You are on page 1of 6

Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications

National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

Robust Fuzzy Control for Inverted Pendulum with


Model Uncertainty and Output Constraint via LMI
Stability Analysis
Gwo-Ruey Yu Shun-Min Wang
Department of Electrical Engineering, Graduate Institute of Opto-Mechatronics,
National Chung Cheng University National Chung Cheng University
Chia-Yi, Taiwan, ROC Chia-Yi, Taiwan, ROC
e-mail: ieegwoyu@ccu.edu.tw

Abstract—This paper presents a robust controller design uncertainty usually appears in the plant, actuators and
method for inverted pendulum with model uncertainty and sensors of control system. T-S fuzzy control system with
output constraint via Particle Swarm Optimization (PSO) and basic stable condition and robust stability condition can let
Takagi-Sugeno (T-S) fuzzy control approach. First, the T-S us find the robust fuzzy controller.
fuzzy model is established for the inverted pendulum.
According to the concept of parallel distributed compensation, The principal frameworks of T-S fuzzy control are the T-
LMI stability conditions are proposed to guarantee the T-S S fuzzy model and parallel distributed compensation (PDC)
fuzzy control system designed via PSO is globally stable. The [7]. We can utilize T-S fuzzy model to represent complicated
PSO-based T-S fuzzy controllers are designed such that the nonlinear system. With the development of science and
inverted pendulum is robust against model uncertainty under technology, the nonlinear control system became more
output constraint. Three examples demonstrate the complicated. The future of T-S fuzzy model is that those
effectiveness of this design. complicated nonlinear system can be represented as fuzzy
blending of linear systems by fuzzy IF – THEN rules [8].
Keywords—T-S fuzzy control, output constraint, model When T-S fuzzy model was constructed, the parallel
uncertainty, PSO. distributed compensation (PDC) offers the design of fuzzy
controller. The design procedure of PDC is that each rule is
I. INTRODUCTION designed from the corresponding rule of T-S fuzzy model.
Recently, fuzzy control systems have rapidly developed Based on PDC, the designed T-S fuzzy controller shares the
in engineering applications [1]. Stability is one of the most same fuzzy membership with the T-S model in the premise
important concepts concerning the properties of nowadays parts. Next, we can use linear matrix inequality (LMI) tool
control system. Nowadays, the literatures about stability box to solve the stability problem.
analysis and design of fuzzy control systems have been From the conventional design of LMI-based T-S fuzzy
extensively published [2]. However, it is still a difficult control, the Lyapunov matrix P and feedback gains F can be
problem to analyze stability of nonlinear fuzzy control obtained simultaneously. Nevertheless, the feedback gains F
systems. There are some useful stability analysis approaches solved via the LMI toolbox is too conservative and the
[3-5] which have been published to solve the problem. control performance cannot be assigned. As the result, the
Takagi-Sugeno (T-S) fuzzy mod-el-based control is one of authors proposed the design of feedback gains F searched via
the valuable approaches. The history of Takagi-Sugeno (T-S) the PSO to find the better performance first. Furthermore, we
fuzzy model-based con-trol can be trace back to three derived the relaxed stability conditions to guarantee the PSO-
decades ago. In 1985, Takagi-Sugeno published a new type based T-S fuzzy system with model uncertainty is globally
of fuzzy model presentation in [6]. The issue of stability stable. Particle Swarm Optimization is a robust stochastic
analysis for T-S fuzzy control systems has been investigated optimization approach based on the movement and
widely within nonlinear systems. The biggest difference intelligence of swarms. PSO applies the concept of social
between T-S fuzzy control and conservative fuzzy control is interaction to solve complicated problems. It uses a number
that T-S fuzzy control provides a strict mathematical of particles which constitute a swarm moving around in the
derivation and a complete analysis process. search space looking for the best solution. The main
Recently, there are some valuable stability techniques contributions of this paper are two parts. First, we used PSO
which are based on nonlinear stability theory have been algorithms to search T-S fuzzy feedback gains. Second, we
reported. As the result, the stability analysis of nonlinear proposed a theorem of robust stability conditions such that
fuzzy control system became simpler than before. Generally, the PSO-based T-S fuzzy controller owns robustness against
we can find an infinite number of feasible controllers if the model uncertainties.
plant is stabilize. Nevertheless, there are always uncertainties
associated with practical control applications. Model

18
Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications
National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

II. PSO-BASED T-S FUZZY CONTROL p

Recent years, fuzzy control systems have rapidly wi ( z (t )) = ∏M ij ( z j (t ))


developed in engineering applications. In this section, author j =1
will introduce the basic concepts of T-S fuzzy control ⎧ r
approach. The principal frameworks of T-S fuzzy control are
the T-S fuzzy model and parallel distributed compensation

⎪ wi ( z (t )) > 0,
⎨ i =1
(PDC). The approach employs the T-S fuzzy model to ⎪w ( z (t )) ≥ 0, i = 1,2..., r
⎩ i
represent the complicated nonlinear sys-tem. Furthermore,
PDC is used to design a model-based fuzzy controller. In this ⎧ wi ( z (t ))
⎪hi ( z (t )) = r ≥ 0,
paper, the stability analysis and control design problems of a
complicated nonlinear sys-tem can be effectively solved via


⎨ i =1

wi ( z (t ))
linear matrix inequality (LMI) tool box and Particle Swarm ⎪ r
Optimization (PSO).

⎪ hi ( z (t )) = 1,
⎪⎩ i =1
A. T-S Fuzzy Model
hi ( z (t )) is the normalization of the membership function.
Nonlinear systems could be represented as T-S fuzzy
model by IF-THEN rules. The framework of T-S fuzzy
model could be represented as follows [8]: B. PSO-based T-S Fuzzy Controller
Parallel distributed compensation (PDC) is a mod-el-
based design method proposed by Kang and Sugeno.
Model Rule i: According to the PDC design procedure, the control rules are
IF z1 (t ) is M i1 and … and z p (t ) is M ip designed from the correspond rules of the fuzzy sys-tems.
Based on the PDC, the T-S fuzzy controller can be
⎧ x& (t ) = A i x(t ) + Bi u (t ), constructed as follows:
THEN ⎨ i = 1,2,..., r (1)
⎩ y (t ) = Ci x(t ),
Control Rule i:
Here, z1 (t ) ,…, z p (t ) are known premise variables, r is IF z1 (t ) is M i1 and … and z p (t ) is M ip
the number of model rules, M ij is the fuzzy set, x(t ) ∈ R n is THEN u(t ) = −Fi x(t ) , i = 1,2,..., r (4)
n×n
the state vector, and u(t ) ∈ R is the input vector; Ai ∈ Rm
,
n×m The entire fuzzy controller could be represent as follows:
Bi ∈ R q× n
and Ci ∈ R are constant matrices. Each linear
consequent equation is a subsystem. By the model rule, the
r
fuzzy system could be represented as follows:
∑ w ( z(t ))F x(t )
i =1
i i

r
u (t ) = − r


i =1
wi ( z (t )){A i x(t ) + B i u (t )} ∑ w ( z(t ))
i =1
i
x& (t ) = r r
∑ w ( z(t ))
i =1
i =− ∑ h ( z(t ))F x(t )
i i (5)
i =1
r
= ∑ h ( z (t )){A x(t ) + B u(t )}
i =1
i i i (2)
The design procedure of the fuzzy controller is to find the
controller gains Fi . By using PSO algorithms, we can search
r

∑ w ( z (t ))C x(t )
i =1
i i
the better feedback gains Fi .
y (t ) = r By substituting (5) into (2), the closed-loop system can

i =1
wi ( z (t )) be represented as

∑ h ( z (t ))C x(t )
r r
=
i =1
i i (3) x& (t ) = ∑∑ h ( z(t ))h ( z(t ))( A − B F ) x (t )
i =1 j =1
i j i i j (6)

where z (t ) = [ z1 (t )... z p (t )]
C. Stability Analysis
Lemma 1: The PSO-based fuzzy control system de-
scribed by (6) is globally asymptotically stable if there exist

19
Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications
National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

a positive definite matrix such that that (7) and (8) are 1
satisfied [8]. Δ ai (t ) ≤ (12)
γ ai
Δ ai (t ) = ΔTai (t ) (13)
G iiT P + PG ii < 0 (7)
1
⎛ G ij + G ji ⎞ T ⎛ G + G ji ⎞ Δ bi (t ) ≤ (14)
⎜ ⎟ P + P ⎜ ij ⎟≤0 γ bi
⎜ 2 ⎟ ⎜ 2 ⎟
⎝ ⎠ ⎝ ⎠
i < j s.t. hi ∩ h j ≠ φ (8) Δ bi (t ) = ΔTbi (t ) (15)

for all i. By the rules, the nonlinear system could be


where
represented as fuzzy model as follows:

G ij = Ai − Bi F j r

From the above stability conditions, we can assure the


x& (t ) = ∑ h ( z(t )){(A + D
i =1
i i ai Δ ai (t )E ai )x(t )
PSO-based fuzzy control system is globally stable. + (Bi + Dbi Δbi (t )Ebi )u(t )}. (16)

D. Output Constraint Condition


Lemma 2: Assume that the initial condition is known. B. Robust PSO-based Stabilization Conditions
The constraint is enforced at all times if the following
The following theorem presents robust PSO-based
LMIs are satisfied [8].
stabilization conditions which can ensure the T-S fuzzy
system with model uncertainty is globally stable.
⎡ 1 x(0)T ⎤ (9) Theorem 1: The PSO-based feedback gains Fi can
⎢ ⎥≥0
⎣ x(0) P ⎦ stabilize the fuzzy system (11) and maximize the norms of
⎡ P C iT ⎤ the uncertain blocks ( i.e, minimize γ ai and γ bi ) if the
≥0 (10)
⎢ 2 ⎥ following equations are satisfied, where α i and β i > 0 are
⎣C i λ I ⎦
design parameters:

III. ROBUST PSO-BASED T-S FUZZY CONTROL minimize r

∑ {α γ + β i γ bi }
2 2

Even though the above-mentioned PSO-based techniques γ , γ , X , M 1 ,..., M r , Y0


2
ai
2
bi i =1
i ai

ensure stabilization and output constraints. There are always


model uncertainties in practical control applications. subject to
Consequently, the robustness of practical control
applications research will be more significant. In this section,
authors will introduce the PSO-based T-S fuzzy control with P > 0, Y0 ≥ 0, M 0 ≥ 0, Sˆii − M < 0, (17)
model uncertainties [8]. Tˆij − 2Y < 0, i < j s.t. hi ∩ h j ≠ φ , (18)

A. Fuzzy Model with Uncertainty



(
⎡ ( Ai − Bi Fi )T P + P ( Ai − Bi Fi ) ) * * * * ⎤

The necessary step to deal with nonlinear systems with ^ ⎢
T
Dai P −I 0 0 0 ⎥
model uncertainty is to combine uncertainty blocks with S ii = ⎢
⎢ D T
bi P 0 −I 0 0 ⎥

,
Takagi-Sugeno fuzzy model. By the way, the nonlinear ⎢ Eai 0 0 − γ 2ai I 0 ⎥
system with model uncertainty could be represented as ⎢ 2 ⎥
⎣ − Ebi Fi 0 0 0 − γ bi I⎦
follows:

⎡⎛ (Ai − Bi F j )T P + P (Ai − Bi F ji ) + ⎞ ⎤
Plant Rule i ⎢⎜ ⎟ PD
⎢⎜ (A j − B j Fii )T P + P (A ji − B ji Fii ) ⎟
ai PDbi PDaj PDbj E aiT − F jT EbiT E ajT − FiT EbjT ⎥

⎢⎝ ⎠
IF z1 (t ) is M i1 and … and z p (t ) is M ip ⎢

D T
ai P
DbiT P

0
I 0
−I
0
0
0
0
0
0
0
0
0
0
0 ⎥
0 ⎥

⎢ ⎥
−I
THEN x& (t ) = ( Ai + Dai Δ ai (t ) E ai ) x (t )
T
ˆ
Tij = ⎢ D aj P 0 0 0 0 0 0 0 ⎥,
⎢ T
Dbj P 0 0 0 −I 0 0 0 0 ⎥
⎢ ⎥
+ ( Bi + Dbi Δ bi (t ) E bi )u(t ), i = 1,2,..., r
⎢ − γ ai2 I 0 ⎥
(11) ⎢
E ai
− Ebi F j
0
0
0
0
0
0
0
0 0
0
−γ I
2
0
0
0 ⎥

⎢ bi
⎢ E aj 0 0 0 0 0 0 − γ aj2 I 0 ⎥
⎢ ⎥
⎣⎢ − Ebj Fi 0 0 0 0 0 0 0 − γ bj2 I ⎦⎥
where the uncertain blocks satisfy
M = block − diag ( M 0 0 0 0 0),

Y = block − diag (Y0 0 0 0 0 0 0 0 0 ),

20
Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications
National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

IV. TWO RULE MODELING FOR INVERTED PENDULUM Here,


SYSTEM WITH MODEL UNCERTAINTY
In this section, the authors present the numerical ⎡ 0 1⎤ ⎡ 0 ⎤
simulation results obtained by using the proposed method to A1 = ⎢ g ⎥ B1 = ⎢ a ⎥
illustrate the difference within various stabilization ⎢⎣ 4l / 3 − aml 0⎥⎦ ⎢⎣− 4l / 3 − aml ⎥⎦
conditions. ⎡ 0 1⎤ ⎡ 0 ⎤
A2 = ⎢ 2g ⎥ B2 = ⎢ aβ ⎥
A. Dynamic Equations of the Inverted Pendulum ⎢ π(4l / 3 − amlβ 2 ) 0⎥ ⎢− 4l / 3 − amlβ 2 ⎥
⎣ ⎦ ⎣ ⎦
We used an inverted pendulum on a cart with model
uncertainty to consider the problem of balancing and swing- ⎡ 0 ⎤ ⎡ 1 ⎤
up. The dynamic equations of the inverted pendulum are Da1 = ⎢ g ⎥, Δ a1 (t ) = ⎢ sin(t )⎥,
shown as follows: ⎢⎣ 4l / 3 − aml ⎥⎦ ⎣ γ a1 ⎦
⎡ 0 ⎤ ⎡ 1 ⎤
Da 2 = ⎢ 2g ⎥, Δ a 2 (t ) = ⎢ sin(t )⎥,
x&1 (t ) = x 2 (t ) ⎢ π(4l / 3 − amlβ 2 ) ⎥
⎣ ⎦ ⎣ γ a2 ⎦
x&2 (t ) =
g sin( x1 (t )) − amlx22 (t ) sin( 2 x1 (t )) / 2 − a cos( x1 (t ))u (t )
, E a1 = E a 2 = [1 0]
4l / 3 − aml cos 2 ( x1 (t ))
(22) ⎡ 1 ⎤ ⎡ a ⎤
Δ b1 (t ) = ⎢ sin(t )⎥, Eb1 = ⎢− ,
⎡ 0 1 ⎤ ⎣ γ b1 ⎦ ⎣ 4l / 3 − aml ⎥⎦
'
⎡ x1 ⎤ ⎢ − amlx2 (t ) sin(2 x1 (t )) ⎥ ⎡ x1 (t ) ⎤
⎢x ⎥ = ⎢ g sin( x ( t )) 2 ⎥⎢ ⎥ ⎡ 1 ⎤ ⎡ aβ ⎤
Δ b 2 (t ) = ⎢ sin(t )⎥, Eb 2 = ⎢−
1
⎣ 2 ⎦ ⎢ x (t )(4l − aml cos 2 ( x (t ))) 4l − aml cos 2 ( x (t )) ⎥ ⎣ x2 (t )⎦ 2⎥
,
⎣ 3 3 ⎦
⎣ γ b2 ⎦ ⎣ 4l / 3 − amlβ ⎦
1 1 1

⎡ 0 ⎤
+⎢ - a cos( x1 (t )) ⎥u (t ) ⎡0 ⎤
⎢ 4l − aml cos 2 ( x (t )) ⎥ Db1 = Db 2 = ⎢ ⎥
⎣⎢ 3 ⎦⎥
⎣1 ⎦
1

where x1 (t ) denotes the angle of the pendulum from the and β = cos(88°) .
vertical (in radians) , x1 (t ) ∈ (−π 2 , π 2) and x 2 (t ) is the
angular velocity; g = 9.8 m s 2 is the gravity constant, m is V. COMPUTER SIMULATION
the mass of the pendulum, M is the mass of the cart, 2l is the
In this section, we will show some examples to illustrate
length of the pendulum, and u is the force applied to the cart
contributions of this paper. The first example shows that the
(in newtons) ; and a = 1 (m + M ) . We choose m = 2.0 kg, M
difference between PSO-based T-S fuzzy controller and
= 8.0 kg, 2l = 1.0 m. conventional T-S fuzzy controller. The second example
shows that the comparison of robust performance between
B. Fuzzy Model with Uncertainty PSO-based T-S fuzzy controller and conventional T-S fuzzy
In order to use the PDC approach, we must represent the controller.
dynamics of the nonlinear plant as a fuzzy model by a
Takagi-Sugeno fuzzy model first. Due to the inverted A. Case 1
pendulum system is uncontrollable when x1 = ± π 2 , we In this case, we consider the inverted pendulum system
approximate the nonlinear system with model uncertainty by with model uncertainty. Then, we construct the fuzzy
the two-rule fuzzy model as follows: controllers respectively via LMI approach and PSO. The
feedback gains and the eigenvalues of common positive
matrix P of the T-S fuzzy controller are shown as follows.
Rule 1
IF x1 (t ) is about 0, LMI-based feedback gains
THEN x& (t ) = ( A1 + Da1 Δ a1 (t ) E a1 ) x (t ) F1 = [− 1243.2 − 359.4]
+ ( B1 + Db1 Δ b1 (t ) E b1 ) u(t ), i = 1,2 (23) F2 = [− 2632.8 − 773.8]
Rule 2 Positive matrix P
IF x1 (t ) is about ± π 2 ( x1 < π 2 ) , ⎡40.2735 12.4305⎤
P=⎢ ⎥
THEN x& (t ) = ( A2 + Da 2 Δ a 2 (t ) E a 2 ) x (t ) ⎣12.4305 4.0985 ⎦
+ ( B 2 + Db 2 Δ b 2 (t ) E b 2 ) u(t ), i = 1,2 (24) PSO-based feedback gains
F1 = [− 2347.7 − 227.2]

21
Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications
National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

F2 = [− 740.1 − 2908.2] 16

Positive matrix P 14
T-S stable
T-S stable by PSO

⎡1.7348 0.0171⎤
P = 1011 × ⎢ ⎥
12

⎣0.0171 0.0064⎦ 10

Fig. 1 shows the fitness function of PSO. Fig. 2 shows 8

x1(deg)
the PSO particles of F1 . Fig. 3 shows the PSO particles of
6
F2 . Fig. 4 shows the responses comparison of state x1 . From
4
Fig. 4, we can clearly see that the PSO-based LMI controller
is better in the case of model uncertainty. 2

0
Maximum of fitness function
0
iterations = 100 -2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
-0.05
Time(sec)

-0.1
Figure 4. The state x1 response
-0.15
fitness function

-0.2

-0.25
B. Case 2
-0.3 In this case, we add the output constraint to limit
-0.35
responses of state x 2 in a reasonable range. We construct the
-0.4
fuzzy controllers respectively via PSO and LMI approach.
The feedback gains and the eigenvalues of common positive
-0.45
matrix P of the T-S fuzzy controller are shown as follows.
-0.5
0 10 20 30 40 50 60 70 80 90 100
Iterations

Figure 1. The fitness function of PSO Robust LMI-based feedback gains


-214 F1 = 10 3 × [− 9.0605 − 3.0786]
-216
F2 = 10 4 × [− 1.0087 − 0.3429 ]
-218

-220 Positive matrix P and eigenvalues


-222
⎡20.0834 6.7915⎤
-224
P=⎢ ⎥
-226
⎣ 6.7915 2.3081⎦
⎡ 0.0103 ⎤
-228
eig P = ⎢ ⎥
-230
⎣22.3813⎦
-232

-234
-2460 -2440 -2420 -2400 -2380 -2360 -2340 -2320 -2300 -2280
Robust PSO-based feedback gains
F1 = 10 3 × [− 2.0745 − 0.1376]
Figure 2. The particles of F1
F2 = 10 3 × [− 0.9730 − 3.1424]
-2750
Positive matrix P and eigenvalues
-2800 ⎡ 5.0085 − 0.0003⎤
P = 10 4 × ⎢ ⎥
-2850 ⎣− 0.0003 0.0015 ⎦
⎡0.0015⎤
-2900
eig P = 10 4 × ⎢ ⎥
-2950
⎣5.0085⎦
-3000 Fig. 5 shows the fitness function of PSO. Fig. 6 shows
the PSO particles of F1 . Fig. 7 shows the PSO particles of
-3050
F2 . Fig. 8 shows the responses comparison of state x1 . From
-3100
-950 -900 -850 -800 -750 -700
Fig. 8, it is obviously the PSO-based LMI controller is better
in the case of output constraint.
Figure 3. The particles of F2

22
Proceedings of 2012 International Conference on Fuzzy Theory and Its Applications
National Chung Hsing University, Taichung, Taiwan, Nov.16-18, 2012

Maximum of fitness function


0 16
uncertainty+constraint
iterations = 100
-500 14 uncertainty+constraint by PSO

12
-1000

10
-1500
fitness function

x1(deg)
-2000
6
-2500
4

-3000
2

-3500 0

-4000 -2
0 10 20 30 40 50 60 70 80 90 100 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Iterations Time(sec)

Figure 5. The fitness function of PSO Figure 8. The state x1 response


-20

-40

-60
ACKNOWLEDGMENT
This research is sponsored in part by NSC under Grant
-80 No. 101-2221-E-194-039-.
-100
REFERENCES
-120
[1] T. H. Lee, H. K. Lam, F. H. F. Leung, and P. K. S. Tam, “A practical
fuzzy logic controller for the path tracking of wheeled mobile
-140 robots,” IEEE Control Systems Magazine, vol. 23, no. 2, pp. 60-65,
April 2003.
-160
-2200 -2000 -1800 -1600 -1400 -1200 -1000 -800 -600 -400 [2] C.-H. Sun and W.-J. Wang, “An improved stability criterion for T-S
fuzzy discrete systems via vertex expression,” IEEE Trans. Syst.,
Man, Cybern. B, Cybern., vol. 36, no.3, pp. 672-678, Jun. 2005.
Figure 6. The particles of F1 [3] K. Tanaka and H. Ohtake, “Guaranteed cost control of polynomial
fuzzy systems,” IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 39,
-500 no. 2, April 2009.
[4] S. Prajna, A. Papachristodoulou, and F. Wu, “Nonlinear control
-1000 synthesis by sum of squares optimization: a Lyapunov-based
approach,” in Proc. Asian Control Conf., Feb. 2004, pp.157-165
-1500 [5] Guoyong Huang, Guoliang Ma, Daobo Wang, “Stabilization control
of switched nonlinear system based on sum of square decomposition,
“ in Proc. 7th World Congress on Intelligent Control and Automation,
-2000 Chongqing, China, June 2008, pp.8040-8044.
[6] K. Tanaka and M. Sugeno, “Stability analysis and design of fuzzy
-2500 control systems,” Fuzzy Sets and Sysr., vol. 45, no. 2, pp. 135-156,
1992
-3000 [7] T. Takagi and M. Sugeno, ‘‘Fuzzy Identification of Systems and Its
Applicationsto Modeling and Control,’’ IEEE Trans. Syst. Man.
Cyber., vol. 15, pp. 116-132, 1985.
-3500
-1000 -900 -800 -700 -600 -500 -400 -300 -200 [8] K. Tanaka and H. O. Wang, “Fuzzy control systems design and
analysis: a linear matrix inequality approach,” John Wiley and Sons,
Inc, USA, 2001.
Figure 7. The particles of F2

VI. COMPUTER SIMULATION


According to our computer simulation, we can get two
points. First, the results show that these derived stability
conditions can be applied to the inverted pendulum system
with model uncertainty and output constraint. Second, we
can construct a better robust controller by combing PSO and
LMI approach.

23

You might also like