This paper, after reviewing the technologies currently in use, describes a proposed model of predicting journey times concentrating on simplifying the structure of predictor as well as enhancing the predictability. Employing was considered of the relation between vehicle detector occupancy and journey time. As the preliminary study the hypothetical vehicle detectors by the use of video camera were placed at two blocks on an 1 km stretch of signalized artery. The results provided the basis for an estimate of journey time from the detector output.
The remaining important problem considered was that of calibrations. Since the relations between detector output and journey time is not only site-specific but also time-varying, the emphasis in calibration exercise should be predominantly based on on-line methods. To meet this requirement the modeling was directed towards the introduction of Kalman filtering theory to update the prediction, parameters.
A final study was made of the validity of the model on an 7.6km stretch of an artery which contains 18 traffic signals and 20 vehicle detectors. The journey time predictions were carried out every five minutes during two different time period of about 200 minutes each. The results highlighted the desirability of using the proposed model.
