On Kalman Filter and Path Hypothesis Performance Assessment in Railway Localization Filter Banks

Kurzfassung

Map-supported Kalman filter (KF) banks are a viable approach for online rail vehicle localization using onboard GNSS, IMU, and velocity sensor data in combination with static digital track network data. The online use dictates that the management of path hypotheses (simply put: connected sequences of tracks treated like a single longer track) and the execution of on-path KF (associated with said path hypotheses) need to be performed in parallel at each time step. The paper briefly introduces the KF bank and some relations to KF and particle filter (PF) approaches. As main contribution, relevant details for the KF and path hypothesis performance assessment in the KF bank are discussed.