Anda di halaman 1dari 9

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


A ship berthing system design with four tug boats

Article  in  Journal of Mechanical Science and Technology · May 2011

DOI: 10.1007/s12206-011-0215-4


14 1,075

4 authors, including:

Young Bok Kim

Pukyong National University


Some of the authors of this publication are also working on these related projects:

Korea Gov. View project

Crane Automation View project

All content following this page was uploaded by Young Bok Kim on 25 November 2014.

The user has requested enhancement of the downloaded file.

Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264
DOI 10.1007/s12206-011-0215-4

A ship berthing system design with four tug boats†

Van Phuoc Bui1, Hideki Kawai2, Young Bok Kim2,* and Kwon Soon Lee3
Department of Control & Mechanical Engineering, the Graduate School of Pukyong National University, Busan, Korea
Department of Mechanical System Engineering, Pukyong National University, Busan, Korea
Department of Electrical Engineering, Dong-A University, Busan, Korea

(Manuscript Received August 6, 2010; Revised December 8, 2010; Accepted December 29, 2010)



In harbor areas, precise ship steering is the most important operation. This requires a set of adequate thrust devices taking into account
surge, sway and yaw motions precisely. However, the effectiveness of actuators during low-speed maneuvering is reduced, making it
necessary to use tugboats to ensure safe berthing. In this paper, we present a mathematical model of a system describing the interaction
between an unactuated ship and tugboats. Thrust allocation is solved by using the redistributed pseudo-inverse (RPI) algorithm to deter-
mine the thrust and direction of each individual tugboat. The main goal of this method is to minimize the power supplied to tugboats and
increase their controllability. The constraints are twofold. First, the tugboat can only exert a limited pushing force, and second, it can only
change directions slowly. Additionally, an adaptive control law is proposed to capture the draft coefficients of the ship, which are known
as uncertainty parameters. The controller guarantees that the ship follows a given path (geometric task) with desired velocities (dynamic
task). The specifications of Cybership I, a model ship, are used to evaluate the efficiency of the proposed method through Matlab simula-
Keywords: Adaptive control; Control allocation; Redistributed pseudo-inverse algorithm; Ship berthing; Ship model

prising that recent research efforts have focused on developing

1. Introduction
intelligent control strategies independent of the dynamic
Based on a marine literature review, ship berthing maneu- model. These include fuzzy control [1, 2] and neural network
vers are considered to be the most complex procedure, with techniques [3, 4]. Although these approaches have the advan-
high pressure for the shipmaster to ensure safe operation. tage of embedded human experience and knowledge about
Compared with other maneuvers such as autopilot for steering, ship behavior in the control strategies, the limits of actuator
position tracking (which includes trajectory tracking and path controllability under dead-slow velocity conditions have not
following), dynamic positioning or station keeping, ship berth- yet been solved. Thus, it is too dangerous to apply these meth-
ing requires the shipmaster to carry out many tasks. When the ods in actual ship berthing operations. Despite the introduction
vessel moves from open seas into confined waters, the ship of new navigation technologies such as the differential global
velocity must be kept at dead slow, which significantly re- positioning system (DGPS) and camera sensing, and advances
duces the controllability of actuators (main propeller, rudder, in propulsion manufacturing, large ship maneuvering in har-
etc.). Furthermore, when the ship comes near a jetty, the bor areas is still done manually with the assistance of tugboats
shipmaster must precisely know the ship’s position, and pre- as shown in Fig. 1.
dict her movement so as to prevent a collision. A large amount To overcome these drawbacks and to develop a fully auto-
of information, including maneuvering conditions, actuator mated solution, we propose a new approach for ship berthing
characteristics, wind effects, wave and current disturbances as using autonomous tugboats. In this paper, we assume that
well as the condition of tugboats, have to be considered. none of the ship’s actuators are used, the ship is thus consid-
For these reasons, automatic berthing approaches have been ered as an unactuated system. The movement of the ship is
investigated since the early 1990s. Given the difficulties in controlled by autonomous tugboats. The mathematical model
capturing changes in hydrodynamic coefficients, it is not sur- of the system describing the interaction between the ship and

four tugboats is presented. Thrust allocation for this over-
This paper was recommended for publication in revised form by Associate Editor
Yeon June Kang actuated system is formulated to determine the thrust and di-
Corresponding author. Tel.: +82 51 629 6197, Fax.: +82 51 629 6188 rection of each tugboat. We consider the allocation as an op-
E-mail address:
timization problem and solve it by using the RPI algorithm [5].
© KSME & Springer 2011
1258 V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264

α1 α4
( x1 , y1 ) ϕ
( x4 , y 4 )

( x2 , y 2 ) α 2

f2 ( x3 , y3 )

Fig. 1. Ship berthing with assistance of tugboats.

One objective is to minimize the power supplied to the tug- Fig. 2. Ship motion with the assistance of four tugboats.
boats and to increase their controllability knowing that a tug-
boat can only exert a limited pushing force and that it can only
⎡cosϕ − sin ϕ 0⎤
change directions slowly. Furthermore, constraints due to
R(ϕ ) = ⎢ sin ϕ cosϕ 0⎥ . (2)
limitations of contact angles between the ship and the tugboats ⎢ ⎥
are also considered. ⎢⎣ 0 0 1 ⎥⎦
Additionally, when the ship moves from open seas to con-
fined waters, its hydrodynamic coefficients change signifi- M ∈ R 3x3 represents a mass/inertia matrix. D ∈ R 3x3 is a
cantly. This, in turn, considerably influences the ship’s han- linear damping matrix assumed to be uncertain and continu-
dling. To overcome this drawback, an adaptive controller that ously differentiable. These matrices can be determined, re-
considers the change of draft coefficients is proposed. spectively, as follows:
The remainder of this paper is structured as follows. In Sec-
tion II, we provide the second order dynamic system of ship ⎡ m − X u 0 0 ⎤
considered in the horizontal plane. The thrust configuration M =⎢ 0 m − Yv −Yr ⎥
matrix is studied through force decomposition analysis. In ⎢ ⎥
⎣ 0 − N v I z − N r ⎦
Section III, the adaptive controller is presented. Control allo- (3)
⎡− X u 0 0 ⎤
cation based on the RPI algorithm is proposed in Section IV.
D=⎢ 0 −Yv 0 ⎥ .
In Section V, the efficiency of the proposed approach is ⎢ ⎥
⎣ 0 0 − Nr ⎦
evaluated through model ship control simulations. Conclu-
sions and plans for future study are summarized and discussed
in Section VI. If we consider the assistance of tugboats, the control input
vector τ = [τ X ,τ Y ,τ Z ] ∈ R 3 (whose components are the surge
force τ X , sway force τ Y and yaw moment τ Z ) is the result of
2. System model
combined efforts of four tugboats as shown in Fig. 2. Vector
The kinematic and linear dynamic equation describing low- τ is defined as
speed maneuvering of an unactuated vessel manipulated by
four external tugboats in the horizontal plane can be written as τ = B(α ) f (4)
follows [6]:
where the vector f = [ f1 f 2 f 3 f 4 ]T ∈ F represents the unidi-
η = R( ϕ )v , rectional thrust produced by each individual tugboat.
Mv + Dv = τ The set of F is described as 0 < fi ≤ f max , ∀i ∈ (1,...,4) .
The geometrical configuration matrix B(α ) ∈ R 3x4 captures
where η = [ x, y ,ϕ ]T ∈ R 3 represents the inertial position (x, the relationship between all four tugboats and the ship. The i-
y) and the heading angle ϕ in the earth-fixed coordinate th column of matrix B(α ) is defined as follows:
frame, v = [u, v, r ]T ∈ R 3 describes the surge, sway and yaw
rates of the ship in a body fixed coordinate frame. The rotation ⎡ cos(αi ) ⎤
matrix R(ϕ ) , which translates the body fixed coordinate ⎢ ⎥
Bi (α ) = ⎢ sin(αi ) ⎥ . (5)
frame into the earth fixed coordinate frame, is defined as
⎢ −l yi cos(αi ) + l xi sin(αi ) ⎥
⎣ ⎦
V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264 1259

Here, the angle αi defines the force direction of the i-th tugboat. stability analysis, the filtered tracking error, r (t ) ∈ R 3 x1 , is
It is measured clockwise and is relative to the x-axis of body introduced as
fixed coordinate frame. The location of the i-th contact point in
the body fixed coordinate system is at (l xi , l yi ) . The control r = e + Ke . (11)
input vector τ can thus be expressed in the form of the geomet-
ric configuration matrix B (α ) and thrust vector f by: Here, K ∈ R 3 x 3 is the constant control gain. It is defined as a
diagonal positive matrix.
T The time derivative of r(t) can be obtained as follows:
⎡ cα1 sα1 −l y1cα1 + l x1sα1 ⎤ ⎡ f ⎤
⎢ ⎥
⎢ cα 2 sα 2 −l y 2 cα 2 + l x 2 sα 2 ⎥ ⎢ f 2 ⎥
τ=⎢ ⎢ ⎥ (6) r = η
 − η
d + Ke . (12)
cα sα3 −l y 3cα 3 + l x 3 sα3 ⎥ ⎢ f 3 ⎥
⎢ 3 ⎥ ⎢ ⎥
⎢ cα 4 sα 4 −l y 4 cα 4 + l x 4 sα 4 ⎥⎦ ⎣ f 4 ⎦ By substituting Eq. (12) into Eq. (7), the open-loop dynamics

for the filtered tracking error signal r (t ) can be expressed as
where s αi = sin(αi ) and cαi = cos(αi ). follows:

M *r = τ * − D*η − M *η
d + M * Ke
3. Adaptive control design
= τ * − R(ϕ ) Dv + R(ϕ ) MS (ϕ ) RT (ϕ )η − M *η
d + M * Ke
The primary control objective is to design the control input
vector such that the unactuated vessel is forced to follow a = −Y (ϕ , v )Θ + τ * + R(ϕ ) MS (ϕ ) RT (ϕ )η − M *η
d + M * Ke,
desired trajectory with an uncertainty of draft coefficients of (13)
the damping matrix D . The controller development is based
on the assumption that all states of the vessel are measureable. where the regression matrix Y (ϕ , v ) and the unknown pa-
To simplify the development of the controller design, the rameter vector Θ are defined by the following expression:
system model presented in Eq. (1) is rewritten as
⎡u 0 0 ⎤ ⎡ Xu ⎤
 + D*η = τ *
M *η (7) Y (ϕ , v ) = R(ϕ ) ⎢⎢ 0 v 0 ⎥⎥ , Θ = ⎢⎢ Yv ⎥⎥ . (14)
⎢⎣ 0 0 r ⎥⎦ ⎢⎣ N r ⎥⎦
where the transformation v = RT (ϕ )η and v = RT (ϕ )η 
− S (ϕ ) RT (ϕ )η are utilized. The skew symmetric matrix
S (ϕ ) ∈ R 3 x 3 is given by Based on the open-loop dynamics of the filtered tracking er-
ror, the transformed control input vector τ * is specified to be
⎡ 0 −ϕ 0 ⎤
S (ϕ ) = ⎢⎢ϕ 0 0 ⎥⎥ .
 (8) τ * = Y (ϕ , v )Θˆ − R(ϕ ) MS (ϕ ) RT (ϕ )η + M *η
⎢⎣ 0 0 0 ⎥⎦ − M * Ke − e − K r r ,

The transformed system matrices M * ∈ R 3 x 3 , D* ∈ R 3 x 3 where K r is defined as a positive definite, diagonal gain
and τ * ∈ R 3 x1 are calculated, respectively, as follows: matrix. The update law, based on the projection presented
below, is defined to generate the bounded parameter estimate
M * = R(ϕ ) MRT (ϕ ) , vector Θˆ (t ) [7]:
D* = R(ϕ )( DRT (ϕ ) − MS (ϕ ) RT (ϕ )), (9)
⎧0 if Θˆ = Θ, − Y T (ϕ , v ) r > 0,
τ = R(ϕ )τ .
 ⎪ ⎪
Θˆ = ⎨0 if Θˆ = 0, − Y T (ϕ , v )r < 0, (16)
⎪ T
⎪⎩ −Y (ϕ , v ) r otherwise.
We noticed that the transformed system matrix M is non-
negative matrix. We describe the position and orientation of
the desired trajectory in the Earth-fixed coordinated frame by Here, Θ denotes the upper bound values for the draft coeffi-
the vector ηd = [ xd , yd ,ϕ d ]T . Without any loss of generality, cients (assumed to be known). The closed-loop dynamics of the
the selected trajectory is assumed to be both sufficiently filtered tracking error signal r (t ) can be obtained as follows:
smooth and bounded ηd , ηd , η d ∈ L∞ . The tracking error de-
noted by e (t ) ∈ R 3 x1 is defined as M * r = −Y (ϕ , v )Θ − e − K r r (17)

e = η − ηd . (10) where Θ = Θ − Θˆ denotes the difference between the actual

and estimated draft coefficients.
In order to simplify the error signals and to facilitate the The non-negative control Lyapunov function is chosen to
1260 V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264

The geometric configuration matrix B(α ) is then calcu-

lated based on the given α i . Finally, the thrust vector f is
chosen for minimal power consumption.
4.2 Solution for varying direction

Determining suitable directions for each tugboat is achieved

by using extended control force decomposition. The thrust of
Fig. 3. Control allocation for unactuated ship using thrusts of tugboats.
each tugboat is separated into two elements relative to x- and
y-directions within the body-fixed coordinate frame. The
analyze the stability of the system: thrust vector f is then represented by the vector f ′ as follows:

1 1 1 f ′ = [ f1x , f1 y , f 2 x , f 2 y , f3 x , f3 y , f 4 x , f 4 y ,]T (21)

V = e T e + Θ T Θ + r T M *r . (18)
2 2 2
where f ix = f i cos α i and f iy = f i sin α i . The geometric con-
By differentiating Eq. (18) and substituting Eqs. (11) and (17) figuration matrix B (α ) is then extended to give B′ :
into the derivative of V(t), we obtain
⎡ 1 0 ... 1 0⎤
V = −e T Ke − r T K r r . (19) ⎢ ⎥
B′ = ⎢ 0 1 ... 0 1⎥. (22)
⎢ −l y1 lx1 ... −l y 4 lx 4 ⎥⎦
Eq. (19) gives the non-positive time derivative of the ⎣
Lyapunov function candidate. Based on the Lyapunov stabil-
ity, it is possible to conclude that the control system is asymp- The vector f ′ is computed using the Moore Penrose
totically stable. Therefore, the tracking error and its derivative pseudo-inverse matrix [10], a special case of the pseudo-
will converge to zero in a finite amount of time. inverse matrix (presented later in more detail):

4. Formulation of control allocation τ c = B′f ′ ⇒ f ′ = B′* τ c (23)

4.1 Control allocation problem

where B′* = B′T ( B′B′T ) −1 . The direction of the tugboats can
In this paper, the control allocation is formulated as shown then be found with:
in Fig. 3. It determines the direction α i and the required force
f i for each individual tugboat from the desired control input ⎡α if α i < max(α min ,α i −1 − ∆tα ),
vector τc , which is produced from the controller. In this case, ⎢
⎛ f ⎞
the control allocation is optimized [8, 9] with the constraints α i = ⎢⎢ tan −1 ⎜ iy ⎟ if α ≤ α i ≤ α ∀i ∈ (1,...,4), (24)
⎢ ⎝ f ix ⎠
of contact angles, slowly varying direction and limited push-
ing force such that ⎢⎣α if α > min(α max ,α i −1 + ∆tα ).
(1) τ c − τ a (α i , f i ) is small to minimize the error between
the actual thrust and the desired signal from the controller. Note that with the limitation on the rate of direction change,
(2) τ a (α i , f i ) is small to minimize the power is supplied to we can decrease the jump in the direction of a tugboat at each
tugboats. sample.
(3) α i (t ) changes slowly to match the dynamic response of
tugboats and to minimize the wear and tear on the thrust 4.3 Solution for limited pushing force
The approach presented in this paper can be outlined as de- In this paper, the control force optimization problem is
scribed below: solved by using the RPI approach. This approach is a con-
First, calculate the angles α i to determine the direction of strained optimization technique. The objective of minimizing
the tugboats. The generated angles must satisfy the power supplied to tugboats can be written as follows:
α min ≤ α ≤ α max . This limitation should be chosen so as to
avoid any slips. Furthermore, the angles should be subject to min J = min (f + c)T W(f + c) (25)
the slowly varying constraint α i < α . These requirements
f f 2
can be combined as α ≤ α i ≤ α , where subject to

⎧α = max(α min ,α i −1 − ∆tα ), τ c − B (α ) f = 0

⎨ (20) (26)
⎩α = min(α max ,α i −1 + ∆tα ). f min ≤ fi ≤ f max ,
V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264 1261

where W ∈ R nxn is the weighting matrix and c ∈ R n is the

offset vector. To solve this problem, we find the Hamiltonian

H = ( f TWf + c TWf + f TWc + c TWc ) +
2 (27)
+ξ ( B (α ) f − τ c ),

where ξ ∈ R n is an undetermined Lagrange multiplier. Tak-

ing the partial derivative of H and setting the results to zero,
we obtain the following relations:
Fig. 4. Cybership I [11], supply vessel scale 1:70.
∂H 1 1
= Wf + (cTW )T + Wc + (ξB (α ))T = 0
∂f 2 2 (28) dynamics using the mathematical model along with the con-
⇒ Wf = −Wc − B (α ) ξ , T T trol approach described above.
Cybership I [11], scale 1:70, which is a model of an off-
and shore supply vessel with four thrusters in the configuration as
shown in Fig. 4, is used in the simulation. It is noted that these
∂H actuators are not used and the motion of ship is done by tug-
= B(α ) f − τ = 0 ⇒ B(α )W -1Wf = τ c boats. The model ship has a mass of 17.6[kg] and a length of
∂ξ (29)
1.19[m]. The center of gravity is located at xg=-0.04[m]. This
⇒ B(α )W [ −Wc − B(α ) ξ ] = τ c .
-1 T T
is also the origin in the body fixed coordinate system. Hydro-
dynamic coefficients of the ship are described as follows:
Solving Eq. (29), we find that
⎡19[kg] 0 0 ⎤
ξ T = −( B (α )W -1 B(α )T ) −1 (τ c + B(α )c ). (30) ⎢
M =⎢ 0 35.2[kg] -0.7[kg ⋅ m 2 ] ⎥⎥ ,
⎢⎣ 0 -0.7[kg] 1.98[kg ⋅ m 2 ]⎥⎦
Substituting Eq. (30) into Eq. (28), we obtain the following:
D =diag {4[kg/s], 6[kg/s], 1[kg ⋅ m 2 /s]}.
Wf = −Wc + B (α ) ( B (α )W B (α ) ) ( τ c + B (α )c )
T −1 T −1

⇒ f = −c + W −1B (α )T ( B (α )W −1B (α )T ) −1 (τ c + B (α )c ). The configuration of the tugboats around the ship are de-
scribed as
If we set B* = W −1B (α )T ( B (α )W −1B (α )T ) −1 , Eq. (31) is
simplified as follows: (l1x , l1 y ) = (0.41, −0.15), (l2 x , l2 y ) = (−0.41, −0.15),
(l3 x , l3 y ) = (−0.41,0.15), (l4 x , l4 y ) = (0.41,0.15).
f = −c + B* ( τ c + B (α )c ). (32)
The slowly varying direction constraint emphasizes that the
Note that if W is the identity matrix, B* is called the set of initial directions α10 ,α 20 ,α 30 ,α 40 considerably affect
Moore–Penrose pseudo-inverse matrix. the direction and control force of the tugboats. In this simula-
After solving the force distribution problem using Eq. (32) tion, varying direction constraint is:
with c initially a zero vector, if no element of the thrust vec-
tor f exceeds the minimum or maximum value, the process π
α = [rad/s] . (35)
stops. However, if one of the elements exceeds the limits, the 90
problem is solved again with Eq. (31) modified as follows:
(1) The zero vector is set to all the elements of the i-th column Constraints about limitation of pushing force and contact
of matrix B(α ) , which corresponds to the position of the angle of each tugboat are chosen as follows:
saturated fi.
(2) The i-th element of vector c is set as the negative of the
saturated value. f min = 0, f max = 0.5[N],
π 5π
α1min = α 2 min = , α1max = α 2 max = , (36)
5. Simulation results 6 6
The primary focus of the simulation is to investigate the −5π −π
α 3 min = α 4 min = , α 3 max = α 4 max = .
performance of the controlled system, as well as tugboats 6 6
1262 V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264

Fig. 7. Ship motion for berthing.

Fig. 5. Subset of moments attainable using the redistributed pseudo-

inverse (RPI) algorithm.

Fig. 8. Ship heading response.

Fig. 6. Planning route for ship berthing.

Fig. 5 shows the attainable moment subset ( M φ ) in the

case α1 = α 2 = π / 3 and α 3 = α 4 = −π / 3 , namely, the result
of mapping the limited pushing force constraint of all four tug- Fig. 9. Ship velocity response.
boats in 4-D onto the control input vector in 3-D. This subset is a
hyper cube. The subset of moments attainable using the RPI this phase, the directions of tugboats are fixed as
approach ( M π ) is located inside the M φ . The RPI approach α1 = α 2 = π / 2 and α 3 = α 4 = −π / 2 .
exhibits good efficiency with a high ratio between vol- The control input vector τ c is calculated from the proposed
umes M π and M φ . We notice that, if the vector τ c is located adaptive controller. This system is simulated with
inside the volume M π , the error between the actual control input K = diag{1,1,1} and K r = diag{0.3,0.3,0.3}. These matrices
vector τ a and the desired control input vector τ c is zero. are chosen to match the subset of attainable moments as de-
For safety berthing, the planning route is separated into two scribed above.
phases as shown in Fig. 6. The first, where the ship moves Figs. 7-9 show the responses during berthing. It is clearly
from point A to point B is called approaching phase and the shown that good performances achieved in both geometric
second, from point B to C is the berthing phase. task and dynamic task. The designed controller and proposed
In this simulation, firstly, the ship is operated to move on control allocation approach guarantee that the ship follows a
the straight line from starting point A (0,0) where the initial given trajectory as well as achieves the desired velocity. In
heading angle is π / 3 and stop at the point B (10,10) where both phase, the heading angle changes from the initial value to
the desired heading angle is π / 4 . The initial directions of desired value in the limited time by combination of tugboat
tugboats in this phase are α1 = α 2 = π / 3 and thrusts. After that, the heading angle is kept and the ship fol-
α 3 = α 4 = −π / 3 . After that, the ship is maneuvered from the lows defined trajectory with high accuracy. Based on the ship
point B (10, 10) to point C (10, 15) in the berthing phase. In performance, it is ensured that the ship can move to the berth
V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264 1263

without collision between ship and the others located in the

harbor as well as between ship and the berth.
Fig. 10 evaluates the efficiency of the proposed control allo-
cation approach. Furthermore, it shows the forces and moment
supply to ship by combination of four tugboats. Based on this
figure, it is clearly to see that the proposed approach can mini-
mize the error between the actual thrust and the desired signal
from the controller. Additionally, it is shown that, in the ap-
proaching phase, to maneuver ship in the straight line from A to
B, 4 tugboats just maintain the surge force τ x . The sway force
τ Y and yaw moment τ z are reduced to zero to avoid the un-
desired motion in Y direction as well as heading angle variation.
In the berthing phase τ X and τ z are decreased to zero.
Figs. 11 and 12 depict the performance of four tugboats dur-
ing berthing. The resulting thrusts and directions of tugboats
satisfy the constraint about limited pushing force and slow
change direction shown in Eqs. (35) and (36). In the approach-
ing phase, the force supplied to ship is produced from the 4
tugboat thrusts synchronously to pass up the actuator satura-
tion. However, in the berthing phase, the tugboat 1 and 2 are
used as main actuators, tugboat 3 and 4 are just used to avoid
Fig. 10. Thrust force comparison: solid lines depict forces commanded the collision between the ship and berth.
by the controller, dashed line represents actual thrusts supplied to the
6. Conclusion
In this paper, we proposed a new approach for ship berthing
with the assistance of autonomous tugboats. An adaptive con-
troller was presented to take into account the uncertainty of
system parameters. The control allocation was considered as
an optimization problem under the constraints that a tugboat
can only exert a limited pushing force and that it can only
slowly change directions. The efficiency of the proposed ap-
proach was evaluated through using a model ship in a Matlab
simulation. It exhibited good performance and revealed the
possibility of extending these results to future studies by test-
ing a model ship under actual conditions. The combination of
tunnel thrusters and the assistance of one, two or more tug-
boats will be studied to determine suitable solutions for vari-
ous ship berthing situations.
Fig. 11. Tugboat thrusts during berthing.
This research was a part of the project titled “A Development
of Highly Efficient Port Cargo Handling System” funded by the
Ministry of Land, Transport and Maritime Affairs, Korea.

[1] T. Takai and H. Yoshihisa, An automatic maneuvering sys-
tem in berthing, Proc. 8th Ship Control System Symposium, 2,
The Hague (1987) 183-201.
[2] Y. A. Kasasbeh, M. A. Pourzanjani and M. J. Dove, Auto-
matic berthing of ship, Proc. of the Institute of Marine Engi-
neers 3rd International Conference on Maritime Communi-
Fig. 12. Tugboat directions in approaching. cations and Control, London (1993) 10-17.
1264 V. P. Bui et al. / Journal of Mechanical Science and Technology 25 (5) (2011) 1257~1264

[3] Y. Zhang, G. E. Hearn and P. Sen, A multivariable neural Hideki Kawai received his B. S., M. S.
controller for automatic ship berthing, Journal of IEEE Con- and Ph.D. degrees in Information Sys-
trol System, 17 (1997) 31-44. tems Science from Soka University,
[4] K. Hasegawa and T. Fukutomi, On harbour manoeuvring Japan, in 2004, 2006 and 2010, respec-
and neural control system for berthing with tug operation, tively. He is currently a research profes-
Proc. 3rd International Conference on Manoeuvring and sor at Pukyong National University in
Control of Marine Craft, UK (1994) 197-210. Busan, Korea. His research interests
[5] K. A. Bordingnon, Constrained control allocation for sys- include robust image processing and image sensing system
tems with redundant control effectors, Phd. Thesis, Dept. design.
Aerospace & Ocean Eng., Virginia Polytechnic Institute &
State Univ., Blacksburg, VA, 1996.
[6] T. I. Fossen, Marine control system- guidance, navigation, Young-Bok Kim received his B.S. and
rigs and underwater vehicle, Trondheim, Norway, Norwe- M.S. degrees in Maritime Engineering
gian University of Science and Technology, 2002. from National Fisheries University of
[7] J. Esposito, M. Feemster and E. Smith, Cooperative manipu- Busan, Korea, in 1989 and 1991, respec-
lation on the water using a swarm of autonomous tugboats, tively. He then received his Ph.D. de-
IEEE International Conference on Robotics and Automation, gree from Kobe University, Japan in
Pasadena, USA (2008) 1501-1506. 1996. Dr. Kim is currently a Professor at
[8] M. Bodson, Evaluation of optimization methods for control the Department of Mechanical System
allocation, Journal of Guidance Control and Dynamic (25) 4 Engineering at Pukyong National University in Busan, Korea.
(2002) 703-711. His research interests include control theory and application
[9] O. J. Sordalen, Optimal thrust allocation for marine vessels, with Dynamic Ship Positioning and crane control system de-
Journal of Control Engineering Practice (5) 9 (1997) 1223-1231. sign etc.
[10] G. Strang, Introduction to linear algebra, Wellesley Cam-
bridge Press, 2003.
[11] J. P. Strand, Nonlinear position control system design for Kwon Soon Lee received the B.S
marine vessels, PhD. Thesis, Dept of Engineering Cybernet- degree from Chungnam National
ics, Norwegian University of Science and Technology, 1999. University, Daejeon, Korea in 1977 and
the M.S degree from Seoul National
Van Phuoc Bui received B.S in University, Seoul, Korea in 1981. He
Mechatronic Engineering in 2005 at then received his Ph.D. degree from
HoChiMinh University of Technology Oregon State University in Corvallis,
(HCMUT) and M.S in Department of OR, U.S.A. in 1990, all in electrical
Mechanical and Control System Engi- engineering. Dr. Lee is currently a professor at the department
neering at the Graduate School of of electrical engineering at Dong-A University in Busan, Ko-
Pukyong National University (PKNU) in rea since 1982. From 2003 to 2008, he was a director of Na-
Busan, Korea, in 2009. He is currently a tional Research Lab. nominated by Korean government. His
doctoral student at PKNU. His research interests include con- research interests include intelligent control theory and appli-
trol theory with applications to marine control system. cation to industrial field.

View publication stats