LMI Computations for Multiplier Optimization

Ulf Jonsson, Lund University

Robustness problems are considered, where structural information about uncertainty is given in terms of an infinite class of integral quadratic constraints (IQC's). Conditions for robust stability or robust performance is given in terms of the information in the IQC's. The robustness analysis is generally an infinite dimensional optimization problem. We introduce a format for restriction of the optimization problem to a finite-dimensional subspace, which directly leads to numerically tractable robustness conditions in terms of linear matrix inequalities. It is possible to investigate conservativeness by considering a dual problem, which in many cases also is numerically tractable.