Robust Control System Design via Optimal, Classical and Adaptive Techniques

Linear quadratic gaussian (LQG) optimal control techniques can control linear multivariable systems that are precisely known [1,2]. Within the framework of the design, they achieve a best possible performance. However, if the plant is imprecisely known, actual performance can be intolerably poor. The guaranteed gain and phase margins for the case of the nominal plant may evaporate in the presence of unmodelled high frequency lightly damped dynamics, for example. To achieve improved trade-offs between performance of the control system for the nominal plant and robustness to plant uncertainties, classical frequency domain notions can be applied to give a modified stochastic model for the LQG controller design as in [3]. To give additional robustness to cope with time-varying uncertain unstable modes in the frequency band of control action, self-tuning regulator (STR) control techniques as [4,5] can be used to augment the LQG based design as in [6,7].