Traffic assignment in stochastic networks and trip table estimation: Revised route choice behavior assumptions and coordinated solutions

Shihsien Liu, Purdue University

Abstract

In recent years, considerable research has been devoted to modeling and solving the traffic assignment problems in a stochastic network. Nevertheless, the O-D table is still obtained from either survey data or the traffic assignment models in a deterministic network. Survey data collection for O-D table is costly and time consuming, but an O-D table can be soon out of date because of changes in land use or the network. Moreover, the traffic assignment problem has been extended to stochastic network conditions, but the models produced to date either have unproven behavioral assumptions that yield unreasonable solutions or are difficult to implement in large scale networks because of the need to enumerate all reasonable paths. The purposes of this thesis are to investigate the suitability of the modeling in the O-D estimation and in the stochastic traffic assignment problem, and algorithm to obtain the O-D table and traffic assignment flow pattern in a stochastic network. Two model formulations and their algorithms are developed. A bi-level programming method using maximum likelihood and a logit approach produces an O-D table estimate and a calibrated $\Theta$ parameter at the same time. The stochastic traffic assignment formulation is transferred from a static traffic assignment problem extension with the addition of a new tripmaker route choice behavior assumption. The algorithm developed in this thesis can solve the stochastic traffic assignment problem in a large scale network with polynomial computational effort. Two computer models were developed in FORTRAN 77. A mathematical-programming-based exact algorithm for estimation of an O-D table and a heuristic algorithm for stochastic traffic assignment problem were implemented. These algorithms were tested on small networks with promising results. The algorithms were also implemented on a larger network near Purdue Campus. The results verify that drivers do not have perfect road information in the study area and that the network demonstrates congestion effects during peak hour. After a one-way street system was induced in the study area in May 1991, users displayed a more dispersed route choice pattern than before. These findings may indicate a new tool that analysts can use to characterize transportation network structure and driver behavior.

Degree

Ph.D.

Advisors

Fricker, Purdue University.

Subject Area

Civil engineering|Operations research

COinS