Logarithmic barrier methods for solving the optimal output feedback problem

We consider the problem of designing a feedback control law when a complete set of state variables is not available. The resulting nonlinear and nonconvex optimization problem for determining the optimal feedback gain will be solved by a logarithmic barrier approach. The method is tailored to the particular structure of the constant output feedback problem. We consider two algorithms, an Anderson-Moore type and an inexact Newton method, which are embedded in the logarithmic barrier framework. The local and global convergence properties of the algorithms are discussed in detail. Using test examples from optimal output feedback design we can also verify these results numerically. (orig.)SIGLEAvailable from TIB Hannover: RR 1843(95-16) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekDeutsche Forschungsgemeinschaft (DFG), Bonn (Germany)DEGerman