Joint state and parameter estimation for a macro traffic junction model

Abstract

To contribute towards the autonomic properties of traffic light systems, a self-estimation algorithm is presented to jointly estimate the states describing the traffic flow dynamics in a junction under different traffic conditions, with model parameters and measurement and process noise. The novel proposed algorithm, based on the Expectation-Maximization algorithm, makes use of a sliding window over time, to estimate junction traffic conditions in quasi real-time. The proposed model and algorithm are validated on a signalized 3-arm junction within the traffic network of Malta.