Comparative Assessment on Static O-D Synthesis

Abstract

Recognizing the benefits of data and the information it provides to travel demand is pertinent to network planning and design. Technological advances have led the ability to produce large quantities and types of data and as a result, many origin-destination (O-D) estimation techniques have been developed to accommodate this data. In contrast to the abundant choices on data types, data quantity and estimation procedures, there lacks a common framework to assess these methods. Without consistency in a baseline foundation, the performances of the methodologies can vary greatly based on each individual assumption.

This research addresses the need for techniques to be tested on a common framework by establishing a baseline condition for static O-D estimation through a synthetic Vissim model of the Sioux Falls network as a case study area. The model is used to generate a master dataset, representing the ground-truth, and a subset of the master dataset, emulating the data collected from real world technologies. The subset of data is used as the input for the O-D estimation techniques where the input is varied to evaluate the effects of different levels of coverage/penetration of each data type on estimation results. A total of five estimation techniques developed by Cascetta and Postorino (2001), Castillo et al. (2008b), Parry and Hazelton (2012), Feng et al. (2015) and X. Yang et al. (2017) are tested with three data types (link counts, partial traces, and full traces) and two traffic assignment conditions (all-or-nothing and user equilibrium).

The result of this research highlights the uniqueness of each network situation and highlights the outcomes of each approach. The wealth of data does not directly equal better information for every methodology. The insights that each data type provides each estimation technique reveals different results. The findings of this research demonstrate and supports that an established testbed framework supports and enhances future O-D estimation scenarios as it pertains to general O-D estimation and extensions of existing techniques.