摘要：AbstractIn this paper, the origin of the curse-of-dimensionality for state-of-the-art Linear Parameter Varying (LPV) subspace methods is investigated and a novel solution based on tensor regression is presented. It is shown that the curse-of-dimensionality arises because an highlystructured space, spanned by the scheduling sequence at different time steps, is vectorized in order to allow for linear regression. The inherent structure is lost in this process. A novel method based on tensor regression is presented which is consistent and does not suffer from the curse-ofdimensionality. Simulations show that the novel method has superior performance with respect to state-of-the-art LPV subspace techniques by looking at the variance and bias of the estimates.
