Development of A Path Flow Estimator for Inferring Steady-State and Time-Dependent Origin-Destination Trip Matrices

Reliable origin/destination (O-D) data are critical to many applications in transportation planning, design and operations. Because of the high costs of and challenges in obtaining reliable O-D trip matrices from surveys or other direct sampling methods, estimating O-D trip tables from a readily available data source, traffic counts, provides an attractive, economical alternative. This project investigates one such an estimation method and implements it in a user-friendly software tool called Visual PFE TD. The developed O-D estimation tool can be used to obtain both static and dynamic O-D trip tables for traffic simulation studies, project evaluations, and transportation planning in a more streamlined and less time-consuming manner. For example, it has been used to obtain an initial seed matrix for Paramics' O-D estimator to speed up the latter’s O-D estimation process. A logit path flow estimator (LPFE) originally proposed by Michael Bell (1995) is adopted in this research for inferring both steady and time-dependent O-D trip tables. LPFE is chosen because: 1) it incorporates the logit-based route choice model while avoiding several difficulties encountered in the conventional bi-level formulation; 2) it avoids the difficult dynamic traffic assignment problem through decomposes the dynamic O-D estimation problem into a sequence of static problems, yet takes into account of queuing by linking the static problems across time with residual queues which can be carried over from one period to subsequent periods; and finally, 3) it has been validated in a number of scenarios as a potential tool to determine O-D flows and path travel times in various transportation networks. In this research, we extended the original LPFE formulation and improved the efficiency of solution algorithms, implemented both steady-state and time-dependent LPFE in an object-oriented programming (OOP) framework, tested the performance of LPFE using synthetic data and quantify the accuracy and reliability of its O-D trip table estimates. We also developed Visual PFE and Visual PFE-TD, the graphic user interfaces (GUI) for both static and time-dependent LPFE. Our test case studies show that LPFE is able to produce path flows and O-D travel demands that accurately match traffic counts under the logit traffic assignment assumption. We also found that information reflecting the spatial structure of travel demands (e.g., a historical O-D table) is of great value to the improvement of the quality of O-D trip estimates, and that LPFE can still produce satisfying estimates even when traffic counts are only available on a small portion of links, as long as such structural information is maintained in the base O-D table.