How Google’s Green Light AI cuts waiting time at traffic lights

Google Research has come up with an AI system that will not only reduce greenhouse gas emissions from cars but also drop the average driver’s blood pressure: by slashing the time motorists spend waiting at traffic lights.

While traffic lights are a source of frustration for drivers – it’s thought the average motorist spends up to two days a year waiting for them to turn red – they’re also a significant source of carbon dioxide being released into the atmosphere. Last week, Google Research announced the first fruits of a project to make road travel smoother: an AI called Green Light.

In Europe, around 20% of CO2 emissions are related to road transport. And, according to University of Surrey research published in 2015, traffic lights are responsible for significant air pollution.

The research noted traffic intersections were “high pollution hotspots”. Because drivers often have to change their driving behaviour as they approach a traffic light – stopping, slowing down, or revving their engine – the concentration of polluting nanoparticles emitted by cars was found to be 29 times higher at intersections compared to within free-flowing traffic.

How Google Green Light AI works

Green Light is designed to optimise when traffic lights at nearby intersections are set to either stop or go, to create a better flow of traffic by reducing the number of times cars have to stop at traffic lights. Google claims Green Light cut stops at traffic lights by 30% – which translates into a decrease in greenhouse gas emissions of around 10%.

Green Light works by crunching traffic data from its own Maps app to predict when to set a traffic light to stop or go to improve traffic flow.

For example, it analyses data around traffic lights. Data such as how long a traffic light will show a green or red light, or whether a light coordinates with others around it. It also examines information on travel patterns around the lights, such as how long a car will wait at a red light on average.

The AI will then model how traffic light timings should be changed to reduce the number of times fewer drivers need to stop and start at lights, and then provide those recommendations to local traffic engineers.

Engineers can access the recommendations through a custom dashboard and, once changes have been made to a particular intersection, they can use it to view how the alterations have impacted the flow of traffic in the area.

The system was piloted in 2021 and is currently live across 70 intersections in countries including Brazil, the US, India, Israel and Germany.

Related reading

• How Google made AlphaFold 3
• Google Translate doubles language support thanks to AI
• Google upgrades NotebookLM with more sources and faster fact-checking