Multimodal freight origin-destination matrix estimation

This dissertation presents a method for estimating multimodal freight origin-destination (O-D) tables based on partial available information. The data used can include link counts, cordon counts and/or aggregate information on portions of the O-D matrix. The assumption is made that the data available reflects an underlying stochastic user equilibrium (SUE) flow pattern, but may also include errors. The flow pattern is delineated on a supernetwork that represents combined mode split and route-choice decisions. The method does not require path enumeration or an initial/target O-D table.; In addition, an analysis for determining the value of information when estimating freight origin-destination tables is presented. The theory of design of experiments is used as a framework to ascertain the value of three different types of information used in the above-mentioned O-D estimation procedure. Results are examined using analysis of variance, and the marginal benefits of adding more information of a specific type are assessed. A forecasting model is derived for determining the quality of a solution obtained from the estimation methodology, and the results from making predictions with this forecasting model are analyzed. As an additional consideration, the effect of introducing errors in the data observations is explored.; The procedure outlined in this dissertation is a good combination of transportation systems techniques that together with a clever algorithmic design has resulted in a novel way of looking at the O-D matrix estimation problem. The results from the validation exercise of the prediction model support the robustness of the methodology. Finally, the overall value of information study points to interesting policy recommendations that suggest the type and quantity of data to be collected in future efforts.