A comparative study on filtering methods for online freeway traffic estimation using heterogeneous data

The paper addresses the problem of fusing multiple data sources in freeway networks. The incremental Unscented Kalman Filter (UKF) and the Unscented Information Filter (UIF) are developed based on cell-transmission model (CTM) of freeway traffic. The efficiency of the aforementioned methods are compared by applying to a toy network with the synthetic data obtained from microscopic traffic simulation. The results show that both are capable of fusing data with sufficient accuracy, even when only a small fraction of traffic information is provided in the data. However, the UKF works better in the case of correct noise covariances while the UIF has better performance when the covariances are incorrect.