have been deployed in cities like Singapore], where traffic flows at almost all hours of the day. But this is a policy fix, not a technology fix for traffic.

In my mind, this will actually be one of the great tensions with autonomous cars — how much should people be allowed to pay extra in order to receive faster throughput? Singapore’s electronic pricing guarantees that cars always flow, but restricts the people who can affordably drive their car into the city. Cars, which were once the great democratizing force of mobility in American history, may become preserves of the relatively wealthy (at least during popular rush hour times).

So without a strong market mechanism restricting miles driven on roads, we should not expect any improvement in traffic with autonomous cars, although we can expect that there will be greater capacity for existing roads. Hopefully the distinction between the two will be picked up more often by VCs and cheerleading CTOs.

Photo by Blondinrikard Fröberg used under Creative Commons.