Adaptive Traffic Signal Controller Improves Flow Fairness
The research-stage system uses vision models to reduce dominant-approach delays by 59% in simulated traffic demand scenarios.
Updated on Oct. 31, 2026 in Artificial Intelligence
Live Poll
Would you support using adaptive traffic signals even if minor roads face longer wait times?
Researchers have developed an adaptive traffic signal control system that uses a dual-model computer vision pipeline for real-time vehicle and license plate detection. This research-stage framework demonstrates significant efficiency gains in simulated traffic environments by actively balancing signal timing under asymmetric demand.
Why it matters
Traditional adaptive signal controllers often prioritize aggregate throughput, which can inadvertently create long delays on side streets. By focusing on lane-specific density and anti-starvation constraints, this system addresses the distribution of wait times to improve flow fairness.
The system features a YOLOv26 vehicle detector achieving 90.92% mAP@50 and a plate detector at 97.21% accuracy. The pipeline processes traffic frames at 84 to 130 FPS, with the CRNN architecture reaching 96.61% character recognition accuracy.
The players
Eclipse SUMO
An open-source microscopic road traffic simulation suite used for testing traffic management algorithms.
YOLOv26
A deep learning object detection architecture designed for high-speed, frame-by-frame visual processing.
The details
The controller uses a two-stage computer vision process where a YOLOv26 model—an object detection architecture optimized for speed and accuracy—identifies vehicle occupancy to calculate lane density. A second model localizes license plates for identification by a Convolutional Recurrent Neural Network (CRNN), a deep learning model designed for sequential data like text. These inputs inform an adaptive algorithm that optimizes green-light intervals while an anti-starvation constraint prevents any single approach from waiting indefinitely during periods of skewed traffic demand.
Timeline
October 31, 2026: The research findings were published.
The Tech Race
This framework moves beyond traditional aggregate-efficiency metrics by addressing the delay distribution issues inherent in current adaptive systems. It competes with existing commercial traffic management algorithms that primarily rely on inductive loops or stationary cameras to calculate cycle timing.
As this system remains in a simulation-only stage, it currently has no direct effect on public transit or commute times. Future deployments would require integration with existing municipal traffic light hardware and camera infrastructure.
The takeaway
This research provides a scalable benchmark for managing traffic flow in complex, skewed demand environments by leveraging high-speed computer vision. Researchers should monitor whether future iterations move from the Eclipse SUMO simulation environment to pilot testing on physical city traffic grids.
Further reading
Explore the latest developments in vision-based systems at Artificial Intelligence.
More information
View the technical publication criteria via the Creative Commons license terms.
Source note: This article includes information reported by Nature.
Live Poll
Would you support using adaptive traffic signals even if minor roads face longer wait times?
