A SINGULAR VALUE DECOMPOSITION APPROACH TO MULTIVARIABLE GPC

Conference item

Generalized Predictive Control, GPC for short, provides an effective means for the control of scalar systems. This paper explores the properties of the inverse transforms of X, Y and Σ, which depend on the branch point locations of the Hermitian form of G, and subsequently proposes suitable algorithms for the computation of appropriate Laurent series expansions for each of the three matrices. The paper also considers the problem of scaling with the view to minimizing the length of the corresponding weighting sequences. More importantly, it is shown that these sequences are bi-causal and scaling can be used to minimize the anti-causal component.

Authors: Kouvaritakis, B; Rossiter, J; Chang, A; Engineers, I

• Publication status: Published
• Host title: INTERNATIONAL CONFERENCE ON CONTROL 91, VOLS 1 AND 2
• Volume: 332
• Issue: 332
• Pages: 104-109
• Publication date: 1991-01-01
• ISSN: 0537-9989
• ISBN: 0852965095