Regional Dynamic Traffic Assignment for Real World Travel Demand Models--Trying It Out in San Francisco

Over the past few years many travel demand modelers have come to the realization that their static traffic assignment algorithm is inadequate. As the demand portion of our models become more sophisticated, the authors are able to confidently model and compare many policy scenarios. Embarrassingly, within the limits of static traffic assignment we cannot calculate such simple measures as travel time between an origin and destination without a string of caveats attached. Nor can the authors provide volume-to-capacity ratios that traffic engineers won’t laugh at. Unfortunately, many of these simple “skims” that we are so unsure about in static assignment are fed back in to our demand models. How well can our sophisticated demand models do when they are being fed wrong inputs from the start? All signs seem to be pointing to simulation-based Dynamic Traffic Assignment (DTA) as the solution to our static assignment ailments, but no regional modeling agency has (to our knowledge) taken the plunge to actively explore DTA as a possible replacement for static assignment. To date, several examples of simulation-based equilibrium DTA models coupled with regional travel demand models exist. However, extraordinary run times and complexity have diminished their usefulness outside of research or for some very specific projects. This paper shares our experience to date: strategies, approach, and discoveries in applying DTA in San Francisco to evaluate its feasibility to be a part of our regional demand model as we explored various DTA packages.