2005 American Control Conference FrB11.

4
June 8-10, 2005. Portland, OR, USA

Fuzzy Model-Based Robust Controller Design for Hydrofoil

Catamaran
Junsheng Ren, Yansheng Yang, Yunfeng Zheng and Tieshan Li

Abstract— This paper addresses the robust fuzzy controller mathematical descriptions in the interpolation of controller
design for a class of nonlinear system with norm-bounded parameters [5], [6]. Therefore, fuzzy inference system
parametric uncertainties, and its application to attitude con- makes it probable to have the feedback control covering the
troller design for hydrofoil catamaran. First, a robust fuzzy
controller is proposed, which can guarantee that the fuzzy whole operating envelope [5] for hydrofoil catamaran. There
system is asymptotically stable. The sufficient conditions are exists voluminous literature on the subject of making use of
formulated in the form of linear matrix inequalities (LMI). various control techniques to attitude controller design for
The comparison between the new result and previous method hydrofoil catamaran [3], [4]. However, very few papers are
shows that the stability conditions herein possess much less found that report Takagi-Sugeno (T-S) fuzzy system and the
conservativeness. Second, The robust fuzzy controller is ap-
plied to attitude controller design for hydrofoil catamaran. parallel distributed compensation (PDC) control structure,
Based on such a boat, “HC200B-A1”, simulation researches which can be used to control uncertain nonlinear systems,
demonstrate the drawbacks of conventional control strategies to satisfy the special requirements of attitude controller of
are overcome successfully. hydrofoil catamaran.
Generally speaking, there are two kinds of fuzzy logic
I. I NTRODUCTION
controller. One is model-free fuzzy controller, and the other
Hydrofoil catamaran is a kind of high-speed boat, and is model-based fuzzy controller. Controller design based on
composed of twin V-hulls and two or more rectangle hy- T-S fuzzy model has been discussed in [5], [7]–[9]. Fuzzy
drofoils at the bottom of the V-hulls. Both simulation study model-based controller can combine the merits of both
and tank experimental results [1], [2] manifest that waves fuzzy controller and conventional linear theory, and fur-
will lead to roll, pitch, heave motion of large amplitude, thermore guarantee stability in the sense of Lyapunov and
which will discomfort passengers and seamen, and even control performance theoretically. Moreover, linear matrix
destroy cargoes. Therefore, the special attitude stabilizer is inequality (LMI) techniques also make model-based fuzzy
indispensable for hydrofoil catamaran. Various studies on controller design more convenient. Since uncertainties will
control systems for hydrofoil catamaran were studied [3], cause instability and degraded performance, robust fuzzy
[4]. How ever, those traditional control strategies have controller based on T-S fuzzy model had also discussed [8].
several drawbacks. First, they are no longer efficient and However, the stability conditions therein are of much con-
feasible once out of the small neighborhood of the operating servativeness.
point, i.e. design speed. Second, they haven’t taken account The main contributions of this paper are divided into two
of the influences of environmental disturbances. Third, its parts. First, a novel fuzzy model-based robust controller
zero left angles may subject to the fluctuations with time, is proposed for a class of nonlinear system with norm-
because there exists the interferences between fore and aft bounded parametric uncertainties. Because the controller
hydrofoils. design takes accounts of much closer relationship between
Gain scheduling is an effective way of controlling sys- the subsystems of the whole dynamic system, the conser-
tems whose dynamics change with the operating conditions. vativeness of the controller design is reduced efficiently.
One drawback of conventional gain scheduling controller Second, by using the proposed controller scheme, attitude
is that the parameter change may be rather abrupt across controller for hydrofoil catamaran are presented. Simulation
the region boundaries, which may result in unsatisfactory researches manifest that the drawbacks of conventional
or unstable performance across the transition. Fuzzy gain control strategies have been overcome successfully.
scheduling has been proposed that utilizes fuzzy tech- This paper is organized as follows. T-S fuzzy system is
nique to determine the controller parameters. In fuzzy gain constructed in Section II. The less conservative controller is
scheduling strategy, fuzzy inference mechanism is used proposed in Section III. In Section IV, the proposed scheme
to interpolate the controller parameters in the transition is applied to the attitude controller design for hydrofoil
regions. As a result, it obviates the need to use exact catamaran. Some conclusions are collected in Section V.
This work was supported by the Research Fund for the Doctoral
Program of Higher Education under Grant 20020151005 and the National II. C ONSTRUCTION OF T-S F UZZY M ODELS
Natural Science Foundation of China under Grant 60474014.
Junsheng Ren, Yansheng Yang, Yunfeng Zheng and Tieshan Li are The development of T-S fuzzy model and its applications
all with Institute of Nautical Science and Technology of Dalian Mar-
itime University, Dalian, 116026, P. R. China, Fax: +86-411-4727395, have been increasingly accelerated over the last decade. The
[email protected], [email protected]. T-S fuzzy system can be used to approximate the nonlinear
0-7803-9098-9/05/$25.00 ©2005 AACC 4339
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.
system with norm-bounded parametric uncertainties [9],
which is constructed as follows. At the consequent part, fuzzy control rules have linear
state feedback gain. It has been proved that the controller
Plant Rule i : using the PDC scheme is an approximator for any nonlinear
state feedback controller [9]. The overall fuzzy controller
IF z1 (t) is Mi1 and, . . . , and zp is Mip , can be represented as follows
THEN ẋ (t) = (Ai + ∆Ai ) x (t) + (Bi + ∆Bi ) u (t) ,
r


(i = 1, 2, · · · , r.) (1)
u (t) = − hi (z (t))Ki x (t) (6)
i=1
where z (t) = {z1 (t) , z2 (t) , · · · , zq (t)} denote the vari-
Therefore, the design of fuzzy controller is to design local
ables of premise part, Ai ∈ n×n , Bi ∈ m×n , x (t)
feedback gain Ki s. Then, the combination of (4) and (6)
denotes state vector, u (t) denotes control input vector, and
results in the overall closed-loop fuzzy system
Mil denotes fuzzy sets, r denotes the number of IF - THEN
rules, and ∆Ai ∈ n×n , ∆Bi ∈ m×n are the system’s
uncertainty matrices and satisfy Assumption 1.
r
r

Assumption 1: Uncertainty matrices ∆Ai and ∆Bi are ẋ (t) = hi (z(t))hj (z(t)) [Ai − Bi Kj
i=1 j=1 (7)
norm-bounded, and have the following structures:
+ Di Fi (Ei1 − Ei2 Kj )] x (t)
[∆Ai ∆Bi ] = Di Fi (t) [Ei1 Ei2 ] (2)
III. LMI-BASED ROBUST F UZZY C ONTROLLER D ESIGN
where Di , Ei1 and Ei2 are constant real matrices of appro-
priate dimensions, and Fi (t) ∈ i×j is unknown matrix- Before proceeding with the research on stability con-
valued functions with Lebesgue-measurable elements and ditions for the closed-loop fuzzy system (7), some useful
satisfies lemmas are introduced first.
Lemma 1: [10] For any constants ε > 0, and any
FiT (t) Fi (t) ≤ I (3) matrices X and Y with appropriate dimensions, we have
By using the fuzzy inference method with a singleton
fuzzifier, product inference, and center average defuzzifiers, X T Y + Y T X ≤ εX T X + ε−1 Y T Y (8)
the final output of T-S fuzzy model is obtained as Lemma 2: For matrices X, Y with appropriate dimen-
sions, there exists an arbitrary scalar ε > 0, such that
r


ẋ(t) = hi (z(t))[(Ai + ∆Ai )x(t) + (Bi + ∆Bi )u(t)] T
i=1
XY + (XY ) ≤ εXX T + ε−1 Y T Y (9)
(4) Proof: As for (8), let X = AT , then X T = A. From
where (8), we have


r
hi (z (t)) = ωi (z (t)) ωi (z (t)),
p
i=1 AY + Y T AT ≤ εAAT + ε−1 Y T Y
wi (z (t)) = Π Mij (z (t)),
j=1 Let A → X, then we have (9).
and Mij (z (t)) denotes the degree of membership of z (t)
on Mij . The degree of membership satisfies Theorem 1: The closed-loop T-S fuzzy system (7) is
r asymptotically stable, if there exist feedback gains Ki s and
ωi (z (t)) > 0, ωi (z (t)) ≥ 0, i = 1, 2, · · · , r. symmetric positive definite matrix P such that
i=1
Note that for all t, there exists

r ΞTii P + P Ξii − Xii < 0, (i = 1, · · · , r) (10)
hi (z (t)) = 1, hi (z (t)) ≥ 0, i = 1, 2, · · · , r.
i=1 ΞTij P + P Ξij − Xij < 0, (1 ≤ i < j ≤ r) (11)
For PDC scheme, fuzzy controller and fuzzy model (Xij )r×r − S < 0, (1 ≤ i < j ≤ r) (12)
(4) possess the same premises. Then, supposing that all
the states are observable, the i-th controller rule can be where ε is an arbitrary positive number, S is a negative
expressed by semi-definite matrix chosen by designer, (Xij )r×r denotes
matrix with element Xij , and
Controller Rule i :
T
IF z1 (t) is Mi1 and, . . . , and zp is Mip , Ξii = (Ai − Bi Ki ) P + P (Ai − Bi Ki ) + εP Di DiT P
T
THEN u (t) = −Ki x (t) , (i = 1, 2, · · · , r.) (5) + ε−1 (Ei1 − Ei2 Ki ) (Ei1 − Ei2 Ki ) ,
4340
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.
  ⎡ ⎤T ⎡ ⎤
T
Ξij = (Ai − Bi Kj ) P + P (Ai − Bi Kj ) h1 (z(t)) x (t) X11 X12 . . . X1r
⎢ h2 (z(t)) x (t) ⎥ ⎢ X21 X22 . . . X2r ⎥
T
+ (Aj − Bj Ki ) P + P (Aj − Bj Ki ) ⎢ ⎥ ⎢ ⎥
V̇ (t) < ⎢ .. ⎥ ×⎢ .. .. . . . ⎥
⎣ . ⎦ ⎣ . . . .. ⎦
+ εP Di DiT P + εP Dj DjT P
hr (z(t)) x (t) Xr1 Xr2 . . . Xrr
T
+ ε−1 (Ei1 − Ei2 Kj ) (Ei1 − Ei2 Kj ) ⎡ ⎤
 h1 (z(t)) x (t)
T
+ε−1 (Ej1 − Ej2 Ki ) (Ej1 − Ej2 Ki ) 2. ⎢ h2 (z(t)) x (t) ⎥
⎢ ⎥
×⎢ .. ⎥
Proof: Choose the Lyapunov candidate as ⎣ . ⎦
hr (z(t)) x (t)
T
V (t) = x (t) P x (t) Next, substituting (12) into the above inequality yields

where P is symmetric positive definite matrix. ⎡ ⎤T ⎡ ⎤

h1 (z(t)) x (t) h1 (z(t)) x (t)
Then, the time derivative of the Lyapunov candidate V (t) ⎢ h2 (z(t)) x (t) ⎥ ⎢ h2 (z(t)) x (t) ⎥
⎢ ⎥ ⎢ ⎥
along the trajectory of (7) is given by V̇ (t) < ⎢ .. ⎥ S⎢ .. ⎥
⎣ . ⎦ ⎣ . ⎦
V̇ (t) = ẋT (t) P x (t) + xT (t) P ẋ (t) hr (z(t)) x (t) hr (z(t)) x (t)
r r  = xT (t) Sx (t)
T
= hi (t) hj (t) xT (t) (Ai − Bi Kj ) P
i=1 i=1
≤ − |λmax (S)| xT (t) x (t)
T
+ (Ei1 − Ei2 Kj ) FiT DiT P + P (Ai − Bi Kj )
+ P Di Fi (Ei1 − Ei2 Kj )] x (t) where λmax (S) denotes the maximal eigenvalue of negative
semi-definite S. Therefore, the closed-loop fuzzy system (7)
After some manipulations, we have is asymptotically stable.


r  The search for the common matrix P and Ki s nowadays
T
V̇ (t) = h2i (z(t)) xT (t) (Ai − Bi Ki ) P can resort to some efficient numerical methods in terms of
i=1
T
+P (Ai − Bi Ki ) + (Ei1 − Ei2 Ki ) FiT DiT P LMIs. How ever, the conditions are not jointly convex in
+ P Di Fi (Ei1 − Ei2 Ki )] x (t) Ki s and P in Theorem 1. Hence, Theorem 2 is proposed,
r  in which the LMIs are tractable.
T
+ hi (z(t)) hj (z(t)) xT (t) (Ai − Bi Kj ) P Theorem 2: The closed-loop T-S fuzzy system (7) is
i<j
+P (Ai − Bi Kj ) + (Ei1 − Ei2 Kj ) FiT DiT P
T asymptotically stable, if there exist matrices Mi s and sym-
+P metric positive definite matrix N such that
 Di Fi (Ei1 − Ei2 Kj )]/2
T
+ (Aj − Bj Ki ) P + P (Aj − Bj Ki )  
T Ωii ∗
+(Ej1 − Ej2 Ki ) FjT DjT P <0 (13)
Ei1 N − Ei2 Mi −εI
+ P Dj Fj (Ej1 − Ej2 Ki )]/2} x (t) ⎡ ⎤
Ωij ∗ ∗
⎣ Ei1 N − Ei2 Mj −εI ∗ ⎦<0 (14)
From Lemma 2 and Assumption 1, we have
Ej1 N − Ej2 Mi 0 −εI

r  (Yij )r×r − Q < 0 (15)
T
V̇ (t) ≤ h2i (z(t)) xT (t) (Ai − Bi Ki ) P
i=1 where
+εP Di DiT P + P (Ai − Bi Ki ) 
T
+ ε−1 (Ei1 − Ei2 Ki ) (Ei1 − Ei2 Ki ) x (t) Ωii = Ai N − Bi Mi + N T Ai − MiT BiT + εDi DiT − Yii ,

r  Ωij = Ai N + Aj N − Bi Mj − Bj Mi + N T Ai + N T Aj
T
+ 12 hi (z(t)) hj (z(t)) xT (t) (Ai − Bi Kj ) P
i<j − MjT BiT − MiT BjT + εDi DiT + εDj DjT − Yij − YijT ,
+P (Ai − Bi Kj ) + εP Di DiT P  Q is negative semi-definite matrix chosen by the controller
T
+ ε−1 (Ei1 − Ei2 Kj ) (Ei1 − Ei2 Kj ) designer, ∗ denotes the transposed element in the symmetric

T
+ (Aj − Bj Ki ) P + P (Aj − Bj Ki ) position, 1 ≤ i < j ≤ r and I is identity matrix.
Furthermore, feedback gain Ki s and symmetric positive
+εP Dj DjT P  definite matrix P are obtained by
T
+ ε−1 (Ej1 − Ej2 Ki ) (Ej1 − Ej2 Ki ) x (t)
P = N −1 , Ki = Mi N −1 (16)
Substitute (10) and (11) into the above inequality, and we Proof: Multiply (13)-(15) with P −1 both left and right
obtain side. Let N = P −1 , Mi = Ki P −1 , Yij = N Xij N and
4341
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.
 TABLE I
Q = (N )r×r S (N )r×r , by use of Schur complements, we
P RINCIPAL PARTICULARS OF H YDROFOIL C ATAMARAN , HC200B-A1
obtain (13) - (15).

Similar results were presented in [8] as follows. L.B.P (m) breadth(m) d0 (m) ∆ (ton) speed(knot)
Theorem 3: If there exist a symmetric and positive def- 35.84 11.584 3.84 200 40
inite matrix P , some matrices Ki s, and some εij , (i, j = chord (m) span (m) xf 1 (m) xf 2 (m) α1,2 (o )
1, . . . , r) such that the following LMIs are satisfied, then 0.96 8.32 6.56 −15.84 2
the continuous-time T-S fuzzy system (7) is asymptotically
stable via the T-S fuzzy model-based state-feedback con-
troller (6):
the comparisons, it can be concluded that theorem 1 and
⎡ ⎤ 2 can relax the condition of asymptotical stability for T-S
Ψii ∗ ∗ fuzzy system (7), and thus lower the conservativeness of
⎣ E1i Q + E2i Mi −εii I ∗ ⎦<0 (17) the controller design.
DiT 0 −ε−1
ii I
⎡ ⎤ IV. A PPLICATION TO ATTITUDE C ONTROLLER D ESIGN
Υij ∗ ∗ ∗ ∗ FOR H YDROFOIL C ATAMARAN
⎢ E1i Q + E2i Mj −εij I ∗ ∗ ∗ ⎥
⎢ ⎥ To design the attitude controller for hydrofoil catamaran,
⎢ E1j Q + E2j Mi 0 −εij I ∗ ∗ ⎥
⎢ ⎥ the mathematical model of hydrofoil catamaran is estab-
⎣ DiT 0 0 −ε−1
ij I ∗ ⎦
lished for the boat’s heave and pitch motion first [2].
DjT 0 0 0 −ε−1
ij I
<0 ⎧
(18)
⎪ 2 2
⎪m(ξ¨ + U θ̇) =
⎪ Ff i + Ff pi +FH + mg cos θ
where ⎪
⎨ i=1 i=1
2 2
⎪
⎪ Iyy θ̈ = − (xf i − xg )Ff i − (xf pi − xg )Ff pi
⎪
⎪ i=1 i=1
Ψii =QATi + ATi Q + MiT BiT + Bi Mi , ⎩
−(xb − xg )∇ cos θ − 2(xH − xg )LH
Υij =QATi + Ai Q + QATj + Aj Q + MjT BiT + Bi Mj (19)
+ MiT BjT + Bj Mi , where m is the mass of boat, U is the along-ship velocity,
Ff i is the force arising from hydrofoil, Ff pi is the force
and Q = P −1 , Mi = Ki P −1 , where ∗ denotes the from flap, g is gravitational acceleration, i.e. 9.8 m/s2 , FH
transposed elements in the symmetric positions. is the force relevant to boat’s hulls, LH is the lift force
How ever, in comparison with Theorem 3, the results in of the hulls, and Iyy is the inertia of moment about Y-
Theorem 1 and 2 are much more relaxed. Comparisons are axis. θ is pitch angle, and ξ is boat’s elevation. xf i , xpi ,
made as follows. xG , xb and xH are the coordinates of hydrofoil’s lift force,
Example 1: Consider a T-S fuzzy model with two IF- flap’s lift force, gravitational force, buoyant force and hull’s
THEN rules lift force on X-axis, respectively. Simulation researches are
Rule 1: IF x1 (t) is M11 , carried out, based on hydrofoil catamaran, HC200B-A1, the
THEN ẋ(t) = (A1 + ∆A1 ) x(t) + (B1 + ∆B1 ) u(t); principal particulars of which are listed in Table I [1].
Rule 2: IF x1 (t) is M21 , Our T-S fuzzy model for hydrofoil catamaran takes boat’s
THEN ẋ(t) = (A2 + ∆A2 ) x(t) + (B2 + ∆B2 ) u(t), speed Ui as language variable. Its membership functions are
shown in Fig. 1.
where    
2 −10 1 Plant Rule i: IF U is Ui ,
A1 = , B1 = ,
1 0 0 THEN ẋe (t) = (Ai + ∆Ai )xe (t) + (Bi + ∆Bi )u(t);
   
a −10 b
A2 = , B2 = ,
1 1 0 U1 U2 U3 U4 U5
1
and for the simplicity of the comparisons, the uncertainty
matrices ∆Ai and ∆Bi are assumed to be zero matrices.
Degree of membership

0.8

The local feedback gains for the two subsystems are 0.6

both obtained by placing the poles at −1 and −15. Herein, 0.4

parameter a is supposed to vary between 2 and 6. By using 0.2

the method in 3, LMI problem of obtaining common P is
feasible only when b < 6.5. The method in [7] is somewhat 0

relaxed, however, when b > 10, the LMI problem is no 0 5 10 15

speed(knot)
20 25

longer feasible. In contrast, with Theorem 1 and 2, the

LMI problem is always feasible even when b < 107 . From Fig. 1. Membership function

4342
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.
 5 −1.7

−1.75
0

ξ (m)
ξ (m)

−1.8

−5
−1.85

0 20 40 60 80 100 0 20 40 60 80 100
time(sec) time(sec)

10 2.8

2.7
0

θ ( °)
θ ( °)

2.6

−10 2.5

2.4
−20

0 20 40 60 80 100 0 20 40 60 80 100
time(sec) time(sec)

Fig. 2. Time response of HC200B-A1’s attitude, i.e. elevation (ξ) and Fig. 3. Time response of HC200B-A1’s attitude, i.e. elevation (ξ) and
pitch angle (θ) with controller off pitch angle (θ) under control

Controller Rule i:
IF U is Ui , THEN u(t) = Ki x(t).
α̃0 = α0 (1 + 0.2 sin t) (21)
where xe (t) = [ξ˙1e , ξ˙2e , θ̇1e , θ̇2e ] denotes state vector, and
u (t) = [αf p1 , αf p2 ] denotes control input [2]. Wave parameters are that, ahead sea, wavelength is 100
HC200B-A1’s local state-space models in the conse-
meters, period 8 seconds, and wave height 5 meters. The
quents are obtained through global linearization method, at
expected elevation ξ and pitch angle θ are −1.74 meters
5 operating points all covered by the working envelope. The
and 2.65◦ , respectively. Fig. 3 shows the simulation results
coefficient matrices of each subsystem can be found in [11].
under control. Simulation researches manifest that, the
Suppose that all the elements of uncertain matrices ∆Bi
desired attitude can be achieved satisfactorily at the arbitrary
are zero. Because zero lift angles in model (19) contain
speed, and the wave-induced heave and pitch motion can
the interactions between fore and aft foils, they tend to
also be attenuated efficiently in comparison with Fig. 2.
vary with time. Herein, zero lift angles are assumed to
From Fig. 3, conclusion can be drawn that the drawbacks
fluctuate between 20% of their nominal values in this study.
of traditional control strategy is overcome successfully by
Therefore, ∆Ai s are obtained by
use of the proposed fuzzy model-based robust controller in
Section III. Fig. 4 displays the control effects to achieve the
∆Ai = Ai |α0 −Ai |(α0 ×120%) (20)
performance in Fig. 3, i.e. the variations of fore and aft flap
where Ai |α0 denotes Ai with the nominal value of zero lift angle, respectively.
angle α0 , and Ai |(α0 ×120%) denotes Ai at its 120% times
of the nominal value. Arbitrary positive scalar ε and matrix
V. C ONCLUSIONS AND F UTURE W ORKS
Q are chosen as 1 and −3I, respectively. By use of Matlab
LMI Control Toolbox, the local feedback gains Ki s and A. Conclusions
common P are presented in Appendix. To testify the robust
fuzzy controller in Section III, an operating point, e.g. 30 In this paper, a novel robust fuzzy controller design has
knots, about 55.56 km/h, which is selected arbitrarily, which been addressed via fuzzy interpolation of a series of linear
is not one of the above 5 operating points. systems. The design is featured by less conservativeness for
Simulation results are presented with attitude controller a class of nonlinear system with norm-bounded parametric
off in Fig. 2, from which it can be seen that wave-induced uncertainties, in which the relationship between the sub-
heave and pitch motion are too large to tolerate. During systems are much closely taken into account. Furthermore,
the simulation research, zero lift angles fluctuate with time, an example has shown that the effectiveness of the fuzzy
namely model-based robust controller for hydrofoil catamaran.
4343
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.
 30
By use of Matlab LMI Control Toolbox, the local feed-
20 back gain Ki s and common P for “HC200B-A1” are
presented as follows.
10
 
αfp1 ( °)

25.697 112.54 754.52 661.72

0 K1 = ,
22.986 93.243 885.30 751.17
−10  
21.782 96.911 602.55 534.85
−20 K2 = ,
20.026 81.636 764.96 48.45
−30  
0 20 40 60 80 100 18.024 76.828 591.19 510.02
K3 = ,
time(sec) 15.929 62.704 671.21 563.14
 
20 6.7346 9.3968 735.78 575.59
K4 = ,
4.3574 3.1563 559.58 435.31
10  
6.6951 10.457 706.59 549.41
K5 = ;
( °)

0 4.2480 3.4506 537.05 416.46

fp2

⎡
α

9.9375 × 10−3 2.3178 × 10−2

−10 ⎢ 2.3178 × 10 −2
1.2847 × 10−1
⎢
P =⎢ −2
⎢ 2.3178 × 10 −2 1.2847 × 10−2
−1
−20
0 20 40 60 80 100
⎣ −5.2587 × 10 1.5426 × 10
time(sec) −3.8504 × 10−2 −7.1486 × 10⎤ −6

Fig. 4. Time response of fore (αf p1 ) and aft (αf p2 ) flap control angles −5.2587 × 10−2 −3.8504 × 10−2 ⎥
⎥
1.5426 × 10−2 −7.1486 × 10−6 ⎥.
⎥
7.2226 3.7499 ⎦
B. Future Works 3.7499 2.9806
There is still a lot of further work left to deal with, e.g. R EFERENCES
the modelling of six degrees of freedom, how to suppress [1] H. Miyata, Y. Tsuchiya, H. Uotani, H. Nakayama, and H. Kajitani,
the rolling motion of hydrofoil catamaran in rough seas, “Development of a new-type hydrofoil catamaran (2nd report: Design
and etc. of a system with flaps and motion properties),” Journal of the Society
of Naval Architects of Japan, vol. 164, pp. 82–91, 1988.
A PPENDIX [2] J. S. Ren and Y. S. Yang, “Fuzzy gain scheduling attitude control
for hydrofoil catamaran,” in Proceedings of the American Control
The uncertain matrices for HC200B-A1’s state-space Conference, Boston, USA, June 2004, pp. 1169–1174.
models are presented as follows, at 5 operating points. [3] T. Itoko, S. Higashino, Y. Yamagami, and T. Ikebuchi, “The devel-
⎡ ⎤ opment of an automatic control system for a submerged hull and
0 1.0598 × 10−5 0 1.1110 × 10−8 foil hybrid super-high-speed liner,” in Proc. 1st Inter Conf. on Fast
⎢ 0 1.8406 × 10−1 0 1.9297 × 10−4 ⎥ Sea Transportation (FAST’91), Trondheim, Norway, Jun. 1991, pp.
∆AT1 = ⎢ ⎣ 0 4.6447 × 10−1 0 4.8695 × 10−4 ⎦,
⎥ 997–1012.
[4] T. Arii, H. Yamato, T. Takai, and R. Shigehiro, “Development of
0 −1.7514 0 7.7904 × 10−2 motion control system for a foil-assisted catamaran ‘Superjet-30’,”
⎡ ⎤ in Proc. 2nd Inter. Conf. on Fast Sea Transportation (FAST’93),
0 6.1913 × 10−5 0 6.4921 × 10−8 Yokohama, Japan, Dec. 1993, pp. 295–316.
⎢ 0 3.3546 × 10−1 0 3.5166 × 10−4 ⎥ [5] P. Korba, R. Babuška, H. Verbruggen, and P. Frank, “Fuzzy gain
∆AT2 = ⎢ ⎣ 0
⎥, scheduling: Controller and observer design based on Lyapunov
2.3548 0 2.4692 × 10−2 ⎦ method and convex optimization,” IEEE Trans. Fuzzy Syst., vol. 11,
0 −15.098 0 6.5673 × 10−2 no. 3, pp. 285–298, 2003.
⎡ ⎤ [6] S. Tan, C. C. Hang, and J. S. Chai, “Gain scheduling: From
0 6.9025 × 10−4 0 7.2405 × 10−7 conventional to neuro-fuzzy,” Automatica, vol. 33, no. 3, pp. 411–
⎢ 0 1.5138 0 1.5879 × 10−3 ⎥
419, 1997.
∆AT3 = ⎢ ⎣ 0
⎥, [7] E. Kim and H. Lee, “New approaches to relaxed quadratic stability
12.392 0 1.2999 × 10−2 ⎦ condition of fuzzy control systems,” IEEE Trans. Fuzzy Syst., vol. 8,
0 −6.4841 0 7.7885 × 10−1 no. 5, pp. 523–533, 2000.
⎡ ⎤ [8] H. J. Lee, J. B. Park, and G. Chen, “Robust fuzzy control of nonlinear
0 9.1868 × 10−3 0 9.6369 × 10−6 systems with parametric uncertainties,” IEEE Trans. Fuzzy Syst.,
⎢ 0 2.0359 0 2.1357 × 10−3 ⎥
vol. 9, no. 2, pp. 369–379, 2001.
∆AT4 = ⎢ ⎣ 0
⎥, [9] H. O. Wang, J. Li, and K. Tanaka, “T-S fuzzy model with linear
28.627 0 3.0030 × 10−2 ⎦ rule consequence and PDC controller: A universal framework for
0 −6.0000 0 1.0949 nonlinear control systems,” Internat. J. of Fuzzy Systems, vol. 5, no. 2,
pp. 106–113, 2003.
⎡ ⎤
0 4.5721 × 10−2 0 4.7962 × 10−5 [10] I. R. Petersen, “A stabilization algorithm for a class of uncertain
⎢ 0 1.5795 0 1.6569 × 10−3 ⎥
linear systems,” Systems Control Lett., vol. 8, pp. 351–357, 1987.
∆A5 = ⎢ ⎣ 0
⎥; [11] J. S. Ren, Y. S. Yang, and C. J. Zhou, “Design of Takagi-Sugeno
35.247 0 3.6974 × 10−2 ⎦ fuzzy controller using PDC and its application to hydrofoil cata-
0 −7.9072 0 0.79277 maran,” Journal of Nonlinear Studies, vol. 11, no. 1, pp. 5–22, 2004.

4344
Authorized licensed use limited to: National Cheng Kung Univ.. Downloaded on November 21,2024 at 04:41:09 UTC from IEEE Xplore. Restrictions apply.