Appendix A

Linear Matrix Inequality and

Preliminary Lemmas

The goal of this appendix is to recall some important notions on linear

matrix inequality (LMI) in order to render our book self-contained. Mainly,
we define the LMI problem and the related problems like the feasibility
problem (FEAS), minimization of a linear objective under LMI constraints
(MINCX), and the generalized eigenvalue minimization problem (GEVP).

A.I LMI Functions

A linear matrix inequality (LMI) has the form
m

F(z) ~ Fo + LZiFi < 0, (A.I)

i=l

where Z = (Zll"" zm) E IR m is the variable to be determined and the

symmetric matrices Fi E IR nxn , 0 :=:; i :=:; m are given. The inequality symbol
in (A.I) means that F(x) is negative-definite, i.e., v T F(x)v < 0 for all
nonzero v E IR n .
For example, a linear system with the following dynamic
x(t) = Ax(t) (A.2)
where x(t) E 1R2 , A E 1R2x2 , is stable if and only if there exists a symmetric
and positive-definite matrix P > 0 such that
(A.3)
378 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

This problem, in fact, can be solved using the LMI toolbox. Let us now
see how we can put this problem in the form of (A.I). For this purpose, let
A= (a1
~
a2
~
) and suppose that P = (Zl
~
Z2),
~
where ZI, Z2, and Z3

are design parameters. Then,

Therefore, (A.3) is a standard LMI feasibility problem that can be solved

using the LMI toolbox.
There are three kinds of generic LMI problems often referred to in this
book. Related to these generic LMI problems, several functions are in-
tensively used: lmivar, lmiterm, setlmis, getlmis, feasp, mincx, gevp,
mat2dec. Before giving numerical examples, let us give some information
on these functions. The material below is based on the LMI Matlab toolbox
help, from which the descriptions of these functions are reproduced.

x = lmivar(type,struct)
[X,ndec,Xdec] = lmivar(type,struct)

Adds a new matrix variable X to the LMI system currently

described. A label X can be optionally attached to this
new variable to facilitate future reference to it.

Input:
TYPE structure of X:
1 -> symmetric block diagonal
2 -> full rectangular
3 -> other
STRUCT additional data on the structure of X
TYPE=1: the i-th row of STRUCT describes the
i-th diagonal block of X
STRUCT(i,1) -> block size
STRUCT(i,2) -> block type, i.e.,
o for scalar blocks t*I
1 for full block
-1 for zero block
TYPE=2: STRUCT = [M,N] if X is a MxN matrix
TYPE=3: STRUCT is a matrix of same dimension as X
where STRUCT(i,j) is either
o i f X(i,j) = 0
+n if X(i,j) = n-th decision variable
 A.I. LMI Functions 379

-n if X(i,j) = (-1)* n-th decision var

Output:
X Optional: identifier for the new matrix variable.
Its value is k if k-1 matrix variables have already
been declared. This identifier is not affected by
subsequent modifications of the LMI system
NDEC total number of decision variables so far
XDEC entry-wise dependence of X on the decision variables
(Xdec = struct for Type 3)

lmiterm(termID,A,B,flag)

Adds one term to some LMI in the LMI system currently

described. A term is either an outer factor, a constant
matrix, or a variable term A*X*B or A*X'*B where X is
a matrix variable.

IMPORTANT: Because the OFF-DIAGONAL blocks (i,j) and (j,i)

are transposed of one another, specify the term
content of ONLY ONE of these two blocks.

Input:
TERMID 4-entry vector specifying the term location and
nature
Which LMI?
TERMID(l) = +n -> left-hand side of the n-th LMI
TERMID(l) = -n -> right-hand side of the n-th LMI
Which block?
For outer factors, set TERMID(2:3) = [0 0] .
Otherwise, set TERMID(2:3) = [i j] if the term
belongs to the (i,j) block of the LMI
What type of term?
TERMID(4) 0 -> constant term
TERMID(4) X -> variable term A*X*B
TERMID(4) -X -> variable term A*X'*B
where X is the variable identifier returned by
LMIVAR
A value of the outer factor, constant term, or
left coefficient in variable terms A*X*B or
A*X'*B
B right coefficient in variable terms A*X*B or
A*X'*B
FLAG quick way of specifying the expression
A*X*B+B'*X'*A' in a DIAGONAL block. Set
FLAG='s' to specify it with only one LMITERM
380 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

command

setlmis(lmisysO)

Initializes the description of a new LMI system.

To start from scratch, type

setlmis ( [] )

To add on to an existing LMI system LMISYSO, type

setlmis(lmisysO)

See also LMIVAR, LMITERM, GETLMIS.

lmisys = getlmis

Returns the internal representation LMISYS of an LMI

system once this LMI system has been fully described
with LMIVAR and LMITERM.

The internal representation LMISYS can be passed

directly to the LMI solvers or any other LMI-Lab
function.

See also SETLMIS, LMIVAR, LMITERM.

[tmin,xfeas] = feasp(lmis,options,target)
Solves the feasibility problem defined by the system LMIS
of LMI constraints. When the problem is feasible, the
output XFEAS is a feasible value of the vector of (scalar)
decision variables.

Given a feasibility problem of the form L(x) < R(x),

FEASP solves the auxiliary convex program:

Minimize t subject to L(x) < R(x) + t*I

The system of LMIs is feasible iff. the global minimum TMIN

is negative. The current best value of t is displayed by
FEASP at each iteration.

Input:
 A.I. LMI Functions 381

LMIS array describing the system of LMI constraints

OPTIONS optional: five-entry vector of control parameters.
Default values are selected by setting OPTIONS(i)=O.
OPTIONS(l): not used
OPTIONS(2): max. number of iterations (Default=100)
OPTIONS(3): feasibility radius R. R>O constrains
x to x'*x < R 2 A

(Default=le9).
R<O means "no bound"
OPTIONS(4): when set to an integer value L > 1,
forces termination when t has not
decreased by more than l%over the last
L iterations (Default = 10).
OPTIONS(5): when nonzero, the trace of execution is
turned off.
TARGET optional: target for TMIN. The code terminates as
soon as
t < TARGET (Default=O)
Output:
TMIN value of t upon termination. The LMI system is
feasible iff. TMIN <= 0
XFEAS corresponding minimizer. If TMIN <= 0, XFEAS is
a feasible vector for the set of LMI constraints.
Use DEC2MAT to get the matrix variable values
from XFEAS.

See also MINCX, GEVP, DEC2MAT.

[copt,xopt] = mincx(lmis,c,options,xinit,target)

Solves the LMI problem:

Minimize c'*x subject to L (x) < R(x)

where x is the vector of DECISION VARIABLES.

Input:
LMIS description of the system of LMI constraints
C vector of the same size as x (see DECNBR).
Use DEFCX to specify the objective c'*x
directly in terms of matrix variables
OPTIONS optional: five-entry vector of control
parameters. The default value is used when
OPTIONS(i)=O .
OPTIONS(l): relative accuracy required on the
optimal value of the objective
382 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

(Default = 0.01)
OPTIONS(2): max. number of iterations (Default=100)
OPTIONS(3): feasibility radius R. R>O constrains x
to x'*x < R~2. R<O means "no bound"
(Default=le9)
OPTIONS(4): integer value L. The code
terminates when the objective value
has decreased by less than
OPTIONS(l) during the last
L iterations (Default=10)
OPTIONS(5): when nonzero, the trace of execution is
turned off (Default=O)
XINIT optional: initial guess for X ([] if none,
ignored when infeasible)
TARGET optional: target for the objective value.
The code terminates as soon as a feasible x is
found such that
c'*x < TARGET (Default=-le20)
Output:
COPT global minimum of the objective c'*x
XOPT minimizing value of the vector x of decision
variables. Use DEC2MAT to get the corresponding
matrix variable values.

See also FEASP, GEVP, DEFCX, DEC2MAT.

[tmin,xopt] = gevp(lmis,nlfc,options,tO,xO,target)

Solves the generalized eigenvalue minimization problem

Minimize t

subject to the LMI constraints:

C(x) < 0

o < Bj(x) ( j=l, .. ,NLFC )

Aj(x) < t * Bj(x) ( j=l, .. ,NLFC )

Here x denotes the vector of (scalar) decision variables.

The positivity constraints Bj(x) > 0 must be specified
for well-posedness, and the LMIs involving t should be
specified last.
 A.I. LMI Functions 383

Input:
LMIS description of the system of LMI constraints
NLFC number of linear fractional constraints (LMIs
involving t)
OPTIONS optional: five-entry vector of control parameters.
The default value is used when OPTIONS(i)=O
OPTIONS(l): relative accuracy required on TMIN
(Default = 1.Oe-2)
OPTIONS(2): max. number of iterations (Default=lOO)
OPTIONS(3): feasibility radius R. R>O constrains
x to x'*x < R-2 (Default=le8).
R<O means "no bound"
OPTIONS(4): integer value L. The code terminates
when t has decreased by less than
OPTIONS(l) during the last L iterations
(Def ault = 5)
OPTIONS(5): when nonzero, the trace of execution is
turned off.
TO,XO optional: initial guesses for t,x (ignored when
unfeasible)
TARGET optional: target for TMIN. The code terminates as
soon as t falls below this value (DEFAULT = -le5)

Output:
TMIN minimal value of t
XOPT minimizing value of the vector x of decision
variables Use DEC2MAT to get the corresponding
matrix variable values.

See also FEASP, MINCX, DEC2MAT.

decvars = mat2dec(lmisys,Xl,X2,X3, ... )

For given values Xl, X2, X3, ... of the matrix variables
involved in the LMI system LMISYS, MAT2DEC computes the
value DECVARS of the vector of decision variables. This
operation is the converse of that performed by DEC2MAT

Input:
LMISYS array describing the LMI system
Xl, X2, X3, ...
values of the matrix variables. MAT2DEC
accepts up to 20 values. An error is issued
384 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

if some matrix variable remains unassigned.

Output:
DECVARS vector of the decision variable values.

See also DEC2MAT, DECINFO, DEFCX.

A.2 LMI Problems

As we mentioned earlier, there exist three main problems that we have used
extensively in this book. These problems are:
• the feasibility problem
• the linear optimization problem
• the generalized eigenvalue minimization problem.

A.2.1 Feasibility Problem

The LMI Feasibility Problem (FEASP) consists of determining the variable
x E IRm such that F(x) < 0 holds. This problem can be solved using the
function "feasp" of the LMI toolbox. In the rest of this subsection we will
see how to state the problem and how to solve using "feasp" function.
A typical situation for the feasibility problem is the stability test for
dynamical systems. In fact, based on control theory a system with the
dynamic (A.2) is stable if and only if there exists a symmetric and positive-
definite matrix P > 0 such that (A.3) is satisfied.
The goal in this case is to find a matrix P > 0 such that the inequality
(A.3) is satisfied. The "feasp" can be used to solve our problem. The next
example demonstrates the use of such a function.
Example A.I Consider a system described by (A.2) and suppose that the
matrix A is given by A = ( ~
2
!
1 ). Let us now see how we can solve
the stability problem. Based on what we described previously, we can write
a code in LM! Matlab toolbox. To solve our stability problem, type the
following code and save it as, for example, appl.m.

clear all

A = [-2 1; 1 -1];

setlmis ( [] )
 A.2. LMI Problems 385

p = lmivar(l, [2, 1]);

lmiterm([l 1 1 P] ,A,l, 's')
lmiterm([-2 1 1 P] ,1,1)

LMI = getlmis;
[copt xopt] = feasp(LMI);
P = dec2mat(LMI, xopt, P)
Then in Matlab environment, running appl.m yields
P =
0.6587 0.5646
0.5646 1.2233
Therefore, the system under study is stable.
Next, let us consider the stabilization problem of the system
x(t) = Ax(t) + Bu(t) (A.4)
where u(t) is the control input. Using the stability result, a state feedback
memoryless controller u(t) = Kx(t) stabilizes system (A.4) if and only if
there exists a symmetric and positive matrix P such that
P(A + BK) + (A + BK)T P < O. (A.5)
Pre- and post-multiplying both sides of (A.5) by X = p- 1 and letting
Y = KX yields that (A.5) is equivalent to
AX+BY+XA T +yTB T <0. (A.6)
This problem represents another example of a feasible LMI problem. To
see how this result can be used, let us consider the following numerical
example.
Example A.2 Suppose that the system parameters are as follows: A =
(~ ~), B = ( ; ). Obviously, this free system (with u(t) == 0) is
unstable. To compute a stabilizing gain, let us type the following code:
clear all
A = [2 1; 1 1];
B = [1; 2];

setlmis ( [] )
X = lmivar (1, [2, 1]);
Y = lmivar(2, [1,2]);

lmiterm([l 1 1 X],A,l,'s')
lmiterm([l 1 1 YJ ,B,l,'s')
386 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

Imiterm([-2 1 1 X] ,1,1)

LMI = getlmis;
[copt xopt] = feasp(LMI);
X = dec2mat(LMI, xopt, X)
Y = dec2mat(LMI, xopt, Y)
K = Y*inv(X)
Running the above code yields
X=
86.3203 53.3647
53.3647 833.4266

Y=
-320.2027 -521.1294

K=
-3.4599 -0.4037
Therefore, a stabilizing controller is obtained.

A.2.2 Minimization of a Linear Objective under LM!

Constraints
Minimization of a linear objective under LMI constraints (MINCX) is
another interesting problem that we have used extensively. The MINCX
problem is stated as follows:
min CT x
xEIR'"
s.t. F(x) < 0
where C E lRm is a given vector. This problem can be solved using "mincx"
of LMI Toolbox.
To provide an example using "mincx", let us consider an 1-1.2 control
problem. For this purpose let us assume the system dynamics are given by
x(t) = Ax(t) + Bw(t)
{ yet) = Cx(t) (A.7)

where wet) is a white noise disturbance with unit covariance. Suppose that
the 1£2 performance is defined by

111-I."~ = lim lE (~t Jrt y

T (S)Y(S)dS).
o
t--+<Xl

Then, it can be shown that the solution to this problem is given by (see
for example [38]):
111-I."~ = min{tr(CPC T ) : AP + PAT + BB T < O}.
 A.2. LMI Problems 387

Obviously, this optimization problem is equivalent to minimizing tr(Q)

subject to
AP + PAT + BB T < 0 (A.8)
CPC T ::; Q. (A.9)
Using the Schur complement, (A.9) is equivalent to

-Q CP)
( PCT _p < O.
Example A.3 Let us now suppose that the system parameters are

A= (-21 -11) B= ( 1)
2 C (2-1 01)
=

and see how we can use the mincx function to minimize our 1-£2
performance. Let us type the following Matlab code:
%% apped3
clear all

A = [-2 1; 1 -1];
B= [1 ;2] ;
C=[21;-10];

setlmis ( [] )

Q = lmivar(l, [2, 1]);

P = lmivar(l, [2, 1]);
lmiterm([l 1 1 P] ,A, 1, 's');
lmiterm([l 1 1 0] ,B*B');

lmiterm([-2 1 1 Q] ,1,1)
lmiterm([-2 1 2 P],C,l)
lmiterm([-22 2 P],l,l)

lmiterm([-31 1 P],l,l);
lmiterm([-41 1 Q],l,l);

c=[l 0 1 0 0 0];

LMI = getlmis;
[copt xopt] = mincx(LMI,c);

Q = dec2mat(LMI, xopt, Q)
P = dec2mat(LMI, xopt,P)
optrace = copt
388 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

Running the above code yields the following solution:

Q
22.8395 -6.1685
-6.1685 1.6728
P =
1.6684 2.8322
2.8322 4.8348

optrace =
24.5122

A.2.3 Generalized Eigenvalue Minimization Problem

The generalized eigenvalue minimization problem (GEVP) is the third in-
teresting problem used extensively in this book. This problem is stated as
follows:

s.t. F 1 (x) < .\F2 (x)

where F 1 (x), F 2 (x) are two matrices of form (A.I). The GEVP is quasi-
convex with respect to the design parameters x and A, which can be solved
by using "gevp" of the LMI toolbox.
Remark A.I When using 'gevp', F2 (x) should be stated to be positive-
definite.
The decay rate of system (A.2) is defined as the largest "( such that
limt-+oo e')'tIIx(t) II = O. Let us consider a Lyapunov function candidate
V(x(t) = x T (t)Px(t). If we can establish that

dV~(t)) ::; -2"(V(x(t)) (A.lO)

holds for all trajectories, then the decay rate of system (A.2) is at least "(.
Noting that (A.I0) holds if and only if
A T p+PA+2"(P::;O, (A.H)
we conclude that the largest lower bound on the decay rate can be found
by solving the GEVP in P and "(
max"(
p>o
s.t. (A.H).
To solve this optimization problem, let us rewrite (A.H) as

2P::; .!:.(-PA - AT P).

"(
 A.2. LMI Problems 389

Example A.4 Let us now consider a dynamical system with A given by

A = (~2 ~~) and see how we can compute the decay rate. For this
purpose, type the following code and save it.
%% apped4
clear all
A = [-2 .2;2 -5];

setlmis ( [] )
P = lmivar (1, [2, 1]);

lmiterm([-l 1 1 P] ,1,1)
lmiterm([-2 1 1 P] ,-1, A, 's')

lmiterm([3 1 1 P],2,1)
lmiterm([-3 1 1 P] ,-l,A,'s')

LMIs = getlmis;
[copt xopt] = gevp(LMIs, 1);

P = dec2mat(LMIs, xopt, P)
alpha = 1Icopt
Running this code yields
P
1.0e-005 *
0.1254 -0.1937
-0.1937 0.3031

alpha =
1.8720

Therefore, the optimal decay rate of system (A. 2) with A = ( -22 0.2)
-5
is 1.8720, which is, actually, the smallest eigenvalue of A.
Since this book deals with stochastic systems, let us give a numerical ex-
ample for checking the stability of a jump linear system using the feasibility
problem.
Consider a jump linear system with the following dynamics:
x(t) = A(rt)x(t), x(O) = xo,
{ y(t) = C(rdx(t) (A.12)

where x(t) E IR n is the system state, y(t) E IRm is the output ofthe system,
and rt E S = {I, 2,···, N} is a Markov process with infinitesimal generator
A = (Aij),i,j E S,Aii = -L-j#iAij,Aij 2:: O,j 1: i. Based on control
390 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

theory the system (A.12) is said to be stochastically stable iffor any initial
condition (xo, ro), the following holds

1 00
2
E{lIx(t)1I 1(xo, ron < 00.

For system (A.12), we have the following result (see [62]): System (A.12)
is stochastically stable if and only there exist symmetric and positive-
definite matrices P = (P(l),"" P(N)) > 0 such that for all i E S, the
following hold:
N
AT (i)P(i) + P(i)A(i) + L AijP(j) < O. (A.13)
j=l

Example A.5 Let us consider a system with S = {1,2} and A =

(-=r4 !7) and assume that A(l) = (~2 ~;), A(2) = (~ ~;).
Type the following code and save it.

%% apped5
clear all

Al = [-2 0.2;2 -5];

A2 = [1 .2;0 -5];
Lambda = [-44;7 -7];

setlmis ( [] )
Pl = Imivar(l, [2, 1]);
P2 = Imivar(l, [2, 1]);

Imiterm([l 1 1 Pl] ,1,Al,'s');

Imiterm([l 1 1 Pl] ,1,Lambda(1,1));
Imiterm([11 1 P2],1,Lambda(1,2));
Imiterm([21 1 P2] ,1,A2,'s');
Imiterm([2 1 1 P2] ,1,Lambda(2,2));
Imiterm([21 1 Pl],1,Lambda(2,1));
Imiterm([-31 1 Pl],l,l)
Imiterm([-41 1 P2],1,1)

LMIs = getlmis;
[copt xopt] = feasp(LMIs);

Pl = dec2mat(LMIs, xopt, Pl)

P2 = dec2mat(LMIs, xopt,P2)
Running the above code yields the following solution:
 A.3. Lemmas 391

Linear Simulation Results

1.4 ----j ---i-----i-----j-----j----j----j----j----r---

1.2 - -: - --i----i----i----i----r-r--i----r-
Q)
1 ----r-- T---T----i----T----r----r---r---r---
I I I I I I I I I

2 0.8
"0
:.:
I
----~---
I
'-----1-----:-----:-----:-----:-----:-----:----
I I

I
I

I
I

I
I

I
I

I
I

I
I

I
a.
E I I I I I I I I •

----i-----r--- 1-----l----t---1----1----1----1----
I I I I I I I 1 1

<t 0.6

0.4 ----i----i----i- ---i----r--r-r--r--r-

02 ----r---T----i----T --1-- ;----r---r---r---
o ----'--- --'---- -'-- ---'-----'--- _.; .J ..... -I_ - - -

o 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time (sec.)

Figure A.I. Behavior of the system trajectory.

Pi =
0.9673 0.0653
0.0653 0.1389

P2
1.5775 0.0689
0.0689 0.1378

Therefore, the system under study is stochastically stable. In the case of

C(l) = C(2) = (1 0), a trajectory of the system is plotted in Figure
A.i.

A.3 Lemmas
This section presents a certain number of results that are used extensively
in this book in various proofs of the proposed theorems.
The following lemma is critical for casting a nonlinear problem into the
framework of LMI.
Lemma A.I Let X, Y be real constant matrices of compatible dimensions.
Then

holds for any c > O.

Proof: The proof follows from the inequality

o~ (j€XT- ~yT)( j€X - ~Y).

o
392 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

Lemma A.2 (Schur Complement) Let the symmetric matrix M be parti-

tioned as

M= (~ ~)
with X, Z being symmetric matrices. We have
(i) M is nonnegative-definite if and only if either
Z ~o
Y=LlZ (A.I4)
{ X - LlZL! ~ 0

or

{ ~~~L2
Z - LJXL 2 ~ 0
(A.I5)

hold, where L l , L 2 are some (nonunique) matrices of compatible

dimensions.
(ii) M is positive-definite if and only if either
z>0
{ X-YZ-lyT>O (A.I6)
or
X >0
{ Z - yT X-ly > o. (A.17)

Matrix X - YZ-lyT is called the Schur complement X(Z) in M.

Proof: We begin with proving (A.14). Necessity. Clearly, Z ~ 0 is neces-
sary. We must prove the necessity of X - Y Z-l Y T. Letting x be a vector
and partition it as x = ( ~~ ) according to the partitioning of M, we
have
(A.18)
Let X2 be such that ZX2 = O. If YX2 -10, let Xl = -aYx2,a > O. Then
x T Mx = a 2xJyT XYx2 - 2axJ yTYx 2,
which is negative for a sufficiently small a > O. Therefore
XX2 = 0 ==> YX2 = 0, 'r/X2,
which implies
(A.19)
for some (nonunique) L l .
 A.3. Lemmas 393

Since M 2 0, the quadratic form (A.I8) has, for any Xl, a minimum over
X2. Thus differentiating (A.I8) with respect to xI
we have

8(x T Mx) T T
0= T = 2Y Xl + 2ZX2 = 2ZL I Xl + 2ZX2
8X2
whence
(A.20)

Using (A.I9) and (A.20) in (A.I8), we find that the minimum of X T Mx

over X2 and for any Xl is
minx T Mx = xi (X - LlZLi)xl,
X2

which proves the necessity of X - LlZL! 2 o.

Sufficiency. The conditions (A.I4) are therefore necessary for M 20, and
since together they imply that the minimum of XT Qx over X2 for any Xl is
nonnegative, they are also sufficient.
By the same argument, conditions (A.I5) can be derived as those of
(A.I4) are, by starting with X.
(ii) is a direct corollary of (i). This completes the proof of Lemma A.2. 0

Let U E lR. nxk . U.1- is said to be the complement of U if U T U.1- = 0 and

[UU.1-] is of maximum rank (which means that [UU.1-] is nonsingular).
Lemma A.3 Let G, U, V be given matrices with G being symmetric.
(i) Then there exists matrix X such that
(A.2I)

if and only if
uI GU.1- > 0, VJ GV.1- > 0 (A.22)

hold, where U.1-, V.1- are orthogonal complements of U and V,

respectively.
(ii) uI GU.1- > 0 holds if and only if there exists a scalar (7 such that
G - (7UU T > O.
Proof: See [38] pp. 32-33.
Using Lemma A.3, one can eliminate some matrix variables in a ma-
trix inequality, and therefore a nonlinear problem can be cast into the
LMI framework. This reduces the computation burden significantly for the
problem under consideration.
394 Appendix A. Linear Matrix Inequality and Preliminary Lemmas

Lemma AA Let X E lR nxn and Y E lR nxn be symmetric, positive-definite

matrices. Then, there exists a symmetric, positive-definite matrix P > 0
satisfying P = (~ :), p- 1 = (~ : ) if and only if X - y- 1 2: O.
Lemma A.5 (See [214]) Let Y be a symmetric matrix, and H, E be given
matrices of appropriate dimensions, and P satisfy p T P ::; I. Then, we
have
(i) Por any e: > 0, HPE + E T p T H T ::; e:HH T + ~ET E.
(ii) Y + H P E + E T p T H T < 0 holds if and only if there exists a scalar
e: > 0 such that Y +e:HH T +c1ETE < O.
Lemma A.6 (See [209]) Let A, D,~, E be real matrices of appropriate
dimensions with II~II ::; 1. Then, we have
(i) for any matrix P > 0 and scalar e: > 0 satisfying d - EPE T > 0,
(A + D~E)P(A + D~E)T
::;APA T +APET(d-EPET)-lEPA T +e:DD T (A.23)
(ii) for any matrix P > 0 and scalar e: > 0 satisfying P - e:DD T > 0,

(A + D~E)T P-1(A + D~E) ::; AT (P - e:DDT)-l A + ~ET E.

e:

Notes
The proof of the Schur complement lemma is borrowed from [122], of which
another derivation using pseudo inverses can be found in [3]. Elimination
of matrix variables is related to a matrix dilation problem considered in
[61], which was used in control problems in [85, 107, 161, 162, 163]. The
proof of Lemma A.3 is from [38].
Appendix B
Matrix Inversion Formulas

Let A be a square matrix partitioned as follows:

A = (Au A 12 )
A 21 A 22
with A u ,A22 being square matrices. If A u (A 22 ) is invertible, then matrix
~ = A 22 - A 21 Ail A 12 (Li = Au - A12Az21 A 2d is called the Schur com-
plement of Au (A 22 ). Matrix A is invertible if and only if Au and ~ are
nonsingular, or A 22 and b. are nonsingular.
Furthermore, if A is invertible, then
1
( Au A 12 ) -1 = (
A 21 A 22
r1
- ~-lA 21 A-
U
1
- A- U A 12 ~-1
~-1
(B.l) )
and
Li -1
)
- 1
( Au A 12 = ( -~- A 12 A Z21
) -1
A-221A 21 Li-1 (B.2)
A 21 A 22 - r2
where
r 1 = A-111 + A-U 1A 12 ~-lA 21 A-u'1
r 2 = A-221 + A-221A 21 Li -1 A 12 A-22'1
Suppose that the matrices Au and A 22 are nonsingular, then we have
r
[Au - A12Az21 A 21 1 = All + All AdA22 - A21Al11 A 12 1A21Al11 (B.3) r
Appendix C
Kronecker Product

Let A = (aij) E IR pxq and B E IRmxn be two matrices. The Kronecker

product of A and B denoted by A ® B is defined by

The following properties of the Kronecker product are easily established:

l®A=A
(A+B) ®C = A®C +B® C
(A®B)(C®D) = AC®BD
(A®B)T = AT ®B T
(A®B)-I = A-I ®B- I .

Let In denote the n x n unit matrix. The q-dimensional vector which has
value one in the kth element and zero elsewhere is called the unit vector
and is denoted by
398 Appendix C. Kronecker Product

The parenthetical underscore is omitted if the dimension can be inferred

from the context. The elementary matrix

has dimensions p x q, and has value one in the i - kth element and zero
elsewhere. Define the permutation matrix by
p q
q) n, E(qx p)
U pxq -~ "" L.J E(px
L.J "" ik '<Y ki .
i=l j=l

This matrix is square (pq x pq) and has precisely a single one value in each
row and in each column. Let X E ~pxq and let vec(X) denote the vector
formed by stacking the columns of X into one long vector, that is,

Xpl
X12

vec(X) = X22

X pq
Then we have the following properties (see [39] for the proof).
Property 1: vec(X T ) = UpXqvec(X).
Property 2: Let A, B, X be matrices of appropriate dimensions. Then
vec(AXB) = (B T <':9 A)vec(X).
Appendix D
Markov Process

In this appendix we will give some results on stochastic processes to help

the reader understand the material covered in this book. The Markovian
property is recalled and some definitions and equalities are given.

D.l Continuous-Time Finite State Markov Process

Definition D.I Let (0, F, P) be a probability space and {rt, t 2: O} be a
stochastic process taking values in S = {I, 2,·· " N}. Then, {rt, t 2: O} is
said to be a Markov process with state space S if
P(r(t) = ilr(w) : w :s; s) = P(r(t) = ilr(s))
holds for all 0 :s; s :s; t and i E S.
For any i,j E S, let Pij(s, t) denote the transition probability, i.e.,
Pij(s, t) = P(r(t) = jlr(s) = i). Matrix P(s, t) = (Pij(s, t)), i,j E S is said
to be the transition matrix. Under the continuity condition limHo+ P(t) =
I, it follows that for 0 :s; s :s; u :s; t,
Pij(s, t) 2: 0, i,j E S
LPij(S,t) = 1, i E S
jES

Pij(s, t) = L Pik(S, u)Pkj(u, t), i,j E S.

kES
400 Appendix D. Markov Process

The last identity is usually referred to as the Chapman-Kolmogorov

equation.
If the transition probability Pij(s, t) depends only on t - s, {r(t), t ~ o}
is said to be stationary. Otherwise, the process is nonstationary. In this
book, the Markov processes are assumed to be stationary.
For a stationary Markov process, it can be shown [84] that the following
hold
. 1 - Pii(t) \
11m = Ai < 00
t-tO t
. Pij(t) \
11m - - = Ai· < 00.
t-ts t J

Obviously, Ai = L#i Aij. The parameter matrix A = (Aij), i, j E $ with,

by convention, Aii = - Ai gives the infinitesimal generator of the Markov
process {r(t), t ~ o} and AijS are called the infinitesimal characteristics.
The transition probability and the generator matrices of the process satisfy
the following forward Chapman-Kolmogorov equation:

D.2 Strong Markov Property and Infinitesimal

Generator of Stochastic Hybrid Systems
This section addresses the Markov property of a hybrid stochastic system

real interval [-7,0], 7 > °

with time delay. Let C[-7,0] be the space of continuous functions on the
(the time delay in the system), and let x(t)
be a vector-valued stochastic process. Define the process Xt with values in
C[-7,0] by
Xt(S) = x(t + s), S E [-7,0].
Let us define the norm of Xt by IIxtll = sUPsE[_T,ojllx(t + s)ll. Let us con-
sider a hybrid system with N modes and mode-switching governed by a
Markov process {rt, t ~ O} taking values in $ = {l, 2, ... ,N} and hav-
ing infinitesimal generator A = (Aij),i,j E $, where Aij ~ O,j i i, and
Aii = - Lj#i Aij.
Suppose that x(t) is described by the following dynamic:
dx(t) = f(rt,xddt + g(rt,xddB(t), Xo = ¢(.), (D.l)
where f(-,·) : $ x C[-7,0] -+ R n , g(-,.) : $ x C[-7,0] -+ jRnxm; B(t) E
jRm is a standard Brownian process and is assumed to be independent of
{rt, t ~ O}; and ¢O is the initial condition.
This model contains all the models used in the book as special cases.
This section addresses the Markov property of this process. For each given
 D.2. Stochastic Hybrid Systems 401

mode i E S, let !J (i, .), 1 :S j :S n denote the jth component of f (i, .) and
let gjk(i,·) denote the (i,j)th component of g(i, .). Define the vector and
matrix norms as IIf(i, ')11 2 = 2:7=1 fJ(i, '), IIg(i, ')11 2 = 2:7=12:;;'=1 g;k(i, '),
respectively. Assume that the following assumptions hold for every given
i E S.

Assumption D.I !J (i, '), gjk (i, .) are continuous real-valued functions on
q-7,0].
Assumption D.2 In the interval [-7,0], x(t) is continuous with proba-
bility one, is independent of B(s) - B(O), s 2 0, and satisfies IEII¢(s)1I 4 <
00.

Assumption D.3 There is a constant M < 00 and a bounded measure f.1

on [-7,0] so that for any 'l/Jb 1J;2 E q-7, 0],
IIf(i,'l/Jd - f(i,'l/J2)11 + IIg(i,'l/Jd - g(i,'l/J2)11
:S J~T 1'l/J1(S) - 1J;2(s)ldf.1(S) (D.2)
{
Ilf(i,O)11 + Ig(i,O)I:S M.
Assumption D.4 For each positive real number p there is a bounded mea-
sure f.1p on [-7,0] so that for 11'l/J11 :S p and II¢II :S p, (D.2) is valid with f.1p
replacing f.1.

Let C be the collection of open sets in S x q-7, 0] and let B be the Borel
field over C. We suppose that the probability measure space introduced in
the sequel is complete with respect to whatever measures are imposed on
them. Let 0 denote the probability sample space, and w denote the generic
element of O. Define MP'x) and iVP,x) as the least a-fields on 0 over which
x(s), - 7 :::; S :::; t, and x(s), t - 7 :::; S :::; t, are measurable, respectively,
for fixed initial condition r(O) = i, Xo = x. Let P(i,x) be the probability
measure on
M(i,x) - U M(i,x)
- t2:O t .

With this notation defined, we have the following theorem.

Theorem D.I Assume that Assumptions D.l-D.3 hold and let Xo =
¢(.) E q-7,0]. Then {(r(t),xt),t 2 O} is a strong Markov process on
the topological state space {(S x q -7,0], C, B)}.
Proof: To prove the Markov property, we need to check the conditions of
Dynkin [71], pp. 77-80. That is, we need to show that

1) the function p defined by p(t,(i,x),r) = P(i,x)((r(t),xt) E r) for

arbitrary r E E S x q -7,0] is measurable, and

2) P(i,x)((r(t+~),Xt+t» E f1Mt(i,x)) = p(~, (r(t),xt),r) with probability

one.
402 Appendix D. Markov Process

To prove the strong Markov property, it suffices to prove that if we .) is

bounded and continuous on (5 x q-7, 0], then E(i,x)w(r(t),xt) ~ ')'(i, x) is
continuous in (i,x), where E(i,x) is the expectation operator corresponding
to P(i,x)' The details of the proof are omitted here. 0

A real-valued function V(·,·) on q-7,O] is said to be in the domain of

A, the infinitesimal operator, if the limits
. E[V(r(t),xdlr(O)=i,xo=x]-V(i,x) (.)
11m = q Z,X
t~O t
lim E[q(r(t), Xt)] = q(i, x)
t~O

exist pointwise in q -7,0] and i E 5 and the limit is uniformly bounded

in x. Then, we write q(i,x) = AV(i,x).
For example, let us consider the following Lyapunov function for system
(D.1):
V(rt, x(t)) = x T (t)P(rdx(t).
Then,
AV(rt, x(t)) = E(r"x,J [2x T (t)P(rd(J(rt, xd + g(rt, xddB(t))
+ ~tr (g(r t ,xdg T (rt, xd aV~:'2x(t)) ]
N
= 2x T (t)P(rd(J(rt, xt}dt + L Ar,j V(rt, xt)
j=1

+ tr (g(rt,xt)g T (rt,xt)P(rd).
In particular, if m = 1, g(rt, Xt) = g(rdx(t) with g(rt) E IR nxn , then
N
AV(rt, x(t)) = 2x T (t)P(rt)(f(rt, xt)dt + L Ar,j V(rt, xd
j=1

+ x T (t)g T (rdP(rdg(rdx(t).
Let (3 be a random time with E[(3I(r(O),xo)] < 00 and let V(r(t),xt) be
in the domain of A. Then
E[V(r((3), xl3)l(r(O), xo)] - V(r(O),xo)

~ IE [1," AV(r(S),X,)dSI(r(O),xoJj. (D.3)

The above equation is called Dynkin's formula. The deterministic coun-

terpart is the basic formula of the calculus: f(x(t)) - f(x(O)) =
J~ j(x(s))ds.
References

[1] Abou-Kandil, H., Freiling, G., and Jank, G. (1994). Solution and asymp-
totic behavior of coupled Riccati equations in jump linear systems. IEEE
Transactions on Automatic Control, 39(8), 1631-1636.
[2] Abou-Kandil, H., Freiling, G., and Jank, G. (1995). On the solu-
tion of discrete-time Markovian jump linear quadratic control problems.
Automatica, 31(5), 765-768.
[3] Albert, A. (1969). Conditions for positive and nonnegative definiteness in
terms of pseudo inverses. SIAM Journal on Applied Mathematics, 17,434-
440.
[4] Aliyu, M.D.S. and Boukas, E.K. (2000). Finite and infinite horizon Hoc
control for stochastic nonlinear systems. IMA, Journal of Mathematical
Control and Information, 17(3), 265-279.
[5] Aliyu, M.D.S. and Boukas, E.K. (2000). Robust Hoc control for Marko-
vian jump nonlinear systems. IMA, Journal of Mathematical Control and
Information, 17(3), 295-308.
[6] Anderson, B.D.a. and Moore, J.B. (1971). Linear Optimal Control.
Prentice-Hall, Englewood Cliffs, New Jersey.
[7] Basar, T. and Bernhard, P. (1990). Hoc-Optimal Control and Related
Minimax Problems. Birkhiiuser, Berlin.
[8] Benjelloun, K. and Boukas, E.K. (1998). Mean square stochastic stabil-
ity of linear time-delay system with Markov jumping parameters. IEEE
Transactions on Automatic Control, 43(10), 1456-1459.
[9] Benjelloun, K. Boukas, E.K., and Costa, a.L.V. (2000). Hoc-control for
linear time-delay systems with Markovian jumping parameters. Journal of
Optimization Theory and Applications, 105(1), 73-95.
404 References

[lOJ Benjelloun, K., Boukas, E.K., and Yang, H. (1996). Robust stabilizability of
uncertain linear time-delay systems with Markovian jumping parameters.
Journal of Dynamic Systems, Measurement, and Control, 118, 776-783.
[11] Blair, W.P. and Sworder, D.D. (1975). Feedback control of a class of lin-
ear discrete systems with jump parameters and quadratic cost criteria.
International Journal of Control, 21(5), 833-841.
[12J Bolzern, P., Colaneri, P., and de Nicolao, G. (1994). Optimal robust filtering
with time-varying parameter uncertainty. IEEE Transactions on Automatic
Control, 39(3), 623-626.
[13J Bolzern, P., Colaneri, P., and de Nicolao, G. (1996). Optimal robust filtering
with time-varying parameter uncertainty. International Journal of Control,
63(3), 557-576.
[14J Boukas, E.K. (1987). Commande Optimale Stochastique Appliquee aux
Systemes de Production. Ecole Polytechnique de Montreal.
[15J Boukas, E.K. (1993). Control of systems with controlled jump Markov
disturbances. Control Theory and Advanced Technology, 9(2), 577-595.
[16J Boukas, E.K. (1995). Systemes Asservis. Editions de l'Ecole Polytechnique
de Montreal.
[17] Boukas, E.K. and Haurie, A. (1990). Manufacturing flow control and pre-
ventive maintenance, a stochastic control approach. IEEE Transactions on
Automatic Control, 35(7), 1024-1031.
[18J Boukas, E.K. and Liu, Z.K. (1999). Robust stability and H oo control
of discrete-time jump linear systems with time-delay, An LMI approach.
Submitted for publication.
[19J Boukas, E.K. and Liu, Z.K. (2000). Delay-dependent output feedback
stabilization of Markov jump systems with time-delay. Submitted for
publication.
[20J Boukas, E.K. and Liu, Z.K. (2000). Delay-dependent robust stability and
stabilization of uncertain linear systems with discrete and distributed time-
delays. Submitted for publication.
[21] Boukas, E.K. and Liu, Z.K. (2000). Jump linear quadratic regulator with
controlled jump rate. IEEE Transactions on Automatic Control, 45(2).
[22] Boukas, E.K. and Liu, Z.K. (2000). Output feedback guaranteed cost con-
trol for uncertain time-delay systems with Markov jump. Proceedings of
ACe.
[23] Boukas, E.K. and Liu, Z.K. (2000). Robust H oo filtering for poly-
topic uncertain time-delay systems with Markov jumps. Submitted for
publication.
[24] Boukas, E.K. and Liu, Z.K. (2000). Systems with Time-Delay, Stability,
Stabilization, H 00 and their Robustness. A chapter.
[25] Boukas, E.K. and Liu, Z.K. (2001). Deterministic and Stochastic Time-
Delay Systems. Birkhiiuser, Boston.
[26] Boukas, E.K., Liu, Z.K., and Liu, G.X. (2001). Delay-dependent robust sta-
bility and H oo control of jump linear systems with time-delay. International
Journal of Control.
 References 405

[27] Boukas, KK. and Shi, P. (1998). Stochastic stability and guaranteed cost
control of discrete-time uncertain systems with Markovian jumping pa-
rameters. International Journal of Robust and Nonlinear Control, 8(13),
1155-1167.
[28] Boukas, E.K. and Shi, P. (1999). Hoc control for discrete-time linear
systems with Frobenius norm-bounded uncertainties. Automatica, 35(9),
1625-163l.
[29] Boukas, E.K., Shi, P., and Andijani, A. (1999). Optimal inventory-
production control problem with stochastic demand. Optimal Control
Applications and Methods, 20(1), 1-20.
[30] Boukas, E.K., Shi, P., and Benjelloun K. On stabilization of uncertain lin-
ear systems with Markovian jumping parameters. International Journal of
Control, 72, 842-850, 1999.
[31] Boukas, E.K. and Yang, H. (1995). Stability of discrete-time linear systems
with Markovian jumping parameters. Mathematics Control, Signals and
Systems, 8, 390-402.
[32] Boukas, KK. and Yang, H. (1996). Optimal control of manufacturing flow
and preventive maintenance. IEEE Transactions on Automatic Control,
41(6), 881-885.
[33] Boukas, KK. and Yang, H. (1999). Exponential stability of stochastic
Markovian jumping parameters. Automatica, 35(8), 1437-144l.
[34] Boukas, E.K., Yang H., and Zhang, Q. (1995). Minmax production planning
in failure-prone manufacturing systems. Journal of Optimization Theory
and Applications, 87(2), 269-286.
[35] Boukas, E.K., Swierniak, A., and Yang, H. (1999). Suboptimality of a de-
centralized feedback controller. Journal of Dynamic Systems, Measurement,
and Control, 121(2), 305-308.
[36] Boukas, E.K., Zhang, Q., and Yin, G. (1995). Robust production and main-
tenance planning in stochastic manufacturing systems. IEEE Transactions
on Automatic Control, 40(6), 1098-1102.
[37] Boukas, E.K., Zhu, Q., and Zhang, Q. (1994). Piecewise deterministic
Markov process model for flexible manufacturing systems with preven-
tive maintenance. Journal of Optimization Theory and Applications, 81(2),
259-275.
[38] Boyd, S. Ghaoui, L.K, Feron, E., and Balakrishnan, V. (1994). Linear
Matrix Inequalities in System and Control Theory. SIAM, Philadelphia.
[39] Brewef, J.W. (1978). Kronecker products and matrix calculus in system
theory. IEEE Transactions on Circuits and Systems, CAS-25(9).
[40] Cao, Y.Y. and Lam, J. (1999). Stochastic stabilizability and Hoc control
for discrete-time jump linear systems with time-delay. In Proceedings of the
14th IFAC World Congress, Beijing, China.
[41] Cao, Y.Y. and Lam, J. (2000). Robust Hoc control of uncertain Markovian
jump systems with time-delay. IEEE Transactions on Automatic Control,
45(1).
406 References

[42] Chao, G.L., Pering, J.W., and Han, KW. (1998). Robust stability analysis
of time-delay systems using parameter-plane and parameter-space methods.
Journal of Franklin Institute, 335B(7), 1249--1262.
[43] Cao, YY, Sun YX., and Lam, J. (1998). Delay-dependent robust Hoc
control for uncertain systems with time-varying delays. lEE Proceedings
D, Control Theory and Applications, 145(3).
[44] Chang, YH. (1999). Robust regional stability analysis of continuous
time-delay systems. lEE Proceedings D, Control Theory and Applications,
146(4),311-317.
[45J Chen, J.D., Lien, C.H., Fan, KK, and Cheng, J.S. (2000). Delay-dependent
stability criterion for neutral time-delay systems. Electronics Letters,
36(22), 1897-1898.
[46] Chen, YH., Wang, W.J., and Mau, L.G. (1996). Robust stabilization of
large-scale time-delay systems with estimated state feedback. Journal of
Optimization Theory and Applications, 89(3), 543-559.
[47] Cheres, E., Palmor, Z.J., and Gutman, S. (1989). Quantitative measure of
robustness for systems including delayed perturbations. IEEE Transactions
on Automatic Control, 34, 1203-1204.
[48J Chilali, M. and Gahinet, P. (1996). Hoc design with pole placement con-
straints, An LMI approach. IEEE Transactions on Automatic Control,
41(3), 358-367.
[49J Choi, H.H. and Chung, M.J. (1995). Memoryless Hoc controller design for
linear systems with delayed state and control. Automatica, 31(6), 917-919.
[50] Costa, a.L.V. (1994). Linear minimum mean square error estimation
for discrete-time Markovian jump linear systems. IEEE Transactions on
Automatic Control, 39(8), 1685-1689.
[51] Costa, a.L.V. (1996). Mean-square stabilizing solutions for discrete-time
coupled algebraic Riccati equations. IEEE Transactions on Automatic
Control, 41(4), 593-598.
[52J Costa, a.L.V. and Boukas, E.K (1998). Necessary and sufficient condition
for robust stability and stabilizability of continuous-time linear systems
with Markovian jumps. Journal of Optimization Theory and Applications.
[53J Costa, a.L.V., do Val, J.B.R, and Geromel, J.C. (1997). A convex pro-
gramming approach to Hz control of discrete-time Markovian jump linear
systems. International Journal of Control, 66(4), 557-579.
[54] Costa, a.L.V., do Val, J.B.R, and Geromel, J.C. (1999). Continuous-time
state-feedback Hz control of Markovian jump linear systems via convex
analysis. Automatica, 35(2), 259-268.
[55J Costa, a.L.V., Filho, E.a.A., Boukas, E.K., and Marquez, RP. (1999).
Constrained quadratic control of discrete-time Markovian jump linear
systems. Automatica, 35(4), 617-627.
[56J Costa, a.L.V. and Fragoso, M.D. (1991). Necessary and sufficient condi-
tions for mean square stability of discrete-time linear systems subject to
Markovian jumps. In Proceedings of the 9th International Symposium on
Mathematics Theory of Networks and Systems, pages 85-86, Kobe, Japan.
 References 407

[57] Costa, a.L.v. and Fragoso, M.D. (1993). Stability results for discrete-
time linear systems with Markovian jumping parameters. Journal of
Mathematical Analysis and Applications, 179(2), 154-178.
[58] Costa, a.L.V. and Fragoso, M.D. (1995). Discrete-time LQ-optimal control
problems for infinite Markov jump parameter systems. IEEE Transactions
on Automatic Control, 40(12), 2076-2088.
[59] Costa, a.L.v. and Marques, RP. (1998). Mixed H2/ Hoc control of discrete-
time Markovian jump linear systems. IEEE Transactions on Automatic
Control, 43, 95-100.
[60] Datko, R (1974). Neutral autonomous functional equations with quadratic
cost. SIAM Journal Control, 12, 70-82.
[61] Davis, C., Kahan, W.M., and Weingerger, H.F. (1982). Norm-preserving
dilations and their applications to optimal error bounds. SIAM Journal of
Numerical Analysis, 19(3), 445-479.
[62] de Farias, D.P., Geromel, J.C., do Val, J.B.R, and Costa, a.L.v. (2000).
aut put feedback control of Markov jump linear systems in continuous-time.
IEEE Transactions on Automatic Control, 45(5).
[63] de Souza, C.E. and Fragoso, M.D. (1993). Hoc control for linear sys-
tems with Markovian jumping parameters. Control- Theory and Advanced
Technology, 9(2), 457-466.
[64] Doyle, J.C., Glover, K., Khargonekar, P.P., and Francis, B.A. (1989).
State space solutions to the standard H 2 and HOC control problems. IEEE
Transactions on Automatic Control, 34(8), 831-847.
[65] Dragan, V. and Morozan, T. (1997). Global solutions to a game-theoretic
Riccati equations of stochastic control. Journal of Differential Equations,
138(2), 328-350.
[66] Dragan, V. and Morozan, T. (1999). Stability and robust stabilization of
linear stochastic systems described by differential equations with Marko-
vian jumping and multiplicative white noise. Technical report, 17/1999,
Institute of Mathematics of Romanian Academy Science.
[67] Dragan, V., Shi, P., and Boukas, E.K. (1999). Control of singularly
perturbed systems with Markovian jump parameters, an Hoc approach.
Automatica, 35(8), 1369-1378.
[68] Dufour, F. and Bertrand, P. (1994). The filtering problem for continuous-
time linear systems with Markovian switching coefficients. Systems and
Control Letters, 23(5), 453-461.
[69] Dufour, F. and Bertrand P. (1996). An image-based filter for discrete-time
Markovian jump linear systems. Automatica, 32(2), 241-247.
[70] Dugard, L. and Verriest, E.L (Eds). (1997). Stability and control of time-
delay systems. Number 228. Lecture Notes on Control and Information
Science, New York, Springer.
[71] Dynkin, E.B. (1965). Markov Process. Springer, Berlin.
[72] Esfahani, S.H., Moheimani, S.a.R, and Petersen, LR (1998). LMI ap-
proach to suboptimal guaranteed cost control for uncertain time-delay
systems. lEE Proceedings D, Control Theory and Applications, 145(6),
491-498.
408 References

[73J Esfahani, S.H. and Petersen, I.R. (1999). An LMI approach to the out-
put feedback guaranteed cost control for uncertain time-delay systems.
Preprint.
[74J Ezzine, J. and Haddad, A.H. (1989). On the controllability and ob-
servability of hybrid systems. International Journal of Control, 49(6),
2045-2055.
[75] Ezzine, J. and Karvanoglu, D. (1997). On almost-sure stabilization of
discrete-time parameter systems, An LMI approach. International Journal
of Control, 68(5), 1129-1146.
[76] Fattouh, A., Sename, 0., and Dion, J.M. (1998). Hoc observer design for
time-delay systems. In Proceedings of 37th IEEE Conference on Decision
and Control, pages 4545-4546, Tampa, Florida.
[77] Feliachi, A. and Thowsen, A. (1981). Memoryless stabilization of linear
delay-differential systems. IEEE Transactions on Automatic Control, 26,
586-587.
[78] Feng, X. and Loparo, K.A. (1990). Almost sure instability of the random
harmonic oscillator. SIAM Journal on Applied Mathematics, 50(3), 744-
759.
[79J Feng, X., Loparo, K.A., Ji, Y, and ehizeck, H.J. (1992). Stochastic sta-
bility properties of jump linear systems. IEEE Transactions on Automatic
Control, 37(1), 38-53.
[80] Fragoso, M.D. (1988). On a partially observable LQG problem for sys-
tems with Markovian jumping parameters and quadratic cost. Systems and
Control Letters, 10(5), 349-356.
[81] Fragoso, M.D. (1989). Discrete-time jump LQG problem. International
Journal of Systems Science, 20(12), 2539-2545.
[82] Fragoso, M.D., do Val, J.B.R., and Pinto Jr., D.L. (1995). Jump linear Hoc
control, The discrete-time case. Control Theory and Advanced Technology.
[83] Francis, B.A. (1987). A Course in Hoc Control Theory. Springer-Verlag,
New York.
[84] Freedman, D. (1983). Markov Chain. Springer-Verlag, New York.
[85J Gahinet, P. (1992). A convex parametrization of Hoc suboptimal con-
trollers. Proceedings of IEEE Conference on Decision and Control, pages
1742-1729.
[86] Gajic, Z. and Borno, I. (1995). Lyapunov iterations for optimal control
of jump linear systems at steady state. IEEE Transactions on Automatic
Control, 40, 1971-1975.
[87J Gallegos, J.A. (1994). Disturbance rejection of time-delay systems. Inter-
national Journal of Systems Science, 25(6), 1081-1091.
[88J Ge, J.H., Frank, P.M., and Lin, e.F. (1997). Hoc control via output
feedback for state delayed systems. International Journal of Control.
[89] Geromel, J.e., Oliveira, M.e., and Bernussou, J. (1999). Robust filtering
of discrete-time linear systems with parameters dependent Lyapunov func-
tions. In Proceedings of the 38th IEEE Conference on Decision and Control,
Phoenix.
 References 409

[90] Geromel, J.C., Peres, P.L.D., and Bernussou, J. (1991). On a convex pa-
rameter space method for linear control design of uncertain systems. SIAM
Journal on Control and Optimization, 29, 381-402.
[91] Gorecki, H., Fuksa, S., Grabowski, P., and Korytowski, A. (1989). Analysis
and Synthesis of Time Delay Systems. John Wiley and Sons, Warszawa.
[92] Griffiths, RE. and Loparo, KA. (1985). Optimal control of jump-linear
gaussian systems. International Journal of Control, 42(4), 791-819.
[93] Gu, K (1997). Discretized LMI set in the stability problem of lin-
ear uncertain time-delay systems. International Journal of Control, 68,
923-934.
[94] Gu, K (1999). A generalized discretization scheme of Lyapunov func-
tional in the stability problem of linear uncertain time-delay systems.
International Journal of Robust and Nonlinear Control, 9, 1-14.
[95] Gu, K and Niculescu, S.L. (2000). Additional dynamics in transformed
time-delay systems. IEEE Transactions on Automatic Control, 45(3), 572-
575.
[96] Guojun, J. and Wenzhong, S. (2001). Stability of bilinear time-delay
systems. IMA Journal of Mathematical Control and Information, 18, 55-60.
[97] Hale, J.K (1977). Theory of Functional Differential Equations. New York,
Springer-Verlag.
[98] Hale, J.K and Lunel, S.M.V. (1993). Introduction to Functional Differential
Equations. Springer-Verlag, New York.
[99] Hmamed, A. (1991). Further results on the stability of uncertain time-delay
systems. International Journal of System Science, 22(3), 605-614.
[100] Hmamed, A., Benzaouia, A., and Bensalah, H. (1995). Regulator problem
for linear continuous-time delay systems with nonsymmetrical constrained
control. IEEE Transactions on Automatic Control, 40(9), 1615-1619.
[101] Hsiao, F.H. and Hwang, H.D. (1996). Criteria for asymptotic stability of
uncertain multiple time-delay systems. Electronics Letters, 32(4), 410-412.
[102] Hsiao, F.H., Pan, S.T., and Teng, C.C. (1997). Robust controller design for
uncertain multiple time-delay systems. Transactions of ASME Journal of
Dynamic Systems, Measurement, and Control, 119, 122-127.
[103] Huang, Y.P. and Zhou, K (2000). On the robustness of uncertain time-
delay systems with structured uncertainties. Systems and Control Letters,
41, 367-376.
[104] Huang, Y.P. and Zhou, K (2000). Robust stability of uncertain time-delay
systems. IEEE Transactions on Automatic Control, 45(11), 2169-2173.
[105] Ikeda, M. and Ashida, T. (1979). Stabilization of linear systems with time-
varying delay. IEEE Transactions on Automatic Control, 24, 369-370.
[106] Ivanescu, D., Dion, J.M., Dugard, L., and Niculescu, S.l. (2000). Dynami-
cal compensation for time-delay systems, an LMI approach. International
Journal of Robust and Nonlinear Control, 10, 611-628.
[107] Iwasaki, T. and Skelton, R E. (1993). All controllers for the general
H oo control problem, LMI existence conditions and state space formulas.
Automatica, 30(8), 1307-1317.
410 References

[108] Jeung, E.T., Kim, J.H., and Park, H.B. (1996). Robust control for parame-
ter uncertain delay systems in state and control input. Automatica, 32(9),
1337-1339.
[109J Jeung, E.T., Kim, J.H., and Park, H.B. (1998). Hoc-output feedback con-
troller design for linear systems with time-varying delayed state. IEEE
Transactions on Automatic Control, 43(7).
[110] Jeung, E.T., Oh, D.C., Kim, J.H., and Park, H.B. (1996). Robust controller
design for uncertain systems with time delays, LMI approach. Automatica,
32(8), 1229--1231.
[111J Ji, Y. and Chizeck, H.J. (1990). Controllability, stabilizability, and
continuous-time Markovian jump linear quadratic control. IEEE Trans-
actions on Automatic Control, 35(7), 777-788.
[112] Ji, Y. and Chizeck, H.J. (1992). Jump linear quadratic Gaussian control in
continuous time. IEEE Transactions on Automatic Control, 37(12), 1884-
1892.
[113] Jiang, X., Fei, S., and Feng, C.B. (2000). Comments on "bounded real
criteria for linear time-delay systems". IEEE Transactions on Automatic
Control, 45(7), 1409--1410.
[114] Kharitonov, V.L. (1999). Robust stability analysis of time-delay systems,
A survey. Annual Reviews in Control, 23, 185-196.
[115] Kharitonov, V.L. and Zhabko, A.P. (1994). Robust stability of time-delay
systems. IEEE Transactions on Automatic Control, 39(12), 2388-2397.
[116] Kojima, A. and Ishijima, S. (1995). Explicit formulas for operator Riccati
equations arising in Hoc control with delays. Proceedings of the IEEE 34th
Conference on Decision and Control, New Orleans, LA.
[117J Kolmanovskii, V.B. and Myshkis, A. (1999). Introduction to the Theory
and Applications of Functional Differential Equations. Dordrecht, Kluwer
Academic Publishers.
[118] Kolmanovskii, V.B. and Richard, J.P. (1999). Stability of some linear
systems with delay. IEEE Transactions on Automatic Control, 44(5),
984-989.
[119] Kolmanovsky, I. and Maizenberg, T.L. (2000). Stochastic stability of a class
of nonlinear systems with randomly varying time-delay. In Proceedings of
the American Control Conference, pages 4304-4308, Chicago, Illinois.
[120J Krasovskii, N.N. and Lidskii, E.A. (1961). Analysis design of controller in
systems with random attributes-Part 1. Automatic Remote Control, 22,
1021-1025.
[121] Krasovskii, N.N. and Lidskii, E.A. (1961). Analysis design of controller in
systems with random attributes-Part 2. Automatic Remote Control, 22,
1141-1146.
[122J Kreindler, E. and Jameson, A. (1972). Conditions for nonnegativeness of
partitioned matrices. IEEE Transactions on Automatic Control, 17(2),
147-148.
[123] Kung, F.C. and Lee, C.H. (1996). Decentralized robust control design
for large-scale time-delay systems with time-varying uncertainties. JSME
International Journal, Series C, 39(3), 528-533.
 References 411

[124] Lee, C.H. (1998). Simple stabilization criteria and memoryless state feed-
back control for time-delay systems with time-varying perturbations.
International Journal of Control, 45(11), 1211-1215.
[125] Lee, C.H., Li, T.H., and Kung, F.C. (1995). On the robustness of stability
for uncertain time-delay systems. International Journal of Systems Science,
26(2), 457-465.
[126] Li, H., Niculescu, S.I., Dugard, L., and Dion, J.M. (1998). Robust guar-
anteed cost control of uncertain linear time-delay systems using dynamic
output feedback. Mathematics and Computers in Simulation, 45, 349-358.
[127] Li, X. and de Souza, C.E. (1995). LMI approach to delay-dependent robust
stability of uncertain linear systems. Proceedings of the 34th Conference on
Decision and Control, New Orleans, LA, pages 3614-3648.
[128] Li, X. and de Souza, C.E. (1997). Delay-dependent robust stability and
stabilization of uncertain linear delay systems: a linear matrix inequality
approach. IEEE Transactions on Automatic Control, 42(8), 1144-1148.
[129] Lien, C.H. and Hsieh, J.G. (2000). New results on global exponential sta-
bility of interval time-delay systems. JSME International Journal Series
C., 43(2), 306-320.
[130] Lisong, Y. (1996). Robust analysis and synthesis of linear time-delay sys-
tems with norm-bounded time-varying uncertainty. Systems and Control
Letters, 42, 281-289.
[131] Liu, P.L. and Su, T.J. (1998). Robust stability of interval time-delay
systems with delay-dependence. Systems and Control Letters, 33, 231-239.
[132] Loiseau, J.J. (2000). Algebraic tools for the control and stabilization of
time-delay systems. Annual Reviews in Control, 24, 135-149.
[133] Luc, e.G. and Habets, J.M. (1996). On the genericity of stabilizability for
time-delay systems. SIAM Journal on Control and Optimization, 34(3),
833-854.
[134] Luo, M., de La Sen, M., and Rodellar, J. (1997). Robust stabilization of a
class of uncertain time-delay systems in sliding mode. International Journal
of Robust and Nonlinear Control, 7, 59-74.
[135] Mahmoud, M.S. (2000). Robust Control and Filtering for Time-Delay
Systems. Marcel Dekker, New York.
[136] Mahmoud, M.S. (2001). Control of uncertain state-delay systems: guaran-
teed cost approach. IMA Journal of Mathematical Control and Information,
18, 109-128.
[137] Mahmoud, M.S. and Al-Muthairi, N.F. (1994). Design of robust controllers
for time-delay systems. IEEE Transactions on Automatic Control, 39, 995-
999.
[138] Mahmoud, M.S. and Bingulac, S. (1998). Robust design of stabiliz-
ing controllers for interconnected time-delay systems. Automatica, 34(6),
795-800.
[139] Mahmoud, M.S., Terro, M.J., and Abdel-Rohman, M. (1998). An LMI ap-
proach to Hoc-control of time-delay systems for the benchmark problem.
Earthquake Engineering and Structural Dynamics, 27, 957-976.
412 References

[140] Mahmoud, M.S. and Xia, L. (1999). Stability and positive realness of time-
delay systems. Journal of Mathematical Analysis and Applications, 239,
7-19.
[141] Malek-Zavarei, M. and Jamshidi, M. (1987). Time-delay systems, analysis,
optimization and applications. North-Holland Systems and Control Series.
[142] Mariton, M. (1990). Jump Linear Systems in Automatic Control. Marcel
Dekker, New York.
[143] Mariton, M. and Bertrand, P. (1985). Output feedback for a class of
linear systems with stochastic jump parameters. IEEE Transactions on
Automatic Control, 30(9), 898-900.
[144] Moheimani, S.O.R. and Petersen, I.R. (1996). Quadratic guaranteed cost
control with robust pole placement in a disk. lEE Proceedings D, Control
Theory and Applications, 143(1), 37-43.
[145] Moheimani, S.O.R. and Petersen, I.R. (1997). Optimal quadratic guaran-
teed cost control of a class of uncertain time-delay systems. lEE Proceedings
D, Control Theory and Applications, 144(2), 183-188.
[146] Moheimani, S.O.R., Savkin, A., and Petersen, I.R. (2000). Synthesis of min-
imax optimal controllers for uncertain time-delay systems with structured
uncertainty. International Journal of Systems Science, 31(2), 137-147.
[147] Mori, T. et al. (1982). On an estimate of the decay rate for stable linear
delay systems. International Journal of Control, 36(1), 95-97.
[148] Mori, T. and Kokame, H. (1989). Stability of x(t) = Ax(t) + Bx(t - T).
IEEE Transactions on Automatic Control, 34(4).
[149] Mori, T., Noldus, E., and Kuwahara, M. (1983). A way to stabilize linear
systems with delayed state. Automatica, 19(5), 571-573.
[150] Morozan, T. (1983). Optimal stationary control for dynamic systems with
Markov perturbations. Stochastic Analysis and Applications, 1(3), 299-325.
[151] Morozan, T. (1995). Stability and control for linear systems with jump
Markov perturbations. Stochastic Analysis and Applications, 13(1), 91-110.
[152] Morozan, T. (1999). Parametrized Riccati equations and input-output
operators for time-varying stochastic differential equations with state
dependent noise. Studii si Cercetari Matematice, (1).
[153] Naimark, L., Kogan, J., and Zeheb, E. (2000). Stabilizability considera-
tion and design of rational controllers for a class of time-delay systems.
Automatica, 36, 475-480.
[154] Niculescu, S.1. (1998). Hoo memoryless control with an a-stability con-
straint for time-delay systems: an LMI approach. IEEE Transactions on
Automatic Control, 43(5), 739-743.
[155] Niculescu, S.I., de Souza C.E., Dion, J.M., and Dugard, L. (1994). Ro-
bust stability and stabilization of uncertain linear systems with state delay:
single delay case. Proceedings of the IFAC Symposium on Robust Control
Design, Rio de Janeiro.
[156] Niculescu, S.I., de Souza, C.E., Dugard, L., and Dugard, J.M. (1994). R0-
bust stability and stabilization of uncertain linear systems with state delay:
 References 413

single delay case (1). Proceedings of the IFAC Symposium on Robust Control
Design, Rio De Janeiro, pages 469-474.
[157] Niculescu, S.I., Dion, J.M., and Dugard, L. (1996). Robust stabilization
for uncertain time-delay systems containing saturating actuators. IEEE
Transactions on Automatic Control, 41(5), 742-747.
[158J Niculescu, S.I., Verriest, E.I., Dugard, L., and Dugard, J.M. (1997). Sta-
bility of linear systems with delayed state: a guide tour. In Stability and
Control of Time-Delay Systems, L. Dugard and E. I Verriest, Eds, Lecture
Notes in Control and Information Sciences, 228, 1-71, Springer-Verlag,
London,
[159] O'Connor, D.A. and Tarn, T.J. (1983). On stabilization by state feed-
back for neutral differential difference equations. IEEE Transactions on
Automatic Control, 28, 615---618.
[160] Oucheriah, S. (2000). Exponential stabilization of a class of uncertain time-
delay systems with bounded controllers. IEEE Transactions on Automatic
Control, 47(4), 606-609.
[161J Packard, A. (1994). Gain scheduling via linear fractional transformations.
Systems and Control Letters, 22, 79-92.
[162J Packard, A., Zhou, K., Pandey, P., and Becker, G. (1991). A collection ofro-
bust control problems leading to LMls. In Proceedings of IEEE Conference
on Decision and Control, pages 1245-1250.
[163] Packard, A., Zhou, K., Pandey, P., and Leonhardson, J. (1992). Optimal
constant input/output similarity scaling for full-information and state-
feedback control problem. Systems and Control Letters, pages 271-280.
[164J Palhares, R.M., de Souza, C.E., and Peres, P.L.D. (1999). Robust H oo
filter design for uncertain discrete-time state-delayed systems: an" LMI ap-
proach. Proceedings of the 38th IEEE Conference on Decision and Control,
Phoenix.
[165] Palhares, R.M., Peres, P.L.D., and de Souza, C.E. (1999). Delay-dependent
robust Hoo-filtering for convex polytopic uncertain systems via LMI.
In Proceedings of the 38th IEEE Conference on Decision and Control,
Phoenix.
[166] Pila, A.W., Shaked, V., and de Souza, C.E. (1999). H oo filtering
for continuous-time linear systems with delay. IEEE Transactions on
Automatic Control, 44(7).
[167] Pila, A.W., Shaked, V., and de Souza, C.E. (1996). H oo filtering for
continuous-time linear systems with delay. In Proceedings of 13th IFAC
World Congress, San Francisco, USA, pages 49-54.
[168J Rami, M.A. and EI Ghaoui, L. (1996). LMI optimization for nonstan-
dard Riccati equations arising in stochastic control. IEEE Transactions
on Automatic Control, 41(11),1666-1671.
[169] Rami, M.A. and EI Ghaoui, L. (1996). Robust state-feedback stabiliza-
tion of jump linear systems via LMls. International Journal of Robust and
Nonlinear Control, (6), 1015-1022.
[170] Ramos, J.L. and Pearson, A.E. (2000). Output feedback stabilizing
controller for time-delay systems. Automatica, 36, 613-617.
414 References

[171] Razimukhin, B.S. (1960). Application of Lyapunov method to problems in

stability of systems with dealay. Automatica i Telemechanica, 21, 740-749.
[172] Richard, J.P. (1998). Some trends and tools for the study of time-delay
systems. In Proceedings of the IMACS-IEEE Multiconference, CESA '98,
pages 27-43, Hammamet, 'TUnisia.
[173] Rosenblueth, J.F. (2001). Certain classes of properly relaxed delayed
controls. IMA Journal of Mathematical Control and Information, 18,
31-52.
[174] Shaked, D., Yaesh, I., and de Souza, C.E. (1998). Bounded real criteria
for linear time-delay systems. IEEE Transactions on Automatic Control,
43(7), 1016-1022.
[175] Shen, J., Chen, B., and Kung, F. (1991). Memoryless stabilization ofuncer-
tain dynamic delay systems, Riccati equation approach. IEEE Transactions
on Automatic Control, 36, 638-640.
[176] Shi, P. (1996). Robust filtering for uncertain systems with sampled mea-
surements. International Journal of Systems Science, 27(12), 1403-1415.
[177] Shi, P. (1998). Filtering on sampled-data systems with parametric
uncertainty. IEEE Transactions on Automatic Control, 43(7), 1022-1027.
[178] Shi, P. and Agarwal, RK (2001). Robust disturbance attenuation of
fluid dynamics systems with norm-bounded uncertainties. IMA Journal of
Mathematical Control and Information, 18, 73-82.
[179] Shi, P., Agarwal, RK, Boukas, E.K, and Shue, S.P. (2000). Robust Hoc
state feedback control of discrete time-delay linear systems with norm
bounded uncertainty. International Journal of Systems Science, 31(4),
409-415.
[180] Shi, P. and Boukas, E.K (1997). Hoc control for Markovian jumping linear
systems with parametric uncertainty. Journal of Optimization Theory and
Applications, 95(1), 75-99.
[181] Shi, P., Boukas, E.K, and Agarwal, RK, (1999). Control of Markovian
jump discrete-time systems with norm bounded uncertainty and unknown
delays. IEEE Transactions on Automatic Control, 44(11), 2139-2144.
[182] Shi, P., Boukas, E.K, and Agarwal, RK (1999). Kalman filtering for
continuous-time uncertain systems with Markovian jumping parameters.
IEEE Transactions on Automatic Control, 44(8), 1592-1597.
[183] Shi, P., Boukas, E.K, and Agarwal, RK (1999). Robust control for Marko-
vian jumping discrete-time systems. International Journal of Systems
Science, 30(8), 787-797.
[184] Shyu, KK and Chien, YC. (1995). Robust tracking and model following
for uncertain time-delay systems. International Journal of Systems Science,
62(3), 589--600.
[185] Shyu, KK and Yan, J.J. (1993). Robust stability of uncertain time-delay
systems and its stabilization by variable structured control. International
Journal of Control, 57(1), 237-246.
[186] Slemrod, M. and Infante, E.F. (1972). Asymptotic stability criteria for
linear systems of differential equations of neutral type and their discrete
analogies. Journal of Mathematical Analysis and Applications, 38, 399-415.
 References 415

[187] Su, H. and Chu, J. (1999). Robust Hoc control for linear time-varying
uncertain time-delay systems via dynamic output feedback. International
Journal of Systems Science, 30(10), 1093-1107.
[188] Su, H. and Chu, J. (1999). Stabilization of a class of uncertain time-delay
systems containing saturating actuators. International Journal of Systems
Science, 30(11), 1193-1203.
[189] Su, H. (1994). Further results on the robust stability of linear systems with
a single time delay. Systems and Control Letters, 23, 375-379.
[190] Su, T.J. and Huang, C.G. (1992). Robust stability of delay dependence for
linear uncertain systems. IEEE Transactions on Automatic Control, 37,
1656-1659.
[191] Sworder, D.D. (1968). On the stochastic maximum principle. Journal of
Mathematical Analysis and Applications, 24, 627--640.
[192] Sworder, D.D. (1969). Feedback control of a class of linear systems with
jump parameters. IEEE Transactions on Automatic Control, 14(1),9-14.
[193] Sworder, D.D. (1972). Control of jump parameter systems with discontin-
uous state trajectories. IEEE Transactions on Automatic Control, 20(10),
740-74l.
[194] Sworder, D.D. and Chou, D.S. (1985). A survey of some design methods for
random parameter systems. In Proceedings of the 24th IEEE Conference on
Decision and Control, pages 894-898, Ft. Lauderdale, Florida.
[195] Sworder, D.D. and Rogers, R.O. (1983). An LQ-solution to a control prob-
lem associated with a solar thermal central receiver. IEEE Transactions on
Automatic Control, 28(8), 971-978.
[196] Szita, G. and Sanathan, S. (1997). Robust design for disturbance rejection
in time-delay systems. Journal of Franklin Institute, 334B(4), 611-629.
[197] Tadmor, G. (1995). Hoc control in systems with a single input delay. In
Proceedings of ACC, Seattle, Washington, pages 321-325.
[198] Tarbouriech, S. and da Silva, J.M.G. (2000). Synthesis of controllers for
continuous-time delay systems with saturating controls via LMIs. IEEE
Transactions on Automatic Control, 45(1), 105-111.
[199] Trinh, H. (1999). An observer design procedure for time-delay systems.
Control and Intelligent Systems, 27(1), 6-10.
[200] Trinh, H. (1999). Linear functional state observer for time-delay systems.
International Journal of Control, 72(18), 1642-1658.
[201] Trinh, H. and Aldeen, M. (1994). On the stability of linear systems with
delayed perturbations. IEEE Transactions on Automatic Control, 39,1948-
1951.
[202] Trinh, H. and Aldeen, M. (1996). Output tracking for linear uncertain time-
delay systems. In lEE Proceedings D, Control Theory and Applications,
143(6).
[203] Trinh, H., Aldeen, M., and Fernando, T. (1999). An observer for intercon-
nected discrete time-delay systems. Control and Intelligent Systems, 27(2),
70-78.
416 References

[204] Tsay, J.T., Liu, P.L., and Su, T.J. (1996). Robust stability for perturbed
large-scale time-delay systems. In lEE Proceedings D, Control Theory and
Applications, 143(3), 233-236.
[205J Verriest, E.L (1996). Riccati type conditions for robust stability of delay
systems. In Proceedings of MTNS, St. Louis, Missouri.
[206] Verriest, E.L, Fan, K.H., and Kullstam, J. (1993). Frequency domain robust
stability criteria for linear delay systems. In Proceedings of the 32nd IEEE
Conference on Decision and Control, San Antonio, Texas.
[207J Wang, S.S., Chen, B.S., and Lin, T.P. (1987). Robust stability of uncertain
time-delay system. International Journal of Control, 46, 963-976.
[208J Wang, W.J. and Wang, RJ. (1995). New stability criteria for linear time-
delay systems. Control Theory and Advanced Technology, 10(2), 1213-1222.
[209J Wang, Y., Xie, 1., and de Souza, C.E. (1992). Robust control of a class of
uncertain systems. Systems and Control Letters, 19, 139-149.
[210] Won, S. and Park, J.H. (1999). Design of observer-based controller for
perturbed time-delay systems. JSME International Journal, 42(1), 129-
132.
[211] Wonham, W.M. (1971). Random differential equations in control theory.
In Probabilistic Methods in Applied Mathematics, volume 2, pages 131-213,
A.T. Bharucha-Reid, Ed., Academic Press, New York.
[212] Wu, H. and Mizukami, K. (1995). Robust stability criteria for dynamical
systems including delayed perturbations. IEEE Transactions on Automatic
Control, 40, 487-490.
[213] Xiaohong, N. (2000). Robust stability for a type of uncertain time-delay
systems. Applied Mathematics and Mechanics, 21(4), 479-484.
[214] Xie, L. (1996). Output feedback H= control of systems with parameter
uncertainty. International Journal of Control, 63(4), 741-750.
[215] Xie, L. and de Souza, C.E. (1993). Robust stabilization and disturbance
attenuation for uncertain delay systems. In Proceedings of the 193 European
Control Conference, Groningen, The Netherlands.
[216] Xu, B. and Lam, J. (1999). Decentralized stabilization of large scale
interconnected time-delay systems. Journal of Optimization Theory and
Applications, 103(1), 231-240.
[217J Xue, X. and Qiu, D. (2000). Robust H= compensation design for time-delay
systems with norm-bounded time-varying uncertainties. IEEE Transactions
on Automatic Control, 45(5), 606--609.
[218J Yao, YX., Zhang, YM., and Kovacevic, R (1996). Parameterization of
observers for time-delay systems and its application in observer design.
In lEE Proceedings D, Control Theory and Applications, 143(3), 225-232,
1996.
[219J Yao, YX., Zhang, YM., and Kovacevic, R (1997). Functional observer
and state feedback for input time-delay systems. International Journal of
Control, 66(4), 603--617.
[220] Yu, L. and Chu, J. (1999). An LMI approach to guaranteed cost control of
linear uncertain time-delay systems. Automatica, 35, 1139-1159.
 References 417

[221] Yu, L., Chu, J., and Su, H. (1996). Robust memoryless H oo controller design
for linear time-delay systems with norm-bounded time-varying uncertainty.
Automatica, 32(12),1759-1762.
Index

admissible control example, 234, 239

definition, 25 delay-independent stabilization
admissible uncertainties, 22, 179 example, 79
definition, 15, 230 deterministic
time delay, 4
Decomposition Lemma, 393 discrete and distributed time delays
delay-dependent example, 264
'H oo control, 111 disrectization
filtering, 122 example, 65
output feedback, 213 disturbance rejection
robust stability, 79, 241 delay-independent condition, 101,
stability, 39 105,286
stabilizability, 39, 43 example, 105, 139, 148, 154
delay-dependent disturbance rejection formulation, 25
example, 116 maximum delay, 138
delay-dependent robust stability maximum level, 103
example, 83 Dynkin
delay-dependent robust stabilization formula, 402
example, 85, 90
delay-independent feasibility problem
filtering, 118 example, 384, 385
output feedback, 206 Matlab code, 384, 385
robust stability, 71, 231 feasp
robust stabilizability, 71, 231 description, 380
stability, 32, 33 filter
stabilizability, 32, 35 design, 322
delay-independent robust stabilization formulation, 311
420 Index

structure, 312 Kronecker product

filtering definition, 397
definition, 27, 117 elementary matrix, 398
delay-dependent, 122 properties, 397
delay-independent, 118 unit vector, 397
design algorithm, 120 Vee, 398
error, 312
example, 121, 127 lemmas on matrix inequalities,
formulation, 27 391-394
Finsler Lemma, 393 linear state feedback stabilization
forward Chapman-Kolmogorov design algorithm, 370
equation, 400 LMI
definition, 377
generalized eigenvalue problem, 41 example, 377
GEVP feasibility, 384
description, 382 problems, 384
GEVP problem lmisys
description, 388 description, 380
example, 389, 390 lmiterm
Matlab code, 389, 390 description, 379
guaranteed cost lmivar
bound, 162 description, 378
bounded, 344
example, 166, 172 Markov jump systems
formulation, 28, 161 dynamics, 185
output feedback controller, 167 Markovian jump parameters
state feedback controller, 164 definition, 177
guaranteed cost control mat2dec
definition, 343 description, 383
output feedback control, 350 matrix inversion
state feedback control, 346 formula, 395
mean square stability
Hoc definition, 234
norm, 25 delay-independent condition, 235
Hoc control example, 236
definition, 25 mincx
delay-dependent, 111 description, 381
formulation, 25, 100 minimization with LMI constraints
description, 386
infinitesimal operator example, 387
definition, 402 Matlab code, 387
mode-dependent output stabilization
jump linear systems example, 279
definition, 177 mode-dependent stochastic stability
dynamics, 178, 284 example, 226
jump rates, 179 mode-dependent stochastic
probability transitions, 179 stabilization
jump rates example, 228
definition, 400 mode-dependent time delay systems
 Index 421

example, 270, 274 output stochastic stabilization

example, 212, 221
nonlinear feedback stabilization
structure, 371 probability space, 399
nonlinear jump systems
assumption, 362 robust 1£00 control
asymptotic square stability, 362 design, 132
dynamics, 362 formulation, 326
mean square stability, 362 memoryless control, 336
stability condition, 363 norm, 27
nonlinear stability robust 1£00 filtering
stochastic, 363 design, 155
nonlinear stabilization robust delay-dependent stability
example, 375 example, 244
linear state feedback controller, 369 robust delay-dependent stabilization
nonlinear state feedback controller, example, 245
371 robust disturbance rejection
stochastic, 368 delay-independent condition, 133
nonlinear state feedback stabilization memoryless controller, 138
design algorithm, 374 robust internal stability
nonlinear stochastic stability definition, 133
example, 368 robust internal stabilization
norm-bounded uncertainties definition, 133
definition, 70, 179 robust output feedback
form, 229 formulation, 246
robust output feedback 1£00 control
observer-based output stabilization design, 149
example, 56 robust output stabilization
orthogonal complement example, 97, 254
definition, 37 robust stability
output disturbance rejection definition, 25, 70
design algorithm, 109 delay-dependent, 79, 241
example, 110 delay-dependent condition, 80, 241
formulation, 303 delay-independent, 71, 231
output feedback mode-dependent systems, 266
delay-dependent, 213 with discrete and distributed
delay-independent, 206 delays, 256
output feedback 1£00 control robust stabilizability
design, 106 definition, 25
output feedback controller delay-dependent memory state
structure, 304 feedback controller, 86
output feedback disturbance rejection delay-dependent memoryless state
example, 309 feedback controller, 84
output feedback guaranteed cost delay-independent, 71, 231
control delay-independent memory
example, 359 controller, 76
output stabilization delay-independent memoryless
design algorithm, 53, 221 controller, 74
example, 53 robust stabilization
422 Index

definition, 70 stabilizability
delay-dependent condition, 245 definition, 25
delay-dependent memoryless delay-dependent, 39, 43
controller, 244 delay-dependent condition, 200
delay-independent condition, 233 delay-dependent state feedback
discrete and distributed time controller, 43
delays, 263 delay-independent, 32, 35
example, 76 delay-independent output feedback,
memoryless controller, 237 49
mode-dependent systems, 271 delayed state feedback controller,
output feedback control, 274 36
with discrete and distributed deterministic stabilizability, 23
delays, 256 observer-based output feedback, 54
robust stochastic disturbance output feedback, 48
rejection output feedback control, 206
definition, 327 output feedback controller, 91
delay-independent condition, 327 state feedback controller, 36
design algorithm, 340 stochastic, 186
example, 334, 341 with mode-dependent, 222
robust stochastic stability stabilization
delay-independent condition, 231 delay-independent design
robust stochastic stabilization algorithm, 74
definition, 181, 230 example, 45, 48
maximum time delay, 44
Schur complement, 392, 395 state feedback guaranteed cost
setlmis example, 349
description, 380 stochastic
stability time delay, 9
asymptotic stability, 24 stochastic delay-dependent
definition, 23, 24 stabilization
delay-dependent, 39 example, 203
delay-dependent condition, 195 stochastic disturbance rejection
delay-independent, 32, 33 definition, 181, 284
delay-independent condition, 187 delay-dependent condition, 293
delay-independent sufficient design algorithm, 291
condition, 72 example, 289, 292, 302
deterministic stability, 23 maximum delay, 301
discretization method, 57 memoryless controller, 289, 301
example, 35, 42 state feedback controller, 182
internal, 101 stochastic filtering
maximum time delay, 42 example, 323
mean exponential, 190 stochastic process
stochastic, 186, 187 definition, 399
sufficient condition, 33 Markov property, 399
time delay sufficient condition, 39 stochastic robust stability
uniform stability, 24 definition, 230
with delayed control, 203 stochastic stability
with mode-dependent, 222 assumption, 190
zero state equilibrium, 24 definition, 180, 186, 362
 Index 423

delay-dependent condition, 195, uncertain Markov jump linear

198 systems
delay-independent condition, 189 dynamics, 326
example, 190, 192 uncertain time delay systems
mean exponential sense, 186 dynamics, 70, 133
mean exponential stable, 230
mean exponentially stable, 180 Vec
mean square quadratic sense, 191 properties, 398
mean square sense, 186
mean square stable, 180
mode-dependent time delay, 223
stochastic stabilization
control design algorithm, 205
definition, 181
delay-dependent condition, 200
design algorithm, 193, 204
example, 194, 205
maximum delay, 202
mode-dependent time delays, 227
stochastic time delay dependent
example, 199
stochastic uncertainties
norm-bounded, 326

time delay system

abbreviations, 12
classical control, 3
communications, 10
definition, 1
discrete and distributed time
delays, 261
dynamics, 13
failure production systems, 9
feeder in production systems, 8
free, 15
heating system, 6
linear dynamics, 22
metal rolling, 4
multiple delays, 31
nominal, 15
notation, 12
practical examples, 4
stochastic systems with delayed
communications, 11
time-varying, 16
unforced, 15

uncertain jump linear systems

dynamics, 178, 229