2021 Volume 34 Issue 7 Pages 199-207
The projection type iterative learning identification method has several advantages such as: (i) no time-derivatives of input/output signals are required and (ii) it gives unbiased estimations. However, this identification method requires a parameterized model obtained by projecting a tracking error signal onto a finite-dimensional subspace, and the parameterized model must be estimated in advance. The model is called a parameter space representation in this paper. This paper presents an approach for estimating the parameter space representation required for the projection type iterative learning identification method. The proposed method is based on the projection of the estimated error signal onto the finite-dimensional signal subspace whose basis is determined by the closed-loop system with the estimated model. The benefits of the proposed method in comparison with existing method are illustrated with simulation studies.
