Section: New Results

Optimization of mu-analysis parameterization

Participant : Guillaume Sandou [correspondent] .

The robustness against parametric uncertainties can be studied using the structured singular value mu. In that case, a normalization of the uncertain parameters is performed, and the mu analysis provides the larger parallelepiped centered in the nominal and included in the stability domain. However, results depend on the initial normalization. In this study, the normalization is optimized so as to get the largest guaranteed stability domain. The corresponding problem being highly nonlinear, a metaheuristic method, Particle Swarm Optimization, is used for that purpose. An academic and a real life example, namely the pendulum in the cart problem, have been used to prove the viability of the approach.