State estimation in continuous dynamic systems by the ellipsoid method (Q1385769)

The authors solve the problem of state estimation through ellipsoidal techniques for a linear dynamic control system. The current information on the system state is collected on the basis of on-line observations with unknown-but-bounded errors. The first step of the solution given here is to reformulate the problem as the estimation of an unknown initial state vector. For the intersection of the given ellipsoid of initial unknown states with the ``strip'' set produced by the observation process, the outer (with respect to inclusion) ellipsoidal estimate is constructed. The differential equations for the centers and the matrices of the resulting ellipsoid-valued estimating sets are derived. The properties of the estimates (such as convergence of the estimation algorithm) are studied.