A discrete parameter-driven time series model for traffic flow in ITS

Abstract

The traffic condition information is critical to signal control and traffic queue management for intelligent transportation systems (ITSs). This work formulates the dynamic traffic flows by a parameter-driven discrete time series model. A modified Expectation-Maximization (EM) algorithm is proposed to estimate the parameters of the model. The conditional expectation given y in the E step of the EM algorithm is replaced by the marginal expectation. Applications of the model and the algorithm are illustrated by analyzing the traffic flow data. Through the data collected by the detectors, the traffic condition about the network can be estimated, smoothed and predicted. This model can be embedded in a chip for signal control and queue management procedures of ITSs.