Modeling Travel Time Variations on Urban Links in London

摘要

This research focuses on developing an econometric framework to combine data available from differentsources to investigate the causes of traffic congestion and identify the key factors contributing to traveltime variations in Central London. Non-linear latent variable regression models that explicitly accountmeasurement errors in the data are developed in this regard combining data extracted from AutomaticNumber Plate Recognition (ANPR) Cameras and Automatic Traffic Counters (ATC). This proceduresignificantly differs from previous researches on this area which were primarily based on traffic flow dataand ignore measurement errors. The results indicate that non-linear latent variable regression models caneffectively explain travel time variations on a regular day using variables related to vehicle type, trafficdensity, and traffic composition. Test results indicate that the proposed framework for correctingmeasurement errors yield significant improvements over base models where such errors are ignored. Thefindings from our study provide valuable insights on influences of different factors on urban trafficconditions. Further, the model framework is general enough for application in other cases where trafficdata have similar measurement errors.