Statistically robust pseudo linear identification

Statistically Robust Pseudo Linear Identificationlectrical Engineeringis common to assume that the noise disturbing measuring devices is of aGaussian nature. But this assumption is not always fulfilled. A few examples arethe cases where the measurement device fails periodically, the data transmissionfrom device to microprocessor fails or the A/D conversion fails. In these cases thenoise will no longer be Gaussian distributed, but rather the noise will be a mix-ture of Gaussian noise and data not related to the physical process. This possesa problem for estimators derived under the Gaussian assumption, in the sensethat these estimators are likely to produce highly biased estimates in a nonGaussian environment.This thesis devises a way to robustify the Pseudo Linear Identification algo-rithm (PLID) which is a joint parameter and state estimator of a Kalman filtertype. The PLID algorithm is originally derived under a Gaussian noise assump-tion. The PLID algorithm is made robust by filtering the measurements througha non linear odd symmetric function, called the t/J function, and let the covarianceupdating depend on how far away the measurement is from the prediction. In theoriginal PLID the measurements are used unfiltered in the covariance calculation.