Author(s)： Pan Qin, Ryuei Nishii, Tadashi Nakagawa and Takayoshi Nakamoto

J. Math-for-Ind. 2A (2010) 109-114.

• File： JMI2010A-11.pdf (180KB)

Abstract
We consider the modeling problem for time-varying systems by Auto-Regressive models with eXogenous variables (ARX) models. To track the variations of time-varying systems, we propose a new Recursive Penalized Weighted Least Squares (RPWLS) method to estimate the ARX models. Furthermore, by virtue of Generalized Information Criterion, the proper ARX models by RPWLS are selected. Numerical examples are provided to verify the performance of the proposed RPWLS method.

Keyword(s).　 ARX model, Time-varying systems, GIC, Model selection