IMPROVEMENT OF PREDICTION SCHEME OF TRAFFIC FLOW ON URBAN STREETS BASED ON PIECEWISE STATIONARITY

Abstract

Traffic volume and time occupancy, measured by vehicle detectors on streets, plays a very important role in a traffic control system. This paper is directed to the improvement of pridiction scheme using the Kalman filter with a constant term in stead of the exponentially smoothing method, which is the commonly-used technique in traffic control system. Contents are as follows.1) Characteristics of measurement noises.2) System identification procedures for the proposed method, assuming that the traffic variation in a day would be partitioned into three stationary periods.3) Relationship between Kalman's gains and smoothing constants.