Abstract:

Poor adaptability of Kalman filtering algorithm may result in inaccurate prediction of expressway traveltime when the traffic flow between two expressway toll stations is non- stationary.In order to solve this problem,aprediction algorithm based on the equidistant interpolation and the Sage- Husa adaptive Kalman filtering is proposed.In the investigation,first,data from manual toll collection and electronic toll collection are merged together to cal-culate the average travel time.Then,the time series between real- time and historical travel time is reconstructedvia the equidistant interpolation,and a prediction model based on the Sage- Husa adaptive Kalman filtering is con-structed with the help of the least square method.Moreover,a prediction system of expressway travel time is deve-loped and is finally applied to the real- time prediction of the travel time between two toll stations.Case study resultsof an expressway section show that,in the three states,namely,the normal state,the accident state,and the holi-day state,the proposed algorithm is able to restrict the average relative error of all periods or of a random accidentwithin 7.5% or 10%,respectively.

Key words: expressway travel time, toll data, equidistant interpolation, Sage- Husa adaptive Kalman filtering