This report contains the results of a study in which two algorithms have been formulated for estimating the travel time of traffic flow on the basis of measurement data from video technology. Both algorithms are based on statistical processing and aggregation of travel time data of individual vehicles. The first algorithm yields an estimated actual travel time for the traffic flow as soon as there are enough measurements available to achieve a user defined level of statistical reliability. The second algorithm requires a user defined period as an input parameter. It periodically yields a new estimate of the actual travel time for the traffic flow together with its achieved statistical reliability. The influence of the parameters being used in the algorithms on the estimation quality has been analysed. For this reason, the algorithms have been implemented and applied to empirical data. An insight is gained into the performance of the newly developed `video travel time estimator'. For this reason, the second periodical algorithm has been compared to a simple `inductive loop travel time estimator'. Both methods yield highly comparable results in the case of free traffic flow. In congested traffic however, the video estimator provides systematically lower estimations for the average travel time. The video estimator reacts more quickly to strong fluctuations in the traffic flow than the loop estimator. (A)
Abstract