Step One
Traffic Breakthrough: How Google's AI (Project Green Light) is Erasing Gridlock and Cutting Emissions
Step One: Solving the Global Intersection Problem with AI
There are few things that are more frustrating in daily life than being stuck in traffic, but that frustration pales in comparison to the amount of damage sitting in traffic does to the environment. While slow moving traffic is frustrating, sitting at intersections generates up to 29 times more air pollution than highway traffic, and with over 60 million intersections worldwide, it adds up. The problem is most acute on the East and West coasts (Average of 53 intersections/km2 in these areas), where density is high, but the scale of urban congestion is a global crisis. while intersections take up only 2% of average trip time, they produce 25% of the pollution due to idling and accelerating, and from a congestion stand point waiting for turn opportunities, interrupted traffic flow and line creep add to congestion.
It’s a difficult problem and many intersections, either managed by stoplights or roundabouts, are poorly designed, interfere with traffic flow and rely on outdated, expensive or manually collected traffic data. However there is a rich, real-time data set available to combat this: Google Maps.
This data set has:
They then task the Ai to develop adjustments to the traffic light timing that they present to the city’s traffic engineers for approval. If approved, the changes can usually be implemented in a matter of minutes, at which point the system begins to measure the changes and compares them to the previous plan while monitoring current status to alert the engineers if further changes are necessary.
Here’s the amazing part…Google does not charge towns and cities for the service…its free and even an AI hallucination can be easily corrected or spotted by the local engineers even before it has been implemented!
Here's the cost analysis:
Quantifiable Environmental & Time Savings
So, why do we bring this up since it has little to do with our mandate consumer electronics.? We spend time trying to present a more realistic picture of AI and that means slogging through tons of AI hype which currently pervades the technology world. In order to present a more balanced view of AI, we also look for things that AI can actually do far better than we lowly humans. AIs are built to look for and remember patterns in all kinds of data, which almost perfectly matches the objective of optimizing traffic flow. Down the road we see a second step, one where all intersections are monitored in real time with sensors mounted on traffic lights. That data is fed back to an AI system that is able to retime every stoplight on the fly, based on real time data, optimizing actual traffic flow on a real-time basis. Step Three would be when the AI is able to send commands to autonomous vehicles to optimize traffic flow without human intervention. We are still a way away from Step two and Three, but they are certainly not out of the realm of possibility. Chalk one up for AI.
Step One: Solving the Global Intersection Problem with AI
There are few things that are more frustrating in daily life than being stuck in traffic, but that frustration pales in comparison to the amount of damage sitting in traffic does to the environment. While slow moving traffic is frustrating, sitting at intersections generates up to 29 times more air pollution than highway traffic, and with over 60 million intersections worldwide, it adds up. The problem is most acute on the East and West coasts (Average of 53 intersections/km2 in these areas), where density is high, but the scale of urban congestion is a global crisis. while intersections take up only 2% of average trip time, they produce 25% of the pollution due to idling and accelerating, and from a congestion stand point waiting for turn opportunities, interrupted traffic flow and line creep add to congestion.
It’s a difficult problem and many intersections, either managed by stoplights or roundabouts, are poorly designed, interfere with traffic flow and rely on outdated, expensive or manually collected traffic data. However there is a rich, real-time data set available to combat this: Google Maps.
This data set has:
- Over 1 billion mapped kilometers of roads.
- 10 million miles of street views from 250 countries and territories
- Adds over 20 million real time reports on road conditions each day.
- This data comes from speed and location data collected from all iOS and Android users who have the Google (GOOG) Maps application open on their phone while they are in their cars, along with government agency data, road sensor data, and traffic cam information.
They then task the Ai to develop adjustments to the traffic light timing that they present to the city’s traffic engineers for approval. If approved, the changes can usually be implemented in a matter of minutes, at which point the system begins to measure the changes and compares them to the previous plan while monitoring current status to alert the engineers if further changes are necessary.
Here’s the amazing part…Google does not charge towns and cities for the service…its free and even an AI hallucination can be easily corrected or spotted by the local engineers even before it has been implemented!
Here's the cost analysis:
- Traffic Studies $5,000 per intersection
- Infrastructure No new hardware is needed as all of the information is in Google Maps driving data
- Implementation Typically requires local engineers to manually implement complex changes
Quantifiable Environmental & Time Savings
- Reduction in stops 20% annual reduction in unnecessary stops across timed intersections. Some individual intersections have seen reductions between 33% and 34%.
- 13.5% average reduction in delays across newly timed intersections with some improving up to 24%
- An average of 4,000 gallons of fuel saved over the course of a year with the best performing intersection saving up to 14,0000 gallons/year
- Quebec City, Canada: The first Canadian city to launch Project Green Light, optimizing 13 key intersections. This initiative targets residents who spend an average 47 hours annually stuck in traffic. Early results show reduced stop-and-go traffic by up to 30% and a 10% cut in CO2 emissions at these intersections.
- Boston, USA: Project Green Light is live at 114 intersections, about 10% of the city's traffic signals. Boston, ranked eighth worldwide for traffic delays, uses AI recommendations to improve signal timing coordination across multiple intersections, cutting stop frequency and emissions.
- Kolkata, India: Declared an essential component of Kolkata Traffic Police’s efforts to reduce gridlock, Project Green Light optimizes busy intersections for safer and more efficient traffic management.
- Manchester, England: The system provides actionable insights across a network of 2,400 signals, enabling adjustments where there was previously no visibility into potential timing improvements.
So, why do we bring this up since it has little to do with our mandate consumer electronics.? We spend time trying to present a more realistic picture of AI and that means slogging through tons of AI hype which currently pervades the technology world. In order to present a more balanced view of AI, we also look for things that AI can actually do far better than we lowly humans. AIs are built to look for and remember patterns in all kinds of data, which almost perfectly matches the objective of optimizing traffic flow. Down the road we see a second step, one where all intersections are monitored in real time with sensors mounted on traffic lights. That data is fed back to an AI system that is able to retime every stoplight on the fly, based on real time data, optimizing actual traffic flow on a real-time basis. Step Three would be when the AI is able to send commands to autonomous vehicles to optimize traffic flow without human intervention. We are still a way away from Step two and Three, but they are certainly not out of the realm of possibility. Chalk one up for AI.
RSS Feed