Analysis of model orders in human dynamics identification using linear polynomial and Hammerstein-Wiener structures

The identification of a human operator´s dynamic characteristics has been an important research subject for decades. Several solutions have been proposed to obtain the model of the human operator as a controller but usually the methods require separate tests to record suitable data for identification. That is, the models cannot be estimated during normal work. This paper focuses on identification of a linear or quasilinear human operator model based on normal task execution data. The performances of ARX, ARMAX and Hammerstein-Wiener models with different orders are compared in time and frequency domains.