Short-Term Travel Time Prediction Using the Kalman Filter Combined with a Variable Aggregation Interval Scheme

Abstract

Data aggregation interval is important for reliable travel time predictions in probe-based systems. Where sufficient probes exist, a short interval can be used to minimize the time delay. However, in the opposite case, a short interval can cause unreliable travel time predictions due to small probes. Thus, the optimal aggregation interval may vary according to traffic flow conditions. This study suggests a methodology for selecting the optimal aggregation interval which varies according to a characteristic of probe travel time. The superiority of the proposed methodology compared to a conventional fixed interval is verified using DSRC probe data collected on a multilane highway near Seoul, Korea. The Kalman filter is adopted for a travel time prediction technique. As a consequence, the prediction accuracy is enhanced by approximately 40% compared to a fixed aggregation interval under free flow conditions.