Dedicated periodic scheduling sequences for LPV system identification

Abstract

In the past, research has focused on the identification of LPV systems in general form, allowing arbitrary variation of the scheduling sequence. This article proposes to restrict the scheduling sequence to a periodic sequence during the identification experiment, allowing the use of subspace techniques known from LTI system identification to identify a periodically time varying model. Unfortunately this periodic model will have a different state basis at each time step. Therefore we propose an algorithm to convert the periodic model to a uniform state basis, allowing to determine the underlying LPV system matrices. Sufficient conditions on the LPV system for this method to work are derived, and a simulation study is presented.