Upper bounds of large deviations in linear discrete-time systems: the robust statement

The paper is devoted to the study of the important effect of large deviations in linear dynamical systems with nonzero initial conditions. The study of transients is actual and practically significant direction in the linear systems theory. The common Lyapunov quadratic function for the family of systems with uncertainties and the invariant ellipsoids approach are used in the article as main technical tools. All the results obtained are also applicablefor non-stationary uncertainties: the only condition for an uncertainty is its spectral norm constraint. The analysis and design problems are considered, and the upper bounds of deviations for linear discrete-time systems with structured matrix uncertainties are obtained. The obtained results have the form of semi-definite programs, which are easy to solve numerically via standart software packages. Using the technique of linear matrix inequalities, the problem of minimization the magnitude of deviations while stabilizing the system via the linear static state feedback was investigated. Numerical simulations demonstrate the low degree of conservatism of the obtained approach. The results have a great potential for generalizations.