multirate sampled-data systems: all h/spl infin/ suboptimal … · troller design, e.g.,...

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 44, NO. 3, MARCH 1999 537

Multirate Sampled-Data Systems:All Suboptimal Controllers and

the Minimum Entropy ControllerLi Qiu, Senior Member, IEEE, and Tongwen Chen,Senior Member, IEEE

Abstract—For a general multirate sampled-data (SD) system,the authors characterize explicitly the set of all causal, stabilizingcontrollers that achieve a certainH1 norm bound; moreover,they give explicitly a particular controller that further minimizesan entropy function for the SD system. The characterization laysthe groundwork for synthesizing multirate control systems withmultiple/mixed control specifications.

Index Terms—Digital control, H1 optimization, matrix factor-ization, multirate systems, nest operators, sampled-data systems.

I. INTRODUCTION

M ULTIRATE systems are abundant in industry [17]; thereare several reasons for this.

• In multivariable digital control systems, often it is unre-alistic, or sometimes impossible, to sample all physicalsignals uniformly at one single rate. In such situations,one is forced to use multirate sampling.

• In general one gets better performance if one can usefaster A/D and D/A conversions, but this means a highercost in implementation. For signals with different band-widths, better tradeoffs between performance and im-plementation cost can be obtained using A/D and D/Aconverters at different rates.

• Multirate controllers are in general time-varying. Thusmultirate control systems can outperform single-rate sys-tems; for example, gain margin improvement [27], [16],simultaneous stabilization [27], and decentralized control[2], [44].

The study of multirate systems started in the late 1950’s[29], [25], [26]. Early studies were focused on analysis andwere solely for purely discrete-time systems; see also [32].A renaissance of research on multirate systems has occurredsince late 1980 with an increased interest in multirate con-troller design, e.g., stabilizing controller design and parameter-ization of all stabilizing controllers [11], [30], [36], LQG/LQRcontrol [8], [1], [31], optimal control [42], [43], [34],

Manuscript received July 21, 1995; revised March 3, 1997 and May 22,1998. Recommended by Associate Editor, M. A. Dahleh. This research wassupported by the Hong Kong Research Grants Council and the NaturalSciences and Engineering Research Council of Canada.

L. Qiu is with the Department of Electrical & Electronic Engineering, HongKong University of Science & Technology, Clear Water Bay, Kowloon, HongKong.

T. Chen is with the Department of Electrical & Computer Engineering,University of Alberta, Edmonton, Alberta, Canada T6G 2G7.

Publisher Item Identifier S 0018-9286(99)02090-5.

Fig. 1. The general multirate sampled-data setup.

control [42], [43], [10], optimal control [15], and thework in [3], [21], and [38]. With the recognition that manyindustrial control systems consist of an analog plant and adigital controller interconnected via A/D and D/A converters,direct optimal control of multirate systems has been studiedin this sampled-data setting [42], [10], [34]. The existingtechniques for multirate control allow for computationof one controller via a numerical convex optimization[43] or more easily via an explicit design [10]. The purposeof this paper is to characterize in an explicit way the set ofall suboptimal controllers and to find a particularsuboptimal controller which minimizes an entropy function.

In this paper we shall treat a general multirate setup. Forthis, we define the periodic sampler and the (zero-order)hold (the subscript denotes the period) as follows:mapsa continuous signal to a discrete signal and is defined via

maps discrete to continuous via

(The signals may be vector-valued.) Note that the sampler andhold are synchronized at

The general multirate system is shown in Fig. 1. We haveused continuous arrows for continuous signals and dottedarrows for discrete signals. Here, is the continuous-timegeneralized plant with two inputs, the exogenous inputand the control input and two outputs, the signal to becontrolled and the measured signal and are multiratesampling and hold operators and are defined as follows:

......

These correspond to samplingchannels of periodicallywith periods respectively, and holding

0018–9286/99$10.00 1999 IEEE

538 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 44, NO. 3, MARCH 1999

channels of with periods respectively.Here and are different integers and is a real numberreferred to as thebase period. If we partition the signalsaccordingly

......

......

then

is a discrete-time multirate controller, implemented via amicroprocessor; it is synchronized withand in the sensethat it inputs a value from theth channel at times andoutputs a value to theth channel at

In the general multirate setup of Fig. 1, we assume through-out that and are causal and linear. Furthermore,is assumed to be time-invariant and finite-dimensional, and

is assumed to satisfy certain periodic property and to befinite-dimensional.

For periodicity of let be the least common multipleof the sampling and hold indexes,Thus is the least common period for all sampling andhold channels. The multirate controller can be chosen sothat is -periodic in continuous time. For this, weneed a few definitions.

Let be the space of sequences, perhaps vector-valued,defined on the time set Let be the unit timedelay on and the unit time advance. Define the integers

We say is -periodic in real timeif

. . .. . .

This means shifting by time unitscorresponds to shifting by units Thus

is -periodic in continuous time iff is -periodicin real time.

Since is linear time-invariant (LTI), it follows that thesampled-data system in Fig. 1 is-periodic if is -periodic in real time. We shall refer to as thesystem period.We shall assume throughout the paper that is -periodicin real time. With all these assumptions, the controllercan be implemented via difference equations [10]

where causality requires if

Our goal in this paper is two-fold: 1) characterize allfeasible multirate controllers which internally stabilize thefeedback system shown in Fig. 1 and make the-inducednorm less than a prespecified value, such controllers arecalled suboptimal controllers and 2) among allsuboptimal controllers, find one which further minimizes anentropy function. Used with other optimization techniques,such a characterization, like its LTI counterpart [14], [22], isessential in designing control systems with simultaneousand other performance requirements. The minimum entropycontrol, also like its LTI counterpart [33], [23], [24], gives aparticular example of such multi-objective control problem inwhich an analytic solution exists.

Although the overall system shown in Fig. 1 is hybrid(involving both continuous-time and discrete-time signals) andtime-varying, the recently developed lifting technique enablesus to convert the problem into an equivalent LTI discrete-time problem. However, the resulting control problem willhave an undesirable and unconventional constraint on the LTIcontroller due to the causality requirement. This constraintis the main difficulty in designing optimal multirate systems.The recent introduction of the nest operators has proven to beeffective in handling causality constraints in multirate design[10]. The results of this paper will be built on the nest operatortechnique.

We would like to remark here that the results in this paperextend directly to periodic discrete-time systems, i.e., directapplication yields a characterization of all suboptimalsolutions which are periodic and causal; this result has notbeen obtained before.

The paper is organized as follows. The next section reviewssome basic facts about continuous-time periodic systems,introduces the concept of entropy for such systems, andestablishes the connection between the entropy and a linear,exponential, quadratic, Gaussian cost function. Section III ad-dresses topics on nest operators and nest algebra, which are themain tools to handle causality in this paper. Section IV brieflydiscusses the procedure of converting our hybrid probleminto an equivalent LTI problem with a causality constraint.Section V gives a characterization of all suboptimal con-trollers and the minimum entropy controller. The Appendicescontain two long and involved proofs.

Preliminary results in this paper have been presented atseveral conferences: the Asian Control Conference (Tokyo,Japan, 1994), the IEEE Conference on Decision and Control(Florida, USA, 1994), and the International Conference onOperator Theory and its Applications (Manitoba, Canada,1994).

Finally, we introduce some notation. Given an operatorand two operator matrices

the linear fractional transformation associated withandis denoted

QIU AND CHEN: MULTIRATE SAMPLED-DATA SYSTEMS 539

and the star product of and is shown in (a) at the bottomof the page. Here, we assume that the domains and codomainsof the operators are compatible and the inverses exist. Withthese definitions, we have

II. ENTROPY OF PERIODIC SYSTEMS

A multirate system as depicted in Section I is a continuous-time -periodic system. In this section, we review some basicconcepts of periodic systems and introduce the concept ofentropy.

Let and be Hilbert spaces andbe a sequence of bounded operators fromto

Then

is an operator-valued function on some subset ofWe saythat belongs to if is analytic in the openunit disk, and

In this case, the left-hand side above is defined to be thenorm of denoted by the operator is boundedfor almost every and

Now let be a sequenceof Hilbert–Schmidt operators from to The set ofHilbert–Schmidt operators equipped with the Hilbert–Schmidtnorm, is a Hilbert space [19]. Then

is a Hilbert-space vector-valued function on some subset ofWe say that belongs to if

In this case, the left-hand side above is defined to be thenorm of denoted by the operator is

Hilbert-Schmidt for almost every and

Assume andExtending the entropy definition for matrix valued analytic

functions [23], [24], we define the entropy of as

This entropy is well defined. Since is a Hilbert–Schmidt operator at almost every its singularvalues form a square-summable sequence Hence

which converges to some number in (0, 1) due to squaresummability of and the fact that Thisalso shows that is nonnegative.

Lemma 1: Assume andThen

1)

2) for with

and

The proof of Lemma 1 is similar to that for the finite-dimensional continuous-time case [33].

Now let us return to periodic systems. Let be acontinuous-time, -periodic, causal system described by thefollowing integral operator:

We assume that the matrix-valued impulse response ofis locally square integrable, i.e., every element is square

integrable on any compact subset of The periodicity ofimplies and the causality impliesthat if

The local square integrability of guarantees that isa linear map from to the space of locally square-integrable functions of Given an arbitrary positive integer

let

Denote the space of -valued sequences by Define thelifting operator via

......

(a)


This lifting gives an algebraic isomorphism betweenand [42]. We use the obvious norm in

...

where is the norm on Denote by thesubset of consisting of all sequences with

and define the norm on to be the left-hand side of theabove inequality. It is clear that if and only if

and is a Hilbert-space isometric isomorphismfrom to

Now we lift to get The lifted systemcan be described by

where map to via the equation shownin (b) at the bottom of the page. The local square integrabilityof ensures that are Hilbert-Schmidtoperators [46].

For -periodic the lifted system is LTI in discretetime; its transfer function is defined as

So if and its entropycan be defined.

We will define the norm, norm, and entropy ofto be those of respectively. Actually, the norm definedthis way is indeed the -induced norm of [7], [5], [40];the norm has natural interpretations in terms of impulseresponses and white noise responses [6], [28]; the entropy notonly provides an upper bound for the norm as stated inLemma 1, but also has a stochastic interpretation in terms of alinear exponential quadratic Gaussian (LEQG) cost function,similar to the case of matrix-valued transfer functions [18].

To avoid an unnecessary technicality, we will concentrate onfinite-dimensional periodic systems, i.e., thosewith finite-dimensional realizations, or equivalently, those whoselifted transfer functions have only a finite number of poles.(The multirate systems to be studied in Fig. 1 fall in thisclass if both and are finite-dimensional.) Let bea Gaussian white noise with zero mean and unit covariance

on the time interval and the corresponding response:Define an LEQG cost function for as

where means the expectation. The proof of the followingtheorem is given in Appendix A.

Theorem 1: Given a finite-dimensional -periodic systemassume its lifted transfer function satisfies

and Then

Now we are ready to state our control problems associatedwith Fig. 1 precisely.

Given a continuous-time finite-dimensional LTI plantand sampling and hold schemesand

1) characterize all feasible multirate controllers suchthat the feedback system is internally stable and

2) find a particular controller from those obtained in (1)such that the entropy

is minimized.These problems will be solved explicitly in Sections V andVI. Next, we present the required mathematical tool based onnest operators.

III. N EST OPERATORS

In this section, we address some issues on nest operatorsand nest algebra [4], [12], which are useful in the sequel.Our main purpose is to probe further the Arveson’s distanceproblem, that is, we characterize explicitly all nest operatorswhich are within a fixed distance from a given operator; wealso give one such nest operator which minimizes an auxiliaryentropy function. The same problems were also studied in themathematical literature [45], but the solutions are different.Our results, based on the unitary dilation, provide furtherinsight as well as certain numerical advantages; they takeforms which are easily applicable to our control problems athand.

Let be a vector space. Anest in denoted isa chain of subspaces in including and with thenonincreasing ordering

(A nest may be defined to contain an infinite number of spaces,but this generalization is not necessary in the sequel.)

......

...

(b)


Let and be both Hilbert spaces. Denote bythe set of bounded linear operators and abbreviateit as if Assume that and are equipped,respectively, with nests and which have the samenumber of subspaces, say, as above. An operator

is said to be anest operatorif

(1)

It is said to be astrict nest operatorif

(2)

Let and be orthogonalprojections. Then the condition in (1) is equivalent to

and the condition in (2) is equivalent to

Given the nests and the set of all nest opera-tors is denoted and abbreviated if

the set of all strict nest operators is denotedand abbreviated if

If we decompose the spacesand in the following way:

(3)

(4)

then the associated matrix representation ofis

......

...

and means that this matrix representationis (block) lower triangular: if The followinguseful lemmas can be proven readily by using the above matrixrepresentation.

Lemma 2:

1) If and then

2) If and orif and then

3) forms an algebra, called anest algebra.In the rest of this section, we restrict our discussion to finite-dimensional spaces.

Lemma 3:

1) If then is always invertible.2) If and is invertible, then

Lemma 4 (Generalized Factorization): Let

1) There exist a unitary operators on andsuch that

2) There exist and a unitary operatoron such that

Lemma 5 (Generalized Cholesky Factorization):Letand assume is self-adjoint and nonnegative.

1) There exists such that2) There exists such thatThe purpose of the rest of this section is to address the

following two matrix problems; given : 1)characterize all such thatand 2) find, among all characterized in (1), the one whichminimizes Here the entropy of a contractive matrix

is obtained as a special case from the entropy definition ofa contractive Hilbert–Schmidt operator-valued function

These two matrix problems are closely related to and areactually simple special cases of the main problems of thispaper: Characterize all suboptimal controllers and findthe minimum entropy controller.

We shall need some more notation. Withand as before,introduce two more finite-dimensional inner-product spacesand A linear operator is partitionedas

with etc. For nestsin respectively, all with

subspaces, the nests and aredefined in the obvious way. Hence writing

means etc.Theorem 2: Let The following statements

are equivalent.1) There exists such that

2)3) There exists

with and both invertible andsuch that

is unitary.The proof of Theorem 2 is given in Appendix B. This

theorem can be used to solve our first matrix problem.Theorem 3: Let and assume condition 3) in

Theorem 2 is satisfied. Then the set of allsuch that is given by

and (5)


Proof: Since

is unitary and are invertible, it follows from [37] thatthe map

is a bijection from the open unit ball of ontoitself. What is left to show is thatiff The “if” part follows from Lemma 2by noting For the “onlyif” part, assume for some

we need to show that too belongs toFrom

we obtain after some algebra

(6)

Since

it follows that is invertible. Hencefrom (6)

Therefore belongs to by Lemma 2.The characterization in Theorem 3 also renders an easy

solution to the second matrix problem.Theorem 4: Let and assume condition 3) in

Theorem 2 is satisfied. Then the unique which satisfiesand minimizes is given by

Proof: According to Theorem 3, all satisfyingare characterized by (5). Consequently, all resultingare given by

and

By Lemma 1, we obtain

Notice that the second term is independent ofandwhich implies that the third term is zero.

Therefore the minimizing is zero and henceOne implication of Theorem 4 is that althoughin condi-

tion 3) of Theorem 2 is not unique, is uniquely determined.

IV. EQUIVALENT LTI SYSTEMS

Our main problems deal with hybrid time-varying systems.Following [10] and [42], we can reduce the control problem toan equivalent one involving only finite-dimensional LTI sys-tems. In this section we briefly review the reduction process.The detailed justification is referred to [10], [42], and [5]. Ouremphasis here is on the relationship between the entropy ofthe original system and the equivalent LTI system.

We start with a state model of

For an integer define the discrete lifting operatorvia

......

Denote

......

and recall the continuous lifting operator in Section II:Here we take We lift and by defining

and

It is easy to check that and are LTI systems, so they

have transfer functions and By definitions

A state-space realization of can be computed

Due to the causality of and the lifted systemsand have some special structures which can be easilycharacterized using nest operators.

Write

Then

Note that is sampled at Similarly


Fig. 2. The lifted system.

and occurs at For define

if

if

Then the -blocks in the lifted plant satisfy

(7)

(8)

(9)

(10)

and for to be causal

(11)

Hence we have arrived at an equivalent LTI problem, shownin Fig. 2, with plant and controller Note that (7)–(10)give special structures of that can be exploited, whereas(11) is a design constraint on that has to be respected inorder for to correspond to a causal

The signals and in Fig. 2 take values in infinite-dimensional spaces. In other words,are operators with either domain or codomain being infinite-dimensional spaces. To overcome this difficulty, we observethat all these operators except have finite rank.

Due to the particular choice of decomposition of andthe operator takes a lower-triangular Toeplitz form

......

The only block with infinite rank is Our next step isto get rid of this by a linear fractional transformation. Since

the diagonal blocks of

are invariant for any satisfying (11). Thereforeis a necessary condition for the solvability

of our control problem. From now on we assume thiscondition is satisfied.

Define a diagonal operator matrix

...

and a Julian operator matrix

Let

Then it is well known [37] that iff

The relationship between the entropiesis given in the following lemma.

Lemma 6:

Proof: By Lemma 1

Since is a constant operator function

Note that and

whose first term is in and second term inHence

The result then follows.A state-space model of can again be computed

Since is diagonal, i.e., andit follows:

Note that the diagonal blocks of have been cancelledby the linear fractional transformation, resulting in a strictly


Fig. 3. The equivalent finite-dimensional LTI system.

(block) lower-triangular Then the advantage of overis that all operators and are of finiterank. Therefore, if we define

and

then has finite-dimensional input and output spaces and

The nests and induce nests in and in anatural way

Assume that a state-space model ofis

The following structure of is inherited from that of

(12)

(13)

(14)

(15)

In summary, our original hybrid time-varying control prob-lem with plant and controller can be converted intoa finite-dimensional LTI control problem with plant andcontroller as shown in Fig. 3, in the sense that the system inFig. 3 is internally stable iff the system in Fig. 1 is internallystable

and

A state-space model of can be computed from that ofusing the techniques developed in [5]. Any satisfying (11)resulted from the design can be converted into a feasible mul-tirate controller We would like to emphasize, however,

that the finite-dimensional LTI problem has a nonconventionalconstraint on the controller given by (11). This constraint isthe causality constraint. Also, the LTI plant obtained from

will automatically satisfy (12)–(15).In order for the problem for the finite-dimensional LTI

generalized plant to be solvable, we need the following.1) is stabilizable and detectable. Some of the

existing techniques to solve the problem forrequire to satisfy the following additional conditions.

2)for all

3)range for all

First it is shown in [35] that if:1a) is stabilizable and detectable andis

nonpathological with respect to2a) has no unobservable modes on the imaginary

axis, is right-invertible and hasno zero at 0;

3a) has no uncontrollable modes on the imagi-nary axis and is left-invertible;

then:1) is stabilizable and detectable

2)for all

3)for all

Now assume that 1a)–3a) and hence1)–3), are satisfied.Then it follows from the same argument as in [20, Section IV-F] that conditions 1)–3) are satisfied if there exists aninternally stabilizing multirate controller such that

V. ALL SUBOPTIMAL CONTROLLERS

AND THE MINIMUM ENTROPY CONTROLLER

In this section, we first characterize all satisfying thecausality constraint (11) such that the system shown in Fig. 3is internally stable and This problem differsfrom the standard problem only in the causality constrainton and is hence called a constrained problem. Ourstrategy in solving this problem is first to characterize all

such that the system in Fig. 3 is internally stable andwithout considering the causality constraint

(this is a standard problem) and then choose, if possible,from this characterization all those satisfying the causalityconstraint.

Several solutions to the standard problem exist in theliterature. Here we adopt the solution in [22]. Note that itis assumed in [22] that and theseassumptions are not satisfied for the equivalent LTI system

However, they are not essential and the solution in [22]can be modified accordingly by following, e.g., the idea in[39]. Assume the solvability conditions are satisfied, then


all stabilizing controllers satisfying arecharacterized by

is invertible (16)

where is not uniquely given in [22] and

by using Lemma 5 we can always chooseso that

and furthermore, and are invertible.Theorem 5: The constrained problem is solvable iff

the corresponding unconstrained problem is solvable and

(17)

Proof: Obviously, the corresponding unconstrained prob-lem has to be solvable in order for the constrained problemto be solvable. Assume that the unconstrained problem issolvable. Since it follows that

iff

Pre- and postmultiply this by and , re-spectively, to get

It follows from Theorem 1 that in order to haveand we

must have (17). Conversely, if (17) is true, then there existsa constant matrix with such that

Hence

achievesIf the conditions in Theorem 5 are satisfied, then there exists

with and and invertible suchthat

is unitary. Define

It is easy to check thatand are invertible, and

By setting the set (16) canbe rewritten as

is invertible

Now we can state the main result of this paper.Theorem 6: Assume the solvability of the constrained

problem. Then the set of all controllers solving the problemis given by

(18)

Proof: First notice that is always in-vertible if Since

and and are invertible, it fol-lows that iffThen the result follows immediately.

Theorem 6 gives a characterization of all controllersin terms of a linear fractional transformation of an attractive

function satisfying the causality constraint. Clearly, thischaracterization is not unique in general, although the set ofsuch controllers is unique. It is then of interest to explicitly

characterize the nonuniqueness. Let be another matrix

satisfying andsuch that

Then it can be shown using the standard theory on linearfractional transformation (LFT) (see [20, Ch. 4] for example)that

where is a unitary matrix belonging toand is a unitary matrix belonging towhere is the set of adjoints of members in andsimilarly for i.e., and are block diagonal

unitary matrices. In particular, this implies Sinceis a particular suboptimal controller by setting

in (18), we call it the central controller. Notice that thecentral controller with the causality constraint is different fromthe central controller without the causality constraint.

In the rest of this section, we show that the central controllerobtained by setting in (18) is the controller whichminimizes

Now let us go back to the characterization given in [22]. Itis known (see [33] for the continuous-time case) that if allsuboptimal controllers are characterized by (16), then all


suboptimal closed-loop transfer functions are characterized by

is invertible

where

is para-unitary satisfying Clearly we have

and Because of this, the controller with-out the causality constraint which minimizes the entropy

is conveniently given byNotice that gives

where

Consequently, if we characterize the controller using (18), thenall suboptimal closed-loop transfer functions are

where

Since is para-unitary and is unitary, it follows that

is para-unitary. It can be checked that andBy Lemma 1

The last equality is due to the factTherefore, the minimum of is achieved atThe following theorem is thus obtained.

Theorem 7: The minimum entropy controller is given by

That is, the minimum entropy controller (with the causalityconstraint) is given by the central controller (with thecausality constraint).

APPENDIX APROOF OF THEOREM 1

The proof of Theorem 1 follows from the idea in [18] buthas two complications: 1) operator-valued transfer functionsare treated, which requires dealing with random variables inHilbert spaces [41] and 2) signals are defined on timeinstead of which requires treating nonstationarystochastic processes. Since is linear, it follows that isa Gaussian process. Define as the stochastic process on

such that for Then can beconsidered as a Gaussian random variable in the Hilbert space

The covariance operator isthen given by

This shows that Sinceis a contractive Hilbert–Schmidt operator

and is causal, it follows that is a self-adjoint contractivenuclear operator. Let the Schmidt expansion of be

Then can be expressed as


and are independent scalar Gaussian randomvariables with covariance Hence

Now lift to get and lift to get Then isequivalent to and has a matrix representation

......

.. ... .

Let be the leading submatrix of Then

Since has only a finite number of poles, the infinite Hankelmatrix

......

...

has finite rank. Let be the first block rows of anddefine

Notice that is a self-adjoint Toeplitz matrix, as shown in(c) at the bottom of the page, and is the th Fourier coeffi-cient of where Denote byand the singular values of and

respectively assuming ordered nondecreasingly. Then

Since and are all contained init follows that

for some This shows that

Hence by using the operator-valued strong Szego–Widom limittheorem [9, Th. 6.4]

Notice that for

Therefore .

APPENDIX BPROOF OF THEOREM 2

The equivalence of 1) and 2) follows from the Arveson’sdistance formula [12]. That 3) implies 1) is obvious. It remainsto show that 2) implies 3). For this, we need a technical lemma.

Lemma 7: Assume the matrices and of appropri-ate dimensions, satisfy the conditions

Then there exists a matrix satisfying

An explicit formula for such a matrix is

......

.... . .

...

(c)


Proof: It follows from [13] that there exists a matrixsuch that

Among all such characterized in [13] in terms of a freecontractive matrix, the “central” one obtained by setting thefree contractive matrix to zero is

Using this we have

and

The last inequality follows fromTo avoid awkward notation in the proof of Theorem 2, we

redefine

Under the decompositions of and in (3) and (4), we getthe matrix representation shown in (d) at the bottom of the

page. Statement 2) becomes

......

We need to decide for and forThis will be done in the following order: In theth step,determine those blocks in the th row and the th row.

Step 1: Set and choose so that

is a co-isometry. Statement 2) implies that any chosen inthis way is nonsingular.

Step Set and choose the rest ofthe th row so that it is a co-isometry and is orthogonalto all of the previously determined rows. This requires

to be an isometry onto the kernel of the matrix shown in (e)at the bottom of the page. Then set and choose

in such a way so that the matrix shown in (f), at the bottom ofthe page, is a contraction and it is orthogonal to all previouslydetermined block rows. This is possible following Lemma 7,

......

.. ....

......

. . ....

......

. . ....

......

. . ....

(d)

......

.... . .

...

......

......

(e)

......

......

.... . .

...

.... . .

......

......

. . ....

(f)


......

......

.... . .

...

.... . .

......

......

. . ....

(g)

......

.... . .

...

......

......

(h)

condition 3), and the fact that the matrix shown in (g), at thetop of the page, is a co-isometry. Finally determine so that

is a co-isometry. By Lemma 7, any chosen in such a wayis nonsingular.

Step Set and choose the rest of the th rowso that it is orthogonal to all the previously determined rows.This requires

to be an isometry onto the kernel of the matrix shown in (h)at the top of the page.

Finally set

andThe above construction guarantees that the matrix

(19)

is unitary, is invertible, and Theinvertibility of follows from that of and the fact that thematrix in (19) is unitary.

ACKNOWLEDGMENT

The authors would like to thank A. Heunis and V. Solo forhelpful discussions.

REFERENCES

[1] H. Al-Rahmani and G. F. Franklin, “A new optimal multirate control oflinear periodic and time-varying systems,”IEEE Trans. Automat. Contr.,vol. 35, pp. 406–415, 1990.

[2] B. D. O. Anderson and J. B. Moore, “Time-varying feedback laws fordecentralized control,”IEEE Trans. Automat. Contr., vol. AC-26, pp.1133–1139, 1981.

[3] M. Araki and K. Yamamoto, “Multivariable multirate sampled-datasystems: State-space description, transfer characteristics, and Nyquistcriterion,” IEEE Trans. Automat. Contr., vol. 30, pp. 145–154, 1986.

[4] W. Arveson, “Interpolation problems in nest algebras,”J. FunctionalAnalysis, vol. 20, pp. 208–233, 1975.

[5] B. Bamieh and J. B. Pearson, “A general framework for linear periodicsystems with application toH1 sampled-data control,”IEEE Trans.Automat. Contr., vol. 37, pp. 418–435, 1992.

[6] , “TheH2 problem for sampled-data systems,”Syst. Contr. Lett.,vol. 19, pp. 1–12, 1992.

[7] B. Bamieh, J. B. Pearson, B. A. Francis, and A. Tannenbaum, “A liftingtechnique for linear periodic systems with applications to sampled-datacontrol,” Syst. Contr. Lett., vol. 17, pp. 79–88, 1991.

[8] M. C. Berg, N. Amit, and J. Powell, “Multirate digital control systemdesign,” IEEE Trans. Automat. Contr., vol. 33, pp. 1139–1150, 1988.

[9] A. Bottcher and B. Silbermann, “Operator-valued Szeg¨o-Widom limittheorems,”Toeplitz Operators and Related Topics, E. L. Basor and I.Gohberg, Eds. Boston, MA: Birkhauser, 1994, pp. 33–53.

[10] T. Chen and L. Qiu, “H1 design of general multirate sampled-datacontrol systems,”Automatica, vol. 30, pp. 1139–1152, 1994.

[11] P. Colaneri, R. Scattolini, and N. Schiavoni, “Stabilization of multiratesampled-data linear systems,”Automatica, vol. 26, pp. 377–380, 1990.

[12] K. R. Davidson,Nest Algebras. Essex, U.K.: Longman Scientific &Technical, 1988.

[13] C. Davis, W. M. Kahan, and H. F. Weinberger, “Norm-preservingdilations and their applications to optimal error bounds,”SIAM J. Numer.Anal., vol. 19, pp. 445–484, 1982.

[14] J. C. Doyle, K. Glover, P. P. Khargonekar, and B. A. Francis, “State-space solutions to standardH2 andH1 control problems,”IEEE Trans.Automat. Contr., vol. 34, pp. 831–847, 1989.

[15] M. A. Dahleh, P. G. Voulgaris, and L. S. Valavani, “Optimal and robustcontrollers for periodic and multirate systems,”IEEE Trans. Automat.Contr., vol. 37, pp. 90–99, 1992.

[16] B. A. Francis and T. T. Georgiou, “Stability theory for linear time-invariant plants with periodic digital controllers,”IEEE Trans. Automat.Contr., vol. 33, pp. 820–832, 1988.

[17] D. P. Glasson, “Development and applications of multirate digitalcontrol,” IEEE Contr. Syst. Mag., vol. 3, pp. 2–8, 1983.

[18] K. Glover and J. C. Doyle, “State-space formulae for all stabilizing con-trollers that satisfy anH1-norm bound and relations to risk sensitivity,”Syst. Contr. Lett., vol. 11, pp. 167–172, 1988.

[19] I. C. Gohberg and M. G. Kre˘ın, Introduction to the Theory of Lin-ear Nonselfadjoint Operators. Providence, RI: American Math. Soc.,1969.

[20] M. Green and D. J. N. Limebeer,Linear Robust Control. EnglewoodCliffs, NJ: Prentice-Hall, 1995.

[21] T. Hagiwara and M. Araki, “Design of a stable feedback controller basedon the multirate sampling of the plant output,”IEEE Trans. Automat.Contr., vol. 33, pp. 812–819, 1988.

[22] P. A. Iglesias and G. Glover, “State-space approach to discrete-timeH1 control,” Int. J. Contr., vol. 54, pp. 1031–1073, 1991.


[23] P. A. Iglesias, D. Mustafa, and G. Glover, “Discrete timeH1 con-trollers satisfying a minimum entropy criterion,”Syst. Contr. Lett., vol.14, pp. 275–286, 1990.

[24] P. A. Iglesias and D. Mustafa, “State-space solution of the discrete-time minimum entropy control problem via separation,”IEEE Trans.Automat. Contr., vol. 38, pp. 1525–1530, 1993.

[25] E. I. Jury and F. J. Mullin, “The analysis of sampled-data control systemswith a periodically time-varying sampling rate,”IRE Trans. Automat.Contr., vol. 24, pp. 15–21, 1959.

[26] R. E. Kalman and J. E. Bertram, “A unified approach to the theory ofsampling systems,”J. Franklin Inst., no. 267, pp. 405–436, 1959.

[27] P. P. Khargonekar, K. Poolla, and A. Tannenbaum, “Robust control oflinear time-invariant plants using periodic compensation,”IEEE Trans.Automat. Contr., vol. 30, pp. 1088–1096, 1985.

[28] P. P. Khargonekar and N. Sivashankar, “H2 optimal control forsampled-data systems,”Syst. Contr. Lett., vol. 18, no. 3, pp. 627–631,1992.

[29] G. M. Kranc, “Input-output analysis of multirate feedback systems,”IRE Trans. Automat. Contr., vol. 3, pp. 21–28, 1957.

[30] D. G. Meyer, “A parametrization of stabilizing controllers for multiratesampled-data systems,”IEEE Trans. Automat. Contr., vol. 35, pp.233–236, 1990.

[31] , “Cost translation and a lifting approach to the multirate LQGproblem,” IEEE Trans. Automat. Contr., vol. 37, pp. 1411–1415, 1992.

[32] R. A. Meyer and C. S. Burrus, “A unified analysis of multirate andperiodically time-varying digital filters,”IEEE Trans. Circuits and Syst.,vol. 22, pp. 162–168, 1975.

[33] D. Mustafa and K. Glover,Minimum EntropyH1 Control, LectureNotes in Control and Information Sciences. New York: Springer-Verlag, vol. 146, 1991.

[34] L. Qiu and T. Chen, “H2-optimal design of multirate sampled-datasystems,”IEEE Trans. Automat. Contr., vol. 39, pp. 2506–2511, 1994.

[35] L. Qiu and K. Tan, “Direct state space solution of multirate sampled-dataH2 optimal control,”Automatica, vol. 34, pp. 1431–1437, 1998.

[36] R. Ravi, P. P. Khargonekar, K. D. Minto, and C. N. Nett, “Con-troller parametrization for time-varying multirate plants,”IEEE Trans.Automat. Contr., vol. 35, pp. 1259–1262, 1990.

[37] R. M. Redheffer, “On a certain linear fractional transformation,”J. Math.Phys., vol. 39, pp. 269–286, 1960.

[38] M. E. Sezer and D. D. Siljak, “Decentralized multirate control,”IEEETrans. Automat. Contr., vol. 35, pp. 60–65, 1990.

[39] A. A. Stoorvogel, “The discrete timeH1 control problem with mea-surement feedback,”SIAM J. Contr. Optim., vol. 30, pp. 182–202,1992.

[40] H. T. Toivonen, “Sampled-data control of continuous-time systems withan H1 optimality criterion,” Automatica, vol. 28, no. 1, pp. 45–54,1992.

[41] N. N. Vakhania, V. I. Tarieladze, and S. A. Chobanyan,ProbabilityDistributions on Banach Spaces. Dordrecht, Holland: D. Reidel, 1987.

[42] P. G. Voulgaris and B. Bamieh, “OptimalH1 andH2 control of hybridmultirate systems,”Syst. Contr. Lett., vol. 20, pp. 249–261, 1993.

[43] P. G. Voulgaris, M. A. Dahleh, and L. S. Valavani, “H1 and H2

optimal controllers for periodic and multi-rate systems,”Automatica,vol. 30, pp. 251–263, 1994.

[44] S. H. Wang, “Stabilization of decentralized control systems via time-varying controllers,” IEEE Trans. Automat. Contr., vol. AC-27, pp.741–744, 1982.

[45] H. J. Woerdeman, “Strictly contractive and positive completions forblock matrices,”Linear Algebra and Its Appl., vol. 136, p. 105, 1990.

[46] N. Young,An Introduction to Hilbert Spaces. Cambridge, U.K.: Cam-bridge Univ., 1988.

Li Qiu (S’85–M’90–SM’98) received the B.Engdegree in electrical engineering from Hunan Uni-versity, Changsha, Hunan, China, in 1981, and theM.A.Sc. and Ph.D. degrees in electrical engineer-ing from the University of Toronto, Toronto, Ont.,Canada, in 1987 and 1990, respectively.

Since 1990, he has held research and teaching po-sitions in the University of Toronto, Canadian SpaceAgency, University of Waterloo, and University ofMinnesota. At present, he is an Assistant Professorat the Department of Electrical and Electronic Engi-

neering, Hong Kong University of Science and Technology, Clear Water Bay,Kowloon, Hong Kong. His current research interests include robust control,digital control, signal processing, and motor control.

Dr. Qiu was an Associate Editor of the IEEE TRANSACTIONS ONAUTOMATIC

CONTROL and is currently an Associate Editor ofAutomatica.

Tongwen Chen (S’86–M’91–SM’97) received theB.Sc. degree from Tsinghua University, Beijing,China, in 1984, and the M.A.Sc. and Ph.D. degreesfrom the University of Toronto in 1988 and 1991,respectively, all in electrical engineering.

From October 1991 to April 1997, he was onfaculty in the Department of Electrical and Com-puter Engineering at the University of Calgary,Canada. Since May 1997, he has been with theDepartment of Electrical and Computer Engineeringat the University of Alberta, Edmonton, Canada, and

is presently an Associate Professor. His current research interests includedigital control, digital signal processing, optimal and robust design, involvingespecially multirate systems. He coauthored (with B.A. Francis) the bookOptimal Sampled-Data Control Systems(New York: Springer, 1995).

Dr. Chen is an Associate Editor for IEEE TRANSACTIONS ON AUTOMATIC

CONTROL. He is a registered Professional Engineer in Alberta, Canada.

multirate sampled-data systems: all h/spl infin/ suboptimal … · troller design, e.g.,...

Documents