Calibration of a Stochastic Car-Following Model Using Trajectory Data: Exploration and Model Properties

The objective of this research is to assess a recently formulated car-following model in terms of its performance during congested periods. The main challenge faced in this research is the structure of the car-following model to calibrate. Since the model relies on utility maximization technique with a stochastic choice between different acceleration alternatives, the model equations are analytically intractable where the solutions should be numerically computed. For that reason, for calibration purposes, a nonlinear optimization procedure is used based on a Genetic Algorithm (GA). The objective is to minimize the deviations between the observed trajectories and the simulated driving dynamics when following the same designated leader. Based on the numerical results, this study showed that the GA approach is suitable to calibrate car-following models with complex structures: the adopted calibration method allowed estimating a set of acceptable parameters estimates. Using these estimates, through simulation, the resulting fundamental diagram contained a free-flow region and a congested region. Phenomena like traffic breakdowns and scattering of flow-density data points were observed in the congested region.


  • English

Media Info

  • Media Type: DVD
  • Features: Figures; Photos; References;
  • Pagination: 19p
  • Monograph Title: TRB 88th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01126673
  • Record Type: Publication
  • Report/Paper Numbers: 09-3450
  • Files: TRIS, TRB
  • Created Date: Jan 30 2009 7:51PM