Estimation of urban arterial travel time based on dynamic bayesian network

Abstract

Estimation and prediction of urban arterial travel time is important for the development of transportation field, especially measures and methods of traffic management and control. This article presents a modeling method for estimating and predicting arterial travel time based on Dynamic Bayesian Network. The probability distribution model of travel time conditioned on different traffic states is constructed by combining the traffic flow theory at intersections and Dynamic Bayesian Network. The Cell Transmission Model that was applied to accurately describe the evolution process of state variables and the travel time observations were obtained by trajectory data, then the relationship between the traffic state and travel time was established by Dynamic Bayesian Network. The EM algorithm and particle filter algorithm are used to iterate the static parameters of the road (such as traffic signal parameters) through machine learning, so as to estimate and predict the distribution of arterial travel time in real time under the current traffic state. The data of Changping District of Beijing were selected to verify the validity of the probability distribution estimation of travel time in this paper, then the direction and value of subsequent improvement and application were proposed.