Accurate predictive urban point-to-point peak hour travel times : trial results.

This paper reports the results of a predictive point-to-point travel time information service trial carried out in peak periods over six weeks in urban Melbourne where processed data from a historical and real time traffic database was pre-delivered by cell phone to commuting drivers. The data set of results was examined for the causes and trends of the errors encountered. The low mean signed error of the prediction travel times showed insignificant bias. The important result of the trial predictive travel times was the mean absolute error of 3 minutes 45 seconds for the trips that averaged around 40 minutes.