The prediction model of near future speed fluctuation was constructed, and the model applied real traffic flow on the motorway. The model is consisted of an auto-regressive method. Parameters of auto-regressive model were sequentially identified using a Kalman filtering method. The time lag in predicted traffic flow fluctuation, which was usually appeared in the past studies, was almost eliminated introducing historical traffic flow patterns. The historical patterns have been formulated by accumulation of about three-year vehicle detector data. The result shows that the prediction model proposed here might be enough in practical use in the traffic surveillance and information system on rural motorway. However, the model has some shortcomings, especially number of classifying the historical patterns. A model modified in the future study must do travel time prediction. (A*)
Abstract