Identification and Predictive Control of ARX Model by p Norm Minimization

J. Pekař, J. Štecha, and V. Havlena (Czech Republic)


Identification, Predictive Control, pnorm, ARX model, it eratively reweighted Least Squares, Linear Programming


Real time system parameter estimation from the set of input-output data is usually solved by minimization of quadratic norm errors of system equations - in literature known as least squares (LS) or its modification as total least squares (TLS) or mixed LS and TLS. It is known, that uti lization of p norm (1 p < 2) instead of quadratic norm suppress the wrong measurements (outliers) in the data. In the paper this property is shown for different norms and it is shown that influence of outliers is suppressed if p 1. Also optimal predictive control utilizing p norm minimiza tion of the criterion is developed and simulation results show properties of such control.

Important Links:

Go Back