DECENTRALIZED CONTROL WITH OFF-LINE MULTI-MODEL IDENTIFICATION AND CONTROLLED REGION PARTITION

PERŮTKA, K.; BOBÁL, V.

Abstract

The paper deals with decentralized control of MIMO (multi input multi output) systems employing a set of SISO (single input single output) controllers running simultaneously. The suitably chosen identification algorithm at adaptive decentralized control is crucial for control, because badly chosen method of recursive identification in combination with other influences can change the system behavior in the way we do not want to. Therefore the requirement of some innovative approach to identification in those systems appeared. The whole part of expected course of control is divided into subparts of different lengths according to the original criterion but each subpart has the same length for all subsystems. Each subpart is identified with several methods, the results are modeled and compared with the open-loop response, the correlation coefficient is the criterion which method of identification is finally chosen for the specific subpart, different subsystems may employ different method of identification. The verification was realized in MATLAB environment by simulations.

Coresponding author e-mail: kperutka[at]ft[dot]utb[dot]cz

Session: Robust and Adaptive Control