California Partners for Advanced Transit and Highways

We present a method to predict the time thatwill be needed to traverse a certain stretch of freeway whendeparture is at a certain time in the future. The predic-tion is done on the basis of the current traffic situation incombination with historical data.We argue that, for our purpose, the current situation of astretch of freeway is well summarized by the ‘current statustravel time’. This is the travel time that would result ifone were to depart immediately and no significant changesin the traffic would occur. This current status travel timecan be estimated from single or double loop detectors, videodata, probe vehicles or by any other means.Our prediction method arises from the empirical fact thatthere exists a linear relationship between any future traveltime and the current status travel time. The slope and in-tercept of this relationship is observed to change subject tothe time of day and the time until departure. This naturallyleads to a prediction scheme by means of linear regressionwith time varying coefficients.Keywords— Prediction, Travel Time, Linear Regression,Varying Coefficients.