Why self-driving cars might make traffic worse, not better
By Sam Schwartz
I hate sitting in traffic.
I hate it so much that in 1986 I implemented a computer program that could determine the best way to beat rush hour so I could work a full day (I was traffic commissioner of New York City at the time) and still make it to Shea Stadium for the World Series.
When it was time to go, my unmarked Plymouth Fury would glide effortlessly through the traffic signals on the Northern Boulevard as they turned green in sequence.
(I’m exaggerating a bit here: travel patterns made my signal plan useful for many other fans as well.)
My dislike of bottlenecks and gridlock also inspired my 1993 book Shadow Traffic’s New York Shortcuts and Traffic Tips.
It was specifically about avoiding and outsmarting the notoriously gnarly traffic in New York City in the pre-Waze days.
The year 1993, let alone 1986, seems a lifetime ago, but traffic hasn’t gotten any better.
In fact, it’s gotten worse.
Bill Ford, executive chairman of Ford Motor Company, warns of “global gridlock” by 2050 “if we continue on the path we’re on.”
He adds, “Our infrastructure cannot support such a large volume of vehicles without creating massive congestion that would have serious consequences for our environment, health, economic progress and quality of life.”
The American Highway Users Alliance found that the top thirty bottlenecks in the country are each responsible for more than one million hours of lost time annually.
Drivers stuck on roads with bottlenecks experience delays of about 91 million hours every year, the equivalent of 45,500 person-work-years.
The lost value of time to the economy from congestion just in this handful of locations is upward of $2.4 billion annually.
If the economy continues to improve, the 2015 Urban Mobility Scorecard predicts that by 2020 annual delays per commuter will grow from 42 hours to 47 hours (more than a full week’s work lost); total delays nationwide will grow from 6.9 billion hours to 8.3 billion hours; and the total cost of congestion will jump from $160 billion to $192 billion.
To put this in perspective, even in a “recovering” 2014 housing market, Americans lost $192.6 billion to foreclosures.
With the stakes and costs this high, the enthusiasm around the belief that autonomous technology can end global gridlock is understandable.
“A Single Autonomous Car Has a Huge Impact on Alleviating Traffic” is a provocative headline that refers to a 2017 study done at the University of Illinois at Urbana-Champaign.
The research, led by assistant professor and engineer Daniel Work, demonstrated that placing a single autonomous car in a real-life circular traffic simulation greatly reduces the chance of a “phantom” traffic jam.
An autonomous vehicle with “intelligently” controlled speed reduced the standard deviations of speed in all twenty cars in the circle experiment by about 50 percent.
The number of times brakes were hit by drivers in the circle was reduced from nine hits per vehicle for every kilometer traveled to about two and a half, and sometimes the rate went down to zero.
Although there is some truth to Work’s finding that a speed-controlled car can influence standard deviations in speed, it’s the kind of research that easily can be hyped as supporting the “miracle” of AV technology.
It would be great if the solution to congestion really was as simple as setting a single Autonomous Vehicle into a stream of commuters. But it’s not that easy.
The slowest vehicle on a road always has the biggest impact on other drivers, so Work’s study isn’t saying anything particularly new.
In 2016, I demonstrated for the TV newsmagazine Inside Edition how a phantom jam occurs even when we have one vehicle traveling at a constant speed — contrary to Work’s finding.
The drivers following a vehicle going at a constant speed just don’t have the ability to travel at a perfectly consistent speed themselves, so the spacing between cars begins to vary, with some getting too close to the car in front of them and some farther away.
The close-in vehicles eventually tap their brakes, which sets a “shock wave” flowing through the line of trailing cars.
Nonetheless, I have little doubt that Autonomous Vehicles in sufficient numbers on a freeway will sharply reduce the number of phantom jams.
I also believe that the carrying capacity of freeways will increase with very high percentages of Automonous Vechicles.
But I am less optimistic that we will see such improvements on city streets, where an increase in the sheer number of vehicles, made worse by the attractiveness of AVs, may actually worsen traffic flow.
Thirty-five percent of congestion happens simply because there are too many cars on the road.
The finite amount of space in cities, with roads going in a variety of directions, both straight and circular, limits capacity no matter how efficient and “smart” cars become, and no matter how many lanes are added to roadways.
The physics of this is not hard: vehicles cannot occupy the same space (that’s called a crash) and keep moving. There is no such thing as infinite capacity, and we have to come to grips with that.
Certainly, carmakers want to sell as many Autonomous Vehicles as possible, but congestion disaster will happen if we continue to emphasize the car as private transportation carrying one person on a disproportionate number of individual trips.
The concept of ownership is so critical to whether AVs turn out to be a social enhancement or a detriment.
The consequences of increased traffic congestion could be disastrous if we shortsightedly allow the Autonomous Vehicle industry to dictate the terms of travel, as the auto industry did throughout most of the twentieth century.
If we’re smart and insist on a variety of stakeholders having a say in how traffic laws and policies pertaining to AVs are written, we can cut down on lost time and money and decrease congestion while enhancing transportation.
The right way to introduce AVs so as to reduce gridlock is to integrate them into our transit systems.
The role of transit will evolve as AVs and ride-sharing become more common.
Transit agencies should focus on high-frequency, high-capacity services in dense urban corridors and provide first- and last-mile connections through lower-capacity on-demand driverless shuttles as well as expanded mobility hubs.
City-operated robo-buses, whether Chariot, Via, or some other mmulti-passenger next-generation form of micro-transit, could deliver much more service for the same amount of money.
Smaller buses, operating at a lower cost per mile with less labor, would provide more frequent service at lower fares.
Airports, notably London’s Heathrow, are already doing this with people-movers, as are other services in cities from tiny Morgantown, West Virginia (which figured this out forty years ago) to Abu Dhabi.
In many suburbs, Autonomous Vehicles could link to existing transit lines or to new transit lines that will link trains to buses.
Cities can also work with transit agencies and private companies alike to design and adopt smartcards, open data, and universal apps that will allow riders to find, compare, book, and pay for trips that combine buses, trains, bikes, and ride-sharing vehicles.
Such services will match customers with the most efficient (and healthiest) travel choices.
AVs could also help ease congestion if we create policies around them that limit their negative aspects and reinforce the positive.
Dynamic road-charging — whether on a real-time basis or in an arrangement that varies by time, day, season, origin, destination, number of passengers, and household income — can be done through a combination of congestion pricing, zone pricing, variable tolls, and VMT (vehicle-miles-traveled) and VHT (vehicle-hours-traveled) fees.
For instance, if you want to show your family the Christmas tree at Rockefeller Center in Manhattan while sitting in your vehicle as it slowly rolls by, that trip should cost a heavy premium — say, $25 per block.
Traveling on an outer Manhattan avenue will cost far less, while travel outside the borough might cost very little or nothing at all.
If center cities can control the number of private vehicles coming in through congestion pricing, mobility for those remaining in motor vehicles would improve.
Cars, trucks, taxis, Ubers, and buses would all move faster.
That’s why Uber, in late 2017, began airing ads supporting congestion pricing for Manhattan, which would also reduce air and noise pollution and our carbon footprint.
Congestion pricing could create a wonderful opportunity for cities to add bike lanes, widen and add pedestrian paths and spaces, license outdoor cafés, plants rows of trees, and create parks, all to the benefit of residents and visitors alike. To me, this would be a win-win-win all around.
On-demand real-time mass transit is another benefit of smart traffic planning. Imagine a transit system that knows when a high school football game is ending, a concert will be over, or how many people are likely to be at a particular stop.
Smart mass transit will also reduce the need for fleets of large buses; instead, a variety of passenger-capacity vehicles will be configured to respond to the needs to travelers.
A forty-person bus would not need to be dispatched to pick up a handful of people.
AVs can help meet the demand for living and working in proximity to amenities through mixed-use neighborhoods, whether in urban areas or small towns and suburbs.
To do this, communities would have to incentivize mixed-use development, overhaul parking requirements (I’d expect a greater than 50 percent reduction, and maybe much higher), and reevaluate new public transit projects.
When shopping and work zones are easy to access and close by, the need for personal cars is greatly reduced.
If ride-sharing becomes mainstream — and more convenient, efficient, and pleasant and cheaper than owning a personal vehicle — we may see a dramatic decrease in congestion everywhere.
Commuters may be more willing to forgo using their own cars if ride-sharing has been able to customize destinations.
I proved the congestion-busting payoff of ride-sharing in 1980 during the New York City transit strike. We imposed an occupancy restriction, the temporary rule during the strike that cars could use the capacity-challenged Long Island Expressway (LIE) only if they had a minimum of three passengers.
To get a report on conditions during the strike, I had traffic engineers in helicopters (in those days we didn’t have Waze, INRIX, or drone technology) reporting back to the command center via walkie-talkies on traffic flow.
The media’s predictions were dire, and frankly, so were my own as a representative of the city: the LIE would be a parking lot! If so, traffic would receive a grade of F on the level of service.
What happened? Traffic on the LIE got an A during the morning rush hour of day one of the transit strike — traffic was free-flowing and moving as fast as if it were 3:00 a.m. — because we had tripled per vehicle occupancy on the highway.
Clearly, we now have a similar chance to free our cities of congestion if we get occupancy close to two per vehicle instead of a little more than one.
There is an antidote to congestion, and I witnessed it firsthand in April 1980.
If we are smart — if we dramatically increase car occupancy to at least two per car and maintain other modes of transit, privilege pedestrians over cars, and increase walking, biking, and train use — we will solve the problem of congestion.
We will put an end to the argument that “we can’t do nothing about the weather and the traffic.” But the politics is not easy.
Samuel Schwartz is the author of No One at the Wheel: Driverless Cars and the Road of the Future, from which this article is adopted.