A ROBUST MODEL FOR ESTIMATING FREEWAY DYNAMIC ORIGIN-DESTINATION MATRICES

Abstract

The purpose of this study is to develop an effective model and algorithm for estimating dynamic Origin-Destination demands for freeways. The primary challenge for this research subject lies in the fact that the number of unknown parameters is always more than the number of observable data, especially for a large network. Hence, the estimated O-D patterns may result in a large variance and insufficient reliability for use in practice. Besides, most existing approaches are grounded on the assumptions that a reliable initial O-D set is available and traffic volume data from detectors are accurate. However, in most highway network systems, both types of critical information are either unavailable or subjected to a significant level of measurement errors. To deal with those critical issues, this study has developed a set of dynamic models and solution algorithms for estimating freeway dynamic O-D matrices. The first extended model formulations can capture the speed discrepancy among drivers with an embedded travel time distribution function and the derivable interrelations between time-varying ramp and mainline flows. These formulations also feature their best use of the available mainline information and travel time function, and hence substantially increase the system observability with fewer parameters. The second component is an iterative algorithm that can be used to provide a reliable estimate of the initial O-D set, which is often unavailable in practice. The proposed algorithm first divides the network into small sub-networks to reduce the number of unknown variables, and recursively compute the O-D proportions for each sub-network to well capture the relations between the O-D demands and the input information. To deal with the constraints that the available data usually contain measurement errors, this research has developed an interval-based model for estimating dynamic freeway O-D demands. This component includes a set of formulations that converts each model input as an interval with its boundaries based on the prior knowledge. This study has performed sensitivity analyses and explored their potential for real-world application with the I-95 freeway corridor in Maryland. The numerical results under various traffic scenarios have indicated the promising properties of the proposed models and algorithms.