| Article ID: | iaor19972394 |
| Country: | Netherlands |
| Volume: | 13 |
| Issue: | 1 |
| Start Page Number: | 73 |
| End Page Number: | 85 |
| Publication Date: | Jan 1997 |
| Journal: | International Journal of Forecasting |
| Authors: | Van Arem Bart, Van Der Vlist Martie J.M., Muste M., Smulders Stef A. |
| Keywords: | forecasting: applications |
In order to produce traffic predictions up to 1h ahead for a motorway network, it is essential to produce reliable estimates for the travel time over motorway sections. This article describes a model for estimating travel times on motorway sections 3-5km in length. The model uses measurements from inductive loop detectors and is based on a linear input-output ARMA model representation. The model is evaluated using results from a field study. The evaluation data set covers traffic situations with borh ‘normal’ congestion and congestion due to accidents or incidents. The observed and estimated travel times are very close in cases of normal congestion. Travel time estimates in cases concerning accidents showed large deviations, but in these cases there were no travel time observations to verify their accuracy.