Fast Tensor Product Model Transformation for Higher Dimensional Problems

Tensor Product (TP) model transformation method was proposed recently as an automated gateway between a class of non-linear models and linear matrix inequality based control design. The core of the TP model transformation is the higher order singular value decomposition of a large sized tensor, which requires high computational power that is usually out of a regular computer capacity in case of higher dimensionality. This disadvantage restricts the applicability of the TP model transformation method to linear parameter varying state-space models with smaller number of state values. The aim of this paper is to propose a modification of the TP model transformation. The proposed version needs considerable less computational effort. The paper also presents a 6 dimensional problem to show that the execution of the original TP model transformation fails in higher dimension, but the modified transformation leads to solution.