AASRI Procedia 3 ( 2012 ) 2 7

2212-6716 2012 The Authors. Published by Elsevier B.V.

Selection and/or peer review under responsibility of American Applied Science Research Institute
doi: 10.1016/j.aasri.2012.11.002


2012 AASRI ConIerence on Modelling, IdentiIication and Control
The Non-Iragile Controller Design Based on Quadratic
PerIormance Optimization
Yang Changwei
a
, Chen Jie
b
*
a
The 365 institute, Northwestern Polvtechnical Universitv, Peoples Republic of China
b
School of Aeronautics, Northwestern Polvtechnical Universitv, Peoples Republic of China

Abstract
This paper presents a non-Iragile controller design method based on system quadratic perIormance optimization. For the
additive controller gain variations, the necessary and suIIicient conditions Ior the existence oI non-Iragile state Ieedback
controller are given and transIormed to the LMI problems, which simpliIies the solutions to obtain non-Iragile state
Ieedback controllers. The Ilight control simulation results prove the reliability and validity oI the method.

2012 The Authors. Published by Elsevier B.V.
Selection and/or peer review under responsibility oI American Applied Science Research Institute

Kevwords: non-Iragile control, robust control, quadratic perIormance optimization, Linear Matrix Inequality (LMI);
1. Introduction
System modelling can never describes the engineering plant exactly, and lots oI scholars showed interesting
on the system optimal problems with system and external uncertainty |ZHOU et al., 1996, YU et al., 2002,
WU et al., 2006|. The Guaranteed Cost Control presented by Chang |Sheldon et al., 1972| tackled this
problem, and can be seemed as the linear quadratic optimal problem`s extension to system uncertainty. But
such Ieedback controller which tackles the uncertainty oI controlled plant and the closed-loop perIormance


* Corresponding author. Tel.: 13389254083; Iax: 02988492344.
E-mail address. shuimujiemail.nwpu.edu.cn.
Available online at www.sciencedirect.com
2012 The Authors. Published by Elsevier B.V.
Selection and/or peer review under responsibility of American Applied Science Research Institute
3 Yang Changwei and Chen Jie / AASRI Procedia 3 ( 2012 ) 2 7

optimization needs to be implemented exactly. This demand seems validity, but actually, because oI the
inevitable uncertainty in advanced controller operation, it`s impossible in the actual engineering. The
controller uncertainty is composed by environment temperature variation, the digital device`s word capacity,
data transIorm error, etc.
As the analysis and synthesis oI closed system which is composed oI plant and controller, the robust control
approach has the consideration oI the control Iragile problem. And the IruitIul results oI robust control theory
are helpIul to tackle the non-Iragile problem. Theoretical research and engineering practice show that, iI we
divide the uncertainty oI plant and controller, the robust and non-Iragile are interactive. When the robust Irom
the closed system to the uncertainty is stronger, then the closed system is Iragile or sensitive to the uncertainty
oI controller. Keel et al., 1997 concludes that the Iragile oI controller will be worse as the order oI controller
become higher. So it`s important to consider the controller`s uncertainty during the controller design and
synthesis, e.g. non-Iragile control.
Over recent years, many remarkable achievements have been achieved in developing advanced non-Iragile
controller |Famularo et al., 1998, Yang et al., 2000|, but the common approaches with operator theory or
Riccati inequality are complex. According to the additive controller gain uncertainty, this paper presents the
non-Iragile controller synthesis problem based on the optimal perIormance control, which controller will be
tackled using the LMI toolbox in Matlab

soItware. A mathematical simulation oI Ilight controller design is

presented to show the validity oI the method above.
2. Problem statements
Lemma 1 |WU et al., 2006| For given proper dimension matrixY , Dand E , in which the Y is symmetric,
0
T T T
Y DFE E F D + + < ,
is existed Ior all the matrix F which satisIy
T
F F I < , iI and only iI there exists the constant 0 > and
1
0
T T
Y DD E E

+ + < .
Lemma 2 |WU et al., 2006| With the given symmetric matrix
11 12
12 22
T
S S
S
S S

=


,
in this matrix,
11
S ,
22
S is nonsingular invertible matrix, then the three conditions below is equivalent:
(1) 0 S < ;
(2)
11
0 S < ,
1
22 12 11 12
0
T
S S S S

<
;
(3)
22
0 S < ,
1
11 12 22 12
0
T
S S S S

< ;
Consider linear continuous time-invariable equation
( ) ( ) ( ) ( ) ( ) x t A A x t B B u t = + + + , (1)
where ( ) x t

is the state vector, ( ) u t

is the control input vector, A, and B is the matrices with

proper dimension. The matrices uncertainty satisIy
[ [ [ [
1 2
( ) ( ) ( ) A t B t DF t E E ^ ^ =
. (2)
The optimal controller ( ) ( ) u t Kx t makes the perIormance index
( ) ( ) ( ) ( )
T T
0
J x t Qx t u t Ru t dt

l
=
l
l
]
, (3)
reach the minimum.
For the given quadratic perIormance index and system uncertainty, the closed system with state Ieedback
controller ( ) ( ) u t Kx t is optimal, iI the symmetric matrix P and K exist which satisIy
4 Yang Changwei and Chen Jie / AASRI Procedia 3 ( 2012 ) 2 7


1 2 1 2
( ) ( )
T
T
P A BK DF E E K A BK DF E E K P K RKQ0
, (4)
and then system perIormance upper bound is
0 0
T
J x Px

.
Consider the controller has parameter perturbation
0 0
( )
k k k
K K KK D F t E , (5)
the non-Iragile controller design approach with optimal perIormance can be given as Iollows.
3. Main results
3.1. Non-fragile controller design with robust performance optimal
Theorem 1 For system (1) with uncertainty (2), the necessary and suIIicient condition to the optimal
perIormance index (3) with non-Iragile Ieedback control strategy is that, iI there exists positive deIine matrix
X , W , and real constant 0 > that satisIy the LMIs below
1 2
1 2 2
1
1
2
(AX BW) ( ) ( )
( )
0
( ) ( )
T T T T T
k k
T
k
k
T T T
k k k
k
AX BW DD E X E W X W BD XE
E X E W I 0 0 E D 0
X 0 Q 0 0 0
W 0 0 R D 0
BD E D 0 D I 0
E X 0 0 0 0 I

+ + + + +

+

<

(6)
Moreover, iI the Ieasible solution Ior the inequality (6) above exists, then the control strategy
1
( ) ( ) u t WX x t

= is the controller Ior the system (1), and the perIormance upper bound is
0 0
T
J x Px

.
Proof. According to the Eq. (4), the necessary and suIIicient condition Ior minimizing the perIormance index
(3) with system uncertainty is
1 2 1 2
( ) ( ) ( )
T T T
Y PDF E E K E E K F PD + + + +
1 2 1 2
( ) ( ) ( )
T T
Y PD PD E E K E E K 0 + + < ,
where ( ) ( )
T T
YQ K RK P A BK A BK P + + + + + .
1 2
1 2
1
1
( ) ( ) ( )
( )
T T T T
P A BK A BK P PDD P E E K I K
E E K I 0 0
M 0
I 0 Q 0
K 0 0 R

+ + + +

+

= <

Substitute the Eq. (3) into the equation above, then we can get the results by Lemma 1
1 2 0 0
1 2 0
1
1
0
( ) ( ) ( )
( )
T T T T
P A BK A BK P PDD P E E K I K
E E K I 0 0
M
I 0 Q 0
K 0 0 R

+ + + +

+

=

[ ]
2
2
( ) ( )
k k
k T T T T
k k k k k k
k k
PBD E
E D 0
F E 0 0 0 F PBD E D 0 D
0 0
D D



+ +





5 Yang Changwei and Chen Jie / AASRI Procedia 3 ( 2012 ) 2 7

0 0 1 2 0 0
1 2 0
1
1
0
( ) ( ) ( )
( )
T T T T
P A BK A BK P PDD P E E K I K
E E K I 0 0
I 0 Q 0
K 0 0 R

+ + + +

+

[ ]
2 2 1
T
T
k k k
k k
k
k k
PBD PBD E
E D E D 0
E 0 0 0
0 0 0
D D 0

+ +




By the Lemma 2, M 0 < is equal to
0 0 1 2 0 0
1 2 0 2
1
1
0
2
( ) ( ) ( )
( )
( ) ( )
T T T T T
k k
k
k
T T T
k k k
k
P A BK A BK P PDD P E E K I K PBD E
E E K I 0 0 E D 0
I 0 Q 0 0 0
0
K 0 0 R D 0
PBD E D 0 D I 0
E 0 0 0 0 I

+ + + + +

+

<

.
Multiply the inequality above right and leIt with
{ }
1
, , , , , diag P I I I I I

respectively.
0 0 1 2 0 0
1 2 0 2
1
1
0
2
( ) ( ) ( )
( )
( ) ( )
T T T T T
k k
k
k
T T T
k k k
k
A BK X X A BK DD X E E K X XK BD XE
E E K X I 0 0 E D 0
X 0 Q 0 0 0
0
K X 0 0 R D 0
BD E D 0 D I 0
E X 0 0 0 0 I

+ + + + +

+

<

.
Let
1
X P

= ,
0
K X W = , then the Eq. (6) can be got. II the Ieasible solution to the LMIs (6) is existed, the
control strategy
1
( ) ( ) ( ) u t W X x t

is the non-Iragile state Ieedback controller.
3.2. Non-fragile controller design based on Linear Quadratic Regulator (LQR)
II the system uncertainty doesn`t exist, then this problem regresses to the non-Iragile control with
conventional linear quadratic perIormance index, e.g.
A 0 , B 0
Theorem 2 For system (1), the necessary and suIIicient condition to the optimal perIormance index (3) with
non-Iragile Ieedback control strategy is that, iI the positive deIine matrix X , W exist, and real constant 0 >
that satisIy the Iollowing LMIs
1
1
( ) ( )
0 0 0
0 0 0
( ) 0 0
0 0 0
T T T
k k
T
k
T T
k k
k
AX BW AXBW X W BD XE
X Q
W R D
BD D I
E X I

+ +

<

, (7)
6 Yang Changwei and Chen Jie / AASRI Procedia 3 ( 2012 ) 2 7

then
1
K KW

is the Ieedback control solution Ior control system (1).

Proof: According to the Eq. (4), the necessary and suIIicient condition Ior minimizing the perIormance index
(3) with system uncertainty is
1
1
( ) ( )
T T
0 0 0
0
A BK PP A BK I K
L I Q 0
K 0 R

+ +

[ ]
( )
( ) ( ) ( ) ( )
T
k k
T T T
k k k k k
k
PBD E
0 F t E 0 0 0 F t PBD 0 D
D 0




+ +

1
1
( ) ( )
0 0 0
0 0 0
( ) 0 0
0 0 0
T T T
0 0 0 k k
0 k
T T
k k
k
ABK P P A BK I K PBD E
I Q
K R D
PBD D I
E I

+ +

<

.
Multiply the inequality above right and leIt with
{ }
1
, , , , diag P I I I I

respectively.
1
1
( ) ( )
0 0 0
0 0 0
( ) 0 0
0 0 0
T T T
0 0 0 k k
0 k
T T
k k
k
X ABK A BK X X XK BD XE
X Q
K X R D
BD D I
E X I

+ +

<

.
Let
1
X P

= ,
0
K X W = , then the Eq. (7) can be got.
4. Simulation examples
Consider the Iollowing Ilight model
( ) ( ) ( ) x t Ax t Bu t = + , (8)
0.0316 6.9739 1.0087 9.8036
0.002 0.6256 0.9128 0
0.0027 0.4 0.8845 0
0 0 1 0
A




=



,
0
0.0873
4.1063
0
B

=



,
the corresponding system states are: Ilight velocity J , attack angle , pitch rate q , and pitch angle ,
aircraIt system input is elevator angle
e
.
System controller addictive gain perturbation can be written as the Eq. (5)
[ ] 1
k
E =
,
[ ] 0.1 0.1 0.1 0.1
k
F =
,
( ) 1
k
t

Using the inequality (7) above, the controller can be solved as
[ ] 2.3685 83.6093 3.536 45.6491 K =

For examining the control strategy, the Ilight system response with controller is given. Fig. 1 to 4 show
the state curves with perturbation.
7 Yang Changwei and Chen Jie / AASRI Procedia 3 ( 2012 ) 2 7


Fig. 1. (a) Ilight velocity V response curves; (b) angle oI attack u response curves

Fig. 2. (a) pitch rate q response curves; (b) pitch angle 0 response curves
5. Conclusions
Using the Quadratic perIormance optimization method, the control stability problem with controller
additive gain perturbation is discussed, and then the stability theorem and state Ieedback control approach are
derived in this paper. In Section 4, the Ilight system perIormance oI aircraIt has been veriIied by mathematical
simulation to validate the controller design approach.
References
|1| Zhou K, Doyle J C, Glover K. Robust and optimal control. New Jersey: Prentice Hall Englewood CliIIs, 1996.
|2| Sheldon. S. L. Chang. Adaptive Guaranteed cost control oI systems with uncertain parameters. IEEE Trans
on Automatic control, 1972;17:474-483.
|3|L.H. Keel, S .P. Bhattacharyya, Robust, Fragile or Optimal? Proceedings oI the American Control
ConIerence Albuquerque, New Meixco, June 1997:1307-1313.
|4|D. Famularo, P. Dorato, C. T. Abdallah, W. M. Haddad, and A. Jadbabaie. Robust non-Iragile LQ
controllers: the static state Ieedback case. International Journal oI control, 1998, 1109-1113.
|5| G.H. Yang, J.L. Wang, C Lin. H control Ior linear systems with additive controller gain variations,
International Journal oI control, 2000;73:1500-1506.
|6| YU Li. Robust control. Beijing: tsinghua university press, 2002.
|7| WU Ming, GUI Wei-hua, HE Yong. Modern robust control. Changsha: Central South university press,
2006.