Bayesian fault detection method for linear systems with outliers

A novel approach for monitoring the accuracy of the Bayesian estimate of linear Gaussian state-space model is introduced, based on the monitoring of the propagation of the errors in the Kalman filter algorithm. The effect of the sensor errors on the Kalman filter estimate is explicitly computed and compensated. A marginalized particle filter is used to compute the posterior distribution of the sensor errors. Using a target tracking simulation it is shown that the proposed method has improved performance over the standard detection-identification-adaptation (DIA) method.