Contractive Method for Identification of Nonlinear Regressors in Dynamic Systems
Rodolfo Horacio Rodrigo, Héctor Daniel Patiño, Gustavo Alejandro Schweickardt · 2020 IEEE Congreso Bienal de Argentina (ARGENCON) · 2020
In this work, a method is exposed that allows the calculation of the parameters of non-linear regressors, of dynamic systems with intervals. The proposed method is a contractive method applicable to homeomorphic transformations. This requirement is met by many regressors used to model nonlinear dynamic systems. It allows to apply a model that can establish and make compatible four aspects: prediction horizon, uncertainty, data required for adjustment and validity domain. In other words, capturing the trend with a validity domain greater than that obtained with an adjustment of a greater number of parameters and less uncertainty. It has a robust character of its convergence. As a counterpart, it is a costly method in computational terms, but this circumstance is saved by its characteristic of parallelizable algorithms. In this way, it can be applied in a control system, with massive shutdown hardware.