INVESTIGATION OF CONTROL TECHNIQUES FOR A FREE GYRO STABILIZED MIRROR SYSTEM

Abstract

This project investigates the modelling and control of a free gyro stabilized mirror system. This system is a typical multi-input multi-output (MIMO) servomechanical nonlinear system which has two axes. The control of such a system is rather complex because there are strong cross-couplings among the axes. In general, the traditional control techniques such as Propotional-Integral-Derivative (PID) method cannot handle the system nonlinearities satisfactorily. In this thesis, Lyapunov Direct method is exploited for controller design, in addition, neural network modelling and adaptive neural network control are proposed. Lyapunov Direct method is widely used for the control of nonlinear systems. In this project, different Lyapunov function candidates are developed and based on them various control laws are derived for the gyro stabilized mirror system. Comprehensive numerical simulation studies are to be carried out and comparison results of different controllers are presented. Asymptotic tracking performance is guaranteed. Simulation results demonstrate the effectiveness of the controllers. Next, radial basis function (RBF) Neural Network modelling and adaptive tracking control for a free gyro stabilised mirror system are investigated. Several kinds of RBFs such as Gaussian, Hardy's multiquadric, Inverse Hardy's multiquadric, are employed for the RBF Neural Networks. Asymptotic tracking performance is guaranteed, and at the same time uniformly stable adaptation is achieved. Simulation results demonstrate the effectiveness of the adaptive controller. Furthermore, the proposed controller are implemented on the real plant and excellent actual implementation performance are obtained. This NN control law can control the gyro system very well with very limited knowledge of the system dynamics.