Mixed discrete/continuous specifications in sampled-data \(H^2\)-optimal control (Q1129708)

The authors propose a new approach to the solution of the MDCS (mixed discrete/continuous specifications) \(H^2\) problem. The advantage of this approach lies in the simultaneous treatment of both continuous-time and discrete-time (hence mixed) performance specifications. The sampled-data \(H^2\) problem is treated directly in the lifted domain in terms of the infinite dimensional parameters of the lifted plant. The intermediate steps are avoided and all computations required are finite-dimensional. A new approach for handling the so-called causality constraints directly in terms of controller parameters is proposed. A closed-form state-space solution to the constrained \(H^2\) problem is derived and is shown to have the same computational complexity as for the unconstrained one. Finally, a novel method for the computation of matrices involving infinite-dimensional integral operator is proposed. It is demonstrated that all the computations required for the solution to the MDCS \(H^2\) problem involve computations of only two matrix exponentials.