Efficient implementation of separable least squares for the identification of composite local linear state-space models
| Authors: | Borges José, Instituto Superior Técnico, Portugal
Verdult Vincent, Delft University of Technology, Netherlands
Ayala Botto Miguel, Instituto Superior Técnico, Portugal |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Nonlinear System Identification I |
|---|
| Keywords: | Identification, Nonlinear Systems, State-Space Models, Least Squares Algorithm, Efficient Algorithms |
|---|
Abstract
In this paper the efficient implementation of separable least squares identification of nonlinear systems using composite local linear state-space models is discussed. A full parametrization of system matrices combined with projected gradient search is used to identify the model. This approach has proven to be feasible, by reducing the iterations of the optimization algorithm and resulting in a better numerical behavior. Further enhancements to this approach result from using numerical tools, such as the QR-decomposition, and efficient approaches to compute the separable least squares matrices and gradients of the cost functions. By means of imulations it is shown that the use of these tools reduce the computation times for the identification problem.
