On the lack of monotonicity of Newton-Hewer updates for Riccati equations (Q2665637)

This note investigates the monotonicity properties of iterative methods for solving the discrete-time algebraic Riccati equation (DARE). There are several classical iterative methods for solving the DARE in the literature, including the ones proposed in [\textit{P. E. Caines} and \textit{D. Q. Mayne}, Int. J. Control, I. Ser. 12, 785--794 (1970; Zbl 0205.15902)], [\textit{G. Hewer}, ``An iterative technique for the computation of the steady state gains for the discrete optimal regulator'', IEEE Trans. Autom. Control 16, No. 4, 382--384 (1971)], algebraic methods [\textit{P. Lancaster} and \textit{L. Rodman}, Algebraic Riccati equations. Oxford: Clarendon Press (1995; Zbl 0836.15005)], and semi-definite programming [\textit{V. Balakrishnan} and \textit{L. Vandenberghe}, IEEE Trans. Autom. Control 48, No. 1, 30--41 (2003; Zbl 1364.90244)]. In particular, these iterative methods generate a sequence of positive-definite matrices which converges to the solution of the DARE. Authors provide a set of counterexamples for the monotonicity of the Newton-Hewer method [Hewer, loc. cit.] for solving the discrete-time algebraic Riccati equation in dynamic settings, drawing a contrast with the Riccati difference equation [\textit{P. E. Caines} and \textit{D. Q. Mayne}, Int. J. Control, I. Ser. 12, 785--794 (1970; Zbl 0205.15902)].