Parameter Identification in Linear Error Equations: Guaranteeing Output Error Boundedness

In this paper, we tackle the classical problem of estimating the parameters of an algebraic linear parameter model with the objective of solving the long-standing problem of guaranteeing boundedness of the output error independently from the growth of the regressors. Two solutions are presented. The first solution provides global results under the assumption that the time derivative of the regressor is available. The other solution disposes of the knowledge of the derivative of the regressor, and yields results that are valid in a semi-global sense, under the assumption that the regressor has a bounded growth. Simulation results provides an illustration of the proposed techniques in comparison with standard unnormalized and normalized gradient laws.