摘要：AbstractThe major bottleneck of Linear Parameter Varying (LPV) subspace identification methods is the exponentially scaling parameter count during the first estimation step. State-ofthe-art subspace methods which perform well for Linear Time Invariant (LTI) systems, result in this bottleneck when they are extended to LPV systems. This has motivated us to search for LTI methods which retain favourable parameter counts when extended to LPV systems. In this paper, we propose a method based on tensor regression which exploits the additional structure of the underlying problem. The proposed method can be used to obtain consistent estimates with variance comparable to estimates from state-of-the-art methods, and is expected to be extendible to LPV systems while avoiding exponentially scaling parameter counts.
