Hide/Show Apps

Outlier robust filters and their multiple model extensions

Şahin Bozgan, İlknur
Kalman filter (KF), which is an algorithm that is utilized to estimate unknown variables based on noisy measurements, has been successfully employed in many applications such as navigation, control, signal processing and target tracking. It is the optimum Bayesian filter in terms of mean square error (MSE) for linear Gaussian state-space models (SSMs). However, in many real world applications, the performance of KF degrades due to the presence of outliers in noises. Motivated by this problem, several algorithms have been proposed to provide robustness towards outliers. In this thesis, existing outlier robust filters are investigated regarding their theoretical derivations, validity of their assumptions, and performances. Furthermore, multi-model extensions of the filters are derived and the merits of the algorithms are illustrated in simulations.