The wild traffic hack that could pave the way for self-driving cars
Autonomous vehicles are stupid — so an engineer has figured out a way to make streets smarter.
Arif Qazi / Insider
Someday, maybe, you’ll be driving like usual, and you’ll stop at a red light, as usual. You’ll take the opportunity to sneak a peek at the car next to you, and the guy behind that wheel — if there’s a wheel at all — will be watching Netflix. His car will be driving itself, as will many of the others around you. It’ll be a bumper-to-bumper jumble of robots and humans.
If all the cars were robots, maybe you wouldn’t even need the traffic light. The self-driving vehicles would just ping their vectors at each other, and collective computation would choreograph their passage through the intersection in a nimble, mechanical ballet. But throw a few meatbags in there jockeying for pole position, and — well, every flicker from green to red becomes a car crash waiting to happen. You might be able to get out of the way of a testosterone-fueled human driver in an SUV, but the self-driving cars will never see him coming.
Unless, that is, you’re in the digital paradise conjured by Ali Hajbabaie, a civil engineer at North Carolina State University. Hajbabaie envisions a state of complete harmony between humans and machines, a chimeric intersection where we all just get along. It relies on a single technological trick — an ingenious hack of the traffic ecosystem. When the light changes, it’s not red, or yellow, or green. It’s white — and white means: “Robots, go! Humans, follow!”
“If we’re talking about this new kind of infrastructure, a fourth phase of light, you might say it solves a problem. But the real question is: solves a problem for whom?”
The big idea here is that self-driving cars aren’t the problem. It’s not their fault that they crash into police cars, manifest weird flocking behavior, and go on strike. What autonomous vehicles need, the thinking goes, are smarter roads. Our streets are designed for human drivers. Hajbabaie wants to rewire the traffic infrastructure to control the flow of both people and robots. “The end goal,” he says, “is to bring connected self-driving-car technology into traffic-control systems, to make intersections safer and more environmentally friendly.”
Lots of traffic lights and their timing are already computer-controlled. But in Hajbabaie’s futuristic vision, cars — especially the self-driving ones — would wirelessly communicate with those computers, with one another, and maybe even with people’s cellphones. That’s known as “Vehicle-to-Everything,” or V2X, and it’s the key to Hajbabaie’s plan. His “white-phase intersections” would trigger robot cars to take the lead. They would enter the intersections first, updating one another on their trajectories and altering course as needed to avoid collisions. Human drivers would trail behind in what traffic-flow theorists call “platoons,” comforted that the white traffic light was telling everyone — and everything — when and where to move. And if things start to go off the rails, whether it’s the humans or the robots messing up, the whole thing just defaults back to red light/green light.
In simulations, the benefits of white lights kick in when as few as 10% of the cars on the road are robots. (Hajbabaie and his team don’t just create digital models — they also have 15 toy-size cars, each about a foot long and equipped with cameras and lidar, that they can put through white-phase paces.) “In our tests, we observe 3% or so reduction in travel time,” Hajbabaie says. As more self-driving cars enter the market, speed through Hajbabaie’s intersections goes up and fuel consumption goes down. Greenlight the white light, and we’re merging onto the road to the future.
Red/green vs. blue/orange
Maybe — but it’s not quite that simple. The elaborate regulatory system that has risen up around cars is a product of societal priorities and policies. Green lights and speed bumps — all the furniture of our built environment — are as much politics and culture as design and tech. That’s why it can feel impossible to build a protected bike lane on a popular shopping street or permit a parklet outside a restaurant, much less build a new trolley line or a high-speed rail system. A radically redesigned traffic light? Good luck getting that approved by your city transportation department before automobiles are replaced by flying cars.
The color of Hajbabaie’s robot light, to be clear, is arbitrary. He doesn’t actually care whether it’s white. That’s just for convenience in his team’s papers. “We want to let human-driven vehicles know that self-driving cars are controlling the intersection,” he says. “If it’s hard for someone to distinguish between colors, we don’t want to put a color there that’s problematic.”
In fact, red means stop and green means go only because people foisted those meanings on them. As the historian Clay McShane wrote, early traffic-light engineers chose red and green because railroad signals used them. Railroads got the idea from ships; coastal lighthouses used red (the most transparent stained-glass color in 1806) so mariners could tell lighthouses apart from the sea, and green was the color the British admiralty decreed for the starboard side of ships in the 1850s. The thing is, about 8% of men of European descent and 5% of men of Asian descent have trouble distinguishing red and green. (The incidence is lower in women.) But in 1923, when a traffic engineer proposed switching the 500 or so traffic lights in the US to blue and yellow, his colleagues nixed the idea. They thought drivers wouldn’t be able to make the switch, and worried that acknowledging the problem would make them look dumb.
Maybe a truly multimodal intersection could have as many colors as an LED screen can manage, one for every kind of thing that might pass through it — with an algorithm making way for drones, delivery trucks, robot cars, people using wheelchairs, unaccompanied 6-year-olds, ducklings, whatever.
A few hyperlocal radicals tried to defy traffic standardization. In Manhattan, Broadway had the familiar red/green setup, but Fifth Avenue used orange lights to mean go, green lights for the cross streets to go, and red to mean caution. No problem, except that Fifth and Broadway intersect at Madison Square. There, McShane wrote, “chaos would eventually ensue.” Fifth Avenue, like the rest of the world, soon yielded to the red/green hegemony.
What ultimately shaped all the new traffic rules, as with every question of public policy, was power and privilege. “The way our roads are built — the things that changed the 1890s city to the kind of city we have today — a lot of that came out of conflict between the rights and responsibilities of different kinds of road users,” says Cameron Roberts, a sustainability and transportation researcher at the University of Wisconsin. The people who could afford cars were a rich and powerful constituency. They eventually got what they wanted: speed and “freedom.” People build roads, new tech fills them up, and then the road builders have to figure out how to respond. That’s how we wound up redesigning cities that favored cars and suburban sprawl. In the battle between the Model T and pedestrians, the pedestrians never stood a chance.
The same will be true of proposals like white traffic lights for self-driving cars. “If we’re talking about this new kind of infrastructure, a fourth phase of light, you might say it solves a problem,” Roberts says. “But the real question is: solves a problem for whom? And what interests and power do they have to motivate that in practice?” If self-driving cars ever move beyond shared services like taxis and deliveries, the first vehicles are going to be owned by wealthy people. Should they get a privileged lane, or a special traffic light devoted to their convenience?
“There’s going to be a certain social set using these things, and they will be much like the earliest motorists — powerful and well-connected,” Roberts says. “I would not have a lot of confidence that something like this would be implemented in a way that is considerate of the needs of pedestrians, cyclists, and public-transit users.”
The world of tomorrow
For the moment, cities are in no danger of being overrun by self-driving cars. The technology seems to have stalled. Investor reports still quote bullish numbers for the eventual size of the autonomous-vehicle market, but the Google-derived robot-car company Waymo, once valued at about $200 billion, now idles at about $30 billion — in line with its competitor Cruise. “I would love a self-driving car, even for double the price of my car,” says Greg Shill, a law professor at the University of Iowa who specializes in transportation. “But the tech is just not there. These things don’t exist, and you can’t even persuade investors that they’re going to exist soon.”
Yet even as a hypothetical, there’s something delightfully science-fictional about the idea of slapping a fourth light on traffic signals to make way for robots. (I think this actually happens on the reliably predictive TV show “Futurama,” but I can’t find the episode.) It’s the kind of thing you’d see on a planet covered by a single giant city, swarming with flying cars. Maybe a truly multimodal intersection could have as many colors as an LED screen can manage, one for every kind of thing that might pass through it — with an algorithm making way for drones, delivery trucks, robot cars, people using wheelchairs, unaccompanied 6-year-olds, ducklings, whatever.
As a thought experiment, the idea of a white-light intersection also forces us to reconsider the automotive infrastructure we take for granted. Our traffic regime is an old one, originally designed for horses and carriages, not for F-150s or self-driving vehicles. In recent years, cities have begun to change their car-centric ways, carving up streets and changing the rules of the road to favor bikes and pedestrians and public transportation over gas-guzzling, climate-warming, street-clogging vehicles. Do we really want to reverse all that progress and re-reshape our surroundings for robots?
Hajbabaie knows his sci-fi-sounding proposal is up for debate. It assumes the communication among cars will be nearly perfect and instantaneous, which will ring false for anyone who has tried to get their printer connected to their WiFi. V2X, the network of vehicle interconnectivity it would depend on, has never really worked. And Hajbabaie’s team has yet to publish its findings on what happens when they introduce pedestrians into their simulations. (Yikes.)
“We are engineers,” he says. “All we can do is do a lot of education and outreach, show what kinds of changes people can expect to see. Show the benefits, show the drawbacks, be honest, and let them decide.” It’s not just a matter of how well the new technology performs. White traffic lights signal more than whether to stop or go — they also signal our values, and illuminate the kind of world we want to build.
Adam Rogers is a senior correspondent at Insider.
