Stability of Parameters Estimated by the Output Least Square Method

Abstract

The stability of the output least squares parameter estimation technique is investigated. The maximum size of the admissible parameter set is given, in order to ensure uniqueness and stability of the estimated parameter in presence of model and measurement errors. A trade off is shown to exist between the size of the admissible parameter set and the magnitude of the model and measurement errors. For finite dimensional parameters, the theory is shown to ensure local identifiability of a parameter as soon as the mapping parameter → measurements is injective.