A Finite Time Convergent Least-Squares Modification of the Dynamic Regressor Extension and Mixing Algorithm

The recently proposed Dynamic Regressor Extension and Mixing (DREM) algorithm can be used to estimate the parameters of structured uncertainties contained in the mathematical model of a plant. In order to provide an adaptation that is less sensitive to the unavoidable mismatch between a plant and its model a least-squares based modification of the DREM estimator is proposed in this paper. The modified estimator yields significantly better estimation results as illustrated by the conducted real-world experiment and its parameter estimates also converge within finite time.