Observers for linear time-invariant systems

Abstract

In optimal control theory, the design of a state feedback control law requires the availability of the entire state vector. However, in most of the control systems the measurements cannot provide the entire state vector. Then, the non-available state variables must be estimated. The estimated state variables are combined with the already available state variables to be substituted in the feedback control law. In this thesis, the observers that are designed to estimate the non-available state variables are considered. In the deterministic case, both for the continuous-time and the discrete-time systems, special emphasis is given to the design of minimal order observers. The stochastic optimal reduced order observer-estimator and the suboptimal minimal order observer are suggested as alternatives to the Kalman filter. These altertiative designs are compared and the inter-relationships are discussed. Furthermore, a computer package programhas been developed for computer aided design of such observers for practical implementation. The user must only supply the necessary data to obtain the values of the parameters of the observer of interest.