Estimation of traffic variables using a linear model of traffic flow

Abstract

Estimation of traffic velocity and the number of vehicles on adjacent sections of a limited access highway is examined. The method evaluated is based upon application of Kalman Filtering Methods to a linear state variable model of traffic flow. The estimator utilizes velocity and flow measurements at selected points along the highway. The flow measurement is a nonlinear function of the state variables and necessitates linearization about the one step ahead prediction of the state (extended Kalman Filter) or about nominal values of the state variables. It is shown that performance using Lincoln Tunnel data is comparable in either case to that of methods previously reported and provides a substantial savings in storage requirements. Also demonstrated is the fact that flow at an internal measurement point may be deleted from the observation vector without a serious effect on performance. This would arise, for example, if control of traffic were to be exercised at such a point.