1. Home
  2. TRID
  3. View Record
https://nap.nationalacademies.org/catalog/27432/critical-issues-in-transportation-for-2024-and-beyond

Deciding on Moving to Activity-Based Models (or Not)

The Michigan Department of Transportation (MDOT) is in the midst of its multiyear Model Improvement and Quality Program to develop better travel demand models in the State of Michigan. MDOT is in the process of determining whether any or all of the next set of travel models in the State should be tour or activity based. This paper describes the work performed related to the evaluation of advanced (tour/activity-based) modeling for possible use in Michigan. A literature review of documentation of existing models and models currently under development was conducted, and agencies currently using or developing activity- or tour-based models were interviewed. The findings were provided to MDOT to determine whether advanced models would be advantageous and if so the best approach for implementing them. This information takes into consideration all data sources currently available along with the technical feasibility of creating, running, and maintaining such a model. The research concludes that development of advanced travel models used by transportation planning agencies in the U.S. has evolved to the point where most modern models use the same basic approach. The activities of individuals in a synthetic population are microsimulated using a series of logit choice models. At least five activity purposes are considered, and auto ownership, (usual) workplace and school location, daily activity pattern, tour- and trip-level destination choice, tour- and trip-level mode choice, and tour- and trip-level time-of-day choice are modeled. Time-of-day is analyzed using time periods of one hour or shorter. Traffic assignment procedures are similar to those used in conventional aggregate trip-based models. The main areas of difference seem to be in the placement of the tour-level time-of-day choice component in the overall model sequence and the level at which household interactions are considered. The process for developing an activity-based model typically takes two to three years and costs an agency about U.S. $600,000 to $800,000 in consultant fees. This schedule and budget can be larger or smaller depending on how the model development process is organized. The U.S. models have been implemented using a variety of proprietary modeling software packages as well as custom programs written specifically for the application of these models. While some of the earlier models were applied on single processor computers, the newer models are taking advantage of distributed processing. The time typically required for a model run is about a day although this varies with the complexity and size of the model. While things have not always gone smoothly, the overall experience of agencies in implementing activity-based models has been positive. The perceived benefits of these models for these agencies outweighed the drawbacks, and most do not regret their decision to use these types of models.

Language

  • English

Media Info

  • Media Type: DVD
  • Features: References; Tables;
  • Pagination: 17p
  • Monograph Title: TRB 88th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01128601
  • Record Type: Publication
  • Report/Paper Numbers: 09-1916
  • Files: TRIS, TRB
  • Created Date: May 19 2009 7:48AM