Data-driven placement of centroid connectors in dynamic traffic assignment
Recent technological advances allows transportation engineering professions to collect, share, and handle unprecedented quantities of data, which has the potential to transform current transportation planning paradigms. In the immediate future, data can be used to improve the precision and capabilities of existing transportation network modeling frameworks. Parcel data is a large, readily available data source that represents the location of public lands, businesses, and residences and is frequently used by government and businesses for land use and zoning decisions. This thesis looks at the viability of using parcel data to inform static traffic assignment (STA) and dynamic traffic assignment (DTA) connector placement in a medium sized network in the Austin, TX region.
Simulation-based DTA models are particularly sensitive to the topological detail of the traffic network, including the location of centroid connectors. Traditional centroid connector placement strategies may lead to excessive congestion and unrealistic traffic patterns, while manual network refinement is prohibitive in large regional models. In this thesis, parcel-level data is used to both allocate travel demand between two sub-regions in each considered traffic analysis zone and to select appropriate nodes for the centroid connector placement. Numerical experiments suggest that the proposed approach better approximates both corridor travel times and traffic counts throughout the network, with improvements of more than 40 percent in travel time estimation accuracy, and 12 percent in traffic count estimation. Additionally, the scenarios that best matched count and travel time data were the scenarios that had the highest average parcel density per entry/exit node, indicating that parcel data is an acceptable proxy for high demand points in the network.
When applied in STA, the results were not quite as promising. Although this methodology was able to improve the utilization of lower capacity links, the results ultimately did not better resemble volume count data. However, this does represent a simple, transparent, and data-driven approach for centroid connector placement in static traffic assignment that performs as well as traditional methods.