On the solution of a nonlinear martrix equation for MIMO symmetric realizations (Q1105525): Difference between revisions

In a previous paper [\textit{L. Fortuna, A. Gallo} and \textit{G. Nunnari}, J. Franklin Inst. 325, 143-153 (1988; Zbl 0628.93028)] a new type of quadratic matrix equation for SISO systems is introduced: (1) \(AP+PA+BC- PBCP=0\). A particular solution \(\bar P\) of equation (1) is studied which carries dual information as regards the contribution of the state variables in terms of Kalman filtering and LQG control. In this paper the previous results are extended to MIMO symmetric realizations. In particular the following main propositions are given: For MIMO symmetric realizations (A,B,C) we have: (2) \(\bar P^ 2=P_{1+}P_{2+}\) with \(\bar P\) with \(\bar P=R^{-1}P_{2+}\) where the matrices \(P_{1+}\) and \(P_{2+}\) are respectively the positive definite solution of the Kalman filtering algebraic equation and the control algebraic equation and R is the well-known matrix defined such that \(RA=A^ TR\) and \(B=R^{-1}C^ T\). If the system (A,B,C) has extended symmetry, relation (2) is also true since \(\bar P\) is the solution of the nonlinear matrix equation \(AP+PA+BVUC-PBVUCP=0\) where V and U are respectively matrices with orthogonal columns and rows. Relation (1) is generally not true for non-symmetrical systems.