Adjustable neural network controller: application to a large segmented reflector

Abstract

A neural network controller (NNC) whose parameters are adjusted online is presented to control a class of multivariable linear systems. The plant to be controlled is assumed to be square (p inputs, p outputs) and almost strictly positive real (ASPR). The NNC is applied to a linear model of a large segmented space reflector and simulation results are presented. The ASPR condition is a strong condition, in general, but for the specific application of interest, i.e. control of flexible structures, the ASPR condition can be satisfied by an appropriate combination of the output variables (positions and velocities). As compared to other adaptive NNC reported in the literature, the proposed NNC is simpler and more suitable for real time applications.