Route travel time estimation based on seasonal model and Kalman filtering algorithm

Travel time is a key factor affecting driving behavior in urban area. The majority of existing research focused on the link or section based travel time estimation and assumed that drivers generally choose the ideal shortest or fastest path. Hence, it's difficult to acquire accurate time estimation such as delays occurred in signalized intersections. This paper analyzes the travel time between OD pairs and estimates the total travel time based on historical data by using the Kalman filtering algorithm. Considering an ordinary Kalman filtering algorithm is not enough to capture the characteristic of periodicity, a seasonal Kalman filtering algorithm is proposed for the further modeling and optimization. Finally, the floating car data from three continuous days in December, 2011 (Shenzhen, China) were obtained for an empirical study. The results indicate that in comparison to the traditional SARIMA time series model and the ordinary Kalman filtering algorithm, the proposed model captures the periodicity and time variations of total travel time, and thus has higher accuracy and fitness. Compared to the results from SARIMA and the ordinary Kalman algorithms, the mean absolute errors (MAE) of the total travel time from the ordinary Kalman filtering predictions decrease 37% and 52%, respectively. The other two error related indexes, namely, the root mean square error (RMSE) and the maximum relative error (MRE) both decrease significantly, which consequently indicates the effectiveness of the proposed method, and further verifies the modelling capability of the seasonal Kalman filtering algorithm.