Application of Structural Measures with the Quadratic Optimization of Stochastically Disturbed Linear Systems.

Quadratic optimization of linear dynamic systems often presents the difficulty that the given practical problem specification can only inadequately be represented by the theoretical problem formulation. By a modified expression for the cost functional, all variables relevant for control, which may be represented by a linear combination of the state and input vector, may be weighted directly. Then a physically comprehensible formulation of the optimization problem is achieved, which also directly assists a physically sensible choice of the weighting matrices. In addition, a structural analysis of the controlled system by means of special structural measures accompanying the controller design provides information about couplings in the system, which can directly be included in the design. The example of an active vehicle suspension shows that a controller design may be carried out in a few iteration steps, which incorporates all given technical boundary conditions.