PREDICTION OF DENSITIES BASED ON SCARCE TRAFFIC FLOW INFORMATION

Abstract

Traffic fluctuations are always evident in highways or urban arterial networks that consist of some signalized or unsignalized intersections. Traffic conditions may change as a result of changes in peak timing flows, miscellaneous incidents, variable weather etc. A constant challenge of traffic engineers and professional peoplethat are closely related to traffic control and management remains the identification of parts in which the traffic situation changes and the provision of information about traffic parameters. Prediction of density parameter in short time intervals is important in lots of traffic modelling and control strategies of freeways and urban arterials. For more, the possession of short time density values for particular parts of the freeway segment, plays an importation role on providing drivers with information about events or traffic incidents. Not always the traffic flow amounts are possible to be measured in any part of the segment we are interested in. Thus maycome due to the lack of detector coverage or detecting defects even if it exists. The purpose of this paper is twofold. First is the development of a discrete model so called Cell Transmission Model (CTM) [1,2] that is analogue with approximation of the LWR hydrodynamic model of traffic flow. The second one is the integration of the Kalman Filter [3] to the mentioned model, in order to increase the accuracy of the modeled traffic densities. A Kalman filter (KF) is a recursive algorithm that uses only the previous time-step’s prediction with the current measurement in order to make an estimate for the current state. KF does not require previous data to be stored or reprocessed with new measurements. At everyiteration, the KF minimizes the variance of the estimation error, making it an optimal estimator if linear and Gaussian conditions are satisfied. In order to highlight the difference between accuracies of the predictions of the densitiesobtained by pure CTM model and by application of the Kalman Filter on it, a short highway segment with simple composition is chosen as object of study. The segment comprises of a ramp and the number o lanes are the same during its entire length.