A support vector machine-based method for LPV-ARX identification with noisy scheduling parameters

In this paper, we present a method that utilizes support vector machines (SVM) to identify linear parameter-varying (LPV) auto-regressive exogenous input (ARX) models corrupted by not only noise, but also uncertainties in the LPV scheduling variables. The proposed method employs SVM and takes advantage of the so-called “kernel trick” to allow for the identification of the LPV-ARX model structure solely based on the input-output data. The objective function, as defined in this paper, allows to consider uncertainties related to the LPV scheduling parameters, and hence results in a new formulation that provides a more accurate estimation of the LPV model in the presence of scheduling uncertainties. We further demonstrate the viability of the proposed LPV identification method through numerical examples, where we show that higher best fit rate (BFR) can be achieved under realistic noise conditions using the proposed method compared to the method initially proposed in [6].