Adaptive Estimation of Time-Varying Parameters using DREM

In this paper we present a method for estimating time-varying parameters in a linear regression equation. We combine local polynomial regression with dynamic regressor extension and mixing to independently estimate the parameters. During local polynomial regression, a time-varying parameter is approximated by locally constant polynomial coefficients. We propose to use the Bernstein basis instead of the commonly used monomial basis to improve numerical conditioning. A simulation example shows that our proposed estimator has improved performance compared to a similar method and allows a higher polynomial order.