Disturbance attenuation via strictly positive real compensators. (Q1428212)

The authors describe conditions for designing strictly positive real (SPR) compensators ensuring closed-loop \(H_\infty\) performance for a linear system. Due to their inherent robustness, SPR compensators are frequently used in aerospace applications (e.g., large flexible structures). If the design of SPR compensators of the same order as the open-loop plant can be expressed as a linear matrix inequality (LMI), a convex optimization problem for which powerful numerical solvers have been developed, this is not the case when enforcing an additional performance specification, such as an \(H_\infty\) norm bound. The design problem then becomes a difficult non-convex bilinear matrix inequality (BMI) problem. Moreover, performance is ensured only by enforcing an additional algebraic constraint on Lyapunov matrices. In this paper, the authors propose to alleviate the algebraic constraint on Lyapunov matrices by introducing a scalar scaling parameter over which optimization can be carried out. Bilinearity of the matrix inequalities is handled by alternatively freezing a set of variables while optimizing over the remaining free variables, a classical heuristic which can sometimes provide useful results.