No label defined (Q4945441)

The main objective of the author is to show explicitly a general class of control problems that reduce to a standard problem (SP): Find positive definite (possibly structured) matrices \(X\) and \(Y\) satisfying linear matrix inequalities (LMIs) and additional rank constraints. The class is defined by the closed-loop specifications given in terms of the well-posedness of feedback systems and includes the \(H_\infty,\) the positive real, nominal/robust/gain-scheduled, continuous-time/discrete-time control problems as special cases. It is shown that the existence of a controller, yielding a closed-loop system whose well-posedness can be proved by a \(P\)-separator, can be checked by solving the SP. A fundamental fact to be revealed here is that the troublesome rank constraints in the SP disappear and the resulting problem becomes convex if the controller is allowed to use sufficient information on the plant. This fact can also anticipated from Packard's result, which showed that the gain-scheduled from \(H_\infty\) control problem can be reduced to an LMI optimization problem if the controller is allowed to be scheduled by the same (or a greater) number of scalar parameters from the plant. The obtained results extend Packard's result to a more general class of control problems including the \(H_\infty\) control problem as a special class. The generality of the author's formulation also allows to interpret the rank constraint arising in the reduced-order control problem as a lack of sufficient plant information. For the entire collection see [Zbl 0932.00034].