Treating Uncertain Demand Information in Origin-Destination Matrix Estimation with Traffic Counts

摘要

This paper investigates the problem of estimating dynamic origin-destination (O-D) trip matrices in large-scale transportation networks, using automatic link traffic counts and uncertain prior demand information, to facilitate the real-time monitoring of network traffic conditions through intelligent transportation systems. The paper proposes and evaluates the performance of two algorithms to estimate the temporal distribution of trip departures used to produce dynamic O-D matrices, through several test experimentations in a real case study. The accuracy and the computing speed of the proposed algorithms is investigated in relation to different bounds, wherein trip departure estimates are sought to be included, levels of link count availability and network sizes. Results show that an optimization algorithm proposed for the fast estimation of trip departure rates with incorporated lower and upper bound constraints lead to dynamic O-D matrices with considerably increased reliability, in comparison to the algorithm that intrinsically excludes demand constraints.