EVALUATION OF A MODEL FOR THE REAL-TIME SHORT-TERM PREDICTION OF TRAFFIC CONDITIONS ON MOTORWAY NETWORKS

Within the DRIVE II project DYNA, a Statistical Traffic Model (STM) has been developed, aimed at producing real-time predictions of traffic flow on motorways. As route choice is not explicitly modelled, the applicability of the STM is restricted to the very short term predictions (5 - 20 minutes ahead). As a result of several problems, the STM did not perform well in the evaluation that was carried out in the DYNA project. Some of the problems were related to the available traffic data that was used as input to the prediction model. Others had to do with the model itself. For this reason, a further evaluation of the STM was carried out last year, where most of the previously encountered problems were avoided. An extensive data set of a 16 km stretch of motorway was available, consisting of flows, speeds and occupancies per lane per minute. In addition, some floating car travel time measurements were available. It turned out that the calibration of the STM still poses problems. As a consequence the evaluation results still are of an intermediate nature, and leave room for improvement. The general impression with the STM so far is that it is possible to outperform naive and historical type of predictions and that reasonable predictions of travel time variation during queue build-up and dissolving are possible. Further work needs to be done on the calibration methodology and on improving the model equations. These and other questions will be addressed in the DRIVE III project D'ACCORD.