Block Krylov subspace methods for large algebraic Riccati equations (Q1418851): Difference between revisions

The author presents numerical algorithms for approximate solutions to large algebraic Riccati equations. Both, continuous-time (\(A^TX + XA - XBB^TX + C^TC = 0\)) and discrete-time (\(X = A^TXA - A^TXB(R+B^TXB)^{-1}B^TXA + C^TC\)) cases are considered, where \(A,\, X \in \mathbb R^{n \times n}\), \(B \in \mathbb R^{n \times p}\), \(C \in \mathbb R^{s \times n}\) and \(R \in \mathbb R^{p \times p}\) with \(R=R^T>0\) assuming that the matrices \(B\) and \(C\) are of full rank and \(s \ll n, \, p \ll n\). Iterative projection algorithms onto block Krylov subspaces are proposed in order low rank approximate solutions to the stabilizing solution of large continuous-time or discrete-time algebraic Riccati equations to be produced. Some theoretical results like deriving an upper bound of the norm of the error and a perturbation result are given. Finally, the author present several numerical examples for illustrating the effectiveness of his approach for large and sparse algebraic Riccati equations and also for an illustration of the importance of the proposed deflation procedure. The paper is interesting for numerical mathematicians dealing with iterative solution methods and algorithms for large and sparse systems of linear equations and particularly for those who are interesting in modelling and numerical solution in control theory (e.g. linear-quadratic regulator problems, \(H_{\infty}\) or \(H_2\)-control, model reduction problems, etc.).