On dynamic uncertainty estimators

The paper is concerned with state predictors that include a disturbance dynamics capable of estimating the disturbance to be rejected. The disturbance dynamics is driven by an unknown input signal, the uncertainty input, which is the output of a dynamic feedback driven by the model error (plant minus model output). As an extension of classical observers, the paper shows the advantage of designing a dynamic feedback. A dynamic feedback can be designed to fit the structure of the uncertainty and has the advantage of increasing the relative degree of the state-predictor transfer function, without penalizing the sensitivity itself. A higher relative degree increases the high-frequency rejection rate, thus impeding neglected dynamics components of spilling through control signals to the detriment of stability. A simulated case study shows the better performance of dynamic estimators.