Travel time prediction on urban network by integrating stop-line detector data, signal controller data and probe information
In urban network, detectors are generally used for signal control. If the data from detectors are also used for travel time estimation and fed back to the traffic management and information systems then it will be a significant contribution for managing urban network. For instance: the model can be applied for performance evaluation of the system and Level of Service of different intersections. It can also be applied for ITS applications such as, Advanced Traveler Information Systems and Public Transport Priority System. This paper presents the methodology for real-time travel time estimation on signalized urban network which is based on the classical principle of cumulative plots at the entrance and exit of the link. In urban environment there are potential sources of deviations in the cumulative plots, for instance: a) mid-link sources and sinks (e.g. parking) b) detector counting error c) unknown proportion of counts for the study link from detector on shared-use lane at the upstream intersection, etc. The deviations in cumulative plots are accumulative and if such issues are ignored then deviations can grow exponentially with time. In the proposed model, cumulative plots are accurately estimated through analytical modeling and above mentioned issues are addressed by developing the following techniques for different traffic flow conditions. Detector data and signal controller data is integrated with the following constraints at the stop-line detector location: a) Detector counts aggregated over signal cycle period are less than the lane capacity for under-saturated condition; and are close to lane capacity for saturated and oversaturated conditions; b) Cumulative arrival at the end of each signal green interval is equal to cumulative departure for under-saturated condition; and is more than cumulative departure for over-saturated condition. For saturated or over-saturated conditions, accuracy is enhanced by data fusion with probe vehicle. Probe vehicle provides time when the vehicle is at upstream and downstream intersections. This information is used as a reference for adjusting cumulative plots generated by the integration of detector data with signal controller data. The result from model testing under control environment for scenarios with 10% demand for midlink source indicates that, the 5th percentile of the accuracy for travel time estimation is generally more than 95% for under-saturated condition with no probe; and for saturated or over-saturated conditions with 3% probe. This model has overcome the issues related to models solely based on cumulative plots or probe vehicles with significant improvement in accuracy and has the potential for real-time travel time estimation on urban network. The model is to be validated with the real data obtained from a study site with 12 consecutive intersections along 1.5 km area at Lucerne city. Signal controller and detector data is obtained from VS-PLUS controllers installed in the area. Actual travel time is obtained from manual number plate survey. The final paper will contain the results of the validation.