A Game Theoretic Approach to Linear Systems with L2-bounded Disturbances

The aim of the present study is to construct a state feedback controller for a given linear system that minimizes the worst-case effect of an L2 -bounded disturbance. Our setting is different from the usual framework of H -theory in that we consider nonzero initial conditions. The situation is modeled in a game theoretical framework, in which the controller designer acts as a minimizing player, and the uncertainty as a maximizing player. We show that a saddle-point equilibrium exists and find an optimal controller.