Particle.news

Download on the App Store

Data-Driven Traffic Model Outpaces Conventional Approach, Real-Time Trials Planned

Calibrated from routine congestion observations, D3M simulates citywide ripple effects with less data.

Overview

  • Peer-reviewed research in Chaos details D3M by Kadir Has University researchers Toprak Firat and Deniz Eroğlu.
  • Rather than fixed flow equations or trip-level inputs, the model tunes parameters directly from real traffic measurements.
  • In synthetic benchmarks, D3M was more accurate than a conventional model and ran up to three times faster.
  • Tests on real-world data showed it represented traffic conditions in Istanbul, London and New York City.
  • The team plans real-time operational tests as the next step toward practical forecasting and planning use.