Model Transformations for State-Space Self-Tuning Control of Multivariable Stochastic Systems.

摘要

A long division method is developed in this article for finding the similarity transformation matrix, which transforms the estimated observable state to the controllable state, and for transforming the estimated left matrix fraction description (MFD) to the right MFD, without utilizing the realized high-dimensional system matrices. The relationships between the realized system matrices in the state-space descriptions and the quotients and remainders of the long division of two polynomial matrices in the MFDs are developed. The proposed new procedure reduces the computational difficulties arising in the development of the state-space self-tuner to be more amenable to on-line adaptive control of multivariable stochastic systems. Keywords: Reprints; Control theory; Algorithms; Mathematical technique; Linear systems; Multivariable systems. (KR)