A city partnership with Google’s artificial intelligence powered Project Green Light has reduced stop-and-go traffic at the intersection of Huntington Avenue and Opera Place by more than 50%.

By using AI to find traffic patterns in Google Maps driving data, Green Light provides rapid recommendations for traffic signal timing to Boston Transportation Department (BTD) traffic engineers.

Due to the program’s success in the Fenway and at two other intersections in Roxbury and Dorchester, the city has announced it will be expanding the program to yet to be determined intersections.

“BTD traffic engineers will continue to assess [Green Light] recommendations for safety, feasibility and effectiveness to determine if they could be implemented for all signalized intersections,” a spokesperson for the mayor’s office said in an emailed statement.

Green Light has been providing signal recommendations for the Huntington Avenue and Opera Place since February, when the partnership began.

Boston, consistently ranked high in the U.S.’s worst traffic cities, is only the second in the U.S. to implement the AI-powered research project, following Seattle. The program has been operating internationally since 2021 and includes 13 cities around the world.

The U.S. Department of Transportation recommends that traffic signal timings be reviewed every three to five years but with Green Light, traffic signals can be updated to reflect moment to moment traffic needs in as little as 5 minutes, where a city has the tech and staffing capabilities.

“The biggest difference between Green Light and the regular operation of a traffic system is that usually traffic engineers need a way to tell how much traffic they have at their intersections,” Matteus Vervloet, Green Light’s product manager, said in an interview. “If we can make these decisions faster, we can help [city engineers] have the confidence to make these decisions.”

Typically, gathering the data needed to adjust signal lights has required expensive monitoring equipment or physical vehicle counts. At this time, in its research phase, the program is provided to the city at no cost.

Green Light uses data gathered from Google Maps to create an aggregated picture of driving trends. Google AI then analyzes the data for inefficiencies in signal light timing that create unnecessary stop-and-go traffic.

Inefficiencies are then flagged to city traffic engineers through recommendations on an online interface, such as whether to add additional green time or coordinate two intersections. Receiving only the recommendations and the results, Google Maps user data is protected from the city.

Green Light measures success by counting “split failures,” or how often a vehicle waiting at a red light is unable to continue through the intersection at the next green light. Due to the higher fuel cost of accelerating and stopping a vehicle repeatedly, air pollution can be 29 times higher at a city intersection than on open roads. Green Light reports an average reduction in emissions by 10% at monitored intersections.

So far, the intersections where Green Light has been implemented do not contain bike lane or trolley-specific traffic signals, as the program has only studied vehicular traffic.

“We currently focus on areas where we think there’s not going to be a negative impact on other modes of transportation,” Vervloet said. “The city also has their own plans that they can compare and review. That's why we’re still partnering with real human traffic engineers.”