Abstract
The assignment procedure is a powerful tool in the modeling process for the evaluation of land-use and transportation system alternatives. It is generally believed that traffic assignment is highly dependent upon the trip matrix generated from the preceding modeling phases. How valid is this viewpoint? This research was undertaken to investigate the sensitivity of the traffic assignment to different trip matrices generated from various constraints. The main research problem was divided in to the following three subproblems: 1) to determine whether there is a significant difference between the assigned volumes obtained from the various trip matrices; 2) to compare assigned volumes with ground counts; and 3) to examine the results using different traffic assignment techniques. The comparisons of trip matrices indicated that there were significant differences between these matrices prior to entering the assignment process. However, the differences in the assignment results were not nearly as great as the differences in the matrices. This gives strong indication as to the power of the assignment process, due to its aggregative nature, to overcome inadequacies from the preceding modeling phases. It was concluded that traffic assignment was slightly more sensitive to the trip length frequency (TLF) constraint than the constraint of the row and column totals. This indicates that as long as an accurate trip length frequency is used, assigned volume will very closely match the volume from the fully modeled trip matrix and/or counted volume. Also, it is expected that the TLF constraint in a large study area will be much more effective than the constraint of the row and column totals. In addition to the all-or-nothing assignment, "stochastic" multipath, incremental capacity-restraint, and equilibrium assignments were applied using the total trip matrices (internal plus external trip tables). These assignment results were compared to ground counts. It was concluded that the incremental and the equilibrium assignments represented a slight improvement from the all-or-nothing and the "stochastic" multipath assignments. However, the difference in results was not significant enough when using capacity constraint to warrant the extra cost such as link capacity data and computer run time involved in the incremental and the equilibrium assignments.
Chang, Duk-Myung (1988). A sensitivity analysis of traffic assignment. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -771517.