Abstract

The iterative algorithm ICCALPV for state space identification of linear parameter varying (LPV) multivariable systems is here proposed. It involves three steps: i) LPV-MIMO model reformulation with affine parameter dependence for optimal future outputs prediction to calculate its state space and orthonormal basis through the past and present-future subspaces intersection via conditional canonical correlation analysis. ii) With the nominal state vector from the first step, the LPV state space model with affine parameter dependence is reformulated to obtain a new state vector and the Kalman gain. iii) For the second step estimated state vector, iterative calculations of LPV innovative state space model with affine parameter dependence are made. An application to the iterative identification of a LPV-MIMO system with three scheduling parameters is presented.