Robust Multivariable Controller Design via Implicit Model-Following Methods

摘要

This study applies the concept of implicit model-following to the design of a multivariable control system based on the Linear system model, Quadratic cost, and Gaussian noise process assumptions of optimal control. The design objective is to achieve improved controller robustness in the face of uncertainties arising from the differences between the simplified linear model used for controller design, and a more realistic truth model representation which includes higher-order dynamics, parameter variations, and nonlinearities. This report introduces the concept of implicit model-following and reviews the formulation of its incorporation into the design of a controller that consists of a Command Generator Tracker employing a Proportional-plus-Integral inner-loop regulator and, if required, a Kalman Filter to provide state estimates for feedback control (CGT/PI/KF controller). The theoretical background for the application of matrix singular value analysis to the study of controller robustness is presented, and a computer program is developed to apply this type of analysis of the general CGT/PI/KF controller configuration. (Author)