The first stoplight was a complete disaster.
Installed in London in 1868, the gas-powered traffic signal exploded, killing the policeman who was responsible for operating it. In the 150 years since, the stoplight evolved from a dangerous novelty to a crucial part of our world. Today, there are some 12,460 traffic signals in New York City alone. Still, some urban planners and scientists say the stoplight is nearing the end of its useful life as urban infrastructure—and that its design flaws waste millions of hours of our time unnecessarily.
This week, a team from MIT published a study in the journal PLoS One examining a radical proposal: Get rid of the stoplights completely. Led by Senseable City Lab‘s Carlo Ratti and Paolo Santi of the Ambient Mobility Lab, the paper proposes something called a “slot-based” intersection, or SI, where cars and infrastructure communicate through an algorithm that choreographs a graceful dance of vehicle platoons through an intersection.
It sounds like madness at first glance, but slot-based network design has already populated other industries. A great example comes from airlines, as Ratti and Santi point out. Take Southwest—instead of letting people line up all at once to board a plane, the airline divides people into six batches, each of which boards at an explicit time.
The process may irk passengers—especially middle-seat suckers—but it’s actually among the fastest ways to board a plane. It’s a form of “slot-based” scheduling, which is already in use everywhere from air traffic control to business management. The basic idea is that actors in a system are grouped into batches, and the speed of their movement is carefully controlled to move them more efficiently through a space.
It’s known as the “slower-is-faster effect,” as Santi and Ratti explain, pointing out that while slot-based design has appeared in other industries, it hasn’t arrived in traffic design yet. They think that time has come, and in their paper, they demonstrate how it could double the efficiency of intersections and cut delays “to almost zero.”
What would their system look like in practice? As Santi explains, a slot-based intersection would rely on two things. First, a sensor-equipped car would need to communicate its trajectory—right, left, or straight—to a central algorithm controlling the intersection, which would group it into a “batch” of other cars going in the same direction. Second, the central software system would need to be able to control the speed of each platoon, using cruise control-style software that already exists in most cars, to limit the speed of your car as it moved through the intersection.
While the MIT team says this system would be at its most efficient with completely autonomous cars, it’s feasible with today’s auto tech, too. “In terms of what kind of technology you would need, you won’t need to wait 20 years,” Santi says. “We don’t need autonomous driving. It’s actually much simpler.”
While the paper doesn’t delve deeply into the technical details of the system, the hardware sounds similar to the Connective Vehicle Pilot Program currently being tested by the Department of Transportation. It could work something like this: Imagine you’re driving home, heading into the busy intersection near your house that’s always snarled with traffic. When you signal a left turn, your car’s onboard computer sends a ping to the central network controlling the intersection. The software responds with the optimal speed for your car to travel based on your trajectory—and your car’s system sets a limit on how fast you can go based on that number, not unlike cruise control. You join a platoon of other cars turning left, and this batch of cars gently arcs through the intersection, while other platoons heading different directions either speed up or slow down to accommodate it.
The problem isn’t that stoplights are poorly designed—it’s that there’s a very tangible limit to how efficient they can be. It’s all about the yellow light.
The earliest stoplights didn’t have yellow, because they didn’t need it: These primitive systems were controlled a police officer sitting in a booth nearby. In the 1920s, a Detroit policeman named William Potts came up with the idea of an “amber” light that would alert drivers that the light was about to change. In turn, this eliminated the need for a human operator because it communicated everything a driver needed to know automatically.
It was a brilliant design, but it’s also subject to diminishing returns. You can increase the number of light changes at an intersection and greatly speed up the flow of traffic, the authors write. But with each change, you add another “setup” phase, aka a yellow light, which decreases overall efficiency at the intersection.
What Santi and Ratti are proposing is a super-intelligent piece of software that could take the basic model of a stoplight—cycling between stop and go—and speed it up, decreasing the wasted time of yellow lights and increasing the number of cars that can move through an intersection, even as it slows them down to better coordinate their movements. It would dissolve our idea of the stop-and-go traffic system by adapting to the needs of a network in real time, they explain. The computational analysis of their model revealed that it could make light-free intersections twice as efficient as conventional ones.
The biggest challenge of adopting a slot-based traffic design may not be with the statistical modeling or technology behind it—it may be with the human behavior sitting behind the wheel. “Of course, giving up control is something that’s very difficult for humans,” Santi says.
To tackle the momentous cultural shift that a slot-based traffic system would require us to make, he envisions an incremental testing strategy, equipping a small area of a city with the system, not unlike the way other researchers are testing similar ideas, such as this four-block-long part of Chicago that’s gone completely free of lights, sidewalks, and curbs. The concept could also be tested in vehicles that are already operating semi-autonomously within a closed area or campus. It could even be applied to bikes or walkers who are trying to cross busy intersections. For now, the MIT team is in talks with several cities to apply the idea, though they can’t reveal where.
“From a technological standpoint, there are no big hurdles to implement this idea,” Santi says. The real hurdle may be in the radical behavioral shift that will be required on the part of drivers. Just as the widespread adoption of autonomous cars has run up against unpredictable drivers, the biggest enemy of autonomous intersections may be humans.