Boston will test AI traffic light system meant to improve BPS bus times

“The less time kids spend on the bus, the more time they can spend either at school or home, learning or resting so they can maximize their learning,” the statement said.

Using AI systems to prioritize public transit buses through intersections has successfully trimmed minutes off of bus trips in other places where it has been deployed, including a regional transit service in Portland, Ore. AI was also tested last year by the MBTA and the City of Boston to shorten travel time for two heavily used public transit bus routes along Brighton Avenue.

This appears to be the first time a city the size of Boston has used AI to help cut the travel time of public school buses, according to the city and experts contacted by the Globe.

The test, which would be conducted during the coming school year, is being funded with part of a $50,000 grant from Bloomberg Philanthropies’ 2025 Mayors Challenge. Boston is among 50 cities worldwide that received the grant to spur government innovation. Bloomberg said 25 of those cities will receive $1 million in January to help support their projects.

Peter Furth, a professor at Northeastern University’s College of Engineering whose research includes traffic signal control, said if the system is successful, it could shave a few minutes off each school bus trip. And those savings add up quickly, he said.

“Getting rid of these big delays can improve reliability,” Furth said. “The buses in their first round are more likely to end on time, so that means they can start the second round on time.”

The similar program in Portland slashed about six minutes from the total travel time of a nine-mile bus route that serves about 15,000 people daily, according to Peter Koonce, a transportation official for the city of Portland.

“It’s really trying to reduce the costs of services for people,” Koonce said in an interview. “It can be a win-win if done with the right amount of management and control of the technology.”

Transportation experts like Furth and Stacy Thompson, a former executive director of Livable Streets Alliance, cautioned that the technology is not a silver bullet for Boston’s long-standing school transportation issues. An AI system on its own can’t overcome a traffic jam stretching for several blocks, or a delivery van parked in a travel lane, for example, they said.

The technology is one tool, Thompson said, that can be used in concert with other steps, like creating more dedicated bus lanes, and redesigning schedules around school start and end times to help ease pressure on the fleet.

“It’s a tool in the toolkit to solve the problem,” she said. “It’s not a panacea.”

It’s not the first time Boston has tried to improve the on-time rate of its school bus fleet, which currently transports more than 20,000 students on 640 buses. The district’s fleet makes about 3,000 runs each day that includes more than 10,000 stops.

In 2017, BPS tried a computer model, designed by MIT researchers and intended to improve the arrival time of buses, but less than half of buses showed up on time for the first day of school. The following year, more than half of school buses were late for the first day.

A school district improvement plan, brokered by the state and Mayor Michelle Wu in 2022, required BPS to dramatically improve its bus service to 95 percent within three years. By the end of the agreement, which expired on June 30, the fleet had nearly hit the target, reaching 94 percent.

With the automated traffic signal system, the goal would be to reach that 95 percent figure, the officials said in their statement. The city and BPS said they were “inspired” by the MBTA’s transit signal project along Brighton Avenue.

In the MBTA test, buses equipped with GPS devices transmit real-time location data to the city’s traffic management center. That data inform MBTA’s vendor, LYT, which uses machine-learning technology to identify the bus, predict its arrival time, and determine when it’d need a green light to safely pass through an intersection.

The MBTA reduced the travel time of two of its most heavily used bus routes — the 57 and 66 buses — which combined transport about 15,000 people daily. Starting in July 2024, the agency used the technology to activate green lights along Brighton Avenue when those buses approached intersections at Allston Street, Harvard Avenue, and Linden Street.

During the last half of 2024, the agency said those buses shaved a combined 110 minutes from their travel time each day, and reduced the amount of time waiting at red lights by one-fifth, according to an agency spokesperson.

Boston officials will choose a few critical intersections which see frequent delays, and will aim for an 8 percent reduction in travel time of school bus trips most impacted by congestion, the statement said.

While the automated system would make adjustments to existing signal timing, safety constraints would be imposed by city engineers. Humans can override the system at any time.

For parent Haval Mohamed, who has two children at the Dever Elementary School, news that Boston would seek another way to improve bus service made him optimistic. He believes the traffic signal technology would benefit both BPS families and the city at large.

“We are enhancing the traffic management flow for the entire city,” Mohamed said.

Follow him on Bluesky at iamjohnhilliard.bsky.social. He can also be reached on Signal at john_hilliard.70 or email him at [email protected].