Innovations in Travel Demand Modeling, Volume 1: Session Summaries (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

capacity using artificial links, splitting a link in two, and using dummy links. • The analysis conducted in this study indicates that speed modeling is intertwined with model calibration. The results suggest that insensitive speed–flow equations give less accurate queue delays, but they tolerate inaccu- rate capacities, dummy links, and inaccurate link flows. Conversely, sensitive speed–flow equations give more accurate queue delays, but cannot tolerate inaccurate capacities and flows. • To obtain more accurate queue delays, more accu- rate speed–flow equations should be used in combina- tion with accurate capacities, coding to distinguish dummy links, and peak- period analysis. In the future, using DTA with simulation— including programs such as Dynasmart(P), DynaMIT(P), or Dynameq— may address some of the limitations identified in this study. A COMPARISON OF STATIC AND DYNAMIC TRAFFIC ASSIGNMENT UNDER TOLLS IN THE DALLAS–FORT WORTH REGION Stephen Boyles, Satish Ukkusuri, S. Travis Waller, and Kara Kockelman Stephen Boyles discussed the use of static and dynamic assignment models in the Dallas–Fort Worth region to analyze congestion pricing alternatives. He described a study comparing the use of three models: traditional sta- tic traffic assignment (STA), the TransCAD approxima- tor, and VISTA’s simulation- based dynamic traffic assignment (DTA). Volume 2 includes a paper on this topic.3 The following points were covered in his presentation. • Use of DTA models provides the capability to account for time- varying properties of traffic flow. Although differences exist among DTA models in how traffic flow is modeled and how the mathematical pro- gram is described, all DTA approaches provide the abil- ity to model traffic flow changes over time. DTA models require more input data than STA models, including time- dependent travel demand data. DTA models also introduce other issues, such as ensuring first- in- first- out queuing disciplines. The use of DTA models requires substantial computational time when applied to a major metropolitan area with large networks, such as the Dal- las–Fort Worth region. • A comparison was conducted in the Dallas–Fort Worth area using three different traffic assignment mod- els. The first model was a traditional STA model. The second model was the TransCAD approximator, which uses analytical, link performance- function- based approximation to DTA. The third model was VISTA, which uses a simulation- based DTA approach. • The traditional STA models use a steady- state approach, with no concept of time. STA models use total demand in a single time period. STA models include link performance functions. The TransCAD DTA approxi- mator is an add- in to the TransCAD software. It is based on an iterative algorithm. It uses link performance func- tions to calculate vehicle delay, which is a major differ- ence from the VISTA model. The link performance functions are less computationally intensive, and the approximator runs faster than VISTA. It does not model traffic flow at the same level of detail, however. • VISTA is network- enabled software that integrates temporal network data and models for a wide range of transportation applications. It is based on a cell trans- mission model (CTM) that divides links into smaller cells, which can be modeled individually at fine resolu- tion of approximately 5 to 10 seconds. A key feature of CTM is that flows are explicitly prohibited from exceed- ing capacity. If demand for a cell exceeds the available capacity, queues form to maintain flow less than capac- ity. This ability to model queues in a more realistic man- ner is a main attraction of VISTA. • The parameters used by the models are different. The link performance function used by STA and the DTA approximator requires that the capacity and free- flow time for each link be specified. The two calibration pa - ram eters must also be specified. The CTM requires the jam density and the length of each cell to be specified. The two parameters indicating the slopes of the flow- density curve when flow is increasing or decreasing with volume must also be specified. • Comparing STA and DTA is not easy because of the fundamental differences in the modeling approaches. The presence of clearance intervals in DTA bias travel times is low compared with static assignment. Clearance intervals account for vehicles that depart near the end of the model period and arrive at their destination beyond the model period. No additional vehicles are assigned during these intervals, but vehicles remaining on the net- work are allowed to complete their trips. The result is that some links experience flows for longer periods of time than in STA, effectively increasing link capacities. This issue does not occur with STA because of the inabil- ity to distinguish when vehicles depart and assume a steady- static condition. • The three approaches were applied to analyze toll alternatives in the Dallas–Fort Worth metropolitan area. A total of 92 links (of the 56,574 total links) were tolled in this application. A 3-hour morning peak period from 6:00 a.m. to 9:00 a.m. was used in the analysis. This 40 INNOVATIONS IN TRAVEL DEMAND MODELING, VOLUME 1 3 See Boyles, S., S. Ukkusuri, S. T. Waller, and K. Kockelman. A Comparison of Static and Dynamic Traffic Assignment Under Tolls in the Dallas–Fort Worth Region. Volume 2, pp. 114–117.