Kadir has university develops fast traffic algorithm expected to improve real-time traffic prediction accuracy

Everyone hates traffic congestion. Big cities, in particular, suffer from an excess of vehicles, making a simple short trip within the city turn into a long journey during peak hours. The problem partly stems from the extreme complexity of the transportation system, where even a small change in one part of the system can trigger a chain reaction that alters the entire city's traffic patterns. Urban planners often find it difficult to foresee all the possible cascading effects when trying to improve the local transportation network.

Image Source: American Physical Society

According to foreign media reports, two researchers, Toprak Firat and Deniz Eroğlu, from Kadir Has University in Istanbul, have developed a more efficient and flexible traffic modeling algorithm, as published in the journal "Chaos." The motivation to address this issue goes beyond academic curiosity. Toprak Firat stated, "We live in one of the most congested cities in the world, namely Istanbul. The traffic problem here is not only an academic subject but also a part of daily life, which gives us strong motivation."

The existing traffic flow algorithms require detailed travel information and rely on hard-coded rules to determine how vehicles pass through intersections, but this leads to overly rigid algorithms, which researchers hope to avoid. To this end, they have developed a model called the Data-Driven Macroscopic Mobility Model (D3M), which relies solely on simple observational data routinely collected by urban planners, such as street congestion levels.

"We do not use fixed equations to describe fluid dynamics, but instead calibrate model parameters directly based on real traffic data," explained Firat. "This allows D3M to adjust behavior patterns according to the actual observed conditions of each city, making it more flexible and realistic compared to models that use hard-coded assumptions."

Researchers validated the model using synthetic benchmarks and real traffic data from London, Istanbul, and New York City. Benchmark tests showed that the accuracy of the D3M model surpasses traditional models and can run up to three times faster. In real-world tests, it accurately represented the diverse traffic conditions of these significantly different cities.

Faster simulation speeds and simplified data requirements provide urban planners with tools to design higher-quality smart cities.

Deniz Eroğlu stated: "The key breakthrough is that cities can now run complex traffic simulations without the need for expensive data collection. Urban planners can test 'hypothetical' scenarios, such as temporary closures due to accidents or maintenance, allowing them to anticipate traffic impacts before investing millions in construction."

The real impact may directly benefit urban residents, as real-time traffic predictions can make commuting easier.

Eroğlu said: "Imagine a system that not only responds to local traffic conditions but also simulates how congestion spreads throughout the city in complex and often unexpected ways. Congestion in one part of the network can trigger bottlenecks kilometers away, not due to local crowding but as a chain reaction from changes in traffic flow. Our model captures this dynamic mechanism, providing system-level forward-looking predictions rather than piecemeal responses."

The research team plans to test the model in a real-time operational environment with the goal of applying traffic forecasting to real cities as soon as possible.