Boston Uses AI to Improve Stop-and-Go Traffic

Let’s face it. AI isn’t as cool or helpful as we thought it would be. Instead, AI has become a bit of a nuisance. Initially, we were worried about job losses. But how many of you now add “Reddit” to the end of your search engine entries? Or get incorrect orders at the drive-thru from humans and machines? Or are even worse at video games? Boston, however, seems to have cracked the AI code, as traffic congestion and emissions in the city have decreased since the machine learning was brought in.

Using AI software, the city was able to analyze traffic patterns and traffic light timing data from hundreds of intersections within Boston. Engineers from the Boston Transportation Department then implement the software’s recommendations to optimize traffic flow.

According to the mayor’s office, traffic lights have been changed at four intersections in the Fenway-Kenmore, Mission Hill and Jamaica Plain neighborhoods. In a statement, city officials noted a 50 percent reduction in stop-and-go traffic at two prominent intersections: Huntington Avenue & Opera Place and Armory Street & Green Street. The former is a busy corridor that carries cars and trains and also crosses a university; the latter is farther away but near parks, businesses and a transit center.

Even though the area is home to renowned researchers from MIT and Boston Dynamics, this AI traffic project is a partnership with Google Research’s Green Light Project. Using AI and Google Maps, the system measures everything that could affect the smoothness of city driving, including waiting times at traffic lights, vehicle speeds, stop-start patterns, and more.

The research data is then used to create an optimization plan to improve traffic while reducing emissions. Green Light City partners around the world have experienced a 10% reduction in emissions due to reduced waiting and idling times.

Because Green Light is in the research phase, the service is free to partner cities without any additional software or hardware. Although it is a globally available transportation program, only two cities in the U.S. are currently research partners: Boston and Seattle.

In Boston, the impact is already being felt – and in a good way.

“It provides our traffic engineers with critical data to make sub-second signal adjustments, which can help reduce congestion on the corridor,” said Jascha Franklin-Hodge, Boston’s streets chief. (Yes, that’s a real title at Boston Public Works and the Department of Transportation.)

The city hopes the initial benefits will grow as it collects more traffic data and expands the program.

“It has a lot of personalization and real human analysis,” said Stacey Thompson, executive director of Liveable Streets Alliance. “But even just having that starting point helps us target in a much more effective way.”