Model conversions of uncertain linear systems via the interval Padé approximation method (Q1908836)

Given a continuous time system with state space triple \((A, B, C)\), one can obtain an equivalent discrete time system with state space triple \((G, H, C)\) by sampling the continuous system with a constant period \(T\). The relations between \((A, B)\) and \((G, H)\) involve the matrix exponential \(\exp (AT)\). The latter can be approximated by a Padé approximant of type \((1/1)\) or \((2/2)\). For uncertain systems, as considered in this paper, the matrices \((A, B)\) are uncertain. Perturbations \((\Delta A, \Delta B)\) will result in the corresponding perturbations \((\Delta G, \Delta H)\). The problem is to estimate the uncertainties in \((G, H)\) in terms of the uncertainties in \((A, B)\) and conversely. To this end, interval arithmetic is used in combination with the Padé approximation method for the exponential, to get tight bounds of these perturbations. A numerical example shows the efficiency of the proposed method.