Delay-dependent robust \(H_{\infty}\) control of uncertain linear systems with time-varying delays (Q814066): Difference between revisions

Consider the delay differential equation \[ \dot{x}(t) = Ax(t) + A_dx(t-\tau_1(t)) + B u(t) + B_d u(t-\tau_2(t)) +E w(t), \] where \(x\), \(u\), and \(w\) denote the state, input, and disturbances, respectively. The \(\tau_i\)'s denote the time varying time delay. Together with this differential equation, there is the measurement \[ z(t) = Cx(t) + C_dx(t-\tau_1(t)) + D u(t) + D_d u(t-\tau_2(t)) + F w(t). \] The system matrices \(A, \ldots, F\) are constant and lie in the polytopic set \({\mathcal P}\). Furthermore, it is assumed that the functions \(\tau_i\) are bounded, and \(0\leq \dot{\tau}_i(t) <1\). Under these systems assumptions, the authors want to construct a state feedback \(u=Fx\) such that for all (possible) delays and systems matrices, the closed loop system is stable and the \(H_{\infty}\)-norm from \(w\) to \(z\) is less than \(\gamma\). Using linear matrix inequalities (LMI), they obtain a sufficient condition. These LMI's involve a.o.\ the corners of the system polytope. For the example of \textit{E. Fridman} and \textit{U. Shaked} [Int. J. Control 76, No. 1, 48-60 (2003; Zbl 1023.93032)], their result is less conservative. Similar results are obtained for delay difference equations. For this class of systems, the condition on \(\dot{\tau}_i(t)\) can be removed.