Calibration of a Macroscopic Traffic Simulation Model Using Enhanced Simultaneous Perturbation Stochastic Approximation Methodology

Simulation, a popular and widely used method for studying stochastic, complex real-world systems, can accurately represent those conditions when its parameters are effectively calibrated. Previous studies on calibration generally focused on minimizing the sum of relative error between the observed data from a certain period of time in a typical day and the simulation output for the same period. This static approach can be explained as calibration with data obtained at one point in time. However, this type of calibration approach cannot capture a realistic distribution of all possible traffic conditions and may produce inaccurate calibration results. This paper proposes a calibration methodology based on the Bayesian sampling approach. Instead of a single demand matrix and corresponding observed traffic conditions that represent a specific point in time, this calibration methodology uses randomly generated demand matrices and corresponding traffic conditions from an observed statistical distribution of these variables. The goal of using input values generated from an observed distribution of demands is to accurately represent a wide range of all likely demand conditions observed at a facility. Moreover, at each iteration, the proposed calibration methodology reestimates optimal parameters by using a stochastic optimization algorithm known as the simultaneous perturbation stochastic approximation (SPSA) algorithm. A macroscopic simulation model of a portion of I-880 in California based on the cell transmission approach is calibrated with the proposed methodology. The proposed enhanced SPSA algorithm outperforms a simple SPSA algorithm based on several case scenarios studied as part of this paper. Future work will focus on calibrating larger networks and microscopic simulation models with the same methodology.

• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Lee, Jung-Beom
  • Ozbay, Kaan
• Conference:
  • Transportation Research Board 87th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2008-1-13 to 2008-1-17
• Date: 2008

Language

• English

Media Info

• Media Type: DVD
• Features: References; Tables;
• Pagination: 18p
• Monograph Title: TRB 87th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

• TRT Terms: Algorithms; Bayes' theorem; Calibration; Macroscopic traffic flow; Operations; Perturbations; Statistical analysis; Stochastic processes; Traffic simulation
• Subject Areas: Administration and Management; Data and Information Technology; Highways; Operations and Traffic Management; Planning and Forecasting; I72: Traffic and Transport Planning;

Filing Info

• Accession Number: 01099148
• Record Type: Publication
• Report/Paper Numbers: 08-2964
• Files: TRIS, TRB
• Created Date: May 21 2008 7:05AM