Keith Barry leads home and car coverage at Reviewed.com, and is a regular contributor to Wired's Autopia blog. His work has been featured in several publications including Car and Driver, USA Today, and the Boston Globe. He frequently rides the 96 bus from his home in Medford, Massachusetts.
The humble traffic signal is gaining some new responsibilities.
Drive your Model T through a major intersection a hundred years ago, and you'd likely encounter a policeman directing traffic. Ten years later, that officer would've been replaced by a traffic light. As the number of cars on the road increased, cities needed a way to keep cars from crashing into each other. With its dogmatic procession of green, yellow, and red, the automated traffic light did the trick.
A few adjustments were made in the decades that followed, such as pedestrian-walk sensors, but the basic concept remained the same: Signals were for safety's sake. But today's traffic engineers are starting to give the humble signal new responsibilities—programming them to not only react to the flow of traffic, but also to predict driver behavior. The signals of the not-so-distant future may help cities cut congestion without adding lanes or building new roads.
A glimpse into this future begins in Utah, home to one of the most advanced state-wide traffic signal systems in the United States. Traffic planners have long used historical data to create signal plans that optimize "green time" to improve traffic flow, with sophisticated systems using different plans for different times or day or sensors to detect certain vehicles waiting at a light. In most places, these signal plans are revised every three to five years, but the Utah Department of Transportation (UDOT) can adjust nearly any signal in the state within 30 seconds.
The system is possible thanks to the state's superior infrastructure. Spurred by the 2002 Winter Olympics, Utah began connecting traffic signals across the state using a fiberoptic network installed in partnership with utility companies. Today, UDOT has data from over a thousand closed-circuit cameras, and can remotely control over 80 percent of the traffic lights in the state.
Mark Taylor, signal engineer for UDOT, clearly sees Utah as a template that other states should follow. "Our senior managers here understand the huge benefit-to-cost ratio that really good signal-timing gives," he said. "You just can't get that building roads or building other infrastructure." Indeed, he estimates that money spent on signaling can pay back at a ratio of 40 to 1—not surprising considering that waiting in traffic saps as much as $120 billion each year from the U.S. economy, mostly due to lost productivity.
Of course, regardless of the benefits, most cities and towns can't rely on a top-down, statewide signal interface, because it's expensive to build and labor-intensive to operate. That's where Stephen Smith comes in. A professor at the Robotics Institute at Carnegie Mellon University in Pittsburgh, he researches artificial intelligence, and is currently working with the city of Pittsburgh to install the first city-wide network of "smart" traffic signals.
Instead of using humans to monitor and react to traffic flow, the new signals use radar sensors and cameras to detect traffic, and sophisticated algorithms to instantly adjust signals based on real-time conditions. "Each intersection builds a plan that optimizes local traffic flow," says Smith. "Once it does that, it communicates its outflows to its downstream neighbors."
Smith's team installed nine smart signals back in 2012 and saw instant results. Travel times along the corridor with the new signals were reduced by 25 percent, idle time fell by 40 percent, and vehicle emissions dropped by 20 percent. The system is also scalable for cash-strapped cities, says Smith, because you can install the signals one intersection at a time as funding becomes available. By the end of this year, Pittsburgh will have 49 smart signals.
Artificial traffic management is an impressive feat, but the Holy Grail of signal planning is a light that stems congestion before it even starts. Carolina Osorio is working toward that goal at MIT. An assistant professor of civil and environmental engineering, Osorio creates elaborate computer models of traffic flows to predict how drivers will behave, including how they will react to changes in signal patterns.
For instance, common sense might lead some cities to create signal plans that prioritize traffic flow along main arteries. But Osorio says such plans might eventually lead to more congestion, as drivers flock to a newly decongested road through a process known as induced demand. "If I decrease travel times along an arterial, it might become an interesting path for people who didn't used to take it," she said.
Osorio's algorithms can also automatically program signals to react to particular issues that a city prioritizes—such as improving pedestrian flow, or preventing traffic from "spilling back" onto highway ramps. Whatever a city's priorities, Osorio's models create signal plans that achieve them. And thanks to apps and navigation systems that feature real-time traffic data, Osorio can now study how a driver might react when her navigation system alerts her to congestion ahead.
"You now have actual real-time maps of how traffic conditions are changing throughout the city," she says. "Hence, we're developing better understanding of what conditions you're reacting to."
Eventually, traffic signals won't have to predict traffic flow at all. Instead, they'll be able to simply ask cars where they're going, and change plans accordingly. So-called vehicle-to-vehicle (V2V) communication uses dedicated short-range communication technology that lets cars "talk" to each other and to traffic signals.
A V2V-compatible signal could relieve congestion by prioritizing emergency vehicles, large groups of concertgoers, or a behind-schedule bus that's loaded with riders. It could also let drivers know which roads have the most green lights. According to the Department of Transportation, V2V compatibility could be standard on all new cars as soon as 2017.
Audi is already testing prototype vehicles that inform drivers when lights are about to change. Called Traffic Light Assist, the system can even tell a driver to slow down or speed up to make a light, all in the name of reducing congestion and vehicle emissions. According to the German automaker, the technology may be commercialized soon.
In the (slightly) more distant future, when vehicles become fully autonomous, the familiar red, yellow, and green lights may even become a historical curiosity, as vehicles themselves decide when to stop and when to go. But for now, the humble traffic signal is holding it down as one of the best ways to reduce congestion.