State Estimation via Switching Observer for Systems with Outliers

Abstract

This paper proposes a method for robust state estimation for systems with irregular sensing failure. It is often experienced that non-contact sensors give abnormal signals (outliers) due to sensing failure. To compensate their effect, the proposed estimator switches observer gains to select normal signals based on the current estimation of the sensor signals. Stability of the error system is guaranteed in terms of switched Lyapunov functions. Both numerical and experimental results show the effectiveness of the proposed method.