Bus travel time prediction based on time-varying adaptive Kalman filter method

Abstract

To avoid the efficiency reduction of transit signal priority (TSP) control caused by inaccurate prediction of bus travel time, a time varying adaptive Kalman filter model is proposed in this paper. To present the transit speed fluctuation characteristics caused by various traffic factors, a model to calculate the dynamic factor is built based on weighted moving average time series method. With the introduction of dynamic factor, a time-varying adaptive Kalman filter model is established to predict bus travel time. This model is compared with other classical ones in experiment. The results show that the mean absolute percentage error (MAPE) of prediction is 2.52%, which is better than the basic Kalman filter model and time series model. Therefore, this method could not only consider the transit speed fluctuation but also significantly eliminate the detection deviation, which contributes to accurate prediction of bus travel time in transit signal priority control.