An Hinf Filter for Parameter Estimation with Enhanced Performance

Abstract

In this paper, we address the problem of parameter estimation in the context of classical linear regression forms when persistency of excitations condition do not hold and under the presence of noise. It is well known that when PE conditions are not met, parameter convergence to the true parameter values is hindered, which can further complicate the task of system identification. In order to cope with such adversities, here the robust Hinf filter is designed along with the DREM technique to enhance the performance of the filter when insufficient excitation are presented. Comparison analysis with classical estimators such as the gradient and Kalman Filter are performed showcasing the advantages of the proposed method.