Scholarly article on topic 'Numerical Optimization Design of Dynamic Quantizer via Matrix Uncertainty Approach'

Numerical Optimization Design of Dynamic Quantizer via Matrix Uncertainty Approach Academic research paper on "Mathematics"

CC BY
0
0
Share paper
Academic journal
Mathematical Problems in Engineering
OECD Field of science
Keywords
{""}

Academic research paper on topic "Numerical Optimization Design of Dynamic Quantizer via Matrix Uncertainty Approach"

Hindawi Publishing Corporation Mathematical Problems in Engineering Volume 2013, Article ID 250683,12 pages http://dx.doi.org/10.1155/2013/250683

Research Article

Numerical Optimization Design of Dynamic Quantizer via Matrix Uncertainty Approach

Kenji Sawada and Seiichi Shin

Department of Mechanical Engineering and Intelligent Systems, The University of Electro-Communications, 1-5-1 Chofugaoka, Chofu, Tokyo, 182-8585, Japan

Correspondence should be addressed to Kenji Sawada; knj.sawada@uec.ac.jp Received 12 July 2013; Accepted 12 September 2013 Academic Editor: Zidong Wang

Copyright © 2013 K. Sawada and S. Shin. 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.

In networked control systems, continuous-valued signals are compressed to discrete-valued signals via quantizers and then transmitted/received through communication channels. Such quantization often degrades the control performance; a quantizer must be designed that minimizes the output difference between before and after the quantizer is inserted. In terms of the broadbandization and the robustness of the networked control systems, we consider the continuous-time quantizer design problem. In particular, this paper describes a numerical optimization method for a continuous-time dynamic quantizer considering the switching speed. Using a matrix uncertainty approach of sampled-data control, we clarify that both the temporal and spatial resolution constraints can be considered in analysis and synthesis, simultaneously. Finally, for the slow switching, we compare the proposed and the existing methods through numerical examples. From the examples, a new insight is presented for the two-step design of the existing continuous-time optimal quantizer.

1. Introduction

With the rapid network technology development, the networked control systems (NCSs) have been widely studied [1-10]. One of the challenges in NCSs is quantized control. In NCSs, the continuous-valued signals are compressed and quantized to the discrete-valued signals via the quantizer of the communication channel, and such quantization often degrades the control performance. Hence, a desirable quantizer minimizes the performance error between before and after the quantizer insertion.

Motivated by this, researchers [11-14] have provided optimal dynamic quantizers for the following problem formulation in the discrete-time domain. For a given plant P, synthesize a "dynamic" quantizer Qd such that the system Zq composed of P and Qd in Figure 1(a) "optimally" approximates the plant P in Figure 1(b) in the sense of the input-output relation. The obtained quantizer allows us to design various controllers for the plant P on the basis of the conventional control theories. Also, this framework is

helpful in not only the NCS problem but also various control problems such as hybrid control, embedded system control, and on-off actuator control.

When we consider controlling a mechanical system with an on-off actuator, first the controlled object and its uncertainties are usually modeled in the continuous-time domain. Second, the model and its uncertainties are discretized to apply the above dynamic quantizer. However, the discretization sometimes results in uncertainties more complicated than those in the original model and creates undesirable complexity in robust control. The continuous-time setting quantizer is more suitable for the robust control of the quantized system than discrete-time one. Thus, our previous works [15,16] have considered the continuous-time setting, while a number of the discrete-time settings have been studied by others [11-14]. In these works, it is assumed that the switching process of discretizing the continuous-valued signal is sufficiently quick relative to the control frequency and only the spatial determination (quantized accuracy) is considered as the quantization effect. This is

u Qd v p

(a) Discrete-valued input system

u z*

Plant (b) Usual system

Figure 1: Two systems.

because the switching speed of the continuous-time delta-sigma modulator for wireless broadband network systems is from 1 MHz to 100 MHz [17,18].

On the other hand, the above assumption is essentially weak in the case of the slow switching such as the mechanical systems with on-off actuators [19]. For the slow switching, we need to consider the quantization effect on both the switching speed and the spatial constraints in continuous time. For example, Ishikawa et al. proposed a two-step design of a feedback modulator [20]: (i) the control performance of the modulator is considered under only the spatial constraint, and (ii) the modulator is tuned in terms of the switching speed constraint. However, the structure of the modulator is more restricted than that of the dynamic quantizer and the obtained modulator is not always optimal. Therefore, the dynamic quantizer under temporal resolution (switching speed) and spatial resolution constraints has still to be optimally designed. The simultaneous consideration of the two constraints is the particular challenge we address in this paper.

We propose a numerical optimization method for the continuous-time dynamic quantizer under switching speed and quantized accuracy constraints. To achieve the method, this paper solves the design problem via sampled-data control framework that has so far provided various results for networked control problems [7-9]. We refer to the previous work on optimal dynamic quantizer design [11, 12] and consider the basic feedforward system in Figure 1(a). In addition to the invariant set analysis [21, 22] similarly to our previous works [13, 15, 16], this paper utilizes a matrix uncertainty approach [23, 24] that is proposed in a sampleddata control framework. Although the obtained results can be more conservative than those in the previous works on continuous-time dynamic quantizer [15, 16] from the viewpoint of the class of the exogenous input and the applicable plants, both temporal and spatial resolution constraints can be addressed in analysis and synthesis, simultaneously. For the fast switching case, the proposed conditions converge to the corresponding conditions of our previous works. Finally, for the slow switching, we compare the proposed and existing methods [15,16] through numerical examples. In particular, a new insight is presented for the two-step design of the existing continuous-time optimal quantizer.

Notation. The set of nxm (positive) real matrices is denoted by R"xm (R+Xm). The set of nxm (positive) integer matrices is denoted by N"xm (N+Xm). We denote by L^ the set

Figure 2: Continuous-time dynamic quantizer with switching speed h.

of piecewise-continuous functions of p-dimensional finite vectors such that ra-norm of its functions is finite. 0nxm and Im (or for simplicity of notation, 0 and I) denote the nxm zero matrix and the m x m identity matrix, respectively. For a matrix M, MT, p(M) and omax(M) denote its transpose, its spectrum radius, and its maximum singular value, respectively. For a vector x, x{ is the ith entry of x. For a symmetric matrix X, X > 0 (X > 0) means that X is positive (semi) definite. For a vector x and a sequence of vectors X := [x1,x2,...}, ||%|| and ||X|| denote their ra-norms, respectively. Finally, we use the "packed" notation ( £ | ^ ) := C(sl - A)-1B + D.

2. Problem Formulation

Consider the discrete-valued input system Zq in Figure 1(a), which consists of the linear time invariant (LTl) continuous-time plant P and the quantizer v = Qj(u). The system P is given by

X A B X

Z C 0 V

where % e R", z e R?, u e Rm, and v e Rm denote the state vector, the measured output, the exogenous input, and the quantizer output, respectively. The continuous-valued signal u is quantized into the discrete-valued signal v via the quantizer Qd. We assume that the matrix A is Hurwitz; that is, the usual system in Figure 1(b) is stable in the continuous-time domain. The initial state is given as x(0) = x0.

For the system P, consider the continuous-time dynamic quantizer v = Qj(u) with the state vector Xq e R"Q as shown in Figure 2. Its switching speed he R+ (or its temporal resolution) is determined by the operator HS, which converts

the continuous-time signal g into the low temporal resolution signal g as follows:

HS : g g : g(kh + d) = g[k], g[k]=g (kh), k = 0,1,2,3,..., d e [0,h).

That is, Mq = HSuq. S is the ideal sampler with the sampling period h and H is the zero-order hold operator. The spatial resolution of the quantizer Qd is expressed by the static quantizer q : Rm ^ dNm with the quantization interval d e R+, that is,

' = q (HSu

Mq = U + VQ

and the continuous-time LTI filter Q is given by

Xq bq] Xq

.VQ. cq 0 eQ.

Note that q is of the nearest-neighbor type toward ->x> such as the midtread quantizer in Figure 3 (mq;) - Mq;|| < d/2 where qt and Mq; are the ith row of q and Mq) and the initial state is given by %q(0) = 0 for the drift free of Qd [11,12].

Remark 1. In synthesis, our previous works [15, 16] ignored the operator HS. In implementation, however, the continuous-time quantizer needs the switching process discretizing the continuous-valued signal. Of course, the applicable interval of switching depends on controlled objects such as narrowband or broadband networked systems and mechanical systems with on-off actuators. Therefore, it is important to consider the operator HS in synthesis.

For the system Zq in Figure 1(a) with the initial state x0 and the exogenous input u e L™, z(t, x0, Qd(u)) denotes the output of z at the time t. Also, for the system in Figure 1(b) without Qd, z*(t,x0,u) denotes its output at the time t. Consider the following cost function:

(x0,u)eR"xL™ teR+u{0)

sup \\z(t,Xo,Qd (U))-Z* (t,Xo,t

If the quantizer minimizes J(Qd), the system Zq "optimally" approximates the usual system P in the sense of the input-output relation. In this case, we can use the existing continuous-time controller design methods for the system in Figure 1(b) without considering the quantization effect. When the controlled object and its uncertainties are modeled in the continuous-time domain, therefore, the continuous-time quantizer can introduce robust control of the continuous-time setting directly, while the discrete-time quantizer requires discretization of the whole control system. Our previous works [15, 16] proposed an optimal dynamic quantizer for the cost function J(Qd) for the fast switching case h = 0. That is, only the spatial deterioration has been considered.

Figure 3: Midtread quantization.

On the other hand, the simultaneous consideration of the temporal and spatial resolution constraints is the problem we address in this paper. To consider the temporal resolution constraint caused by the operator HS, this paper modifies the cost function as follows:

Jhs (q,

sup sup sup \\z (kh + d, x0, Qd (u))

(x0,u)eR"xL™ 0e[O,h) fceN+u{0)

-z* (kh + d, x0

Fixing 9 = 0 ignores the output error between before and after the quantizer is inserted over the kth sampling interval and leads to the cost function setting that is utilized for the discrete-time optimal dynamic quantizers [11-14]. Therefore, the optimal quantizer for JHS(Qd) minimizes the output error between the systems in Figures 1(a) and 1(b) in terms of the input-output relation under the temporal and spatial resolution constraints.

Motivated by the above, our objective is to solve the following continuous-time dynamic quantizer synthesis problem (E): for the system Zq composed of P and Qd with the initial state x0 e R" and the exogenous input u e I™, suppose that the quantization interval d e R+, the switching speed h e R+, and the performance level ye R+ are given. Characterize a continuous-time dynamic quantizer Qd (i.e., find parameters (nQ, Aq, Bq, Cq)) achieving Jhs (Qd) < y.

This paper proposes continuous-time quantizers in terms of solving the problem (E) on the basis of invariant set analysis and the sampled-data control technique, while other researchers [11,12,14] have proposed the discrete-time optimal ones.

Remark 2. The cost function setting of this paper is more complicated than the existing continuous-time and discrete-time cases [11-16], so this paper considers the basic feedforward system composed of P and Qd similar to the previous works on optimal dynamic quantizer design [11,12].

£q := V - u.

Remark 3. The plant P is restricted to be stable because of the feedforward structure, while the existing results can address unstable plants. To remove this restriction, we need to consider a feedback system structure similar to existing ones [13-16]. This is our future task.

3. Main Result

3.1. System Expression. In this subsection, we consider the system expression for the quantizer analysis. Define the quantization error e as

:= q (uQ) -uQ = V -u,

From the properties of the quantizer q and the operator HS,

e(kh + d) = e[k] e

k = 0,1,2,..., 9 e[0,h), (8)

holds where e[k] = e(kh). Then, one obtains

v (kh + d) = Vq [k]+u[k]+e[k],

where vQ[k] = VQ(kh) and u[k] = u(kh) for k = 0,1,2,..., 9 e [0,h). In this case, by using the sampled-data control technique, the following lemma holds.

Lemma 4. Denote by x the state vector of the usual system in Figure 1(b) and define the signals as follows:

IT -T TV = \ v - X XQ]

Zp := z - z .

For the cost function JHS(Qd), the difference between z(kh + 0,xo,Qd(u)) andz*(kh + 9,x0,u)fork = 0,1,2,..., 9 e [0,h) is given by the following system:

Ç[k+1]= A [k]+ Be[k] ¡■h

\ eA{h-r) Bu(kh + T) dr

\ eAQ(h-T)BQu(kh + T) dr

zp (kh + 9) = C (9)Ç [k] + D (9)e [k]

-C\ eA{h-r)Bh(kh + T)dr,

where £[0] = 0, u(kh + t) := u[k] - u(kh + t); the matrices A, B, C(9), and D(9) are defined as follows:

eArdvBCr

0 eAih + \ eÄQTdcBnC,

eAr<dcB

eÄQXdxBr

C(9) :=

D (9) := C I eATdr. Jo

Proof. See Appendix A.

CeAe C\ eArdTBCr

We focus on u(kh + t) of Z. For the operator HS and the signal ue L™,

lim \\[(I-HS)u] (f)\\ =0

holds. This implies that we cannot ignore the temporal resolution constraint on the cost function JHS(Qd) even if h ^ 0. On the other hand, low-pass prefiltering rectifies this situation [25]. In fact, for the stable LTI system P,

lim \\[(I-HS)Fu] (i)\\ = 0,

A F BF

holds. For the evaluation of the cost function JHS(Qd), then this paper utilizes

u = HSFr, re L

as the exogenous input. Note that u e L^ if stable P is strictly proper and r e L^.Forthe signal (15), u[k] = u(kh+9) (k = 0,1,2,..., 9 e [0,h)) holds, so the terms of u(kh + t) in (11) are eliminated. Then, Z is rewritten as

Ç[k+1] = At, [k]+ Be[k], Zp (kh + 9) = C (9)$ [k] + D (9)e [k]

Also, this paper solves the following synthesis problem (E'): for the system Zq composed of P and Qd with the initial state x0 e R" and the exogenous input u e L^ in (15), suppose that the quantization interval d e R+, the switching speed h e R+, and the performance level ye R+ are given. Characterize a continuous-time dynamic quantizer Qd (i.e., find parameters (nQ,Aq,Bq,Cq)) achieving JHS(Qd) < y.

3.2. Quantizer Analysis. The quantization error e of (16) is bounded as mentioned earlier. The reachable set and the invariant set characterize such a system with bounded input. Consider the LTI discrete-time system given by

Ç[k+1] = AÇ[k] + Bw [k],

where Z e and w e Rm denote the state vector and disturbance input, respectively. We define the reachable set and the invariant set.

Definition 5. Define the reachable set of the system (17) to be a set Rm which satisfies

Ç e Rn

3k e N+ 3w [•] e W,

Ç [k] = 'Bw [i]

W := {we Rm :wTw< l}.

Definition 6. Define the invariant set of the system (17) to be a set x which satisfies

Ç e x, we'

+ Bw e x.

The analysis condition can be expressed in terms of matrix inequalities as summarized in the following proposition [22].

Proposition 7. Consider the system (17). For a matrix 0 < P = PT e R"<x"<, the ellipsoid E(P) := tf e R"< : ^%rP% < 1} is an invariant set if and only if there exists a scalar a e [0,1 - p(A)2] satisfying

A 'PA - (1 - a) P

ATPB BTPB - alm

Note that the ellipsoidal set E(P) covers the reachable set RTO from outside. Define the set E := [w e Rm : e = (^md/2)w satisfies (7)} and rewrite the system (16) as

$[k+1]= A [k]+ &w[k],

zp (kh + d) = C (9)$[k] + D (9)w[k],

where £ = (^md/2){„ zp = (^md/2)zp, e = (^md/2)w, and we E. The left multiplication of ZHS with -^md/2 leads to ZHS. The relation E c W clearly holds because eTe < md /4 and the set W is an independent bounded disturbance without the relation (7). That is, the reachable set of ZHS with w e E is no larger than that of ZHS with the disturbance we W.

Then, this paper utilizes the reachable set to estimate the influences of the quantization error and the invariant set to characterize the cost function /HS(Qd) by substituting A = A and B = B into (20). Move on to the matrix exponential e of C(9) and D(9) in (21), which is rewritten as

eA0 = I + Q (9) A, Q (9) := I eArdr.

Along with this, zp of (21) is also rewritten as (kh + 9)

= [C + CQ (9) A CQ (9) BCQ] % [k] + CQ (9) w [k] = (C + CQ (9) <D)Ç[k]+CQ. (9)w[k], C:=[C 0], DD := [A BCQ].

In addition, from the properties of Rœ and E(P),

H (Qd)

< sup sup \\C (9)Ç+ D (9)w\\

< sup sup \\C(6)îî+ D (6)w\

0e[o,h)JgRra>®eW

< sup sup \\C (9)Ç + D (9)w\

0e[°,h)£eE(P),5TeW

(: E c '

< sup sup \\(C + CQ(9)D)Ç\

0e[°>fi);E£E(P)

Rœ c E (P)) Vmd

+ sup sup ||CQ (9) w\

0e[ 0 ,h)ûJeW

holds. Similarly to our previous papers [13, 15,16], by using the L1 control technique in [21], we provide the sufficient conditions for computing y1 e R+ and y2 e R+ of (24) as follows:

P + DT Q(9)T CT

' + cq (9) D

Q(9)TCTCQ (9) < y?In, V9 e [0, h).

Remark 8. For the inequalities (25) and any vectors % e R" and w e Rm, we have

Z^Pl, T (CT + DDTQ(9)TCT) (C + CQ(9) 9)% yi%

^ ^¡^(C + CQ (9) 9)T (C + CQ (9) DD) % < yffP^ (v Schur complement)

and wTQ(9)TCTCQ(9)w < ylwTw. Then, we see that (24) holds if le E(P) and we W.

The inequalities (25) are difficult to test since we need to find P, y1, and y2 satisfying (20) and (25) for infinitely many values of d e [0, h). Then, using the matrix uncertainty technique [23, 24], we consider their sufficient conditions, which are easy to compute. Considering 0.(9) in (22) as a matrix uncertainty, we introduce the following lemma regarding the matrix exponential [26, 27].

where the matrices <&h and Th are defined by

eAr drBCr

1 reArdTA 1 \ hi o hj<

0 h\0 eÄQTdr(A q + BQCQ)

Lemma 9. For the matrix 0.(9) in (22),

amax (0(9))<8(9)<8(h), V9e[0,h), holds where

eKA)9 - !

hj( M'

eAr drB

eAiTdrBr

S (9) :=

Proof. See Appendix A.

... , ^(A)=0, p(A)

9, p (A) = 0,

Denote by Z the system Z without operator HS. In this case, the system Z is given by

^ (A) := max

X : X e eig

(A + At)-

L:Ç(t) = AÇ(t) + Be(t), where

Proof. Since amax(eAd) < e^{A)e (see [26]),

2 ) A := A BCQ , B := B

0 A Q + BqCQ BQ.

Î9 c9

0 ^max {eAß)dß<\o dß

holds.

(29) □

By using Lemma 9 and the S-procedure [23, 28, 29], the sufficient condition analyzing the cost function JHS(Qd) of the system Zq can be expressed in terms of matrix inequality as summarized in the following theorem.

Theorem 10. Consider the system Zq composed of P and Qd with the initial state x0 e R" and the exogenous input u e L^ in (15). For the quantization interval d e R+ and the switching speed h e R+, the upper bound ofthe cost function JHS (Qd) is given by

Jhs (Qd)<Y(h)

s.t.y(h) := (y + amax (C)S(h))

zp (t) = CÇ(t), (34)

Regarding the definition of ^(t), see Lemma 4. An advantage the condition (31) over conditions (20) is that it can be used for a small h without numerical difficulty. This idea comes

from [23, 24]. In the limit of h ^ 0, (1/h) j* eAhdr ^ I

and (l/h)j0h eAvhdT ^ I hold, so ®h ^ A and rh ^ B hold. In the same limit, from 8(h) ^ 0, conditions (31) and (32) converge to the analysis conditions of the continuous-time dynamic quantizer for the system Z in [15,16]. On the otherhand, forasmall h, eAh ^ I, eAih ^ I, j* eAh dr ^ 0

and j0h eAqhdT ^ 0 hold, so A and B (A and B) are close to identity and zero matrices, respectively, and the left side of (20) is close to zero.

In numerical computation, it is appropriate to fix the structure of S such that S0(9) = 0(9)S holds. For example, we can set S = sgIn, sg e R+ and this setting leads to the following optimization problem (Aop):

@=@T>0,S=seI„>0,1/2h-p(A)2/2h>ak>0,y>0

if there exist 0< Q = <f e r("+"q)x("+"q), 0 < S = St e R ah e [0, 1/2h - p(A)2/2h] and ye R+ satisfying

s.t. (31) and (32).

®h@ + ml +2ah Q rh

-2ahIm ^h

9SQ y2Iq - S (h) CSC1

^W)QD 5 0

SQ(9) = Q(9)S, V9e[0,h),

When scalar ah is fixed, the conditions in Theorem 10 are linear matrix inequalities (LMIs) in terms of the other variables. Using standard LMI software and the line search of ah, we can obtain an upper bound of JHS(Qd).

3.3. Quantizer Synthesis. The problem (Aop) suggests that the quantizer synthesis problem (E') is reduced to the following nonconvex optimization problem (OP):

e=@T>0,S=se In>0,AQ,BQ,CQ,1/2h-p(A)2/2h>ak>0,Y>0

s.t. (31) and (32). That is, if (OP) is feasible, (E') is feasible.

From the matrix product such as (1/h) JQ eAc>hdr and Aq+BqCq in (31), the synthesis condition is difficult to derive from Theorem 10 unlike the continuous-time case without the operator HS in [15,16]. Thus, we fixed the parameters as follows:

nn = n,

An = A

Bn = B.

The structure (38) does not severely limit the synthesis because Aq and Bq of the continuous-time dynamic quantizer for the system Z in [15,16] are also (38). See Appendix B. In other words, Xq = Axq + BeQ estimates the quantization influence on the system P. Along with this, we fix Q of (31) as follows:

Y = Y' >0, V = V' >0.

The structure (39) also does not impose a severe limitation on the synthesis because an appropriate choice of the quantizer state coordinates allows us to assume that Q has the special structure for the full order case Hq = n [30].

Under some circumstances (38) and (39), we obtain the following synthesis condition.

Theorem 11. Consider the system Zq composed of P and Qd with the initial state x0 e Rn and the exogenous input u e L™ in (15). Suppose that the quantization interval d e R+, the switching speed h e R+, and the performance level ye R+ are given. For a scalar ah e [0, l/2h], there exist a continuous-time dynamic quantizer Qd achieving (30) if one of the following equivalent statements holds.

(i) There exist matrices 0 < Q = QT e r("+"q)x("+"q), o < S = ST e Rnxn and a dynamic quantizer Qd satisfying (31), (32), and (38).

(ii) There exist matrices 0 < Y = YT e Rnxn, 0 < V = VT e Rnxn, W e Rmxn, 0 < S = ST e Rnxn

satisfying

& Ah + &Ah + 2«h&P &Bh Vh&Ah '

&Th -2ah!m Vh&Th < 0, (40)

Vh&Ah Vh&Bh -&P .

&c y2Iq -8(h) CSC1 0

Sn(9) = n(9)S, y9e[0,h),

where &P :=

&Ah := &Bh :=

(AY + BW) (AV + BW) Wh (AV + BW) Wh (AV + BW)

&c := [CY CV]

&Dh := [^8(hj(AY + BW) ^8(h)(AV + BW)]. In this case, such a quantizer parameter is given by

An = A, Bn = B, Cn = WY-1. (43)

nn = n,

Proof. We fix Q as shown in (39) and introduce the change of variables W = CqY. Hence, (31) and (32) result in (40) and (41). Also, designing Cq yields «h e [0,l/2h] because p(A) is determined by CQ and [0, l/2h- p(A)2/2h] c [0, l/2h]. □

In the limit of h ^ 0; Wh converges to I and 8(h) converges to 0; then conditions (40) and (41) also converge to the synthesis condition of the continuous-time dynamic quantizer for the system Z in (34). Also, by setting S = sgIn for Theorem 11, the quantizer synthesis problem (E') is reduced to the following optimization problem (Sop):

y=y>QT,V=VT>Q,S=sei„ >Q,W,1/2h>afc >0,y>0

s.t. (40) and (41).

If (Sop) is feasible, (E') is feasible. Therefore, a continuous-time dynamic quantizer considering both spatial and temporal resolution constraints is obtained from Theorem 11.

Remark 12. To consider numerical optimization analysis or synthesis of a quantizer as shown in (Aop) and (Sop), we need the signal assumption (15) in Theorems 10 and 11. On the other hand, for the high speed switching such that h is very small, the assumption (15) ensures that solutions to the problem (E') converge to our previous results [15, 16]. Therefore, the results of this paper partly include our previous results [15,16] although each class of exogenous signals and plants is restricted.

4. Discussion

For the slow switching, we compare the proposed method and existing continuous-time quantizer [15,16]. Consider the system Zq. The plant P is the stable minimum phase LTI system:

x(t) z(t)

-3 3 0 0-2 2 1 1 0

x(t) v (t)

In the case without the operator HS, an optimal form of the continuous-time quantizer Q? [15,16] is given by (B.2).

Time t(s)

Time t(s)

(a) (b)

Figure 4: Time responses of Zq with the proposed quantizer for h = 0.01.

Time t(s)

234 Time t(s)

(a) (b)

Figure 5: Time responses of Zq with (B.2) for h = 0.01.

See Appendix B. The continuous-time quantizer Q°f and its performance are parameterized by the free parameter f e R+. For the simulation, we consider a two-step design for Q°dp; we first set f and second insert the operator HS in the obtained Q°f. Also, the achievable performance of JHS(Qf) is calculated by (Aop).

For the comparison, we set the switching speed h = 0.01 [s] and the quantization interval d = 2. First, we set f = 50 and then obtain Q°f with yc = 0.102. Also, y(h) = 0.219 for JHS(Q°f) is obtained from (Aop). Second, we solve the problem (Sop) and obtain y(h) = 0.219 and the matrix Cq = [-19.26 -22.72]. In this case, both quantizers can approximate well. Figures 4 and 5 illustrate the simulation results of the time responses of Zq with the proposed quantizer and the quantizers Q°f in (B.2). The initial state x0 = [0 0]T and the input u(t) = sin nt + cos 0.7nt are given. In Figures 4 and 5, the thin lines and the thick lines are for the conventional system in Figure 1(b) and the system Zq in Figure 1(a), respectively. We see that the controlled outputs of the discrete-valued input systems with the dynamic quantizers approximate those of the usual systems even if the quantized outputs are applied. Also, the two controlled outputs approximated by both quantizers are exactly the same.

Next, we consider the case h = 0.1. In this case, the two controlled outputs approximated by the two quantizers differ. (Aop) for Q°f is infeasible. From (Sop), on the other hand, we obtain y(h) = 0.949 and the matrix CQ = [-1.1321 -3.097]. Figures 6 and 7 illustrate the simulation results on the time responses of Zq with (B.2) and the proposed quantizer in the same fashion. We see that z(t) of the usual plant P is

approximated by z(t) of the system Zq with the proposed quantizer, while z(t) of the system Zq with (B.2) diverges. From this example, we see that the proposed method can address the spatial resolution and the temporal resolution issues, simultaneously. Also, Theorem 10 verifies whether the quantizer Q°f is applicable to the given switching speed setting.

Remark 13. In the above numerical experiments, the proposed quantizer is designed and the quantizer Q°f is analyzed for u = HSF(sin nt + cos 0.7nt), while the time responses of the quantizers are simulated for u = sin nt + cos 0.7nt. That is, this is the conservativeness caused by the signal assumption (15). However, we see that the above results verify the effectiveness of the proposed method even if the signal conservativeness exists.

Here, we focus on the eigenvalues of A for the system Z with Q°f. The eigenvalues for f = 50 and h = 0.1 are {0.741,0.819, 0.550, -4.21} and then A is unstable in the discrete-time domain. From Theorem 10, (Aop) is infeasible if p(A) is bigger than 1 (in other words, A is unstable). {0.741,0.819} are the eigenvalues of eAh. That is, eAch + j0h eAvrdTBQCQ(=:. AQ(h)) for Q°dp is unstable. Then, we consider the case in which f = 3 and h = 0.1 (yc = 0.707) such that Aq(K) is stable. The corresponding eigenvalues are {0.637, 0.354}. In this case, y(h) = 1.085 for JHS(Q°f) is obtained from (Aop).

From the above results, the existing continuous-quantizer in [15, 16] may be suitable for a two-step design such that AQ(h) is stable via the parameter f. In terms of the upper

2 1 0 -1 -2

2 1 0 -1 -2

2 1 0 -1 -2

X \

< ! \

j j

Time t(s)

Time t(s) (b)

Figure 6: Time responses of Zq with (B.2) for h = 0.1.

2 1 0 -1 -2

234 Time t(s)

234 Time t(s)

(a) (b)

Figure 7: Time responses of Zq with the proposed quantizer for h = 0.1.

bound of cost function (B.3), first, let us consider the problem (P-1): maximize f for Q°f such that

3XT = X>0 s.t. AQ(h)TXAQ (h)-X< 0, (46)

where the parameters (Aq, Bq,Cq) of Aq(h) are given by (B.2). Thisproblem is LMIfor thelinesearchof /.For h= 0.1, its solution is f = 18.9 (yc = 0.173). However, y(h) = 113.148 for JHS(Q°J) is obtained from (Aop). By using Theorem 10, next, let us consider the problem (P-2):

e=eT>0,S=seIn>0,1/2h-(p(A)2)/2h>ak>0,f>0,y>0 (47)

s.t. (31) and (32),

where the parameters (Aq, Bq, Cq) of Wh and ^ are given by (B.2). This problem is LMI for the plane search of f and ah. For h = 0.1, its solution is f = 5.715 (yc = 0.397) and then y(h) = 0.949 for JHS(Q°dp) is obtained. This performance is about the same as that of the proposed quantizer. Therefore, we see that Theorem 10 is also helpful for the two-step design of the existing continuous-time quantizer [15,16] even if the tractable optimization method instead of the plane search remains an issue for future work. Such a method correlates the parameter f with the switching speed h, so its insight is expected not only to result in a new two-step design but also to clarify the relationship between the discrete-time and continuous-time dynamic quantizers. Of course, important future topics also include considering the quantized feedback control system with unstable plants and generalizing the exogenous signal for the evaluation of the cost function.

5. Conclusion

Focusing on the broadbandization and the robustness of the networked control systems, this paper has dealt with the continuous-time quantized control. We have proposed numerical optimization methods analyzing and synthesizing the continuous-time dynamic quantizer on the basis of the invariant set analysis and the sampled-data control technique. The contributions of the proposed method can be summarized as follows.

(i) Both the temporal and spatial resolution constraints can be simultaneously considered, whereas Ishikawa et al. [20] considered the two constraints step-by-step and we [15,16] previously ignored the temporal constraint in synthesis. As a result, the proposed method is applicable to both the slow and fast switching cases.

(ii) The maximum output difference for each sampling interval is proven to be evaluated numerically via the matrix uncertainty approach, while the existing results [11-14] evaluate that only for each sampling instance.

(iii) The analysis and synthesis conditions are given in terms of BMIs. However, the quantizer analysis and synthesis problems are reduced to tractable optimization problems.

(iv) The new insight is presented for the existing continuous-time quantizer design [15, 16].

Also, this paper has clarified the following areas for future work.

(i) Because of the feedforward structure, the plant is restricted tobe stable. Toaddressunstablesystems, we

need to propose design methods for feedback control systems.

(ii) The sensors and actuators are distributed in the networked control system [31], so it is necessary to design multiple (decentralized) quantizers rather than a centralized quantizer similar to the existing ones [14,16].

(iii) The class of exogenous signals evaluating the cost function is restricted. To avoid this conservativeness, it is necessary to propose the equivalent discrete-time expression instead of (11) using adjoint operator similar to sampled-data control [25, 32].

(iv) For networked control applications, it is important to consider the time-varying sampling period, time delay, packet loss, and so on similar to [6-10]. For example, the works [6, 8] using the LMI technique address time-varying sampling period and time delay, so it is expected that our method using the LMI technique also extends to such problems.

Then, z(kh + d,x0,Qd(u)) for k = 0,1,2,..., 9 e [0,h) is expressed by the discretized system of Zq:

x[k+l] xQ [k+ 1]

= A | x[^] + Be[k]

Xq [k] J h

\ eA(h-r)Bu[k] dr

eAi(h-T) BQu(kh + T) dr z (kh + 9) = C (d)\X[1!i]+ D (9)e[k]

[xq [k]

+ c\ eA(h-T)Bu[k] dr 0

Also, z*(kh + 9,x0,u) for k = 0,1,2,..., 9 e [0,h) without the quantizer is given by

Appendix A. Proof

The proof of Lemma 4 is as follows.

x[k+1]

= eAhx [k] + \ eA(h-T)Bu (kh + r) dr

z* (kh + 9)

= CeAd x[k]+c\ eA(h-T) u(kh + r) dr,

Proof. z(kh + 9) for k = 0,1,2,..., 9 e [0,h) (which is the behavior of z(t) over the kth sampling interval) is given by the discretized system of P:

x[k+1]= eAhx [k] + \ eA(h-T)fiv (kh + r) dr

z (kh + 9) = CeAe x [k]+c\ eA(0-T) Bv (kh + r) dr,

where x[k] = x(kh). v(t) is given by (9) and Vq[&] is given by the discretized system of Q:

Xq [k+ 1]

.Anh . ri.i . „Anr,

= (eA+ \0 ^^drfiQCQlxQ ^

+ \ eA^{h-r) BQh(kh + r) dr,

[k] + \ eAQtdrBQe [k]

Vq [k] = CqXq [k],

where XQ[k] = XQ(kh). This is because eQ is given by eQ = ^q + e + u - u, Vq = hsvq, u = HSu. (A.3)

where x(0) = x0 and x[k] = x(kh). From Zq and P^, we obtain the system Z in (11). □

For the proof of Theorem 10, we use the S-procedure [23, 28, 29].

Lemma A.1. For the real matrices L, M, and N of appropriate size, the inequality

L + MAN + NT ATMT > 0

holds for any matrix A such that amax(A) < S if and only if there exists a matrix S = S > 0 such that

L-SMSMT ^8NT ^SN S

>0, SA = AS.

Then, the proof of Theorem 10 is as follows.

Proof. We use Lemmas 9 and A.1 with

M = A = £1(9).

P 9§T

n=\d o

In this case, for the inequalities (25), we obtain their sufficient conditions as follows:

y1 < y s.t. -

y2Iq - 8 (h) CSC

■f8(hjD 0

sa (6) = a (0) s,

^sJhD7 0 s

ve e [0,h)

72 <^max (c)S(h) s.t. a(ef cTca (9)

<amax(C)2S(h)2In, V9e[0,h).

Then, the upper bound of JHS(Qd) is given by (30). By substituting A = I + h<&h and B = h^ into (20), we obtain

A1 PA -(1-a) P BTPA

TPB - alm

(I + h®h)TP (I + h®h) -(1-a) P h(l + h®h)TPrh

htfP (I + h®h)

®lP + P®h + a/hP Prh rlP -a/h!n

h^P^ -alm

P [Vk®h Vhrh]<0.

(A.10)

By Schur complement [29], (A.10) is equivalent to (31) where Q = P-1 and ah = a/2h. Also, (A.9) is equivalent to (32) where Q = P-1. □

B. Continuous-Time Dynamic Quantizer [15,16]

For the system Z in (34) without the operator HS, we consider the following non-convex optimization (OP'):

_ min yc

P>0,p(s/)>a>0,A q,BQ,CQ,Yc>0

~f p+ps$+2(xp pS&

-2<xlm 5T-

The dynamic quantizer without the operator HS is obtained from (OP') and J(Qf) < yc is achieved. Also, an optimal form of the continuous-time quantizer [15,16] is given by

xQ = Axq + B(v -u) v = q(-(CB)-1C(A +fl)xQ + u),

where its achievable upperbound of J(Q°f) is characterized by

inf Yc =

-°max (CB) ,

(f-\0\) (B.3)

g = max {] (A), v (A - B(CB)-1C (A + fl))},

where v(A) := max|Re(A) : X e eig(A)}. The continuous-time quantizer Q°dP and its performance are parameterized by the free parameter f e R+. Note that the larger values of f not only provide the better approximation performance, but also switch the outputs v, more quickly. In other words, the quantizer from (OP ) results in the switching that is too fast and is sometimes not applicable to the slow switching case.

Acknowledgments

The authors would like to thank the reviewers for their valuable comments. This work was partly supported by Grant-in-Aid for Young Scientists (B) no. 24760332 from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

References

[1] W. S. Wong and R. W. Brockett, "Systems with finite communication bandwidth constraints. II: stabilization with limited information feedback," IEEE Transactions on Automatic Control, vol. 44, no. 5, pp. 1049-1053,1999.

[2] N. Elia and S. K. Mitter, "Stabilization of linear systems with limited information," IEEE Transactions on Automatic Control, vol. 46, no. 9, pp. 1384-1400, 2001.

[3] M. Fu and L. Xie, "The sector bound approach to quantized feedback control," IEEE Transactions on Automatic Control, vol. 50, no. 11, pp. 1698-1711, 2005.

[4] J. Baillieul, "Special issue on networked control systems," IEEE Transactions on Automatic Control, vol. 49, no. 9, pp. 1421-1423, 2004.

[5] P. Antsaklis and J. Baillieul, "Special issue on technology of networked control systems," Proceedings of the IEEE, vol. 95, no. 1, pp. 5-8, 2007.

[6] D. Yue, E. Tian, Z. Wang, and J. Lam, "Stabilization of systems with probabilistic interval input delays and its applications to networked control systems," IEEE Transactions on Systems, Man, and Cybernetics A, vol. 39, no. 4, pp. 939-945, 2009.

[7] W. P. M. H. Heemels, A. R. Teel, N. van de Wouw, and D. Nesic, "Networked control systems with communication constraints: tradeoffs between transmission intervals, delays and performance," IEEE Transactions on Automatic Control, vol. 55, no. 8, pp. 1781-1796, 2010.

[8] B. Shen, Z. Wang, and X. Liu, "Sampled-data synchronization control of complex dynamical networks with stochastic sampling," IEEE Transactions on Automatic Control, vol. 57, no. 10, pp. 2644-2650, 2012.

[9] W. P. M. H. Heemels, M. C. F. Donkers, and A. R. Teel, "Periodic event-triggered control for linear systems," IEEE Transactions on Automatic Control, vol. 58, no. 4, pp. 847-861, 2013.

[10] K. Okano and H. Ishii, "Networked control of uncertain systems over data rate limited and lossy channels," IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, vol. E96-A, no. 5, pp. 853-860, 2013.

[11] S. Azuma and T. Sugie, "Optimal dynamic quantizers for discrete-valued input control," Automatica, vol. 44, no. 2, pp. 396-406, 2008.

[12] S. Azuma and T. Sugie, "Synthesis of optimal dynamic quantizers for discrete-valued input control," IEEE Transactions on Automatic Control, vol. 53, no. 9, pp. 2064-2075, 2008.

[13] K. Sawada and S. Shin, "Synthesis of dynamic quantizers for quantized feedback systems within invariant set analysis framework," in Proceedings of the American Control Conference (ACC '11), pp. 1662-1667, July 2011.

[14] Y. Minami, S. Azuma, and T. Sugie, "Optimal decentralized sigma-delta modulators for quantized feedback control," IEICE Nonlinear Theory and Its Applications, vol. 3, no. 3, pp. 386-404, 2012.

[15] K. Sawada and S. Shin, "Synthesis of continuous-time dynamic quantizer for quantized feedback systems," in Proceedings of the 4th IFAC Conference on Analysis and Design of Hybrid Systems (ADHS '12), pp. 2488-2493, 2012.

[16] K. Sawada and S. Shin, "On numerical optimization design of continuous-time feedbacktype quantizer for networked control systems," in Proceedings of the 8th IEEE International Conference on Automation Science and Engineering (CASE '12), pp. 11401145, 2012.

[17] K. Reddy and S. Pavan, "Fundamental limitations of continuous-time delta-sigma modulators due to clock jitter," IEEE Transactions on Circuits and Systems I, vol. 54, no. 10, pp. 2184-2194, 2007.

[18] K. Matsukawa, Y. Mitani, M. Takayama, K. Obata, S. Dosho, and A. Matsuzawa, "A fifth-order continuous-time delta-sigma modulator with single-opamp resonator," IEEE Journal ofSolid-State Circuits, vol. 45, no. 4, pp. 697-706, 2010.

[19] M. Koike and Y. Chida, "Vibration control design considering quantization error for an on-off control system including input time-delay," in Proceedings of the 50th Annual Conference on Society ofInstrument and Control Engineers (SICE '11), pp. 760765, September 2011.

[20] M. Ishikawa, I. Maruta, and T. Sugie, "Practical controller design for discrete-valued input systems using feedback modulators," in Proceedings of the European Control Conference (ECC '07), Kos, Greece, July 2007.

[21] R. Sanchez-Pena and M. Sznaier, Robust Systems Theory and Applications, John Wiley & Sons, 1998.

[22] H. Shingin and Y. Ohta, "Optimal invariant sets for discrete-time systems: approximation of reachable sets for bounded inputs," in Proceedings of the 10th IFAC/IFORS/IMACS/IFIP Symposium on Large Scale Systems, Theory andApplications (LSS '04), pp. 401-406, 2004.

[23] Y. Oishi and H. Fujioka, "Stability analysis of aperiodic sampleddata control systems: an improved approach using matrix uncertainty," in Proceedings of the 19th International Symposium on Mathematical Theory of Networks and Systems (MTNS '10), pp. 1147-1152, July 2010.

[24] Y. Oishi and H. Fujioka, "Stability and stabilization of aperiodic sampled-data control systems using robust linear matrix inequalities," Automatica, vol. 46, no. 8, pp. 1327-1333, 2010.

[25] T. Chen and B. Francis, Optimal Sampled-Data Control Systems, Springer, 1995.

[26] G. H. Golub and C. F. van Loan, Matrix Computations, Johns Hopkins Studies in Mathematical Sciences, Johns Hopkins University Press, 3rd edition, 1996.

[27] H. Fujioka, "Command shaping for sampled-data servo systems with constraints," Transactions of the Institute of Systems, Control and Information Engineers, vol. 19, no. 7, pp. 284-289, 2006 (Japanese).

[28] M. Fu and N. E. Barabanov, "Improved upper bounds for the mixed structured singular value," IEEE Transactions on Automatic Control, vol. 42, no. 10, pp. 1447-1452,1997.

[29] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory, vol. 15 of Studies in Applied Mathematics, SIAM, 1994.

[30] I. Masubuchi, A. Ohara, and N. Suda, "LMI-based controller synthesis: a unified formulation and solution," International Journal of Robust and Nonlinear Control, vol. 8, no. 8, pp. 669686,1998.

[31] J. A. Stankovic, "When sensor and actuator networks cover the world," ETRI Journal, vol. 30, no. 5, pp. 627-633, 2008.

[32] Y. Hayakawa, S. Hara, and Y. Yamamoto, type problem for sampled-data control systems—a solution via minimum energy characterization," IEEE Transactions on Automatic Control, vol. 39, no. 11, pp. 2278-2284, 1994.

Copyright of Mathematical Problems in Engineering is the property of Hindawi Publishing Corporation and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.