TY - JOUR
T1 - A class of complex nonsymmetric algebraic Riccati equations associated with H-matrix
AU - Dong, Liqiang
AU - Li, Jicheng
N1 - Publisher Copyright:
© 2019 Elsevier B.V.
PY - 2020/4
Y1 - 2020/4
N2 - In this paper, based on the study of fluid flow models modulated by a Markov chain, we propose a new class of complex nonsymmetric algebraic Riccati equations (NAREs) associated with H-matrix. Under the assumption that the diagonal elements of the associated H-matrix are on a same ray in the complex plane, we generalize the definition of the extremal solution of an H-matrix algebraic Riccati equation (HARE) whose associated H-matrix has positive or negative diagonal elements, and show that the extremal solution exists and is unique. Besides, we show that the basic fixed-point iterative methods are linearly convergent and the Newton's method is quadratically convergent when they are applied to search for the extremal solutions of the HAREs. We also give out criteria for choosing suitable parameters such that the three existing doubling algorithms can deliver the extremal solutions of the HAREs, and show that the doubling algorithms are quadratically convergent. More importantly, we also propose a preprocessing technique to transform an HARE under a certain assumption into a new HARE under the aforementioned assumption. Numerical experiments show that our methods and preprocessing technique are effective.
AB - In this paper, based on the study of fluid flow models modulated by a Markov chain, we propose a new class of complex nonsymmetric algebraic Riccati equations (NAREs) associated with H-matrix. Under the assumption that the diagonal elements of the associated H-matrix are on a same ray in the complex plane, we generalize the definition of the extremal solution of an H-matrix algebraic Riccati equation (HARE) whose associated H-matrix has positive or negative diagonal elements, and show that the extremal solution exists and is unique. Besides, we show that the basic fixed-point iterative methods are linearly convergent and the Newton's method is quadratically convergent when they are applied to search for the extremal solutions of the HAREs. We also give out criteria for choosing suitable parameters such that the three existing doubling algorithms can deliver the extremal solutions of the HAREs, and show that the doubling algorithms are quadratically convergent. More importantly, we also propose a preprocessing technique to transform an HARE under a certain assumption into a new HARE under the aforementioned assumption. Numerical experiments show that our methods and preprocessing technique are effective.
KW - Basic fixed-point iterative method
KW - Doubling algorithm
KW - Extremal solution
KW - H-matrix algebraic Riccati equation (HARE)
KW - Newton's method
KW - Preprocessing technique
UR - https://www.scopus.com/pages/publications/85075383490
U2 - 10.1016/j.cam.2019.112567
DO - 10.1016/j.cam.2019.112567
M3 - 文章
AN - SCOPUS:85075383490
SN - 0377-0427
VL - 368
JO - Journal of Computational and Applied Mathematics
JF - Journal of Computational and Applied Mathematics
M1 - 112567
ER -