Sunteți pe pagina 1din 5

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO.

5, MAY 2014 1355

New Results on Output Feedback Control for Linear sults have been obtained. In [11], based on a special congruence trans-
Discrete-Time Systems formation, sufficient static output feedback control conditions are
presented, which exploits more degrees of the Lyapunov matrix. By
Xiao-Heng Chang, Member, IEEE, and considering a matrix equality condition about the Lyapuov matrix, suf-
Guang-Hong Yang, Senior Member, IEEE ficient LMI conditions are given for designing static output feedback
controllers in [12]. In [13], by introducing auxiliary matrix vari-
ables with diagonal structure, dynamic output feedback controllers
were designed for discrete-time linear systems based on the LMI tech-
Abstract—This note investigates the problem of output feedback
control for linear discrete-time systems. Three types of controllers
nique. For the problem of observer-based control, [14] and [15]
are considered, namely, static output feedback controllers, dynamic output gave LMI-based design conditions by using the equality condition as
feedback controllers, and observer-based output feedback controllers. New [12]. Though these conditions in [12]–[15] are of LMIs, the require-
design conditions for the three type of output feedback controllers are intro- ments for auxiliary matrix variables to be with diagonal structure or
duced in terms of unified linear matrix inequality (LMI) representations, equality conditions are strict, which might result in conservative de-
which guarantee the prescribed performances of the closed-loop sys-
tems. In contrast to the existing LMI conditions for designing the output sign results.
feedback controllers, the improvement of the proposed results over This note studies the problem of three types of output feedback
the existing ones is shown by theoretical proof and numerical example. control for linear discrete-time systems. New design conditions
Index Terms—Discrete-time systems, control, linear matrix inequal- for the three type of output feedback controllers are given by unified
ities (LMIs), output feedback. LMI representations, under which the prescribed performances
of the closed-loop systems are guaranteed. By theoretical proof, one
can be shown that the proposed design conditions include the LMI
I. INTRODUCTION results given in [12]–[15] as special cases. A numerical example is
given which clearly demonstrate the less conservatism of the proposed
During the past two decades, the control problem has attracted design results.
great attention from both academic and industrial communities. A great
number of results on control have been reported in the open lit- Notations
erature. Most of the control studies are given in terms of state In symmetric block matrices, the symbol is used as an ellipsis for
feedback which follow the assumption that the system states are avail- the terms that are introduced by symmetry. is the identity matrix with
able for controller implementations (see, for instance, [1]–[4]). How- appropriate dimension.
ever, the assumption is not true in many practical cases since the states
are often unavailable [5]–[10]. For the output feedback control
problem, it can be considered through three approaches. The first one II. PROBLEM STATEMENT AND PRELIMINARIES
is called static output feedback control. This one is interesting Consider a linear discrete-time system described by the following
to reduce real time computational cost when implementing practical state-space equation:
applications. The second approach to address the problem of output
feedback control is to use a dynamic output feedback compen-
sator. This kind of feedback control is a good way to improve the (1)
closed-loop transient response. Finally, the third way is the so-called
observer-based control. This is based on the introduction of a state
where is the state variable, is the control input,
observer and is interesting when the state is not entirely available from
is the measurement output, is the disturbance,
measurements.
is the controlled output. , , , , , , , and
The output feedback control problems for linear systems have
are system matrices with appropriate dimensions. Without loss of gen-
been extensively studied in the past decades, and many important re-
erality, we assume that is of full column rank through this note (the
same assumption can be found in [12]–[15]).
Manuscript received December 08, 2012; revised June 22, 2013; accepted In this note, we will consider the following three types of output
September 10, 2013. Date of publication November 06, 2013; date of current feedback controllers:
version April 18, 2014. This work was supported by the Funds of National
Science of China under Grants 61104071, 60974043, and 61273148, the A) static controller
Program for Liaoning Excellent Talents in University, China under Grant
LJQ2012095, the Foundation for the Author of National Excellent Doctoral (2a)
Dissertation of PR China under Grant 201157, the Fundamental Research
Funds for the Central Universities under Grant N110804001, and the Funda- where is the controller gain.
mental Research Funds of State Key Laboratory of Synthetical Automation
for Process Industries under Grant 2013ZCX01. Recommended by Associate
B) dynamic controller
Editor T. Zhou. (Corresponding Author: X.-H. Chang.)
X.-H. Chang is with the College of Engineering, Bohai University, Jinzhou, (2b)
Liaoning, 121013, China, and also with the Key Laboratory of Manufacturing
Industrial Integrated Automation, Shenyang University, Shenyang, Liaoning, where is the controller state, , , , and
110044, China (e-mail: changxiaoheng @163.com; changxiaoheng@sina.com). are the controller gains.
G.-H. Yang is with the College of Information Science and Engineering,
Northeastern University, Shenyang, Liaoning, 110004, China, and also with C) observer-based controller
the State Key Laboratory of Synthetical Automation for Process Industries,
Northeastern University, Shenyang, Liaoning, 110004, China (e-mail: yang-
guanghong@ise.neu.edu.cn).
Digital Object Identifier 10.1109/TAC.2013.2289706 (2c)

0018-9286 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1356 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO. 5, MAY 2014

where and are the estimated state and Lemma 4 Dynamic Output Feedback Control [13]: Consider
estimated output, respectively, is the observer gain, is the the system (1) ( ) with the controller (2b) and give a scalar .
controller gain. If there exist matrices , , and such that
By (1) and (2), the closed-loop system is described as

(3)

where is an augmented state vector, , , and are aug-


mented matrices consisting of system matrices, while is a controller
matrix containing only designed controller (observer) gain matrices. (7a)
A) for the static case and
(7b)
(4a)
the prescribed performance is guaranteed with
B) for the dynamic case and
, where , , ,

, , , and is a nonsingular
(4b)
matrix satisfying , .
Lemma 5 Observer-Based Output Feedback Control [14], [15]:
Consider the system (1) ( ) with the controller (2c) and give a
C) for the observer-based case and scalar . If there exist matrices , , , , and such that

(4c)

Our aim is to design the output feedback controller (2) such that
the closed-loop system (3) meets performance requirement. The
following preliminary lemmas are need to derive the new results.
Lemma 1 [16]: For matrices , , and with appropriate (8a)
dimensions, the following inequality holds: (8b)

the prescribed performance is guaranteed with


and .
Lemma 2: For matrices , , , and with appropriate dimensions
and scalar , let there be the following condition
III. MAIN RESULTS
(5a) To solve the output feedback control problem, we recall the
following result which will be needed in proof of our main results.
then, we have Lemma 6 [17]: Consider the closed-loop system (3) and give a
scalar . If there exist matrices and such that
(5b)

Proof: Pre- and post-multiplying (5a) by the full row rank matrix
and its transpose, respectively, the inequality (5b) can be (9)
obtained.
In order to show the contrasts between the existing output feedback
control results and the proposed ones, we also recall the following the prescribed performance is guaranteed.
existing results, which can directly be obtained from [12]–[15]. In the existing results for output feedback control [12]–[15],
Lemma 3 Static Output Feedback Control [12]: Consider the based on the condition in (9), a coupling (i.e., product) term of system
system (1) ( ) with the controller (2a) and give a scalar . If matrices, auxiliary matrix variable (or Lyapuov matrix) and controller
there exist matrices , , and such that matrix is brought, it is difficult to obtain LMI-based design conditions.
To solve the problem, auxiliary matrix variables to be with special
structure or equality conditions have to be used in the above litera-
(6a) ture, it may lead to significant design conservativeness. In this note, the
problem of three types output feedback control for linear discrete-
time systems are studied. Different from the results given in [12]–[15],
(6b) the proposed ones eliminate the coupling term, it makes the LMI for-
mulation of design conditions easier. The new results are given in the
the prescribed performance is guaranteed with . following theorems.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO. 5, MAY 2014 1357

Theorem 1: Consider the closed-loop system (3) and give a scalar then the prescribed performance is guaranteed, where
. For known scalar parameters and , if there exist matrices
, , , , and such that , and are the same as in (10), is

defined in (7a). Furthermore, the controller matrix is given by


.
Proof: Similar to (11)–(13) in the proof of Theorem 1, the in-
equality (16) implies that
(10)
then the prescribed performance is guaranteed, where is
defined in (9) and . Furthermore,
the controller matrix is given by . (17)
Proof: We are about to prove the conclusion using Lemma 6. Sup-
pose that inequality (10) holds. The feasible solution of this inequality
implies that , , and matrices and are nonsingular.
Using Schur complement to (10), which leads to

which is equal to (14).


(11) Remark 1: Theorems 1 and 2 present new results for output feed-
back controllers design of linear discrete-time systems. It is nec-
essary to mention that the proposed design conditions (10) and (16)
where
are of LMIs that can be easily and effectively solved via LMI con-
.
trol toolbox [18]. In addition, in the design procedure, we can search
By Lemma 2 with , , optimal values of the scalar parameters and to reduce the conser-
vatism of the solutions. One way to address the search issue is to first
, and , we
solve the feasibility problem of the LMIs (10) and (16) using LMI con-
have
trol toolbox and obtain a set of initial scalar parameters. Then, applying
a numerical optimization algorithm, such as the program fminsearch in
the optimization toolbox of Matlab, a locally convergent solution to the
problem is obtained. In [19], the algorithm has been used and proved
to be effective.
(12) Remark 2: From (3), we can see that the closed-loop system with
Based on Lemma 1 for , one gives three type of output feedback controllers has been generalized a uni-
fied form, it implies that the proposed conditions in Theorems 1 and 2
are suitable for designing the three type of output feedback con-
trollers. Specifically, the design conditions in Theorems 1 and 2 will be
(13) applied to compare the existing results for different controllers, that is
for the static and observer-based output feedback controllers with
the condition in Theorem 1, while for dynamic output feedback
controllers with the condition in Theorem 2.
By defining , one can readily obtain from (13) that In what follows, we will study the relationships among the proposed
results and the LMI conditions in Lemmas 3–5. The following the-
(14) orem shows that the results suggested herein include the ones given
by [12]–[15] as special cases, it is helpful to obtain a conclusion that
the new results are less conservative than the existing LMI conditions
For a scalar , note that for designing output feedback controllers.
implies that . By (14), one has Theorem 3: If the conditions given in Lemmas 3–5 hold, the condi-
tions in Theorems 1 and 2 also hold.
(15) Proof:
A) For static output feedback control:
For static output feedback control, the design condition (6)
Pre- and post-multiplying (15) by and its transpose, re- in Lemma 3 can be rewritten as follows:
spectively, the inequality in (9) can be obtain.
Theorem 2: Consider the closed-loop system (3) and give a scalar
. For known scalar parameters and , if there exist matrices (18)
, , , , and such that

Since the above condition is satisfied, which implies that


(16) , and

. Then there exist large


1358 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO. 5, MAY 2014

enough and small enough such that the following TABLE I


matrix inequality holds: COMPARISON OF FOR STATIC OUTPUT FEEDBACK CONTROL.

where
(19)
By defining , , , ,
and applying Schur complement, the inequality (10) in Theorem TABLE II
1 can be obtained. COMPARISON OF FOR DYNAMIC OUTPUT FEEDBACK CONTROL.
B) For dynamic output feedback control:
For dynamic output feedback control, pre- and post-multi-
plying (7a) by and its transpose, respec-
tively, we have

(20)

TABLE III
COMPARISON OF OBSERVER-BASED OUTPUT FEEDBACK CONTROL.
where

On the other hand, from (7b) and ,


we can know that IV. EXAMPLE
In this section, a numerical example is presented to illustrate the ef-
fectiveness of the proposed design results. In the example, compar-
ison between our results and the existing ones is given for static output
feedback control, dynamic output feedback control, observer-
based output feedback control, respectively.
Similar to (19), by defining , , Consider a linear discrete-time system in the form of (1) with
, , , and applying Schur com-
plement, the inequality (16) in Theorem 2 can be obtained.
C) For observer-based output feedback control:
For observer-based output feedback control, the design con-
dition (8) in Lemma 5 can be rewritten as a new appearance in
the from (18) with . Then, the proof simply fol-
lows the proof of A) static output feedback control.

Remark 3: In [6], the static output feedback control problem for where a positive number is be chosen to be 0.5, 1.2, 1.7, and 1.9.
linear systems is investigated, where slack matrix variables with lower- For this system, we apply Theorem 1 in this note and Lemma 3 to de-
triangular structure are introduced and LMI-based design conditions sign static output feedback controllers, Theorem 2 in this note and
are given. From the proof of Theorem 3 (part B), we can easily know Lemma 4 to design dynamic output feedback controllers, Theorem
that the design condition in Theorem 2 also includes the result given in 1 in this note and Lemma 5 design observer-based output feedback
[6]. controllers, respectively. Tables I–III summarize the minimum values
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 59, NO. 5, MAY 2014 1359

of the performance by using our design conditions and the [10] J. Lam and S. Zhou, “Dynamic output feedback control of dis-
ones in [12]–[15] for the three type of output feedback controllers. crete-time fuzzy systems: A fuzzy-basis-dependent Lyapunov function
approach,” Int. J. Syst. Sci., vol. 38, no. 1, pp. 25–37, Jan. 2007.
By the comparison in Tables I–III, we can see that the output feed-
[11] G. I. Bara and M. Boutayeb, “Static output feedback stabilization
back controllers design conditions proposed in this note are much with performance for linear discrete-time systems,” IEEE Trans.
less conservative than the results given in [12]–[15]. Autom. Control, vol. 50, no. 2, pp. 250–254, Feb. 2005.
[12] C. A. R. Crusius and A. Trofino, “Sufficient LMI conditions for output
V. CONCLUSION feedback control problems,” IEEE Trans. Autom. Control, vol. 44, no.
5, pp. 1053–1057, May 1999.
In this note, the problem of output feedback control for linear [13] K. H. Lee, J. H. Lee, and W. H. Kwon, “Sufficient LMI conditions
discrete-time systems is studied. New LMI conditions are given, under for output feedback stabilization of linear discrete-time systems,”
which static output feedback controllers, dynamic output feedback IEEE Trans. Autom. Control, vol. 51, no. 4, pp. 675–680, Apr. 2006.
controllers, and observer-based output feedback controllers [14] C.-H. Lien, “Robust observer-based control of systems with state per-
turbations via LMI approach,” IEEE Trans. Autom. Control, vol. 49,
are designed. Theoretic proof is given to show that the proposed results
no. 8, pp. 1365–1370, Aug. 2004.
can include the existing ones as special cases. By a numerical example, [15] C.-H. Lien, “ non-fragile observer-based controls of dynamical
one can see that the new results significantly relax the conservativeness systems via LMI optimization approach,” Chaos, Solitons and Frac-
of the existing ones. Furthermore, the developed results in this note can tals, vol. 34, no. 2, pp. 428–436, Oct. 2007.
be used to handle other output feedback control design problems [16] K. Zhou and P. P. Khargonekar, “Robust stabilization of linear systems
with norm-bounded time-varying uncertainty,” Syst. Control Lett., vol.
such as robust control [20]–[24] and non-fragile control [25]–[28] for 10, no. 1, pp. 17–20, Jan. 1988.
linear discrete-time systems. [17] S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
Inequalities in Systems and Control Theory. Philadelphia, PA: SIAM,
ACKNOWLEDGMENT 1994.
[18] P. Gahinet, A. Nemirovski, A. J. Laub, and M. Chilali, LMI Control
The authors would like to thank the reviewers for their helpful com-
Toolbox. Natick, MA: The MathWorks, 1995.
ments. [19] L. Xie, L. Lu, D. Zhang, and H. Zhang, “Improved robust and
filtering for uncertain discrete-time systems,” Autom., vol. 40, no. 5,
REFERENCES pp. 873–880, May 2004.
[1] L. Xie and C. E. de Souza, “Robust control for class of uncertain [20] D. C. W. Ramos and P. L. D. Peres, “An LMI condition for the robust
linear time-invariant systems,” IEE Proc. Control Theory Applicat., stability of uncertain continuous-time linear systems,” IEEE Trans.
vol. 138, no. 5, pp. 479–483, Sep. 1991. Autom. Control, vol. 47, no. 4, pp. 675–678, Apr. 2002.
[2] P. Gahinet and P. Apkarian, “A linear matrix inequality approach to [21] D. C. W. Ramos and P. L. D. Peres, “A less conservative LMI condi-
control,” Int. J. Robust Nonlin. Control, vol. 4, pp. 421–448, 1994. tion for the robust stability of discrete-time uncertain systems,” Syst.
[3] S. Xu and C. Yang, “ state feedback control for discrete singular Control Lett., vol. 43, no. 5, pp. 371–378, Aug. 2001.
systems,” IEEE Trans. Autom. Control, vol. 45, no. 7, pp. 1405–1409, [22] V. J. S. Leite and P. L. D. Peres, “An improved LMI condition for ro-
Jul. 2000. bust -stability of uncertain polytopic systems,” IEEE Trans. Autom.
[4] I. Yaesh, S. Boyarski, and U. Shaked, “Probability-guaranteed robust Control, vol. 48, no. 3, pp. 500–504, Mar. 2003.
performance analysis and state-feedback design,” Syst. Control [23] L. Lu, R. Yang, and L. Xie, “Robust and control of discrete-
Lett., vol. 48, no. 5, pp. 351–364, Apr. 2003. time systems with polytopic uncertainties via dynamic output feed-
[5] M. C. De Oliveira, J. C. Geromel, and J. Bernussou, “Extended back,” in Proc. Amer. Control Conf., Jun. 2005, pp. 4315–4320.
and norm characterizations and controller parametrizations for [24] J. C. Geromel, R. H. Korogui, and J. Bernussou, “ and robust
discrete-time systems,” Int. J. Control, vol. 75, no. 9, pp. 666–679, Jun. output feedback control for continuous time polytopic systems,” IET
2002. Control Theory Appl., vol. 1, no. 5, pp. 1541–1549, Sep. 2007.
[6] J. Dong and G.-H. Yang, “Static output feedback control synthesis [25] L. H. Keel and S. P. Bhattacharyya, “Robust, fragile, or optimal?,”
for linear systems with time-invariant parametric uncertainties,” IEEE IEEE Trans. Autom. Control, vol. 42, no. 8, pp. 1098–1105, Aug. 1997.
Trans. Autom. Control, vol. 52, no. 10, pp. 1930–1936, Oct. 2007. [26] L. Li and Y. Jia, “Non-fragile dynamic output feedback control for
[7] C. Lin, Q.-G. Wang, T. H. Lee, Y. He, and B. Chen, “Observer-based linear systems with time-varying delay,” IET Control Theory Appl.,
fuzzy control design for T-S fuzzy systems with state delays,” vol. 3, no. 8, pp. 995–1005, Aug. 2009.
Autom., vol. 44, no. 3, pp. 868–874, Mar. 2008. [27] X.-H. Chang, Takagi-Sugeno Fuzzy Systems Non-Fragile H-Infinity
[8] C. Scherer, P. Gahinet, and M. Chilali, “Multiobjective output-feed- Filtering. Berlin, Heidelberg: Springer, 2012, pp. 1–163.
back control via LMI optimization,” IEEE Trans. Autom. Control, vol. [28] X.-H. Chang and G.-H. Yang, “Nonfragile filtering of continuous-
42, no. 7, pp. 896–911, Jul. 1997. time fuzzy systems,” IEEE Trans. Signal Process., vol. 59, no. 4, pp.
[9] J. Yu, J. Tan, H. Jiang, and H. Liu, “Dynamic output feedback control 1528–1538, Apr. 2011.
for Markovian jump systems with time-varying delays,” IET Control
Theory Appl., vol. 6, no. 6, pp. 803–812, Apr. 2012.

S-ar putea să vă placă și