Smooth factorizations of matrix valued functions and their derivatives (Q1183049)

This paper discusses numerical factorization of matrix valued functions for application to the numerical solution of differential algebraic equations with time dependent coefficients and, in particular, Riccati equations of the form (DARE) \(E^ T\dot X(t)E=E^ TX(t)E+A^ TX(t)E+Q(t)-E^ TX(t)W(t)X(t)E\) with \(E=E(t)\), \(A=A(t): [t_ 0,t_ f]\subset\mathbb{R}\to\mathbb{R}^{n,n}\) and \(Q(t),W(t): [t_ 0,t_ f]\to\mathbb{R}^{n,n}\), which arise from linear quadratic control problems. Initially the authors discuss numerical computation of smooth transformations of time dependent matrix pencils \(\alpha E(t)-\beta A(t)\), with \(E(t)\), \(A(t)\) as above. It is shown how to obtain smoothness of the matrix factors and to find numerical values of their derivatives. The results are applied to DARE, for which a numerical algorithm is developed and its use illustrated on a simple numerical example.