ESTIMATION OF TRAFFIC VARIABLES USING POINT PROCESSING TECHNIQUES

An alternative approach to estimating aggregate traffic variables on freeways--spatial mean velocity and density--is presented. Vehicle arrival times at a given location on a roadway, typically a presence detector, are regarded as a point or counting Poisson process whose rate is a function of the state of the traffic at every instant of time. Moreover, the traffic state is modeled as a finite-state Markov chain. A sequential point process filter, optimum in the mean-squared error sense, is designed to estimate the state from observations of the vehicle arrival-time sequence. Different possibilities for incorporating potential additional information such as speed and headway are explored. Parameter values for the underlying Markov chain are obtained via a maximum likelihood estimator. Qualitative behavior of the proposed algorithms is studied with simulated traffic flow data from both macroscopic and microscopic models.

  • Corporate Authors:

    Massachusetts Institute of Technology

    Electronics Systems Laboratory
    Cambridge, MA  United States 

    Transportation Systems Center

    55 Broadway, Kendall Square
    Cambridge, MA  United States  02142
  • Authors:
    • Lopez-Lopez, R
    • HOUPT, P K
  • Publication Date: 1978-5

Media Info

  • Pagination: 80 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00181381
  • Record Type: Publication
  • Source Agency: National Technical Information Service
  • Report/Paper Numbers: DOT-TSC-RSPA-78-9 Final Rpt.
  • Contract Numbers: DOT-TSC-849
  • Files: NTIS, TRIS
  • Created Date: Oct 12 1979 12:00AM