How Google uses AI to reduce stop-and-go traffic on your route — and fight fuel emissions

Jul 29, 2024

[[read-time]] min read

Ari Marini

Keyword contributor

At nan commencement of 2020, a squad wrong Google Research was asked to research caller ideas for investigation projects that focused connected accelerating ambiance mitigation. “We were looking into each kinds of large ideas, from cultivated nutrient to power to aerial pollution,” Dotan Emanuel, a package technologist connected nan team, says.

At nan meal array 1 evening, Dotan shared immoderate of those large ideas pinch his family — and nan speech soon pivoted to a vexation acquainted to galore of us: “My woman Osnat said, ‘Why don’t you do thing astir postulation lights? We guidelines astatine them for nary bully reason,’” he recalls.

Road proscription is responsible for important world and municipality greenhouse state emissions. It’s particularly problematic astatine metropolis intersections wherever contamination tin beryllium 29 times higher than connected unfastened roads, and astir half of these emissions travel from postulation accelerating aft stopping. With millions of postulation lights crossed nan world, nan standard of nan problem was immense — and if Google could do thing to reside it, truthful was nan opportunity.

“My first thought was that we can’t do thing astir postulation lights,” Dotan says. “But erstwhile it comes to research, nan astir fascinating challenges dishonesty successful nan unknown.”

With their curiosity sufficiently piqued, Dotan and his squad dug into nan mechanics of postulation engineering. They recovered that while immoderate magnitude of stop-and-go postulation is unavoidable, a information tin beryllium prevented by optimizing postulation ray timing. To do that, cities traditionally needed to either instal costly hardware aliases tally time-consuming manual conveyance counts, neither of which supply complete accusation connected cardinal parameters they need.

“We quickly understood we person a beardown advantage that cities could use from — complete a decade of Google Maps driving trends from crossed nan globe,” Dotan says. “And a fewer weeks later, we had a task connection ready.”

That connection was for Project Green Light, an inaugural that uses AI to make recommendations for metropolis engineers to optimize existing postulation lights and trim stop-and-go emissions. After evaluating dozens of different awesome ideas, Green Light was chosen for its simplicity, scalability and imaginable for impact.

The Green Light squad utilized Google Maps’ driving trends to create an AI exemplary that measures really postulation flows done an intersection, including patterns of starting and stopping, mean hold times astatine a postulation light, and coordination betwixt adjacent intersections. The exemplary identifies imaginable improvements, for illustration shaving disconnected respective seconds from a reddish postulation ray during off-peak hours aliases an opportunity to coordinate betwixt intersections that aren’t yet synced. The city’s engineers past reappraisal those recommendations and tin instrumentality them successful arsenic small arsenic 5 minutes, utilizing their city's existing infrastructure.

“In bid to execute a affirmative ambiance impact, we want to beryllium capable to deploy high-quality Green Light recommendations to galore cities globally and standard fast. So we purposely group up everything to beryllium elemental and lightweight — cities don’t request to put successful immoderate dedicated package aliases hardware integrations,” says Green Light Program Manager Alon Harris. “We conscionable stock our recommendations pinch nan city, and past they measure them and return action.”

Since their first aviator successful 2021, nan squad has tested much and much intersections, developed much meticulous predictions and took Green Light connected nan roadworthy to much than a twelve cities crossed nan world, including Rio de Janeiro, Seattle, Bengaluru, and astir recently, Boston. The squad besides developed a broad dashboard to easy stock recommendations and analytics pinch partner cities, while continuing to show for immoderate caller needed changes.

“We connection each metropolis dedicated reports pinch tangible effect metrics, specified arsenic really galore stops drivers saved astatine an intersection complete time. We deliberation that’s going to beryllium a existent inducement to not conscionable instrumentality nan first recommendations, but besides bring Green Light to much intersections,” Alon says.

Today, Green Light is unrecorded successful complete 70 intersections, helping to prevention substance and little emissions for up to 30 cardinal car rides monthly. Early numbers bespeak nan imaginable to trim stops by up to 30% and trim emissions astatine intersections by up to 10%.

The squad is moving to standard Green Light to hundreds of cities and tens of thousands of intersections successful nan adjacent fewer years. And connected that roadworthy to making nan world a small better, they dream helping group acquisition little stop-and-go postulation will dispersed immoderate joyousness on nan way.

“You cognize erstwhile you thrust done a series of 5 greenish lights, and it feels for illustration your fortunate day?” Dotan says. “We want to make that emotion much communal to much group astir nan world.”