Evaluation of origin/destination matrix estimation techniques to support aspects of traffic modeling

A travel demand forecasting model is a representation of the traffic volume on a roadway network. It is one of the most important analysis steps in transportation planning. Predicting the current or future traffic volume on roadways is essential to transportation planners because it enables them to effectively use tax payer’s dollars.;Unfortunately, the money and time required for data collection and model development frequently results in only a little left over for the forecasting phase. Finding ways to increase the speed and reduce costs associated with developing models would be beneficial to all communities, but especially small- and medium-sized communities where funds are limited or nonexistent and models are not always available.;This document presents a methodology for estimating an origin/destination matrix from existing traffic counts and determines the relationship between the number of traffic counts required in a network and the number of origin/destination pairs to build an accurate origin/destination matrix. This relationship is a contribution as it allows for decreases in the time and cost associated with developing models in small- and medium-sized communities.;A small fictitious network was created as an example to verify that the methodology works. Trips were assigned to the network and used as traffic counts. The methodology was employed to estimate the origin/destination pairs using all available traffic counts.;Since most communities do not have traffic counts available on all roadways, research was done to define the number of traffic counts required to accurately estimate an origin/destination matrix. An experimental process was created where scenarios of randomly removing traffic counts were simulated. It was found that roughly 65 percent of the traffic counts needed to be available for the example network.;The methodology was further tested by applying it to an actual city network. The experimental process was repeated. It was found that 75 to 80 percent of the traffic counts were required to accurately estimate the origin/destination matrix.;The relationship of traffic counts to origin/destination pairs revealed that four times more traffic counts than zones in the network were needed for accurate origin/destination matrix estimation.