A PILOT PROGRAM using artificial intelligence in a San Anselmo traffic light system may have reduced vehicle delays by more than 30% at a problem intersection, town officials said.

The intersection of Sir Francis Drake Boulevard, Red Hill Avenue and Center Boulevard, nicknamed The Hub, has been notoriously slow for decades.

The new system consists of adding some additional cameras and installing a small server inside the traffic signal box. The computer vision software uses continuous 3D perception to track the exact position and speed of every vehicle in real time and sends a message to the pre-existing traffic signal controller, adjusting the signal timing to keep traffic flowing efficiently.

“It’s like having a live human being look down the street, and instead of waiting for a timer to change, it looks down the street to see if there’s no cars, or cars coming faster or slower,” said San Anselmo public works director Sean Condry. 

“[The AI system] reduces, apparently, 90 hours a day for the vehicles that are traveling through there,” said Scott Schneider, San Anselmo’s assistant public works director. “That’s 15 seconds per person times 65,000 vehicles per day.” 

That’s the data from the company that created the system, Roundabout Technologies, Condry said, but the town is contracting with a consultant to compare new and old data and verify the company’s findings. 

“We have gotten many positive complaints that it works so well now that people can no longer sit there and text or do emails at the intersection,” Condry said. 

The old system would give everybody a set amount of time, whether there are more cars coming or not. The AI system can differentiate between cars and public buses, and the company will soon be installing additional features that detect pedestrians and cyclists.  

Schneider said the cameras do not make recordings, and they do not include face detection or read license plates. Two to four new cameras are required at each intersection, and a cellular connection allows each AI server to communicate with those at other intersections to coordinate operations. 

An animated modeling video illustrates how artificial intelligence is being used at the intersection of Sir Francis Drake Boulevard and Red Hill Drive in San Anselmo, Calif., in 2025. Instead of relying on basic sensors that detect a car’s presence within a narrow range, the company’s system uses continuous 3D perception to track the exact position and speed of every vehicle in real time. (Roundabout Technologies via Bay City News)

Condry said the system has been up and operating for about a month or two. They turned it on before the start of the school year. They have also installed another new system about a block away on Bank Street.

Ongoing monitoring costs are about $12,600 per year for a contract that covers 24/7 monitoring, support and updates. The cost of installing the cameras and wiring at The Hub cost about $30,000. The system at Bank Street will cost $10,000 for installation and $4,600 per year for monitoring and optimization.   

“They’re not as good as a roundabout,” said Schneider, adding that a roundabout with pedestrian flyover bridge could cost between $40 to $50 million. “A roundabout could probably dramatically decrease wait times and do even better, but this AI has proved to be remarkable in the improvements that we’ve noticed.” 

The town recently received a $250,000 grant from the Transportation Authority of Marin to install the system at the town’s other 10 traffic signals, along with one signal in the town of Fairfax. Those systems should be installed over the next six months, said Condry.

“We might be the first jurisdiction in California that has all of our intersections controlled by AI, but we only have 12 traffic signals,” he said. 

Ruth Dusseault is an investigative reporter and multimedia journalist focused on environment and energy. Her position is supported by the California local news fellowship, a statewide initiative spearheaded by UC Berkeley aimed at supporting local news platforms. While a student at UC Berkeley’s Graduate School of Journalism (c’23), Ruth developed stories about the social and environmental circumstances of contaminated watersheds around the Great Lakes, Mississippi River and Florida’s Lake Okeechobee. Her thesis explored rights of nature laws in small rural communities. She is a former assistant professor and artist in residence at Georgia Tech’s School of Architecture, and uses photography, film and digital storytelling to report on the engineered systems that undergird modern life.