Optimal transfer schemes for switching controllers

We develop two general schemes for reducing the performance degradation caused by a signal substitution at the plant input. Such substitutions arise in modern control systems where we may transfer from manual to automatic, or between alternative controllers as operating points or performance criteria change. The problem reduces to the selection of the initial controller state for the controller which is to be switched in. The first scheme selects a controller state consistent with (hypothetical) signals at the plant input and output which are close (in the sense of a weighted 2-norm) to observed signals. The second scheme (for regulator and step reference problems) minimises directly the weighted plant input and output after the switch with respect to the controller initial state. We consider these schemes in the context of the regulator problem, and the reference tracking problem