A comparative study of robust parameter estimation methods

The problem of robust system identification when no statistical information is available on the noise, but only a bound on it is known, is addressed. A unified presentation of two classes of ellipsoidal outer-bounding algorithms is made, and a comparison with robust projection algorithms is carried out from a theoretical point of view. It is shown that these two families of algorithms can be written in a very similar form, the main differences between them being the definition of the dead zone and the computation of the correction gain. >