Feature Stories

Transportation Network Modeling: Helping to Solve Traffic Congestion Problems

photo of car driving by Greeting From Austin mural at the Roadhouse Relics gallery

Integrated transportation network modeling can provide an understanding of how different kinds of infrastructure networks work together. 

Apr. 21, 2014

Assistant Professor Stephen Boyles is investigating the relationships between transportation networks of different scales/sizes, learning how to quantify these relationships, and discovering the implications for transportation planning.

When modeling transportation systems, engineers often have to make tradeoffs when deciding how much detail to include in the model, and how large of an area to study.  Using too small of an area means missing out on factors affecting the entire region (such as long-distance freight travel), while modeling a large area forces engineers to use simpler and more abstract models because of limitations in data and computer power. The most advanced models can take days to run on large metropolitan areas like Austin or Dallas-Ft. Worth.

Transportation practitioners often use multiresolution modeling or subarea analysis, where several models are used. This means a larger, less detailed model is examined together with a smaller, more detailed version focusing on an area of particular interest (like downtown or a congested freeway corridor).  

However, this is done without any kind of supporting theory.  Engineers ask how the "boundary" should be chosen where two models are stitched together. How exactly do the models interact across these boundaries?

illustrated line map of 7th street area of Austin

Example of the Austin roadway network, including the entire regional model, a subnetwork representing the downtown area, and an even smaller corridor around 7th Street.  

Boyles’ research project, Integrated Multiresolution Transportation Network Modeling, which received a 2013 National Science Foundation CAREER Award, is aimed at understanding these questions, specifically in the context of transportation networks, but also as a tool for understanding how different kinds of infrastructure networks (such as transportation and electricity) work together.  

His hypothesis is that the key to successful performance of multiresolution modeling is the use of
 "soft" boundaries rather than "hard" boundaries.  By "hard" boundaries, it means that the large and small area models are clearly separated; traffic is first modeled on the large region, then those results are "zoomed in" to form the basis for the subarea model.  
"Softening" the boundaries allows more of the interactions between the large and small areas to be captured.  An example of this is changing a driving route based on changes in the subarea. If the streets in downtown Austin are changed in some way (adding a streetcar, changing one way streets to two way, etc.), people commuting to work might change whether they come on I-35 or Loop 1, or make other changes in their route.

To do this, when modeling the small area, the regional network is not deleted, but simplified in a way that allows the main alternate routes to be included:

numbered diagrams

Boyles' early experiments show that this doesn't increase the computation time much (relative to only modeling the subarea, which can be done in seconds or minutes) but greatly increases accuracy relative to omitting all of the regional network.

Over the course of the project, he plans to develop a new network modeling theory to explain these interactions, and apply them to several kinds of transportation network models.