This paper presents a review of recent developments about gray-box identification. Different procedures are grouped based on how they handle auxiliary information in the identification process. Properties of NARMAX (nonlinear autoregressive moving average model with exogenous inputs) polynomial models applied in gray-box identification are presented. This paper discusses how to use auxiliary information about the process static function in order to aid in the model structure selection and parameter estimation. Such ideas are illustrated by means of two numerical examples, one simulated and one that uses real data. In both examples black-box and gray-box models are obtained and compared in order to highlight the main features of gray-box identification.
Non-linear identification; gray-box; NARMAX models; auxiliar information
