Lehman Center for Transportation Research at Florida International University

Incorporating Automatic Feedback into FSUTMS for Equilibrium and Peak Spreading Modeling

  • Sponsor: Florida Department of Transportation
  • Contact: Dr. Albert Gan, 305-348-3116, gan@eng.fiu.edu

  • The traditional four-step travel demand modeling process, on which the Florida Standard Urban Transportation Model Structure (FSUTMS) is based, feeds output from one step to the next as input. One inherent weakness in this sequential process is that the travel impedances used in trip distribution and modal split are usually not consistent with those estimated in traffic assignment. This inconsistency arises because travel impedances are a function of network congestion but the level of congestion is not known during the trip distribution and trip assignment steps. For demand forecasts involving congested networks, the problem is especially severe as travel times and speeds are highly sensitive under congested conditions. The need to incorporate feedback to achieve model equilibrium in the four-step process is well recognized and has been the subject of research for many years.

    One approach to achieving consistent impedances is through a feedback loop that iterates through trip distribution, mode split, and traffic assignment several times until travel impedances stabilize. However, incorporating such a feedback loop into the four-step process is not a straightforward process. Many researchers and practitioners have experienced difficulty achieving convergence in a feedback process when the 'direct method' (i.e., the unaltered output of assignment is used directly as input for a previous step in modeling process) of feedback is used. A number of alternatives to the direct method have been identified as ways to reduce processing time and assure convergence. The Method of Successive Averages (MSA) that averages the volumes from previous iterations and the volumes from the most recent assignment is an example of such alternatives.

    Several more sophisticated models in the FSUTMS family, such as the Southeast Florida Regional Planning Model, include a one-iteration feedback loop to produce congested impedances for use in the second (or final) round of distribution and modal split of home-based work trips. Although this one-iteration feedback loop allows the use of congested impedances in trip distribution and modal split, it is not designed to achieve consistency travel impedances in the four-step process. An automated feedback procedure for consistent travel impedances in FSUTMS does not currently exist. FSUTMS users who desire consistency in their models are limited to performing only the direct method of feedback manually for a very limited number of iterations that may or may not converge. This manual process is not only cumbersome, but the results are also prone to pitfalls and errors. The research and development of a standardized, automated feedback procedure for FSUTMS is long overdue.

    Another potentially useful feedback loop is one that allows the adjustment of 'peak-to-daily' ratio (CONFAC) dynamically to address a well-recognized problem associated with the use of the ratio. Prior to FSUTMS version 5.3, congested impedances were obtained from traffic assignment that uses a single 'peak-to-daily' ratio (CONFAC). This approach is not sufficient to address different peaking characteristics of different facility types, and is the culprit of path skim errors leading to the under- and over-assignments. To solve this problem, FSUTMS version 5.3 adopted a "multiple-CONFAC" structure, allowing the use of different CONFACs for different classes of facilities. This approach established a foundation for better modeling of different peaking characteristics for different classes of facilities. However, two issues remain. First, appropriate CONFACs for different facilities have not been empirically established based on traffic count data, especially the relationship between CONFACs and volumes. Second, applying CONFACs for future forecasts will require iterative fine-tuning to reach the equilibrium between daily link volume and peak spreading. The iterative fine-tuning mandates an automatic feedback loop process be developed and implemented for CONFACs adjustments.

    This project aims to achieve the following objectives to address the stated problems:
    1. To calibrate CONFACs for different facility types based on actual traffic count data.
    2. To establish empirically the relationships between volume (possibly a variation of volume) and CONFAC for each facility type to guide the adjustment of CONFACs.
    3. To develop and implement an automated feedback loops in FSUTMS to achieve consistent travel impedances in the four-step process and to dynamically adjust CONFACs based on the empirical relationships.