A Decoupled Data-Driven Strategy for Estimating Parameters with Nonlinear Dependence

In this work, we introduce a decoupled data-driven methodology for the online estimation of parameters with nonlinear dependence. Common estimation methods for nonlinear systems rely on the incorporation of the mathematical model under study to the estimation structure. Other variational approaches trend to be more computationally expensive. In contrast, data-driven estimation methods, based on the retrospective cost model refinement (RCMR) algorithm, have demonstrated an efficient and adequate ability in the estimation of parameters. However, a better performance is observed when the estimation structure is divided into several independent sub-structures. The proposed decoupled RCMR strategy is explained and compared with the original structure using numerical examples. The results of the numerical examples show that the proposed strategy exhibits a good trade-off between the velocity of convergence and error minimization.