This study examines the accuracy of some basic travel time estimation models that employ simple speed-time relationships: instantaneous model, time slice model, and dynamic time slice model. Speed data are generated from extensively deployed dual loop detectors on freeways. A case study is reported for Melbourne's South Eastern freeway, which relies on real traffic data rather than simulation. Modelled travel times are compared to values measured in a timed number plate survey. The effect of smoothing loop data on the performance of speed-based models is investigated by comparing estimations from wavelets screened data with that from raw data. Other effects on the performance of this type of model, such as network length, speed aggregated time interval, and severity of congestion are also considered. Surprisingly, none of the three basic models accurately estimates travel time under the scenarios investigated. This contradicts the theoretical superiority of the time slice model and the dynamic time slice model. Model refinements are examined to improve their performance. (Author/publisher) For the covering entry of this conference, please see ITRD abstract no. E210528.
Abstract