Kalman filter algorithm for short-term jam traffic prediction based on traffic parameter correlation

Abstract

A Kalman filter model considering the correlation property of traffic flow parameters is proposed to realize network short-term traffic flow prediction under jam traffic. The proposed state-space model of short-term traffic flow prediction is presented by solving the conservation equation using Lax-Wendroff scheme. In addition, the spatial-temporal characteristics of the traffic flow on urban expressway networks and the influence factors of on and off ramp are taken into account for flow rate prediction. The estimation algorithm of the proposed state-space model is designed based on the Kalman filter method. A region expressway network in Beijing is taken as an example to evaluate the performance of the proposed method. The results show that the maximum prediction mean absolute percentage error (MAPE) of the proposed Kalman filter model is less than 10% since the input of the Kalman filter model considers the impacts the spatial-temporal characteristics, and the mean of prediction MAPE is 7.96%. For the same predicted conditions, the mean prediction MAPEs of ARIMA and Elman model are 19.88% and 10.51%, respectively.