After a Quiet Year, Congestion Is Revving Back Up. Are Connected Intersections the Answer to Our Traffic Nightmares?
“Every time you stop a car, bad things happen,” says Gordon Meth. “You’re delaying someone. There are gas emissions. There’s a potential for crashes and accidents. The more you make someone stop, the higher likelihood someone will ram into them.”
Few people understand the complexity of the urban intersection better than Meth, a civil engineer who teaches about traffic management at Rutgers University. Meth has designed more than 300 intersections in the northeast United States and optimized timing cycles for another 800 around the world. He got his start in civil engineering at the age of 21, when he was tasked with monitoring 1,600 intersections in Toronto, Canada. He remembers studying the vehicle flows through these intersections and marveling at how dramatically travel times were affected by variables such as accidents, pedestrians crossing, or left turns. Since then, he’s made a career out of limiting the bad things that can happen at intersections.
“Someone once told me that the only thing you need to know to be a traffic engineer is how many seconds there are in an hour: 3,600,” Meth says. “All the traffic signals in one network divvy up that number; you’re just assigning those seconds to different lights and deciding who gets green when.”
Of course, the real equations are far more sophisticated. In fact, intersections are no longer one-size-fits-all junctions designed to keep vehicles from smashing into one another. They are fast becoming living laboratories of the urban environment, a testbed for emerging technologies designed to speed up travel, draw down emissions, and increase the safety of everyone on the road.
Today, the best approaches to traffic management are intersections outfitted with smart-infrastructure technology that can detect real-time traffic patterns and anticipate upcoming delays, and then work to minimize waiting across the board. In the near-future, traffic-streamlining strategies likely will go one step further, incorporating artificial intelligence and data from signals and vehicles all over the city. Eventually these systems will be connected to fleets of autonomous vehicles, with vehicle-to-everything, or V2X, technology giving self-driving vehicles a “tap on the shoulder” to tell them about stalled traffic, unseen pedestrians, or cyclists crossing a roadway up ahead. Studies have already shown that such shoulder-taps can help lessen congestion and speed travel times across the urban grid.
Managing urban traffic—or in civil engineering terms, managing “conflict demand”—will only become more important in the next few decades. According to the Federal Highway Administration, more than 50 percent of combined fatal and injury crashes occur at or near intersections. And per a 2010 report from the National Highway Traffic Safety Administration, a stopped vehicle was a critical factor in 12.2% of intersection-related crashes. Even though traffic within cities has declined over the last year due to the COVID-19 pandemic, most experts expect it to surge back with the revving up of the world economy. A 2016 UN report predicted that metropolitan areas will become home to up to 60 percent of the world’s population by 2030, and the World Economic Forum has said the number of cars worldwide is expected to double by 2040. With all these people trying to get around, navigating intersections in a timely and safe manner will become critical for keeping cities working smoothly.
Managing the complexity
Rush hour. You’re start-stopping down a four-lane city boulevard cluttered with traffic. After whisking through a few green lights, you finally lose the traffic-light lottery and roll up to a red. At first, the intersection seems like any other in town: it’s surrounded by light posts, traffic signs, and the time-killing signals hovering above. But this particular intersection is much, much more.
That tiny disc attached to the bottom of the traffic light? It’s actually a camera recording images of everything that moves through the intersection. And that metal box next to the light pole on the sidewalk? That’s housing for a computer that’s processing data from the camera. The antenna at the top of the light post? It’s broadcasting traffic data out to all the other vehicles equipped with the requisite hardware to receive it.
This is the state of the art of smart infrastructure right now. And the technology is soon to be improved with integration to autonomous vehicle technology. Recently, Ford Motor Company kicked off a pilot program with the Michigan Department of Transportation, and the Ford subsidiary Quantum Signal AI, to install infrastructure nodes atop a number of traffic signals in the city of Saline, Michigan. Equipped with sensors like radar, lidar, and cameras, the nodes communicate with self-driving vehicles via Cellular V2X (C-V2X) signals to give the vehicles a clearer sense of their upcoming surroundings. This information, combined with the views detected by the vehicle’s own powerful sensors, allow a self-driving system to anticipate the unexpected, from a car approaching an intersection at an unsafe speed, to an unseen pedestrian blocked by a truck. “That’s pretty valuable information that could help improve safety for everyone involved,” says Tony Lockwood, Autonomous Vehicle Manager for Ford Motor Company.
Similarly, the city of Ann Arbor, Michigan will soon be home to more than 20 smart intersections as part of a University of Michigan initiative to demonstrate the safety potential of connected and automated vehicles. Intersections will be fitted with cameras, radar, and infrared sensors, and will send information about object behaviors and speeds to vehicles in the vicinity, triggering onboard warnings when cars are in dangerous situations. A predecessor program called Safety Pilot Model Deployment provided evidence that smart infrastructure has the potential to reduce unimpaired crashes by 90 percent.
Elsewhere, the Toyota Mobility Foundation (TMF) has teamed up with researchers from the Alan Turing Institute, the United Kingdom’s national institute for data science and artificial intelligence, to explore how traffic management systems can become more dynamic through machine learning. In addition to monitoring London traffic in real-time, these systems also use the data they obtain to drive simulations that help algorithms learn how to predict changes over time. “We learned that you can locally optimize something and it doesn’t always lead to globally good performance,” says Neil Walton, a fellow at the Turing Institute. “If you have the system serve the longest queue, it can lead to oscillating behavior and create more traffic. We need to keep looking at how we deal with these systemic issues.”
Some car manufacturers are turning to autonomous vehicles to unlock these problems. In 2017, engineers from Ford went to Milton Keynes, England, to trial a new experimental technology called Intersection Priority Management. This technology used a vehicle-to-vehicle communication protocol that allowed cars to share location, direction of travel, and speed. Proprietary software then analyzed the trajectory of other nearby vehicles (including those not in the pilot) and suggested the best speed for each of the participating vehicles to pass through the intersection without stopping.
Volkswagen followed suit in 2018, rolling out a system with the technical title Car2X WLANp that informs drivers of traffic light sequences as they approach each junction. The idea behind both of these systems was twofold: to improve traffic flow by preventing unnecessary braking or accelerating around intersections, and to make intersections safer for all users.
The role of AI
Elsewhere, traffic management efforts are focused on leveraging AI to make traffic move faster and more safely through intersections. At Siemens Mobility, the Intelligent Transportation Systems (ITS) division is prioritizing data-driven, self-learning approaches to traffic management that process multiple data streams simultaneously. Sitraffic Motion MX, a new Siemens technology rooted in AI, promises an adaptive network that dynamically coordinates green lights in a street network. During a recent pilot project in Germany, using this technology across 24 intersections on a 6-kilometer stretch of road led to smoother traffic flows, fewer emission levels, and less noise pollution. A similar pilot with electronic bikes in Portugal yielded similar results.
The transportation analytics company INRIX has been working with the city of Austin to use cloud-based traffic-signal software to help identify problematic intersections. Using anonymous connected vehicle data, the company can track upwards of 7 percent of all vehicles on city streets, providing a large enough data set to identify traffic patterns on 33 different Austin corridors. The city uses INRIX’s Signal Analytics dashboard to identify intersections plagued by backups and “split-failures,” when numerous vehicles don’t have enough time to pass through a green light before it turns red. Traffic engineers then use the data to reconfigure the timing of signals, alleviating the backups before residents begin to complain.
In Pittsburgh, a company named Rapid Flow has built a smart distributed scheduling system so multiple intersections can talk to each other. A recent pilot program indicated that a distributed scheduling system can reduce congestion significantly. “When deployed right, our solution uses technology that gives visibility to all the overall traffic patterns,” says Rapid Flow CEO Griffin Schultz. “That information can be incredibly valuable.” In fact, a recent pilot of the technology demonstrated a reduction in travel times by 20 to 25 percent, and a drop in the number of stops and wait times by 30 to 40 percent. Another pilot involving self-driving cars has shown even further improvements, reducing travel times by an additional 20 percent.
In England, a startup named Vivacity Labs is engineering perhaps the most sophisticated trial yet. During a recent pilot program, Vivacity Labs set out to identify and classify all vehicles and road users in an attempt to give civil engineers and authorities insight into rush-hour patterns, the most commonly trafficked routes, and where parking spaces are likely to be available. The company installed more than 400 cameras at major intersections around the city of Milton Keynes. These cameras identified individual vehicles and charted how much time it took them to get around the city. Utilizing AI, the system extrapolated the data to predict traffic conditions 15 minutes in advance, with 89 percent accuracy. (The system also has built-in anonymization filters that work to obfuscate the identities of those it records; instead of seeing specific vehicles with specific license plates and pedestrians with faces, it records only shapes and motions.)
Mark Nicholson, Vivacity Labs founder and CEO, said his company could leverage this data to reroute traffic around congestion to maximize mobility. Eventually the technology could empower municipalities to redirect certain types of traffic to meet a variety of objectives, such as emission reduction or noise control.
The bottom line with all of these technologies: As AI enables cities to be more responsive to real-time ebbs and flows of traffic patterns, the more smoothly that traffic can flow through intersections across the board. Ultimately, this type of responsiveness could work to alleviate those nightmares for civil engineers. Whichever approach a city takes, the key to relieving urban congestion starts where most of us get stopped: At the intersection.