On the problem of output feedback control synthesis under uncertainty and finite-time observers (Q416811)

The present paper solves the problem of control synthesis for a linear system of ordinary differential equations with a linear observable output. The disturbances (uncertainty) are present both in the equations of system dynamics and in the observation equations. The differential equations describing the system dynamics contain control parameters that satisfy hard a priori constraints as well. The main problem solved in the present paper is to construct an admissible control that brings the trajectories of system into the neighborhood of the target set for arbitrary a priori unknown admissible disturbances. The problem of control synthesis is separated into two parts. The first part is a finite-dimensional guaranteed estimation problem, which is solved by a finite ``observer'' that evaluates the generalized online position of the system. The second part is an infinite-dimensional control synthesis problem under uncertainty in the class of generalized positions. The both parts use Hamiltonian formalism, i.e. of the corresponding equations of Hamilton-Jacobi type modified for control problems into the Bellman-Isaacs dynamic programming equations. The solution of problem for a linear system is reduced to finite-dimensional procedures, which permit one to replace the straightforward integration of Hamiltonian systems by the computation of set-valued functions that solve both parts of the original problem. The approximate computation of information and weakly invariant sets is carried out with the use of ellipsoidal calculus techniques. The suggested approach permits one to compute approximations to the above-mentioned sets efficiently and solve the control synthesis problem approximately. As an illustration, the problem of damping an oscillation system consisting of two coupled springs is presented. The numerical simulation results are presented for various types of disturbances (stochastic and deterministic) and for various sets of observable state coordinates.