Comparing Recursive Estimators in the Presence of Unmodeled Dynamics

Indirect recursive identification algorithms have been suggested for robust control design in the literature. The purpose of this contribution is to clarify that such algorithms need proper calculation of the update direction in undermodeled situations. An indirect pseudolinear regression is compared to a (more suitable) indirect recursive prediction-error method. Their asymptotic properties are examined through simulations, using a specific control structure.