Can AI help reduce traffic congestion in Boston? The city is partnering with Google to find out. – The Boston Globe

“Traffic might be the biggest headache that you have to deal with every single day,” Boston Mayor Michelle Wu told reporters last Thursday, touting Google’s technological promise to target small traffic hotspots. “We know that even small tweaks can go a long way.”

Project Green Light uses AI and Google Maps’ driving trends to model traffic patterns and also make signal timing recommendations for city traffic engineers to implement, according to the company’s website. For the past five months, the Google team has been analyzing traffic at hundreds of intersections around Boston and providing suggestions for optimizing traffic signals and patterns to minimize time stopped unnecessarily at red lights.

While transportation planners warn that Google’s technology is not a panacea, the technology offers the promise to quickly, albeit often modestly, reduce preventable traffic snarls.

Since the partnership began, Boston has already implemented Google’s suggestions by changing signal timing at intersections in Fenway-Kenmore, Mission Hill, and Jamaica Plain. Once the changes are made, such as keeping a light green in one direction for longer, the Green Light team then analyzes the resulting impact on traffic and provides the city with that data.

Intersections where changes have been made have quickly seen improvements.

At Huntington Ave. & Opera Place and at Amory Street & Green Street, “stop-and-go traffic has been reduced by over 50 percent,” according to the city. Wu touted the statistics as well, saying that the use of the technology to combat congestion “is one piece of something that we know to be a bigger part of the solution.”

In some cases, she added, that solution can be quite simple. “Sometimes it’s just a matter of how long a particular light stays green going one direction in the intersection versus the cross street,” Wu said.

Worldwide, 13 other cities on four continents are using it, including Rio de Janeiro in Brazil, Manchester in England, and Hamburg in Germany. Google is not currently charging its partner cities to use Green Light, and the program does not require cities to purchase hardware.

Early numbers from Google’s analysis of traffic patterns before and after recommended changes were made to traffic signals during tests conducted in 2022 and 2023 indicate a “potential for up to 30 percent reduction in stops,” according to the company. Google says cities using the technology have also seen, on average, a 10 percent reduction in tailpipe emissions at intersections.

While the city’s transportation department continuously monitors traffic with cameras at intersections, enabling it to respond when traffic snags arise, Wu said manual adjustments can’t fully address the big-picture problem.

“In order for us to really think about traffic and fixing it across the entire city, we can’t just go like by light by light and do it,” she said. With the data from Google, “we can then go in and really have a tailored approach … in a much more effective way.”

The technology can be “incredibly effective,” said Stacy Thompson, the executive director of LivableStreets Alliance, a transit advocacy group based in Cambridge.

At intersections where traffic patterns are changing throughout the day, the traffic lights can be taught “how to respond in more real-time,” she said. Or, she added, the technology can help often backed-up intersections cycle traffic through faster.

But it is not a “one-size-fits-all solution” for tackling congestion, Thompson stressed. Problems arise when the technology only focuses on cars, and intersections in the city need to “work for everyone,” including bikers and pedestrians, she said.

“I would actually love to see an expansion of this program that includes things like queue jumps [when a bus gets priority at an intersection] … things that really are also monitoring pedestrian flow,” Thompson said. “And, of course, optimization for the increasing number of bike lanes and bike signals.”

“All need to fit under this smart signals technology,” said Thompson, who added that the city should be transparent about where and how it is using the technology.

Ultimately, these are the goals, said Michael Lawrence Evans, Boston’s director of emerging technology. But some of that technology does not yet exist, and true adaptive signaling is expensive and requires a lot of hardware and maintenance.

“A platform like Green Light was a pretty low barrier way for us to try more frequent signal timing changes based on fresh data,” he said.

Having that new data “to validate the impact of the interventions is really helpful,” said Santi Garces, Boston’s chief innovation officer, but such tools are “not substitutes for this comprehensive policy approach to building better roads.”

Wu also acknowledged that the AI-powered experiment will not solve Boston’s traffic issues.

Her administration has prioritized making buses more reliable, reducing congestion by ramping up enforcement of double parking, and improving street safety and access, such as by expanding the city’s bike lane network. Artificial intelligence is one more part of the arsenal. The big-picture focus is “on trying to make sure that this is as convenient as possible to get around,” Wu said.

Shannon Larson can be reached at shannon.larson@globe.com. Follow her @shannonlarson98.