Table of Contents
International Journal of Computational Mathematics
Volume 2014, Article ID 727093, 8 pages
http://dx.doi.org/10.1155/2014/727093
Research Article

Hermitian Positive Definite Solution of the Matrix Equation

1College of Mathematics and Computational Science, Guilin University of Electronic Technology, Guilin 541004, China
2College of Mathematics and Econometrics, Hunan University, Changsha 410082, China

Received 26 May 2014; Revised 10 August 2014; Accepted 9 September 2014; Published 24 September 2014

Academic Editor: Asai Asaithambi

Copyright © 2014 Chun-Mei Li and Jing-Jing Peng. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

We consider the Hermitian positive definite solution of the nonlinear matrix equation . Some new sufficient conditions and necessary conditions for the existence of Hermitian positive definite solutions are derived. An iterative method is proposed to compute the Hermitian positive definite solution. In the end, an example is used to illustrate the correctness and application of our results.

1. Introduction

In this paper we consider the Hermitian positive definite solution of the nonlinear matrix equation where are complex matrices; and are Hermitian positive definite matrices. Here, denotes the conjugate transpose of the matrix . The nonlinear matrix equation (1) plays an important role in linear optimal and robust control. For instance, the solvability of the discrete-time linear quadratic optimal control problem with , depends on the solvability of (1) in some special cases [14]. Due to the important applications in system and control theory, in the past decades, (1) with has been extensively studied, and the research results mainly concentrated on the following: (a)sufficient conditions and necessary conditions for the existence of an Hermitian solution [59];(b)numerical methods for computing the Hermitian solution [4, 1013];(c)properties of the Hermitian solution [14, 15];(d)perturbation analysis for the discrete algebraic Riccati equation [1618].(e)connection with symplectic matrix pencil [9, 19, 20];(f)connection with stochastic realization and spectral factorization [2123].Nonetheless, (1) with has not been studied as far as we know.

In this paper we study the generalized nonlinear matrix equation (1). Firstly, we transform (1) into an equivalent nonlinear matrix equation. By Sherman-Woodbury-Morrison formula [24, Page 50], we have then

Set then That is, Therefore, the nonlinear matrix equation (1) can be equivalently rewritten as (7). So we first investigate the Hermitian positive definite solution of (7) in Section 2 and then derive some new results on the nonlinear matrix equation (1) by using the matrix transformations (5) in Section 3. Finally, we use an example to illustrate the correctness and application of the results of Section 3.

Throughout this paper, we write if the matrix is Hermitian positive definite (semidefinite). If is Hermitian positive definite (semidefinite), then we write . If an Hermitian positive definite matrix satisfies , we denote by . We use to denote all eigenvalues (each repeated as many times as its algebraic multiplicity) of an Hermitian matrix . The symbol denotes the spectral norm of the matrix .

2. Hermitian Positive Definite Solution of (7)

In this section, we first give some sufficient conditions and necessary conditions for the existence of an Hermitian positive definite solution of (7), and then we construct an iterative method to compute the Hermitian positive definite solution. We begin with some lemmas.

Lemma 1 (see [25]). If , then .

Lemma 2. If (7) has an Hermitian positive definite solution , then

Proof. Since (7) has an Hermitian positive definite solution , that is, then And from Lemma 1 it follows that which leads to that is,

Lemma 3. Suppose that the matrices are nonsingular; if (7) has an Hermitian positive definite solution , then

Proof. Since (7) has an Hermitian positive definite solution , that is, then which implies By (17) and Lemma 1 and noting that the matrices are nonsingular, then we have which leads to that is, By (16) and (20), we have

Theorem 4. Let If then (7) has an Hermitian positive definite solution and , where

Proof. Since and noting that then we have Therefore, the quadratic equation has a positive root , and the quadratic equation has a positive root , where and are defined by (24).
Now we consider the map Obviously, is a convex, closed, and bounded set and the map is continuous on . For arbitrary , since and are positive roots of (28) and (29), respectively, then we have that is, , which implies that . By Brouwer’s fixed point theorem, the map has a fixed point , which is the positive definite solution of (7). The theorem is proved.

Theorem 5. If for all , then (7) has an Hermitian positive definite solution. Furthermore, if then (7) has a unique Hermitian positive definite solution.

Proof. Consider the map Obviously, is a convex, closed, and bounded set and the map is continuous on . If for all , then we have that is, Hence, . By Brouwer’s fixed point theorem, the map has a fixed point in , which is the Hermitian positive definite solution of (7).
For arbitrary , we have Since , then is a contraction map in . By Banach’s fixed point theorem, we know that the map has a unique fixed point in , and this shows that (1) has a unique Hermitian positive definite solution in . Noting Lemma 3, we know that (1) has a unique Hermitian positive definite solution. The theorem is proved.

Nextly, we give an iterative method to compute the Hermitian positive definite solution of (7).

Theorem 6. If there is a real number satisfying then the matrix sequence converges the Hermitian positive definite solution of (7), where the sequence is generated by the following iterative method:

Proof. Consider the iterative method (40). From (39) it follows that
Since , then and which implies that Therefore,
Assume that when , we have ; then which implies that By mathematical induction, we obtain that the matrix sequence is monotone decreasing. Now we begin to use mathematical induction to show that the matrix sequence is bounded below by some Hermitian positive definite solution of (7).
When , by (39) we have Assume that when , we have ; then , and for we have
Therefore, the matrix sequence is a monotonic decreasing sequence and bounded below by some Hermitian positive definite solution , and then it converges to an Hermitian positive definite solution of (1). The theorem is proved.

3. Hermitian Positive Definite Solution of (1)

By the matrix transformations (5), Lemmas 2 and 3, and Theorems 46, we have the following theorems.

Theorem 7. If (1) has an Hermitian positive definite solution , then

Theorem 8. Suppose that the matrices are nonsingular; if (1) has an Hermitian positive definite solution , then

Remark 9. The upper bounds of Hermitian solution for the discrete algebraic Riccati equation have been extensively studied in the past three decades (see [14] and references therein). Compared with previous works, one gives a new upper bound of Hermitian solution but also gives a lower bound in Theorem 8.

Theorem 10. Let If then (1) has an Hermitian positive definite solution , and , where

Theorem 11. If for all , then (1) has an Hermitian positive definite solution. Furthermore, if then (1) has a unique Hermitian positive definite solution.

Theorem 12. If there is a real number satisfying then the matrix sequence converges to the Hermitian positive definite solution of (1), where the matrix sequence is generated by the following iterative method:

4. Numerical Experiments

In this section, we give an example to demonstrate the correctness and application of our results.

Example 1. Consider the following discrete-time LQR optimal control problem: with the additional constraint that , where , and both and are as follows: The optimal control sequence that minimizes the performance index is , where and satisfies the discrete algebraic Riccati equation It is easy to verify that the nonlinear matrix equation (62) satisfies the conditions of Theorem 10; hence, (62) has an Hermitian positive definite solution . Now we use iterative method (58) to compute the Hermitian positive definite solution of (62). Let , and after 6 iterations, we get the positive definite solution and its residual error . Of course it is easy to verify that which confirms the correctness of Lemma 3.
From we have hence, . Then taking we have

5. Conclusion

In this paper, we consider the Hermitian positive definite solution of the nonlinear matrix equation , which plays an important role in linear optimal and robust control. By making use of Sherman-Woodbury-Morrison formula, we transform this equation into an equivalent nonlinear matrix equation. So we can investigate this nonlinear matrix equation by the aid of the equivalent nonlinear matrix equation. Some new sufficient conditions and necessary conditions for the existence of an Hermitian positive definite solution are derived. An iterative method is proposed to compute its Hermitian positive definite solution. In the end, an example is used to illustrate the correctness and application of our results.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

The work was supported by the National Natural Science Foundation of China (nos. 11101100, 11301107, and 11261014) and the Natural Science Foundation of Guangxi Province (nos. 2012GXNSFBA053006 and 2013GXNSFBA019009).

References

  1. D. J. Bender and A. J. Laub, “The linear-quadratic optimal regulator for descriptor systems,” IEEE Transactions on Automatic Control, vol. 32, no. 8, pp. 672–688, 1987. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  2. P. Lancaster and L. Rodman, Algebraic Riccati Equations, Oxford University Press, Oxford, UK, 1995. View at MathSciNet
  3. B. D. Anderson and J. B. Moore, Linear Optimal Control, Prentice Hall, Englewood Cliffs, NJ, USA, 1971. View at MathSciNet
  4. T. Pappas, A. J. Laub, and N. R. Sandell, “On the numerical solution of the discrete-time algebraic Riccati equation,” IEEE Transactions on Automatic Control, vol. 25, no. 4, pp. 631–641, 1980. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  5. D. J. Clements and H. K. Wimmer, “Existence and uniqueness of unmixed solutions of the discrete-time algebraic Riccati equation,” Systems & Control Letters, vol. 50, no. 5, pp. 343–346, 2003. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  6. J. C. Engwerda, A. C. Ran, and A. L. Rijkeboer, “Necessary and sufficient conditions for the existence of a positive definite solution of the matrix equation X+A*X-1A=Q,” Linear Algebra and Its Applications, vol. 186, pp. 255–275, 1993. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  7. A. Ferrante and B. C. Levy, “Hermitian solutions of the equation X=Q+NX-1N*,” Linear Algebra and its Applications, vol. 247, pp. 359–373, 1996. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  8. G. Freiling, V. Mehrmann, and H. Xu, “Existence, uniqueness, and parametrization of Lagrangian invariant subspaces,” SIAM Journal on Matrix Analysis and Applications, vol. 23, no. 4, pp. 1045–1069, 2002. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  9. H. K. Wimmer, “Normal forms of symplectic pencils and the discrete-time algebraic Riccati equation,” Linear Algebra and Its Applications, vol. 147, pp. 411–440, 1991. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  10. P. Benner, Contributions to the Numerical Solution of Algebraic Riccati Equations and Related Eigenvalue Problems, Logos, Berlin, Germany, 1997.
  11. P. Van Dooren, “A generalized eigenvalue approach for solving Riccati equations,” Society for Industrial and Applied Mathematics, vol. 2, no. 2, pp. 121–135, 1981. View at Publisher · View at Google Scholar · View at MathSciNet
  12. L. Z. Lu and W. W. Lin, “An iterative algorithm for the solution of the discrete-time algebraic Riccati equation,” Linear Algebra and Its Applications, vol. 188-189, pp. 465–488, 1993. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  13. A. J. Laub, “A Schur method for solving algebraic Riccati equations,” IEEE Transactions on Automatic Control, vol. 24, no. 6, pp. 913–921, 1979. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  14. R. Davies, P. Shi, and R. Wiltshire, “Upper solution bounds of the continuous and discrete coupled algebraic Riccati equations,” Automatica, vol. 44, no. 4, pp. 1088–1096, 2008. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  15. I. G. Ivanov, “Properties of Stein (Lyapunov) iterations for solving a general Riccati equation,” Nonlinear Analysis: Theory, Methods & Applications, vol. 67, no. 4, pp. 1155–1166, 2007. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  16. M. M. Konstantinov, P. H. Petkov, and D. W. Gu, “Improved perturbation bounds for general quadratic matrix equations,” Numerical Functional Analysis and Optimization, vol. 20, no. 7-8, pp. 717–736, 1999. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  17. J. G. Sun, “Sensitivity analysis of the discrete-time algebraic Riccati equation,” in Proceedings of the 6th Conference of the International Linear Algebra Society, vol. 275/276 of Linear Algebra and its Applications, pp. 595–615. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  18. L. Zhou, Y. Lin, Y. Wei, and S. Qiao, “Perturbation analysis and condition numbers of symmetric algebraic Riccati equations,” Automatica, vol. 45, no. 4, pp. 1005–1011, 2009. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  19. A. Ferrante and B. C. Levy, “Canonical form of symplectic matrix pencils,” Linear Algebra and its Applications, vol. 274, pp. 259–300, 1998. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  20. A. Ferrante and L. Ntogramatzidis, “A unified approach to finite-horizon generalized LQ optimal control problems for discrete-time systems,” Linear Algebra and its Applications, vol. 425, no. 2-3, pp. 242–260, 2007. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  21. A. Ferrante, “A parameterization of minimal stochastic realizations,” IEEE Transactions on Automatic Control, vol. 39, no. 10, pp. 2122–2126, 1994. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  22. A. Lindquist and G. Picci, “Realization theory for multivariate stationary Gaussian processes,” SIAM Journal on Control and Optimization, vol. 23, no. 6, pp. 809–857, 1985. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  23. A. Lindquist and G. Picci, “A geometric approach to modelling and estimation of linear stochastic systems,” Journal of Mathematical Systems, Estimation, and Control, vol. 1, no. 3, pp. 241–333, 1991. View at Google Scholar · View at MathSciNet
  24. G. H. Golub and C. F. Van Loan, Matrix Computations, Johns Hopkins University, London, UK, 3rd edition, 1996. View at MathSciNet
  25. X. Zhan, Matrix Inequalities, Springer, Berlin , Germany, 2002. View at MathSciNet