The problem of rejection of nonrandom bounded exogenous disturbances (also known as peak-to-peak gain minimization) has the long history. It is the subject of l1-optimization theory. However, l_1-optimization technique often leads to high-dimensional controllers and is hard to implement in the continuous-time case. A natural way to overcome these difficulties is to appeal to the invariant sets ideology, in order to reduce complexity and attain the control objectives. Among various possible "shapes" of invariant sets utilized in the research areas above, ellipsoids should be distinguished because of their simple structure and direct connection to the quadratic Lyapunov functions approach. Moreover, the ellipsoidal description allows to exploit the powerful machinery of linear matrix inequalities and semidefinite programming as a technical solution tool. In the talk we address the above mentioned problem by use of the linear dynamical output feedback full-order controller. Up to the authors' knowledge, the design of general dynamical controller for rejection of L-infinity-bounded disturbances remained an open problem. The efficiency of the approach is illustrated via the real-life control problem for the gyroplatform. The approach is also applicable to discrete-time systems and to robust problem formulations.