An Efficient Dynamic Origin-Destination Matrix Estimation Framework

This paper proposes an efficient bi-level framework for the estimation of dynamic Origin-Destination (O-D) matrices for urban road networks. A maximum likelihood model is employed to calibrate the seed O-D matrix as the input of dynamic O-D estimation. The parameters of traffic flow models for road network are then calibrated based on the traffic data from traffic detectors. After these two initial steps, a bi-level dynamic O-D estimation method is developed to obtain a consistent solution between the resulting O-D matrix and traffic measurements. At the upper level of the model, two improvements include introducing adaptive conversion factors and adaptive weighting factors. With adaptive conversion factors, the model is able to handle congested traffic; with adaptive weighting factors, the model can avoid possible divergent problem during O-D estimation. At the lower level of the model, a Simultaneous Perturbation Stochastic Approximation (SPSA) based model is proposed. This model is designed to successively calibrate traffic flow models to save computational time. Adaptive weighting factors are also introduced into the lower level of the model to ensure the simultaneous decreasing of the volume deviation and speed deviation. Dynamic O-D estimation and traffic flow model calibration are performed alternately to reach final convergence. DYNASMART-P is used to dynamically load the traffic demand on the road network in this study. The application of the model in a real freeway segment demonstrates its ability to provide results of satisfactory accuracy within reasonable time and, hence, its potential usefulness to support ITS applications.


  • English

Media Info

  • Media Type: DVD
  • Features: Figures; References; Tables;
  • Pagination: 19p
  • Monograph Title: TRB 89th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01155463
  • Record Type: Publication
  • Report/Paper Numbers: 10-3798
  • Files: TRIS, TRB
  • Created Date: Jan 25 2010 11:56AM