ESTIMATION OF PARAMETERS IN MODELS FOR TRAFFIC PREDICTION - A NEW APPROACH

Parameters for traffic prediction models are usually based on traffic counts at cross sections. The results of these are applied for test runs on the model using parameter values for a similar city where an o-d survey has been made. Corrections are then made on a trial and error basis to get reasonable agreement between traffic volumes counted over cross sections and corresponding data calculated from the model. This report proposes a new approach which uses the counted link volumes as observations of the dependent variable in a regression model. The problem is regarded as non-linear regression, and parameter values are obtained by an iterative least squares algorithm. An example is given. It is shown that estimates of parameters can be made without expensive and lengthy o-d surveys. Where there is no o-d survey, the proposed procedure is very useful in obtaining better estimates and reducing practical work in the estimating procedure. Further studies should be made into optimum sample design for the selection of links for observation, and practical application of the proposed method. /TRRL/

  • Corporate Authors:

    Gothenburg University, Sweden

    Kulturgeografiska Institutionen, Fack
    Gothenburg,   Sweden 
  • Authors:
    • HOEGBERG, P
  • Publication Date: 1975

Media Info

  • Features: Figures; References; Tables;
  • Pagination: 8 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00147655
  • Record Type: Publication
  • Source Agency: Swedish National Road and Traffic Research Institute
  • Report/Paper Numbers: Resh Rpt. 1975:1 Monograph
  • Files: ITRD, TRIS
  • Created Date: Sep 20 1977 12:00AM