AI Puts Traffic Managers in the Driver’s Seat
Efficient automobile traffic is a fundamental city service. When traffic slows, productivity suffers, goods can’t flow, and first responders are delayed.
Yet cities around the globe struggle to manage their streets. The intersection is perhaps the greatest traffic control challenge. The risks are abundant: drivers running red lights, making illegal turns, or dodging pedestrians. So it’s no surprise that 40% of all traffic incidents in the United States occur in intersections.
These challenges will only grow as streetscapes evolve. New transportation modes like autonomous vehicles and shared scooters add complexity to traffic monitoring. Shifting environmental priorities are growing the mix of pedestrians and cyclists, leading to safety concerns. And the uncertain future of transportation means planners need a flexible approach that can rapidly adapt to new circumstances.
The good news: New AI solutions can make intersections smarter.
Data Intelligence Enables Measurable Results
One example of a city using new technology to address these challenges is Bangkok, Thailand. It has implemented GRIDSMART AI-enabled cameras and machine learning to monitor and control hundreds of intersections. With nearly 400 units deployed across the city, it has experienced a decidedly positive impact:
- Traffic delays have been reduced by nearly 25%
- The average queue length is up to 30% lower
- Vehicles running red lights are down 68%
- Nearly 52,000 vehicle commuter hours per year have been saved
- An estimated $855,000 has been saved by the city annually
“Our vision is to improve a billion lives through intelligence,” said Bill Malkes, founder and CEO of GRIDSMART Technologies. “The only way we can accomplish this goal through intelligence is having the ability to process and crunch data.”
The Need for a Smarter Solution
Conventional intersection control systems are limited. They typically track only what’s happening inside an intersection—and not the volume, type, and velocity of vehicles traveling in and out. They also require multiple cameras to view an entire intersection, making most systems challenging to install, configure, and maintain.
To stay ahead of fast-moving transportation trends, city managers and traffic engineers require solutions that don’t just address today’s challenges but tomorrow’s as well. They need to deploy new or upgrade existing systems with the ability to easily add new functionality in the future.
Bottom line? There are three must-haves for effective intersection monitoring, safety, and control:
- Intelligence to handle more vehicles with less delay within a city’s existing infrastructure and roadways
- Open architecture that works with other smart city applications such as air quality tracking
- Easy to deploy, manage, and maintain to reduce resources and costs
This is where the GRIDSMART System—a unique single-camera solution for intersection actuation, traffic data collection, and situational awareness—enters the picture.
Along with GRIDSMART’s iconic Bell Camera, the system uses computer vision algorithms to track all moving objects. It provides real-time data to manage the timing of the traffic lights and improve intersection efficiency and safety. Plus, as shown in Figure 1, the GRIDSMART system offers the benefits of simplicity, flexibility, and transparency.
“On the average, if you look at a GRIDSMART system, we take about 25% of the install time and something less than 25% of the maintenance time, compared to conventional video systems,” said Malkes.
A camera captures a full intersection view, including the center where cars, trucks, motorcycles, bicycles, and pedestrians cross paths. The system tracks the trajectory of this traffic as it approaches, enters, crosses, and leaves the intersection.
This means better traffic pattern accuracy—enabling the adjustment of traffic light cycle intervals in real time. With extensive data analytics, city planners can truly make a positive impact on people’s quality of life and safety.
For example, one city changed a vital intersection to permissive left turns, resulting in 10% more vehicles using the intersection and removing thousands of vehicles a day off side streets.
“Permissive lefts often result in accidents as vehicles pass in front of oncoming traffic. GRIDSMART uniquely sees the center of the intersection so we were able to validate that while traffic throughput increased, accidents did not go up by changing from protected to permissive left-hand turns,” said Malkes. “The city achieved an efficiency improvement without the consequence of an increase in collisions.”
How a Single Camera Manages Complex Intersections
The GRIDSMART solution comprises three main components—the Bell Camera and GS2 Processor, as shown in Figure 2, as well as the GRIDSMART Client.
The Bell Camera is protected by its distinctive bell shape housing, which mitigates the sun’s glare and adverse weather conditions, decreasing maintenance costs.
The fisheye lens provides a unique “horizon to horizon” full intersection view, unlike other cameras that provide only single directional views. The result is more accurate situational awareness and unobstructed incident views. In addition, the camera’s virtual pan-tilt-zoom function enables traffic managers to set up and adjust multiple views.
As vehicles and pedestrians come within range, the camera starts building three-dimensional models of each “object.” Based on preprogrammed logic, the models determine traffic light timing.
The GS2 processor is installed in a standard control cabinet and is directly connected to the Bell Camera. In using GRIDSMART proprietary algorithms, it’s the processor engine where all the computation takes place—running on an Intel® Core™ i5 CPU and an Intel® SSD.
An abundance of detailed traffic statistics, such as the hourly number of left turns or vehicle speed, are uploaded nightly from the GS2 processor to the cloud. This provides traffic operators and engineers access to both current and historical activity and trend reporting.
The third component of the solution—GRIDSMART Client software—is like a traffic management center (TMC) on a laptop. Traffic engineers use the client application to configure, control, and view intersections, as well as monitor run performance reports and synchronize configurations to the Cloud. The client is home to a virtual DVR, which enables a large variety of intersection views.
Finally, the GRIDSMART solution uniquely provides an open API. This means that cities can easily automate retrieval of data and images from their existing traffic control systems.
Malkes said, “We have processed more than 50 million hours of video and counted 43 billion cars in all kinds of environmental conditions worldwide. This requires applying intense vision processing algorithms that necessitate rapid and reliable computing capacity. That's why Intel® has always been the greatest possible partner for us. We focus on doing things traffic related, while Intel keeps giving us more power in a cost-effective manner.”