Salnikova N. Robust algorithms of ellipsoidal estimation with incomplete step.

The dissertation is devoted to the development of new robust procedures of guaranteed and fuzzy ellipsoidal estimation which are workable(efficient) under conditions of unauthenticity of a priori information on controlled plant model and characteristics of a measurement channel. The method of a synthesis of parametric family of usual and fuzzy ellipsoidal algorithms of estimation with varied step parameter is developed in the dissertation. Due to rational step choosing, ellipsoidal observers asymptotically converging to the estimated state vector and providing reception of point estimates, which are optimal on additive criterion over a whole observation interval, are synthesized. On the basis of the method of fuzzy ellipsoidal estimation with incomplete step, robust state observation algorithms for linear dynamic plant under conditions of inexactly known plant parameters, infringement of a priori hypotheses on it, absence of the a priori information on the initial state vector, and also at infringement of a priori hypothesis about absence of noise in the measurement channel, are developed. The problem of state observation of a linear dynamic system operating in continuous time is solved in the class of determined and fuzzy ellipsoidal algorithms with incomplete step. Necessary conditions of convergence of the offered algorithms are obtained.