An estimation filtering for time-varying systems with unknown faults

This paper proposes a discrete time-varying finite memory structure estimation filter with the consideration of unknown faults using only finite observations on the most recent window. The proposed estimation filter is obtained from a linear combination of the non-faulty filtered state estimate and the filtered fault estimate. The non-faulty filtered state estimate is obtained from assuming that there are no faults. The filtered fault estimate is obtained from the residuals of the non-faulty filtered state estimate as observations. Computer simulations are performed for the indoor positioning mechanism in wireless sensor networks to demonstrate the validity of the proposed estimation filter. The proposed estimation filtering based approach shows better performance than the Kalman filtering based approach.