DDNS Discrete Dynamics in Nature and Society 1607-887X 1026-0226 Hindawi 10.1155/2018/9216760 9216760 Research Article On Convergence of Infinite Matrix Products with Alternating Factors from Two Sets of Matrices http://orcid.org/0000-0002-6465-0040 Kozyakin Victor 1 Kulenovic Mustafa R. S. Kharkevich Institute for Information Transmission Problems Russian Academy of Sciences Bolshoi Karetny Lane 19 Moscow 127051 Russia ras.ru 2018 2242018 2018 19 12 2017 18 03 2018 2242018 2018 Copyright © 2018 Victor Kozyakin. 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.

We consider the problem of convergence to zero of matrix products AnBnA1B1 with factors from two sets of matrices, AiA and BiB, due to a suitable choice of matrices {Bi}. It is assumed that for any sequence of matrices {Ai} there is a sequence of matrices {Bi} such that the corresponding matrix products AnBnA1B1 converge to zero. We show that, in this case, the convergence of the matrix products under consideration is uniformly exponential; that is, AnBnA1B1Cλn, where the constants C>0 and λ(0,1) do not depend on the sequence {Ai} and the corresponding sequence {Bi}. Other problems of this kind are discussed and open questions are formulated.

Russian Science Foundation 14-50-00150
1. Introduction

Denote by M(p,q) the space of matrices of dimension p×q with real elements and the topology of elementwise convergence. Let AM(N,M) and BM(M,N) be finite sets of matrices. We will be interested in the question of whether it is possible to ensure the convergence to zero of matrix products, (1)AnBnA1B1,AiA,BiB,for all possible sequences of matrices {Ai}, due to a suitable choice of sequences of matrices {Bi}.

As an example of a problem in which such a question arises, let us consider one of the varieties of the stabilizability problem for discrete-time switching linear systems . Consider a system whose dynamics is described by the equations (2)xn=Anun,AnA,un=Bnxn-1,BnB,where the first of them describes the functioning of a plant, whose properties are uncontrollably affected by perturbations from class A, while the second equation describes the behavior of a controller. Then, by choosing a suitable sequence of controls {BnB}, one can try to achieve the desired behavior of system (2), for example, the convergence to zero of its solutions: (3)xn=AnBnA1B1x0.

As was noted, for example, in [6, 7], the question of the stabilizability of matrix products with alternating factors from two sets, due to a special choice of factors from one of these sets, can also be treated in the game-theoretic sense.

If, in considering the switching system, it is assumed that there are actually no control actions, that is, BnI, then (2) take the form (4)xn=Anxn-1,AnA.In this case, the problem of the stabilizability of the corresponding switching system turns into the problem of its stability for all possible perturbations of the plant in class A, that is, into the problem of convergence to zero of the solutions (5)xn=AnA1x0of (4) for all possible sequences of matrices {AiA}. Convergence to zero of the matrix products AnA1, arising in this case, has been investigated by many authors (see, e.g., [2, 811], as well as the bibliography in ).

The presence of alternating factors in the products of matrices (1) substantially complicates the problem of convergence of the corresponding matrix products for all possible sequences of matrices {AiA} due to a suitable choice of sequences of matrices {BiB} in comparison with the problem of convergence of matrix products AnA1 for all possible sequences of matrices {AiA}. A discussion of the arising difficulties can be found, for example, in . One of the applications of the results obtained in this paper for analyzing the new concept of the so-called minimax joint spectral radius is also described there.

2. Path-Dependent Stabilizability

Every product (1) is a matrix of dimension N×N; that is, it is an element of the space M(N,N). As is known, the space M(N,N) with the topology of elementwise convergence is normable; therefore we assume that · is some norm in it. We note here that since all norms in the space M(N,N) are equivalent, the choice of a particular norm when considering the convergence of products (1) is inessential. Nevertheless, in what follows, it will be convenient for us to assume that the norm · in M(N,N) is submultiplicative; that is, for any two matrices X,Y, the inequality XYX·Y holds. In particular, a norm on M(N,N) is submultiplicative if it is generated by some vector norm on RN; that is, its value on matrix A is defined by the equality A=supx0(Ax/x), where x and Ax are the norms of the corresponding vectors in RN.

Definition 1.

The matrix products (1) are said to be path-dependent stabilizable by choosing the factors {Bn} if for any sequence of matrices {AnA} there exists a sequence of matrices {BnB} for which (6)AnBnA1B10asn.

As an example, consider the case where sets A and B consist of square matrices of dimension N×N, and B={I}, where I is the identical matrix. In this case, Definition 1 of the path-dependent stabilizability of the matrix products (1) reduces to the following condition: (7)AnA10asn,for each sequence {AnA}. As is known, in this case, convergence (7) is uniformly exponential. Namely, the following statement, which was repeatedly “discovered” by many authors, is true (see, e.g., [2, 811]).

Theorem A (on exponential convergence).

Let the set of matrices A be such that for each sequence {AnA} convergence (7) holds. Then there exist constants C>0 and λ(0,1) such that (8)AnA1Cλn,n=1,2,,for each sequence {AnA}.

Our goal is to prove that an analogue of Theorem A (on exponential convergence) is valid for the path-dependent stabilizable matrix products (1).

Theorem 2.

Let A and B be the sets of matrices for which the matrix products (1) are path-dependent stabilizable. Then there exist constants C>0 and λ(0,1) such that for any sequence of matrices {AnA} there is a sequence of matrices {BnB} for which (9)AnBnA1B1Cλn,n=1,2,.

To prove the theorem, we need the following auxiliary assertion.

Lemma 3.

Let the conditions of Theorem 2 be satisfied. Then there exist constants k>0 and μ(0,1) such that for any sequence of matrices {AnA} there is a positive integer kk and a set of matrices B1,,BkB for which AkBkA1B1μ<1.

Proof.

By Definition 1 of the path-dependent stabilizability of the matrix products (1) for each matrix sequence {AnA} there exists a natural k such that (10)AkBkA1B1<1,for some sequence of matrices {BnB}.

Given a sequence {An}, let us denote by k({An}) the smallest k under which inequality (10) holds. To prove the lemma, it suffices to show that the quantities k({An}) are uniformly bounded; that is, there is a k such that (11)kAnk,AnA.

Assuming that inequality (11) is not true, for each positive integer k, we can find a sequence {An(k)A} such that k({An(k)})k. In this case, by the definition of the number k({An}), (12)AmBmA1B11,B1,,BmB,for each positive integer mk-1k({An(k)})-1.

Let us denote by Ak the set of all sequences {AnA}, for each of which inequalities (12) hold. Then {An(k)}Ak and, therefore, Ak. Moreover, (13)A1A2,and each set Ak is closed since inequalities (12) hold for all its elements; sequences {An}Ak, for each positive integer mk-1.

We now note that each of the sets Ak is a subset of the topological space A of all sequences {AnA} with the topology of infinite direct product of the finite set of matrices A. By the Tikhonov theorem in this case A is a compact. Then, each of the sets Ak is also a compact. In this case, it follows from (13) that k=1Ak and, therefore, there is a sequence {A-nA} such that (14)A-nk=1Ak.By the definition of the sets Ak, for the sequence {A-nA}, the inequalities (15)A-mBmA-1B11hold for each m1 and any B1,,BmB which contradicts the assumption of the path-dependent stabilizability of the matrix products (1). This contradiction completes the proof of the existence of a number k for which inequalities (11) are valid.

Thus, we have proven the existence of a number k such that, for each sequence {AnA} and some corresponding sequence {BnB}, strict inequalities (10) are satisfied with k=k({An})k. Moreover, since the number of all products AkBkA1B1 participating in inequalities (10) is finite, then the corresponding inequalities (10) can be strengthened: there is a μ(0,1) such that for any sequence of matrices {AnA} there exist a natural kk and a set of matrices B1,,BkB for which AkBkA1B1μ<1.

We now proceed directly to the proof of Theorem 2.

Proof of Theorem <xref ref-type="statement" rid="thm1">2</xref>.

Given an arbitrary sequence {AnA}, by Lemma 3, there exist a number k1k and a set of matrices B1,,Bk1 such that (16)Ak1Bk1A1B1μ<1.

Next, consider the sequence of matrices {AnA,nk1+1} (the “tail” of the sequence {AnA} starting with the index k1+1). Again, by virtue of Lemma 3, there exist a k2k1+k and a set of matrices Bk1+1,,Bk2 such that (17)Ak2Bk2Ak1+1Bk1+1μ<1.

We continue in the same way constructing for each m=3,4, numbers (18)kmkm-1+kand sets of matrices Bkm-1+1,,Bkm for which (19)AkmBkmAkm-1+1Bkm-1+1μ<1.

Let us show that, for the obtained sequence of matrices {Bn} for some C>0 and λ(0,1), which do not depend on the sequences {An} and {Bn}, inequalities (9) are valid. Fix a positive integer n and specify for it a number p=p(n) such that (20)n-k<kpn.Such p exists, since the sequence {km} strictly increases by construction. We now represent the product AnBnA1B1 in the form (21)AnBnA1B1=DDpD1,where (22)D=AnBnAkp+1Bkp+1,Di=AkiBkiAki-1+1Bki-1+1,i=1,2,,p.Then (23)Dϰn-kpϰk,whereϰ=maxAA,BB1,AB (since the sets A and B are finite, ϰ<). Further, by the definition of the matrices Di and inequalities (19), (24)Diμ<1fori=1,2,,p.Taking into account the fact that, by virtue of (18), for each m, the estimate kmkm is fulfilled, from here and from (20) we obtain for the number p a lower estimate: pn/k-1. And then from the estimates established earlier for D, D1,,Dm, we deduce that (25)AnBnA1B1D·DpD1ϰkμpϰkμn/k-1ϰkμμ1/kn.Hence, putting C=ϰk/μ and λ=μ1/k, we obtain inequalities (9).

3. Path-Independent Stabilizability

Let us now consider another variant of the stabilizability of matrix products (1) due to a suitable choice of matrices {Bi}.

Definition 4.

The matrix products (1) are said to be path-independent periodically stabilizable by choosing the factors {Bn} if there exists a periodic sequence of matrices {B-nB} such that (26)AnB-nA1B-10asn,for any sequence of matrices {AnA}.

It is clear that path-independent periodically stabilized products (1) are path-dependent stabilized.

Theorem 5.

Let A and B be the sets of matrices for which the matrix products (1) are path-independent periodically stabilizable by a sequence of matrices {B-nB}. Then there exist constants C>0 and λ(0,1) such that (27)AnB-nA1B-1Cλn,n=1,2,,for any sequence of matrices {AnA}.

Proof.

Denote by p the period of the sequence {B-n}. Consider the set of (N×N)-matrices: (28)D=D=ApB-pA1B-1:A1ApA.

Since the set of matrices A is finite, set D is also finite. Moreover, by Definition 4 of path-independent periodic stabilization, (29)AnpB-npA1B-10asn,for each sequence {AnA}. Hence, for each sequence {DnD}, there is also (30)DnD10asn.In this case, by Theorem A (on exponential convergence), there are k>0 and μ(0,1) such that (31)DkD1μ<1,D1DkD,or, equivalently, (32)AkpB-kpA1B-1μ<1,A1AkpA.

Further, repeating the proof of the corresponding part of Theorem 2 word for word, we derive from inequalities (32) the existence of constants C>0 and λ(0,1) such that for any sequence of matrices {AnA} inequalities (27) hold.

4. Remarks and Open Questions

First of all, we would like to make the following remarks.

Remark 6.

In the proof of Lemma 3, in fact, we used not the condition of path-dependent stabilizability of the matrix products (1) but the weaker condition that for each matrix sequence {AnA} there exist a natural k=k({An}) and a collection of matrices B1,,BkB for which equality (10) holds. Correspondingly, the statement of Theorem 2 is valid under weaker assumptions.

Theorem 7.

Let the sets of matrices A and B be such that for each matrix sequence {AnA} there are a natural k and a collection of matrices B1,,BkB for which (33)AkBkA1B1<1.Then there exist constants C>0 and λ(0,1) such that for any sequence of matrices {AnA} there is a sequence of matrices {BnB} for which (34)AnBnA1B1Cλn,n=1,2,.

Remark 8.

All the above statements remain valid for the sets of matrices A and B with complex elements.

Remark 9.

Throughout the paper, in order to avoid inessential technicalities in proofs, it was assumed that the sets of matrices A and B are finite. In fact, all the above statements remain valid in the case when the sets of matrices A and B are compacts, not necessarily finite, that is, are closed and precompact.

Comparing the notions of path-dependent stabilizability and path-independent periodic stabilizability, one can note that in the second of them the requirement of periodicity of the sequence {B-n} stabilizing the matrix products (1) appeared. Therefore, the following less restrictive concept of path-independent stabilizability seems rather natural.

Definition 10.

The matrix products (1) are said to be path-independent stabilizable by choosing the factors {Bn} if there is a sequence of matrices {B-nB} such that convergence (26) holds for any sequence of matrices {AnA}.

It is not difficult to construct an example of the sets of square matrices in which the matrix products AnB-nA1B-1 converge slowly enough, slower than any geometric progression. For this, it is enough to put A={I} and B={I,λI}, where λ(0,1), and define sequence {B-} so that the matrix λI appears in it “fairly rare,” at positions with numbers k2, k=1,2,.

Question 11.

Let the matrix products (1) be path-independent stabilizable by choosing a certain sequence of matrices {B-nB}. Is it possible in this case to specify a sequence of matrices {B~nB} (possibly different from {B-nB}) and constants C>0 and λ(0,1) such that, for any sequence of matrices {AnA} for all n=1,2,, the inequalities AnB~nA1B~1Cλn will be valid?

Let us consider one more issue, which is adjacent to the topic under discussion. In the theory of matrix products, the following assertion is known [2, 811]: let A be a finite set such that for each sequence of matrices {AnA} the sequence of norms {AnA1,n=1,2,} is bounded. Then all such sequences of norms for the matrices are uniformly bounded; that is, there exists a constant C>0 such that (35)AnA1C,n=1,2,,for each sequence of matrices {AnA}.

Question 12.

Let finite sets of matrices A and B be such that for each sequence of matrices {AnA} there is a sequence of matrices {BnA} for which the sequence of norms {AnBnA1B1,n=1,2,} is bounded. Does there exist in this case a constant C>0 such that for every matrix sequence {AnA} there is a sequence of matrices {BnA}, for which the sequence of norms {AnBnA1B1,n=1,2,} is uniformly bounded, that is, for all n=1,2,, the inequalities AnBnA1B1C hold?

Conflicts of Interest

The author declares that he has no conflicts of interest.

Acknowledgments

The work was carried out at the Kharkevich Institute for Information Transmission Problems, Russian Academy of Sciences, and was funded by the Russian Science Foundation (Project no. 14-50-00150).

Stanford D. P. Stability for a multi-rate sampled-data system Society for Industrial and Applied Mathematics. Journal on Control and Optimization 1979 17 3 390 399 10.1137/0317029 MR528902 Kozyakin V. S. On the absolute stability of systems with asynchronously operating pulse elements Avtomat. i Telemekh 1990 10 56 63 translation in Automation and Remote Control, vol. 51, no. 10, part 1, pp. 1349–1355, 1991 Sun Z. Ge S. S. Switched Linear Systems: Control and Design 2005 London, UK Springer Communications and Control Engineering 10.1007/1-84628-131-8 Jungers R. M. Mason P. On feedback stabilization of linear switched systems via switching signal control SIAM Journal on Control and Optimization 2017 55 2 1179 1198 10.1137/15M1027802 MR3633780 2-s2.0-85018997995 Lin H. Antsaklis P. J. Stability and stabilizability of switched linear systems: a survey of recent results Institute of Electrical and Electronics Engineers Transactions on Automatic Control 2009 54 2 308 322 10.1109/TAC.2008.2012009 MR2491959 2-s2.0-61349147303 Asarin E. Cervelle J. Degorre A. Dima C. Horn F. Kozyakin V. Ollinger N. Vollmer H. Entropy games and matrix multiplication games 47 the proceedings of 33rd Symposium on Theoretical Aspects of Computer Science, (STACS 2016) 2016 Dagstuhl, Germany Schloss Dagstuhl–Leibniz-Zentrum fuer Informatik 11:1 11:14 LIPICS - Leibniz International Proceedings in Informatics 10.4230/LIPIcs.STACS.2016.11 Bouyer P. Markey N. Randour M. Larsen K. G. Laursen S. Average-energy games Acta Informatica 2018 55 2 91 127 10.1007/s00236-016-0274-1 MR3767722 2-s2.0-84978803469 Daubechies I. Lagarias J. C. Sets of matrices all infinite products of which converge Linear Algebra and its Applications 1992 161 227 263 10.1016/0024-3795(92)90012-Y MR1142737 Zbl0746.15015 2-s2.0-0001219086 Berger M. A. Wang Y. Bounded semigroups of matrices Linear Algebra and its Applications 1992 166 21 27 10.1016/0024-3795(92)90267-E MR1152485 Gurvits L. Stability of discrete linear inclusion Linear Algebra and its Applications 1995 231 47 85 10.1016/0024-3795(95)90006-3 MR1361100 Zbl0845.68067 2-s2.0-33846998909 10.1016/0024-3795(94)00010-7 Hartfiel D. J. Nonhomogeneous Matrix Products 2002 River Edge, NJ, USA World Scientific Publishing Co. Inc Kozyakin V. An annotated bibliography on convergence of matrix products and the theory of joint/generalized spectral radius Institute for Information Transmission Problems Moscow, Russia, 2013 10.13140/2.1.4257.5040 Kozyakin V. Minimax joint spectral radius and stabilizability of discrete-time linear switching control systems https://arxiv.org/abs/1712.06805, 2017