Computational issues in alternating projection algorithms for fixed-order control design

Alternating projection algorithms have been introduced recently to solve fixed-order controller design problems described by linear matrix inequalities and non-convex coupling rank constraints. In this work, an extensive numerical experimentation using proposed benchmark fixed-order control design examples is used to indicate the computational efficiency of the method. These results indicate that the proposed alternating projections are effective in obtaining low-order controllers for small and medium order problems