This thesis is dedicated to finding the best approximation of large-scale models of multiple-inputs, multiple-outputs linear dynamical systems, given an admissible complexity. New algorithms are presented to produce reduced-order models that aim at minimizing the approximation error based on the H
2 and H
∞ norms. For the former, the "Model Function" framework and efficient globalized approaches are introduced. For the latter, a rational-interpolation scheme available for single-input, single-output systems is extended. For all algorithms, numerical implementations are available within the MATLAB toolbox sssMOR.
«
This thesis is dedicated to finding the best approximation of large-scale models of multiple-inputs, multiple-outputs linear dynamical systems, given an admissible complexity. New algorithms are presented to produce reduced-order models that aim at minimizing the approximation error based on the H
2 and H
∞ norms. For the former, the "Model Function" framework and efficient globalized approaches are introduced. For the latter, a rational-interpolation scheme available for single-input, single-outp...
»