See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261132019

The practice of MATLAB simulation in modern control theory course teaching

Conference Paper · October 2012

DOI: 10.1109/ICACI.2012.6463300

CITATIONS READS
4 255

2 authors, including:

Liu Lina
Shandong University of Technology
10 PUBLICATIONS   20 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Liu Lina on 07 September 2016.

The user has requested enhancement of the downloaded file.

2012 IEEE fifth International Conference on Advanced Computational Intelligence(ICACI)
October 18-20, 2012 Nanjing, Jiangsu, China

The Practice of MATLAB Simulation in Modern Control Theory

Course Teaching
Lina Liu and Jian Hu

Abstract—Modern control theory is one of the basic Simulink is an important one of the components, which
professional course for the senior undergraduates, such as provides an integrated environment to the user, such as
automation, electrical engineering and automation specialties. It dynamic systems modeling, simulation and comprehensive
has the characteristics of higher theoretical definition and analysis. In this environment, the user does not need to write a
abstraction, bigger calculated amount, degree of depth and large number of procedures, but only to operate the mouse
complexity, and the students should have well-knit
simply, select appropriate library modules. Then complex
mathematical foundation. The study can be made more
interesting and challenging by introducing the simulation simulation model can be constructed. Simulink has many
technology into course teaching. By applying MATLAB advantages, such as wide adaptability, structure and
language and Simulink simulation tools, combining with the procedure clearer, more precise simulation etc [3], [4].
teaching examples, the system synthesis design problem in In the teaching process, graphical and interactive of the
modern control theory course can be dealt with very easy. A main teaching contents maybe realized easily through the
closed-loop control system with full-order state observer is organic combination of MATLAB language, Simulink and
designed. The design process shows that fundamental changes in the multimedia courseware of modern control theory. To do
modern control theory course teaching have been made by this, the teaching process will be more vivid and visual, and
MATLAB language and Simulink tools. The teaching process is
the learning interest is excited. Meanwhile, it is helpful to
converged to the combination of practical engineering case and
vivid teaching from original dry pure theory teaching. strengthen the comprehension of abstract theory knowledge,
obtain some engineering practice knowledge, and enhance
I. INTRODUCTION analysis ability of the students. In this paper, MATLAB
language and Simulink simulation tools are briefly introduced
M ODERN control theory is one of the basic professional
courses for the senior undergraduates, such as
automation, electrical engineering and automation
into the control system design applications. The practice
proved that fundamental changes have been made by
MATLAB language and Simulink tools in modern control
specialties. It is also suitable for some specialties as a public theory course teaching. The teaching process is converged to
technology course, such as the electrical information class, the combination of practical engineering case and vivid
electro-mechanical class and the economics and management teaching from original dry pure theory teaching.
specialties and so on. The basic task of this course is studying
the methods of linear system analysis and synthesis. It has the II. SYNTHESIS DESIGN OF MODERN CONTROL THEORY
characteristics of higher theoretical definition and
abstraction, bigger calculated amount, degree of depth and The study object of modern control theory is linear system,
complexity, and the students should have well-knit the study purpose is to construct a state space model of the
mathematical foundation [1], [2]. Due to the complexity of linear system firstly, and then analysis and synthesis the
matrix operation and the abstraction of course, it is difficult linear system. The mainly contents are time analysis ǃ
for the students to understand and mastery the contents, even Lyapunov stability analysisǃcontrollability and observability
produce disgusted mood. By introducing the simulation analysisǃstate feedback and state observer design and so on.
software MATLAB into the modern control theory course In modern control theory, two ways are mainly used to
teaching, it maybe helpful to solve these problems. improve the system performance. One is state feedback, the
MATLAB is the abbreviation of Matrix Laboratory, and is other is output feedback. The attributes and feedback function
a business mathematics software of America MathWorks of state feedback are higher than these properties of output
company. It is commonly used in algorithm development, feedback. Therefore, the state feedback method is usually
data visualization, data analysis, numerical calculation and so used to improve the system performances during system
on, mainly contains two parts: MATLAB and Simulink. Now synthesis design. For a completely controllable linear
it has become one of the basic mathematical tools of time-invariant system, the poles of closed loop system can be
automatic control, dynamic simulation, modern control arbitrary assigned via state feedback, so that the desired
theory, intelligent control and other advanced courses. dynamic characteristics can be obtained. This process is
called a poles placement problem. However, in commonly
cases, the system state is not necessarily measurable. This
L Liu, corresponding author, is with the Department of Electrify and makes the realization of state feedback is difficult in physics.
Electronic, Shandong University of Technology, Zibo, 255049,
Shandong,PRC(Phone: 13561645280, E-mail: [email protected]) Therefore, the system states are needed to be reconstructed
J Hu, is with the Department of Electrify and Electronic, Shandong through some known variables. This process is called a state
University of Technology, Zibo, 255049, Shandong, PRC (E-mail: observer design problem. In the traditional teaching process,
[email protected])

978-1-4673-1744-3/12/$31.00 ©2012 IEEE 896

the two methods above need to be analyzed in details, the Where, Xˆ (t ) is the estimate of the actual state X (t ) , and
workload is big. But, the poles placement problem and state
the feedback gain vector is shown in (7).
observer design problem can be realized conveniently by
applying the place function or the Acker function after L = [l1 l2 " ln ]T
(7)
introduce MATLAB software. Meanwhile the designed
closed-loop system can be simulated visually by Simulink C. State Feedback Control System With State Observer
simulation tools. It is clear that if a system is controllable and the states are
available for feedback, then the poles of the system can be
A. Poles Placement Problem placed arbitrarily by introducing a constant feedback control.
The distribution situation of closed-loop poles determines However, in most practical applications, the states are not
the system stability and dynamic quality. Therefore, in the completely accessible and all the designer knows are the
system design, the expected closed-loop poles can be output y (t ) and input u (t ) . A closed-loop state feedback
determined firstly according to the requirement of system control system is shown in Fig. 1, in which the state is
performance. Then the poles of closed-loop system would be estimated by a full-order observer.
assigned at desired positions via state feedback. The poles
placement problem can be described as follows.
Consider the LTI system which is described by
­ X (t ) = AX (t ) + bu (t ) (1)
®
¯ y (t ) = cX (t )
if we introduce the state feedback control
u (t ) = − KX (t ) + r (t ) (2)
into the open-loop system (1), the eigenvalues, or poles, of
the closed-loop system

­ X (t ) = ( A − bK ) X (t ) + bv(t ) Fig. 1. Closed-Loop System via Estimated State Feedback

® (3)
¯ y (t ) = cX (t ) Where, the feedback control law is designed by the
can also be placed arbitrarily on the expected locations, estimated state, it is
s1 , s2∗ , " , sn∗ , and the state feedback gain vector can be
∗
u (t ) = − KXˆ (t ) + v(t ) (8)
constructed as Obviously, the dimension of the closed-loop system in
K = [k1 k2 " kn ] (4) Fig.1 is 2n, i.e. the system has n actual state variables and n
the transfer function of it is shown in (5). estimated state variables. The system can be described by (9).
G ( s ) = c[ sI − ( A − bK )]−1 b (5) ­ª X (t )º ª A − bK º ª X (t ) º ªbº
°« »=« » « ˆ » + « » r (t )

B. Full-order State Observer Design Problem °°«¬ Xˆ (t )»¼ ¬ Lc A − Lc − bK ¼ ¬ X (t ) ¼ ¬b¼
For a completely controllable system, the closed-loop
®
° ª X (t ) º
poles can be placed arbitrary. But due to the states can not be ° y (t ) = [c 0]« ˆ »
measured directly, or the restriction of measuring equipment ¯° ¬ X (t ) ¼ (9)
in economy and fitness of use, the state feedback is difficult to According to the separation principle, the transfer function
realize in physics. One way of overcoming this difficulty is to of system (9) is equal to the transfer matrix of the state
reconstruct the system state. And then a desired state feedback system without using the observer, such as (6). and
feedback can be realized by replacing the real states with the it can be calculated as (10).
reconstruction states. The essence of state reconstruction is to
construct a new system, let the measurable input variable G ( s ) = c[ sI − ( A − bK )]−1 b (10)
u (t ) and output variable y (t ) as its input signals, make its In order to make the dynamic of the state error converge to
zero as well as much faster than the closed-loop dynamic, the
output signal Xˆ (t ) equal to the state X (t ) of original system
expected poles of observer are often chosen as (11).
in some conditions. The output signal Xˆ (t ) is called a Re[λ*i ( A − Lc)] ≤ (2 ~ 3) Re[λ*i ( A − bK )] < 0 (11)
reconstructed state or state estimated value of the original
state X (t ) , and the new system is called a state observer. Where, λ*i ( A − bK ) are the expected poles of the
Assume the states, which are used to feedback above, are close-loop system by introducing the state feedback control.
impossible to be measured. A full-order observer can be
designed to estimate them. The reconstructed system can be III. DESIGN AND SIMULATION EXAMPLE OF CLOSED-LOOP
described by (6). CONTROL SYSTEM WITH STATE OBSERVER
­° Xˆ (t ) = ( A − Lc) Xˆ (t ) + bu (t ) + Ly (t ) A open-loop control system is shown in Fig.2.
®
°̄ yˆ (t ) = cXˆ (t )
(6)

897
 According to the constraint relationship (11) between the
observer poles and the expected poles of closed-loop system,
the observer poles can be set to be s1, 2 = −21 .
Fig. 2. black diagram of a control system

Please design a closed-loop control system with a C. MATLAB Design Of The State Feedback Control Law
full-order state observer, the performance indexes which And Observer
should be satisfied are listed as follows: Input the instructions at the MATLAB command window:
1) Dynamic performance indexes: overshoot: σ % ≤ 5% ; a) % whether the system poles can be placed arbitrarily?
peak time: t p ≤ 0.5s ; band width: ωb ≤ 10rad / s . >>A=[0 1;0 -5];b=[0 1]';
>>Qc=ctrb(A,b); nc=rank(Qc)
2) Steady state performance index: static position error : nc =
e p = 0 (for the step input). 2
The design and simulation process of the closed-loop % poles placement is realized by applying place function
control system with a full-order state observer is illustrated as >>P=[- sqrt(50)+sqrt(-50);- sqrt(50)-sqrt(-50)];
follows. >>K=place(A,b,P)
K=
A. Determine The State Space Description Of Original 100.0000 9.1421
System % inspect whether the poles of the new system are the
According to the system block diagram, we can obtain expected closed-loop poles
1 >>Ac=A-b*K;
X 1 ( s) = X 2 ( s)
s >>D=eig(Ac)
1 D=
X 2 ( s) = U ( s) -7.0711 + 7.0711i
s+5 -7.0711 - 7.0711i
Y ( s) = X 1 (s ) From the results, we can see that the poles of the new
Rewriting them and by introducing inverse Laplace system are the expected poles. Thus, the state feedback gain
transform we can obtain that matrix is
­ x1 = x2 K = [100 9.14]
®
¯ x2 = −5 x2 + u By introducing the state feedback control law (2), the poles
y = x1 of the closed-loop system can be placed at the expected
locations, − 7.07 + j 7.07,−7.07 − j 7.07 .
The differential equations above are expressed in matrix
form notation as b) % whether a full-order observer can be designed?
>>A=[0 1;0 -5];b=[0 1]';c=[1 0];
ª x1 (t ) º ª0 1 º ª x1 (t ) º ª0º
« x (t )» = «0 − 5» « x (t )» + «1»u (t ) >>Qo=obsv(A,c); nc=rank(Qo)
¬ 2 ¼ ¬ ¼¬ 2 ¼ ¬ ¼ nc =
ª x (t ) º 2% calculate the feedback gain matrix by applying acker
y = [1 0]« 1 » function
¬ x2 (t ) ¼ >>P1=[-21 -21]; A1=A';b1=c'; c1=b';
This is the state space description of the original system. >>K=acker(A1,b1,P1);
B. Determine The Expected Closed-loop Poles >>L=K'
>>ALc=A-L*c
In order to satisfy the demand of overshoot: σ % ≤ 5% , L=
optimal damping ratio ξ = 0.707 is chosen, and σ = 4.3% 35
satisfies the demand about overshoot. 225
ALc =
Due to the demand of system band width: ωb ≤ 10rad / s ,
-35 1
ωb = ωn 1 − 2ξ 2 + 2 − 4ξ 2 + 4ξ 4 -225 -5
and , substitutes ξ = 0.707 Thus, the feedback gain matrix of the full order observer is
into it, ωn = ωb ≤ 10rad / s can be deduced. ª 35 º
L=« »
ωn = 10rad / s is selected. ¬225¼
π and the full order observer is
Here, peak time t p = ≈ 0.45s satisfies the 
ωn 1 − ξ 2 Xˆ (t ) = ( A − Lc) Xˆ (t ) + bu (t ) + Ly (t )
demand of peak time t p ≤ 0.5s . ª − 35 1ºˆ ª0 º ª 35 º
=« » X (t ) + « »u (t ) + « » y (t )
The expected poles of the closed-loop system are ¬− 225 − 5¼ ¬1 ¼ ¬225¼
s1, 2 = −ξωn ± jωn 1 − ξ 2 = −7.07 ± j 7.07 Block diagram of the closed-loop system with a full-order
. state observer is shown in Fig.3.

898
 Fig. 5. The Unit Step Response Curve of the Closed-loop System

Form Fig.5, we can see that the overshoot, peak time and
static position error of the closed-loop system with the
full-order state observer all satisfy the design demands, the
expected design target is achieved. The example above is a
typical system synthesis design problem by using MATLAB
and Simulink simulation tools. It can be seen that the poles
placement and observer design process can be vividly and
visually seen through the simulation tools. Meanwhile, the
unit step response curve can be mapped directly. And then,
from the given unit step response curve, various performance
indexes can be directly red out, tedious computation is
avoided, enthusiasm of students and teaching quality are
Fig. 3. Block Diagram of the Closed-loop System with greatly improved.
a Full-order State Observer

D. Simulation Calculation By Applying Simulink Tools IV. CONCLUSION

Open the model editing window, pull all the needed The applications of MATLAB in modern control theory
modules from the related modules to this window, connect teaching are not limited to the synthesis design problem, more
them and modify the module parameters, the simulation additional applications will be infiltrated in the teaching of
diagram of the closed-loop system with a full-order state every chapter. Through creation of teaching situations, using
observer is obtained, as shown in Fig.4. Double-click the MATLAB for the curriculum design and typical problems
scope module, and the simulation results can be got. At the teaching, the quality of teaching is improved, the students'
same time, input the following instructions at the command learning initiative is strengthened, comprehensive practical
window: ability and creative thinking ability are trained,
>>set(0,'ShowHiddenHandles','On') comprehensive quality is improved, thus the teaching goal is
>>set(gcf,'menubar','figure') achieved. To make this course become fully reflect the
the hidden graphics editing tools is opened, edit the graphic modern control law and characteristic, closely combining
and the unit step response curve can be got, as shown in Fig.5. with engineering practice, to cultivate "application type,
compound talents". The course is one of the Shandong
University of Technology class excellent courses and
provincial excellent courses.

REFERENCES
[1] S. Y. Zhang, L. Q. Gao. Morden control thoery, Tsinghua university
publishing house, Beijing, Oct.2006, pp. 56-59,99-102,161-178.
[2] J. Hu, L. N. Liu. Morden control thoery, Defense industry publishing
house, Beijing, May. 2012, pp.27-52.
[3] Z. Y. Zhang, Z. Y. Yang. A course in MATLAB. Beijing university of
aeronautics and astronautics publishing house, Beijing, Aug.006, pp.
173-191.
[4] Y. B. Liang㧘L. H. Zhang. “Applications of Simulink simulation in
modern control systems,” Journal of Zhoukou NormaI University,
Zhoukou, Vo.l28.No.5, pp.42-43,Sep. 2011.

Fig. 4. The Simulation Diagram of the Closed-loop System with

a Full-order State Observer
1.5

• • • • 4.3%

• • • • • 0

0.5

• • • • • 0.45s

-0.5
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
t

899

View publication stats