One way of measuring vehicle travel time along a specified route is to establish registration plate recording stations at both ends of the route. Usually only part of the registration number is recorded, say the last four characters. This leads to random spurious matches, as might mis-recording of number plates. In addition, depending on the exact purpose of the exercise, some correct matches might also be best discarded, e.g. vehicles that stop en route. Unwanted matches, from whatever cause, have only a marginal effect on estimates of flow, but have a massive effect on estimates of travel time. This paper demonstrates the above problem and presents an approach to overcome it. Instead of the imposition of essentially arbitrary bounds on acceptable travel times, robust statistics are used. The approach is developed so as to extract much more information from the data than is currently done. A consideration of the statistical basis leads to the development of confidence intervals for the underlying travel time at a given point in time (clearly this will vary, say, throughout the morning peak). Data from a study in Leeds is used for illustration, and as a means of suggesting how best to put the approach into practice. (Author/TRRL)

  • Availability:
  • Supplemental Notes:
    • This paper was presented at Seminar, Developing Countries, held at the PTRC Annual Summer Meeting, University of Warwick, England, 4-7 July, 1983.
  • Corporate Authors:

    PTRC Education and Research Services Limited

    110 Strand
    London WC2,   England 
  • Authors:
    • FOWKES, A S
  • Conference:
  • Publication Date: 1983

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 141-148

Subject/Index Terms

Filing Info

  • Accession Number: 00382854
  • Record Type: Publication
  • Source Agency: Transport Research Laboratory
  • ISBN: 0-86050-112-4
  • Report/Paper Numbers: Volume P240
  • Files: ITRD, TRIS
  • Created Date: May 30 1984 12:00AM