Google Research

Creating, Calibrating, and Validating Large-Scale Microscopic Traffic Simulation

  • Theophile Cabannes
  • Alben Bagabaldo
  • John Lee
  • Qianxin Gan
  • Ayush Jain
  • Alice Blondel
  • Alexandre Bayen
Transportation Research Board 102th Annual Meeting (2023) (to appear)


The challenges of creating, calibrating, and validating a traffic microsimulation are not apparent until one tries to create their own. Through the development of a traffic microsimulation of the San Jose Mission district in Fremont, CA, this article shares a blueprint for creating, calibrating, and validating a large-scale microsimulation of any city. Codes and data are made openly available for anyone to reproduce the simulation or its creation inside the Aimsun microsimulator. The calibration process enables simulating the movement of 130,000 vehicles through a Fremont subnetwork with more than 4,000 links using a representative 2019 afternoon six-hour demand. Executing the simulation on calibrated data gives a linear regression between the simulated and real data with slope of 0.976 and R^2 of 0.845 across 83 sensors at 15-minute time intervals.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work