Real-time parametric modeling and estimation of urban traffic junctions

Abstract

An online dual estimation algorithm is developed to jointly estimate in real-time traffic quantities such as queue lengths, occupancies and flows, as well as the parameters of a macroscopic model of a signalized junction. These parameters include turning ratios and saturation flows, together with model uncertainties. The proposed novel methodology is based on the Expectation-Maximization algorithm, modified for real-time estimation, with a Kalman filter implementing the expectation step and a multivariate gradient-based approach for the maximisation step. The algorithm is validated by simulating the typical signalized 3-arm and 4-arm junctions. This work is aimed to form a part of the adaptive control loops for traffic light systems that are able to autonomously adjust with changing traffic conditions, so as to ensure efficient vehicle flows.