Iterative learning identification: Dynamic parametrization modeling and comparison

Abstract

AbstractThis article elaborates the iterative learning mechanism for time‐varying system identification, and describes the learning algorithms that could achieve the consistent estimation for time‐varying parameters under persistent repetitive‐excitation conditions. A dynamic parametrization approach, in this article, is presented for modeling and analysis a general class of nonlinear systems. The derivations are conducted to give linear‐in‐the‐parameters models with time‐varying coefficients. The resultant models can be in a unified form, with the aid of the variable difference representation, and the iterative learning least squares algorithm and its variant are applicable for the purpose of parameter estimation. Moreover, a learning control scheme is adopted for demonstrating effectiveness of the dynamically‐parametrized modes, which are simulated and fully compared with the presented numerical results.