Directing Traffic

Beginning in 2021, researchers at Google began piloting an exercise intended to reduce urban traffic congestion called Project Green Light. The premise was fairly simple–the project uses data from Google Maps’ driving trends to create an AI model that measures how traffic flows through an intersection, including patterns of starting and stopping, average wait times at a traffic light, and coordination between adjacent intersections. By changing the timing of those lights, the theory went, congestion could be significantly reduced.

Its potential is intriguing; according to the latest Google publication on it, early numbers indicate it can reduce stops by up to 30% and reduce emissions at intersections by up to 10%. Yet its existing scope is quite tiny (70 intersections in just over a dozen cities worldwide), and the specifics of the way it works are hard to come by (Google did not respond to requests from Communications for comment).

What Project Green Light has done, however, is to offer entree to a robust, if nascent, ecosystem of academic researchers, public sector transportation agencies, and commercial traffic optimization vendors, all working on ways to reduce the time both private and public transportation vehicles spend idling at stoplights and the resulting by-products of pollution and frustration. The emergence of Green Light may finally be what is needed in the wider scope of policy and technology to bring the traffic light out from under a bushel, if you will.

Steve Remias, head of product strategy for signal analytics at Kirkland, WA-based smart transportation data vendor INRIX, said Green Light brings two specific attributes to public knowledge that were more or less missing prior to its launch.

“One piece is ‘well, it’s Google’,’’ Remias told Communications. “Anything Google does is bigger, better, better marketed, and better recognized. I think people are interested in the sense that Google has a data lake that is gigantic, so if they shifted a fraction of 1% of their focus to something, they could absolutely crush it. And I think the other piece Google brought was a little bit more of the ‘we’ll solve your problems for you’ attitude.”

However, Remias said anybody interested in optimizing movement in any municipality needs to take much more into account than merely changing red/green timing on traffic lights to let more cars go through in any given cycle.

“Cities know their traffic and traffic signals and geometries, and the context of everything that is going on around them, better than anyone,” he said. “Google is coming in and saying, ‘we can provide you the answer, the easy button’. I don’t know if I fully agree with that. Certainly, they can help. There’s no doubt they can help. But you need to include the context of things like how many bicycles are on a road, or if there’s a school on the corner with heavy pedestrian traffic. The Number One objective is not always to minimize the lights at intersections. It can be an objective, but it’s not always the Number One objective.”

To illustrate how traffic flow priorities shift, Austin, TX, implemented a pedestrian-friendly scheme for traffic control after special events such as Major League Soccer games and the Austin City Limits music festival. By incorporating vehicular and pedestrian data and adjusting signal timing for special events, the city saved an aggregate 50,400 hours of delay, adding up to $1.76 million in savings.

Remias said creating coherent strategies around all these disparate sources, and developing communications capabilities between signals and transportation agency control rooms, is still a very open proposition.

“You often hear the term ‘rich ecosystem’,” he said. “There are a lot of penetration rates thrown around—‘We have 30% of vehicles’—well, what’s the breakdown? Is it cellphone data? Is it apps? Vehicle telematics?

“Also , how frequently are you getting the data? What is its quality? We at INRIX have a substantial amount of data but, using signal timing as an example, if a truck is giving us a waypoint from their in-vehicle telematics every three minutes, it’s hard to figure out delay at an individual signal.”

Once a delay is suspected and confirmed, it still usually necessitates a visit from an engineer to the signal’s control box to re-time the signal, and those recalibrations usually don’t happen more than every three to five years. It’s a costly, time-consuming norm likely to remain for quite some time.

“This type of research is still in the experimental phase,” smart traffic technology pioneer Henry Liu said. “Most the implementations are on a small scale. There is no technology barrier. All the technology has been resolved.  A lot of this is, we need to try to gain the confidence of the public agencies. The only barrier right now is on deployment.”

Trajectory data is key

Liu, who pioneered the use of digital traffic flow improvement technology almost 20 years ago while at the University of Minnesota, recently used onboard data from General Motors vehicles to demonstrate a vehicle trajectory analysis scheme called a probabalistic time-space diagram. It decreased the delay and number of stops at signalized intersections in Oakland County, MI, by up to 20% and 30%, respectively, using only about 6% of the vehicles that approached those intersections.

As vehicle manufacturers add more connected capabilities to their vehicles, integral sensors could make the adoption of technologies similar in principle to Green Light quicker and more pervasive, without having to rely on motorists’ mobile devices.

Liu and his colleagues studied almost half as many signals (34) in one city as Green Light has mapped worldwide, which demonstrates improved traffic flow on a municipality-wide basis, but Liu told Communications the technology is not location-specific. It can be used anywhere as long as connected vehicles can communicate their trajectories. However, he hopes his team’s next project will improve on the recently published one by allowing for real-time adjustment.

“For the implementation we did with Oakland County, those lights are not centrally connected,” Liu said. “We run our optimization algorithm on our own cloud, generate all the parameters, hand them to the traffic engineers, and they have to go out into the field to make manual adjustments.

“However, after that success we got a lot of interest from other agencies. The Federal Highway Administration just gave us a new project, once again with Oakland County and GM. We are going to do real-time optimization with lights that are connected to control operations. We get data from GM from their cloud, we compute the parameters, optimize them, and send them directly to the county control center, where they are distributed automatically to the light. This is really important for extreme congestion situations.”

Liu said being able to combine control room knowledge of signal status and vehicle trajectory will be an immense boon to motorists and controllers alike.

“Trajectory data is very good in those situations,” he said. “If you are at the intersection and the light is green but you can’t move, we’ll know it’s blocked downstream, and it’s easy for us to identify the situation and it’s easy for us to optimize.”

Combining public and private sector data

Liu’s observation that public agencies need to trust the technology may come to fruition sooner rather than later; in many instances, they already have tracking and communications technologies embedded on their emergency and transit vehicles, and enhancing public safety and delivering cost-effective transportation are high on their task lists.

At least one vendor is already deploying smart traffic solutions for these agencies: Santa Clara, CA-based LYT has thus far concentrated on optimizing public transit, emergency, and snow plow vehicles.

In one example, LYT’s cloud-based, AI-powered transit signal priority (TSP) system utilizes preexisting bus-fleet tracking sensors and city communication networks to dynamically adjust the phase and timing of traffic signals to prioritize their journeys while minimally impacting cross traffic. The Santa Clara Valley Transportation Authority (VTA), in partnership with the City of San José, piloted the company’s  LYT.speed system on 17 intersections in East San Jose from July through December 2019. The technology’s effect on VTA’s route 77 was an overall 18% to 20% travel time improvement, equating to a five- to six-minute reduction in signal delay.

LYT CEO Tim Menard said his company’s technology takes advantage of what might be called the realities of density in transportation infrastructure. “We are living in a time when developed countries are really developed,” he said. “America is finally hitting situations Europe learned a while ago. You can’t build more roads. The houses are there and the infrastructure is there.

“Everybody wants to solve traffic congestion; it’s one of the three things you can run to get elected on, but there is no silver bullet. So LYT came in to demonstrate how to get police, fire, ambulance, and buses and transit where they are going. If we want to get cars off the street, we have to make these competitive; we have to make public services a premier service again.”

LYT might also be in the vanguard on demonstrating how various vendors’ technologies will have to co-exist. It has agreements with Seattle’s transportation department, as well as that of surrounding King County, and one of Google Green Light’s test beds is in Seattle.

“We definitely have talked to Google,” Menard said. While he couldn’t mention specifics yet, he said, “It’s coming together.” (Seattle transportation department officials did not comment.)

INRIX’s Remias said the incentives for multiple levels of government agencies are just too well aligned to let smart traffic optimization lag too long.

“It’s not the Wild West, but it certainly is not as seamless as it could be,” he said. “I know some cities have gotten grants for ‘dashboard fatigue’ for their traffic engineers; they don’t want to have to look at 10 dashboards. I think things will continue to be evolving, and over the next five to 10 years you’ll see more people asking for this.”

Gregory Goth is an Oakville, CT-based writer who specializes in science and technology.