Researchers at Texas A&M University and the U.K.'s University of Edinburgh have developed a system that uses machine learning to optimize the timing of traffic signals to reduce wait times at intersections.
Their approach can successfully train a deep neural network in real time and transfer what it has learned from real-world observations to a different control function that can be understood and regulated by traffic engineers.
The researchers used a simulation of a real intersection and found that optimizing their interpretable controller reduced vehicle delays by as much as 19.4% compared to commonly deployed signal controllers, which are the "brains" of an intersection.
From Texas A&M Today
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA