Computer model could help city planners design better streets and reroute drivers for smoother commutes.
From the Journal: Chaos
WASHINGTON, Sept. 16, 2025 – Everyone hates traffic. Big cities in particular are plagued by an overabundance of vehicles, turning a simple crosstown jaunt into an odyssey during rush hour. Part of the problem is that traffic is incredibly complex, and a small change in one part of the system can have ripple effects that alter traffic patterns throughout a city. City planners attempting to improve local traffic grids can often struggle to foresee all the effects their changes could have.
In Chaos, by AIP Publishing, a pair of researchers from Kadir Has University in Istanbul developed a more efficient and flexible algorithm to model traffic. The duo’s desire to tackle this issue stems from more than just idle curiosity.
“Being based in Istanbul, we live in one of the most congested cities in the world,” said author Toprak Firat. “Traffic isn’t just an academic problem here; it’s part of daily life and that gave us a strong motivation.”
While traffic flow algorithms do exist, they often require detailed trip information and rely on hard-coded rules to determine how vehicles move through intersections. This leads to a rigid algorithm that the authors wanted to avoid. Instead, they developed a model they called the data-driven macroscopic mobility model (D3M), which relies only on simple observations that city planners routinely collect, like how packed the streets are.
“Rather than using fixed equations for flow dynamics, we calibrate the model parameters directly from real-world traffic data,” said Firat. “This allows D3M to adapt its behavior to the observed conditions in each city, making it more flexible and realistic than models with hard-coded assumptions.”
The researchers tested their model on both synthetic benchmarks and real-world traffic data from London, Istanbul, and New York City. In the benchmark tests, the D3M model was more accurate than a conventional model and up to three times faster. In the real-world tests, it could accurately represent the diverse traffic conditions of these very different cities.
The faster simulation speeds and easier data requirements mean city planners have the tools to design better, smarter cities.
“The key breakthrough is that cities can now run sophisticated traffic simulations without needing expensive data collection,” said author Deniz Eroğlu. “Urban planners could test ‘what-if’ scenarios — like temporary closures due to accidents or maintenance — and see the predicted traffic impact before spending millions on construction.”
But the real impact could be felt directly by city residents, who could benefit from real-time traffic forecasting, making their commutes easier.
“Imagine a system that doesn’t just react to traffic locally, but simulates how congestion can spread in complex, often unexpected ways across an entire city,” said Eroğlu. “A jam in one part of the network might trigger bottlenecks kilometers away — not because of local crowding, but due to the ripple effects of shifting flows. Our model captures these dynamics, offering system-level foresight instead of piecemeal reaction.”
The authors are planning to test their model in a real-time operational environment, with the goal of bringing traffic forecasting to real cities soon.
###
Article Title
Data-driven modeling of traffic flow in macroscopic network systems
Authors
Toprak Firat and Deniz Eroğlu
Author Affiliations
Kadir Has University