Mixed H∞ and H2 input design for identification

A framework for input design in identification is presented in this contribution. By introducing a suitable finite dimensional linear parametrization of the input spectrum, the input design problem is reduced to a convex and finite-dimensional program. It is shown that several useful model quality constraints can be incorporated in the framework. In particular, different frequency by frequency constraints can be used. Such constraints arise in, e.g., control applications where it is desirable to be able to guarantee robust stability and robust performance for all models in an identified model set.