Traffic Estimation Based On Point Process Models

Two families of models for the relation between detector data and the flow of traffic on urban streets are developed. The first describes the relation between the queues at traffic signal and data from nearby detectors. The second model headways statistics for urban traffic thereby relating detector data to the passage of platoons of vehicles. Recent results in the theory of point processes then give the non-linear minimum error variance filters/predictors corresponding to these models. It is shown that these optimal estimators are computationally feasible in a microprocessor. The algorithm also performs very well when tested using data derived from UTCS-I traffic simulation.