Article

A dynamic model and a robust

controller for a fully-actuated marine
surface vessel*

Journal of Vibration and Control

17(6) 801812
! The Author(s) 2010
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1077546309346245
jvc.sagepub.com

N Khaled and NG Chalhoub

Abstract
A nonlinear six degree-of-freedom dynamic model is presented for a marine surface vessel. The formulation closely
follows the current literature on ship modeling. It considers the effects of inertial forces, wave excitations, retardation
forces, nonlinear restoring forces, wind and current loads along with linear viscous damping terms. The capability of the
model is shown through its prediction of the ship response during a turning-circle maneuver. The ship model is used
herein as a test bed to assess the performance of the proposed controller. The present study assumes that the ship is
fully actuated and all state variables of the system are available through measurements. A nonlinear robust controller,
based on the sliding mode methodology, has been designed based on a reduced-order version of the ship model. The
latter accounts only for the surge, sway and yaw motions of the ship. The initial simulation results, generated based on
the reduced-order model of the marine vessel, demonstrate robust performance and good tracking characteristics of the
controller in the presence of structured uncertainties and external disturbances. Furthermore, they illustrate the adverse
effects of the physical limitations of the propulsion system on the controlled response of the ship. Next, the same
controller is implemented on the six degree-of-freedom model of the ship. The simulation results reveal tracking
characteristics of the controller that are similar to those observed in the initial results, in spite of significantly larger
modeling uncertainties.

Keywords
Nonlinear robust controller, ship modeling and control of marine surface vessels, sliding-mode controller
Received: 27 October 2008; accepted: 27 July 2009
*This paper was contributed by Professor Raouf Ibrahim

1. Introduction
The dynamic behavior of marine surface vessels is
highly nonlinear. Moreover, it is signicantly inuenced by environmental disturbances induced by
winds, random sea waves and currents. Therefore,
good track-keeping and course-changing characteristics
of the ship can only be achieved by implementing controllers that are robust to both modeling uncertainties
and external disturbances.
Nonlinear control theory has been extensively used
in both track-keeping and course-changing maneuvers
of marine vessels (Fossen, 2000; Pivano et al., 2007).

However, many of these compensators, such as state

feedback linearization techniques (Fossen, 1993; Berge
et al., 1998; Moreira et al., 2007), output feedback controllers and backstepping schemes (Godhavn, 1996;
Fossen and Grvlen, 1998; Strand et al., 1998; Fossen
and Strand, 1999; Pettersen and Nijmeijer, 2001) are
Department of Mechanical Engineering, Wayne State University, Detroit,
MI, USA
Corresponding author:
Nabil G Chalhoub, Department of Mechanical Engineering, Wayne State
University, Detroit, MI, USA
Email: nchalhoub@eng.wayne.edu

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

802

Journal of Vibration and Control 17(6)

model-based schemes. As a consequence, these techniques are susceptible to modeling inaccuracies. On the
other hand, the sliding mode methodology (Slotine and
Li, 1991; Khalil, 1996) enables one to design robust controllers without requiring a full knowledge of the systems nonlinearity (Le et al., 2004). The design of
sliding mode controllers is mainly based on knowing
the upper bounds of the modeling inaccuracies. The performance of these controllers tends to be robust to both
structured and unstructured uncertainties.
The focus of the current work is to develop a nonlinear sliding mode controller to yield a robust performance of the ship during track-keeping maneuvers. The
controller is designed based on a reduced-order model
of the ship, which only considers the surge, sway and
yaw motions. The marine surface vessel is assumed to
be fully actuated and all its state variables are available
through direct measurements.
A six degree-of-freedom nonlinear model for a
marine surface vessel is presented in the next section.
It will be used as a test bed to assess the performance of
the proposed controller. The model follows closely the
existing literature on ship modeling (Newman, 1977;
Fossen, 2005; Perez, 2005). Its formulation accounts
for the wave excitation forces, retardation forces, inertial forces, nonlinear restoring forces, wind and current
loads along with linear viscous damping terms. In addition, the physical limitations of the propulsion system
are accounted for in the model formulation of the ship.
The proposed nonlinear robust controller is discussed
in Section 3. The digital simulations are presented in
Section 4. They illustrate the capability of the nonlinear
six degree-of-freedom model in predicting the ship
response during a turning-circle maneuver. Moreover,
they demonstrate a robust tracking characteristic of the
proposed controller in both the reduced- and the fullorder models of the ship, in spite of signicant modeling dierences between the two cases. Furthermore,
the results reveal the adverse eects of the physical

2. Dynamic model of the ship

The ship is treated as a rigid body having six degreesof-freedom, namely, surge, sway, heave, roll, pitch and
yaw (see Figure 1). Two coordinate systems have been
used. The rst one is an inertial frame fX, Y, Zg whose
origin is located at an arbitrary point on
 the calm
 sea
surface. The second coordinate system, xo , yo , zo , is a
non-inertial, body-xed coordinate system attached to
the ship at point o, which coincides with the center of
oatation of the ship. The xo , zo plane is chosen to
coincide with the vertical plane of symmetry of the ship
hull. The xo  and yo axes are directed towards the
bow and the starboard of the ship.
Following the Society of Naval Architectures and
Marine Engineers (SNAME) convention (1950), both
the position and orientation of the ship are dened with
respect to the inertial frame. However, the ship translational and angular velocity vectors are expressed with
respect to the body-xed frame. The scalar equations,
describing the translational motion of the ship, are
derived from the linear momentum balance. They are
given as

m u_  vr wq  xG q2 r2 yG pq  r_
_ FX
zG pr q

_
m v_  wp ur  yG r2 p2 zG qr  p
xG pq r_ FY

_
m w_  uq pv  zG p2 q2 xG rp  q
_ FZ
yG rq p

where FX , FY and FZ are the components of the resultant force, F, of all externally applied forces on the

ship along the i , j and k directions, respectively.
o o
o
Moreover, the angular momentum balance around
point o will yield the following three scalar equations
governing the rotational motion of the marine vessel:

Ix p_  Ixy q_  Ixz r_  Ixz pq  Iyz q2 Iz rq Ixy pr  Iy qr Iyz r2

myG w_ pv  uq  mzG v_ ur  pw MoX
 Ixy p_ Iy q_  Iyz r_ Ixz p2 Iyz pq  Iz rp Ix pr  Ixy qr  Ixz r2
mzG u_ qw  vr  mxG w_ pv  uq MoY

 Ixz p_  Iyz q_ Iz r_  Ixy p2 Iy pq  Iyz pr  Ix pq Ixy q2 Ixz rq

mxG v_ ur  pw  myG u_ qw  vr MoZ
limitations of the propulsion system on the controlled response of the ship. In conclusion, the current
work is summarized and its main contributions are
drawn.

where M is the resultant of all externally applied

o
moments on the ship. Both F and M reect the eects

o
of wave excitations, retardation forces, wind and
current loads, linear viscous damping terms,

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Khaled and Chalhoub

803

rG
~

p l)
ol
i0 (r

u )
rge
(su

j0
k0

Body-fixed

ro

r*

z0

ay )

q
(pitch )
v

frame

G
r
(yaw )
w
(heave )

(sw

x0

y0

Figure 1. Schematic of the ship hull.

Figure 2. Modified Pierson-Moskowitz wave spectrum.

nonlinear restoring forces along with the control

actions generated by the propeller, the rudder and
side thrusters.
Long-crested sea waves are considered in the current
study. The wave height, h, at an arbitrary point X, Y,
dened with respect to the inertial frame, is commonly
described by (Newman, 1977; Perez, 2005)

determine the wave excitation forcing functions along

with the frequency dependent added mass and wave
damping terms (Newman, 1977; Faltinsen, 1990;
Perez, 2005). In addition, the uid is assumed to
be inviscid, incompressible and irrotational. The
wave excitation forces and moments are computed as
follows

hX, Y, t

650
X

Ai cos!i t "i  kX cos  Y sin 

i1

F jw e t

650 
X
p
Xj , !i  2S!i ! cos!i t "i
i1

3
where Ai and "i are the amplitude and the phase angle
of the ith frequency component of the wave height,
respectively. "i is considered to be a random variable
with a uniform distribution
pbetween 0 and 2. Ai is
determined from 2S!i ! where S! is the wave
spectrum. The latter is assumed to be the Modied
Pierson-Moskowitz wave spectrum. It is dened as
(Perez, 2005)
S!

AS BS =!4
e
!5

where AS 0:312 H21=3 !4o , BS 1:25 !4o and !o is the

modal frequency at which the wave spectrum reaches
its maximum value. The wave spectrum corresponding
to H1=3 5 m and !0 0:69 rad=s is shown in Figure 2.
Moreover, to avoid risking hX, Y, t from repeating
itself, !i is selected randomly in the interval
i  1!, i! (Perez, 2005). It should be noted
that ! is considered to be constant and equal to
0.01 rad/s.
The formulation of the seakeeping problem, which
customarily considers the ship motion to be harmonic
with small amplitudes, has been used herein to

 kX cos  Y sin  j , !i
j 1, . . . , 6

for

where    
is the wave encounter angle.
Moreover, Xj , !i  and j , !i are the magnitude
and phase angle of the force transfer function dened
by the ratio of the wave excitation force inuencing the
jth degree-of-freedom of the ship over the wave amplitude. The six force transfer functions are determined
numerically by using a 3-D potential theory software
WAMIT (Lee and Newman, 2004). It should be mentioned that the latter does not account for the eect of
the ship forward speed.
The frequency dependent added mass, akl !, and
wave damping, bkl ! terms are also computed by
using WAMIT (Lee and Newman, 2004) for a frequency range between 0 and 6.5 rad/sec. The impulse
response kkl t in the kth direction due to a unit velocity
impulse in the lth direction can be related to the wave
damping term, bkl !, as follows (Ogilivie, 1964;
Kristiansen et al., 2005)
2
kkl t


Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Z1
bkl !  bkl 1 cos! t d!
0

804

Journal of Vibration and Control 17(6)

The convolution integral associated with kkl t, based

on an arbitrary velocity term _l in the lth direction, can
be written as
Z1

kkl t   _l  d

for

k, l 1, . . . , 6

This will result in a 6  6 retardation matrix Kt.

Following the procedure outlined by Kristiansen et al.
(2005), the singular value decomposition method was
used to generate a non-minimal state space realization
for a single-input single-output (SISO) system whose
input and output variables are _l t and ykl t, respectively. A model reduction procedure was then implemented to reduce the order of the state space realization
to eight without signicantly compromising its accuracy.
This is illustrated in Figure 3 for the case of k15 t. The
state space representation corresponding to the k, l
entry of the retardation matrix Kt can be described as
kl
_ Akl  kl Bkl _l


ykl Ckl  kl Dkl _l

Figure 3. Curves illustrating the accuracy of the state

space formulation.

In the current study, the seakeeping problem will be inuenced by the actual motion of the ship through the input
term, which is considered herein to be the perturbation in
the lth velocity component of the ship. Therefore, _l is
dened to be the variations around the moving average
value of the instantaneous lth velocity component of the
ship. The moving average is determined by implementing
a forgetting factor, which puts signicantly heavier
weights on recent than on older data of the ship velocity.
Furthermore, the retardation force, F retardation
, representk
ing the memory eect in the P
kth equation of motion of the
ship can be evaluated from 6l1 ykl .
Next, the buoyancy force and moment are computed
based on the instantaneous submerged volume of the
ship with respect to the sea free-surface. These forcing
functions, which are balanced by the ships own weight,
are determined by integrating over the entire submerged volume of the ship. This is done herein by dening a 3-D mesh that partitions the ship hull into
32000 cubes (see Figure 4). The dimensions of each
cube are selected to be 5, 2 and 0.04 m in the i , j
o o
and k directions, respectively. The computation of the
o
instantaneous submerged volume of the ship involves
the evaluation of a degree of submergence, ,
for each block. i corresponding to the ith block is
dened by


i

Zc  h Xic , Yic , t
1

i sat
9
block thickness
2


where Xic , Yic , Zic are the coordinates of the centroid of
the ith block (see Figure 4) and h is the elevation of the sea
free-surface at Xic , Yic . Both the centroid and h
are dened with respect to the inertial frame. The lower
and upper saturation limits are set to 0 and 1, respectively.
Note that i 0 reects the case in which the ith block is
located above the sea free-surface. However, i 1 and
0 5 i 5 1 correspond to total and partial submergence
of the ith block, respectively. The instantaneous submergedP
volume of the ship can now be computed from
Vsub 32000
i1 i Viblock . The coordinates of the center of
buoyancy (CB) are calculated as follows

Figure 4. Centroids of the blocks in the 3-D mesh of the ship.

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Khaled and Chalhoub

805

P32000 
xCB

i1

P32000
yCB

i1

P32000
zCB

i1



xiblock i Viblock
Vsub



yiblock i Viblock
Vsub



i ziblock i Viblock
Vsub

those employed in the calculation of the wind loads

with the exception that they are only being applied on
the submerged portion of the ship. They are given by
fxc 0:5Cxc water V2rc Lpp T
10

11

Tpr =8CQ water V2r D3pr

13

where Vr is the speed of the ow in the blade section

evaluated at 0:7Dpr =2. It is given by
Vr

q

2
V2pr 0:7npr Dpr

14

where npr is the angular velocity of the propeller shaft.

Vpr is the mean value of the uid speed at the entrance
of the propeller disk. It is determined from (Blanke,
1982; Journee and Massie, 2001)


Vpr 1  wf u

15

Typical values for the wake fraction number, wf , are

between 0.1 and 0.4 (Fossen, 1994). It is assumed
herein to be 0.1. Moreover, for a ship maneuvering
task, four-quadrant data on dimensionless thrust, CT ,
and torque, CQ , coecients should be used (Kuiper,
1992; Journee and Massie, 2001; Roddy et al., 2006).
Kuiper (1992) provided such data for a B4-70 openpropeller as a function of the hydrodynamic pitch
angle,
, which is calculated as follows

3
2.5
Righting Arm, GZ [m]

where Vrc is the velocity of the current relative to the

ship. The numerical values for Cxc , Cyc and Czc are
obtained from the OCIMF report (1994).
The propeller is responsible for delivering the thrust
required to keep the ship on track. The present work
considers a xed-pitch, sub-cavitating, Wageningen
B4-70 screw propeller (Journee and Massie, 2001;
Roddy et al., 2006). The propeller thrust, Fth , and the
corresponding torque, Tpr , that should be applied on
the propeller shaft are determined from (Journee and
Massie, 2001)

Fth =8CT water V2r D2pr

where Vrw 10 is the wind velocity relative to the ship

evaluated at 10 m above the calm sea surface. Note that
mzw is applied about an axis perpendicular to the calm
sea surface and passing through the midpoint between
the aft and forward perpendiculars of the ship. The
formulation for determining Cxw , Cyw and Czw coecients are provided by Isherwood (1973).
The formulation used for computing the current
induced forces and yawing moment are similar to

1.5
1

0:7Dpr =2 arctan

0.5

0
0

12

mzc 0:5Czc water V2rc Lpp T

The righting arm curve of the current ship is shown in

Figure 5, which reveals a range of stability of around
84.3o. Furthermore, the hydrostatic moment
 is deter
mined herein with respect to the origin of xo , yo , zo .
Moreover, linear viscous damping forces and
moments

  are introduced as mbu u, mbv v, mbw w,
Ix bp p, Iy bq q, Iz br rT where bu , bv , bw , bp , bq and br
are chosen herein to be 1, 1, 3, 8, 8 and 8, respectively
(Ueng et al., 2008).
The wind resistive forces and yawing moment are
computed as follows (Isherwood, 1973; Fossen, 1994;
Oil Companies International Marine Forum (OCIMF),
1994; Journee and Massie, 2001; Perez, 2005)
fxw 0:5Cxw air AT V2rw 10
fyw 0:5Cyw air AL V2rw 10
mzw 0:5Czw air AL V2rw 10L

fyc 0:5Cyc water V2rc Lpp T

20

40

f [deg]

60

80

Figure 5. Righting arm curve of the ship.

100

Vpr
0:7npr Dpr

16

Note that the present formulation has some provision to

ensure that the propeller thrust, specied by the controller, does not exceed the physical limitation of the ship
propulsion system. This is done by using equation (14)

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

806

Journal of Vibration and Control 17(6)

into equation (13) to substitute Vr by its expression

with respect to npr in both Fth and Tpr . The resulting
expression of Fth will then be implemented to solve for
npr that is required to yield the thrust assigned by the
controller. If npr is less than the maximum speed deliverable by the ships engine then its value will be used;
otherwise, npr is set to its maximum value. Based on the
current value of npr , equation (13) will be used to compute Tpr . Once again, if the calculated value of Tpr is
smaller than the maximum torque that can be generated
by the ships propulsion system then both npr and Fth will
be used in the digital simulation. Otherwise, npr will be
recalculated based on the maximum value of Tpr . The
new npr value will now be used to compute the deliverable thrust, Fdeliv
th . This value will replace the unrealistic
Fth assigned by the controller.
The lift, fL , and drag, fD , forces generated by the
rudder are determined from (Perez, 2005; Journee and
Pinkster, 2002)
fL 0:5CL water V 2rud Arud

17a

0:5CD water V 2rud Arud

17b

fD

Both fL and fD are applied at the center of pressure, CP

(Journee and Pinkster, 2002; Perez, 2005). The numerical values for CL and CD , based on an aspect ratio
AR of 6 and a rudder section between 0.06 and
0.18, are provided by Abbott and Von Doenho
(1958) and Journee and Pinkster (2002). In evaluating
Vrud , the eect of the propeller on the ow heading
towards the rudder has been accounted for by considering an idealized, steady, one-dimensional ow
through the propeller. The latter is modeled by a thin
actuator disk across which, the ow velocity is considered to be continuous while the pressure is assumed to
undergo a sudden change. Based on this simplied
model, one can express Vrud as (Lewis, 1988; Fox and
McDonald, 1992; Perez, 2005)

Vrud

2Fdeliv
th

V2pr
water Adisk

18

In the initial phase of this work, the rudder dynamics

have been ignored. Therefore, the lift force is considered to be directly assigned by the controller and the
drag force is computed according to equation (17b).
The force and the moment exerted on the ship due to
the rudder are (see Figure 6)
f fD io fL jo
rud

and

M r

rud

 oCP

 f
rud

19

d xr
fD

Cp

rud

Vrud

V pr

Or

fL

Mrud

io
~

xo

jo
~

s
yo

Figure 6. Top view of the rudder.

where the component of M , induced by fL , reprerud

sents the control action responsible for yielding the
desired ship heading.

3. Controller Design
The controller is designed based on a reduced-order
model whose formulation is obtained from equations
(1) and (2) by only retaining the second order dierential equations governing the surge, sway and yaw
motions of the ship. In addition, all terms pertaining
to the heave, pitch and roll motions have been deleted
from the selected
equations.
R
R
R Dening the state vector to
be xT u d, v d, r d, u, v, r, the three second

order dierential equations of motion can be converted
to six rst order state equations that can be expressed in
the following compact form:
 
 
x_ f x G x u


 

 

20

 
where f x represents the eects of inertial forces,
 
gravitational
acceleration, buoyancy forces, wave excitation forces, retardation forces, linear viscous
  terms,
wind and current resistive forces. The G63 x matrix

can be expressed as 033 diag g1 , g2 , g3 T .
In the current work, the marine vessel is assumed to
be fully actuated. Its three control variables represent
cont
the propeller thrust, Fcont
th , the side thrust, F th sway , and
the rudder lift force, fL . As a consequence, the control
(20) can now be dened as
hvector, u , in equation
i


cont
Fcont
th , Fth sway , fL . Moreover, all state variables of
the system are assumed to be available through
measurements.

The reduced-order model, described in equation

(20), has been used in the current work to design a
nonlinear robust controller in order to control the
ship motion during maneuvering tasks. The proposed

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Khaled and Chalhoub

807

controller is designed based on the sliding mode methodology.

The latter does not require exact knowledge of
 
f x . As a result, a nominal
model,
 
  based on an
 
approximate expression f^ x for f x , has been used
 
 
in the design of the compensator.
It should be stressed
that the good performance of the controller hinges
upon the full knowledge of the upper bounds of the
modeling uncertainties,
which are dened by
 
    
fi  fi x  f^i x   Fi with i 3, . . . , 6. For the


ship maneuvering problem, the following three sliding
surfaces are selected:
si e_i i ei

i 1, . . . , 3

21

where the ith tracking error is dened to be ei xi  xid .

The control signals are calculated from
1
ui uieq  g1
i ki sgnsi uieq  gi ki satsi =i

i 1, . . . , 3

22

where the equivalent control laws, uieq s, are determined

by setting s_i 0 with i 1, . . . , 3. The switching terms
are responsible for keeping the system on the sliding
surfaces. Their gains, ki s, are determined by satisfying
the following sliding conditions:

1d 2
s ei , t  i jsi j i 1, . . . , 3
2 dt i 

23

To alleviate the chattering problem associated with the

switching terms, the sgnsi functions are replaced by
the saturation functions satsi =i where i is the thickness of the boundary layer surrounding the ith sliding
surface. In the current study, 1 , 2 and 3 are
selected to be 0.1 m, 0.1 m, and 0.01 rad, respectively.

current speed was dened to be 1 m/sec. Moreover,

the Modied Pierson-Moskowitz wave spectrum, discussed in Section 2 and shown in Figure 2, was used
in the simulation.
In assessing the performance of the six degree-of-freedom model, a two-segment maneuver was used. The rst
segment consists of a straight line while the second segment is a turning-circle maneuver with the wave encounter angle varying based on the instantaneous heading of
the ship. The propeller thrust, Fth , was kept constant at
its maximum value of 5107 N, while the rudder angle
of attack, rud , was assigned 0o and 25o values in the
rst and second segments of the maneuver, respectively.
The simulation results are illustrated in Figure 7. The
tactical diameter, DT (see Figure 7), is found to be
340.6 m. This leads to a non-dimensional
tactical
diameter of 2:43 (D0T DT jrud j=35Lpp ), which
agrees with experimental data provided by Lewis
(1988) and Barr et al. (1981) for ships of comparable
geometric dimensions and weight. Furthermore, ships
undergoing turning-circle maneuvers are expected
to exhibit lateral drifts (Lewis, 1988). This fact is
conrmed in Figure 7, which reveals a lateral drift,
LD , of 4.5 m.
Next, the performance of the robust controller is
assessed by considering the reduced-order model
described in equation (20). The desired trajectory is a
straight line parallel to the X-axis. The desired Y-coordinate is kept constant at 10 m while the desired yaw
angle is set to zero. The desired ship velocity prole in
the surge direction involves an acceleration period, a
constant velocity period, and a deceleration period.
The initial conditions of the ship are selected to be
X0, Y0, 0 1 m, 9 m, 0 rad with zero initial
translational and angular velocities. Therefore, the

50

LD

4. Digital simulation results

50

Y [m]

The purpose of the digital simulations is threefold.

First, it focuses on examining the capability of the nonlinear six degree-of-freedom model in predicting the
ship response in maneuvering tasks. Second, it serves
to assess the performance of the proposed nonlinear
robust controller in the presence of signicant structured uncertainties. This was done by implementing
the controller on the reduced-order model. Third, the
controllers performance is examined under both structured and unstructured uncertainties by implementing
it on the six degree-of-freedom model of the ship.
Throughout this study, the simulations results were
generated based on a barge of 100 m in length, 20 m in
beam and 4.8 m in draft sailing in a following sea. The
wind speed was considered to be 30 m/sec and the

150

DT

250

350

100

100

200

X [m]

Figure 7. Turning-circle maneuver of the ship.

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

300

808

Journal of Vibration and Control 17(6)

layers surrounding the sliding surfaces used in the
design of the controller. The magnitude of the errors
is directly impacted by the selected thickness of the
boundary layers.
Next, the performance of the proposed controller
is assessed in the presence of structured and unstructured uncertainties along with signicant external disturbances. This is done by applying the controller on

4
uship

Surge Velocity [m/sec]

3.5

ud

3
2.5
2
1.5
1
0.5
0

20

40
60
Time [sec]

80

100

Figure 9. Desired and actual ship velocity in surge.

Xerror [m]

task of the controller is to ensure that the ship converges to

and tracks the desired trajectory in the presence of modeling uncertainties and external disturbances. It should be
stressed that the nonlinear reduced-order model of the
ship, accounts for the eects of inertial forces, nonlinear
restoring forces, wave-induced excitations, retardation
forces, viscous damping along with wind and current
loads. The results are shown in Figures 8 to 11. The initial
error in the position of the ship has caused the controller to
yield a higher surge velocity than the desired one in order
to force the ship to converge to its desired trajectory. This is
illustrated in the rst two seconds of the results in Figures 8
and 9. However, once the convergence has been achieved
then the controller forced u to converge to ud during the
remainder of the acceleration period (see Figure 9).
Another important aspect of the results stems from the
tracking errors that occur during the constant velocity
period of the maneuver (see Figure 9). Figure 10 reveals
that these errors occur during brief periods where the propulsion system of the ship has reached its physical limitations and the torque applied on the propeller shaft is locked
at its maximum value. During these periods, the controller
loses its ability to provide the necessary corrective action
that is needed to keep the ship on track; thus, allowing the
tracking error to grow. Once the environmental resistive
forces and moments on the ship decrease in magnitude, the
required Fcont
th will decrease and its corresponding Tpr will
fall below the upper bound set by the physical limitations
of the propulsion system. As a consequence, the ship will
regain its controllability characteristic and the controller
will be able to eliminate the tracking error. Note that a
saturated value of Tpr does not necessarily mean that
Fcont
th will be saturated as shown in Figure 10.
Figure 11 demonstrates the capability of the controller in correcting for the sway and yaw tracking errors. It
should be pointed out that the steady-state errors
exhibited in Figures 10 and 11 stem from the boundary

0.5

0
0

x 10

20

40

60

80

100

deliv
F th
[N]

250
Xd

200

50

Tpr [Nm]

X [m]

100

1
150

50

x 10

1
20

60
40
Time [sec]

x 10

Figure 8. Desired and actual ship responses in surge.

70
80

100

70
80

100

2.3

2.2
20

2
80

60
60

1
0

40

20

100

2.258
2.256

0
1
0

150

50

x 10

50
60
40
60
Time [sec]

100

Figure 10. Error in the surge position of the ship along

with the deliverable thrust and its corresponding propeller
shaft torque.

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Khaled and Chalhoub

809
because the same controller was used herein for both
the reduced- and the full-order models of the ship.
Thus, the controller had to perform the same task
with signicantly higher modeling errors in the case
of the full-order model than in the case of the
reduced-order model. Moreover, Figure 15 shows that
the error in the yaw angle remains well within the

4
ud
3.5

Surge Velocity [m/sec]

the six degree-of-freedom model of the ship. The initial

conditions of the ship are selected to be X0, Y0,
Z0, 0, 0, 0 1 m, 9 m, 0 m, 0 rad, 0 rad,
0 rad with zero initial translational and angular
velocities. The desired trajectory is kept the same as
the one used in assessing the controller performance
on the reduced-order model of the ship. Figures 12 to
14 demonstrate the robustness of the controller in the
presence of unstructured uncertainties by yielding a
surge response of the ship that is similar to the one
observed in the case of the reduced-order model.
A comparison of Figures 9 and 13 reveals an increase
in the tracking error of u during the constant velocity
period of the maneuver. The slight deterioration in u
is attributed to the increase in frequency and duration
for which Tpr was locked at its maximum value (see
Figures 10 and 14). Such a behavior was expected

u ship

3
2.5
2
1.5
1
0.5
0
0

80

60
40
Time [sec]

20

100

0.5

Figure 13. Desired and actual velocity in surge generated

based on the full-order model of the ship.
0

2
error [rad]

20

x 10

40

60

80

100

0
1

20

40
60
Time [sec]

80

100

Xerror [m]

Y error [m]

0.5

0
0

Figure 11. Tracking errors in sway and yaw.

x 10

20

40

60

80

100

70
80

100

70
80

100

deliv
Fth
[N]

2
250
Xd
200

150

x 10

1
2.26

2.25

1
0

Tpr [Nm]

X [m]

100
50
0
50

x 10

20

40
40

50

60
60

2
x 10

2.6

2.2
100
150

1
0
0

1
20

40
60
Time [sec]

2
0

2
80

Figure 12. Desired and actual displacements in surge

generated based on the full-order model of the ship.

20

100

40
50
60
40
60
Time [sec]

Figure 14. Error in the surge position of the ship along with
the deliverable thrust and its corresponding propeller shaft
torque generated based on the full-order model of the ship.

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

810

Journal of Vibration and Control 17(6)

0.03

0.02

0.5
0.01

1
0
2

x 10

20

40

60

80

100

q [rad]

Y error [m]

error [rad]

0.01

0
0.02

2
4
0

0.03

20

40
60
Time [sec]

80

100

Figure 15. Tracking errors in sway and yaw generated based on

the full-order model of the ship.

40
60
Time [sec]

80

100

x 10

0.4
0.2

f [m]

Z [m]

20

Figure 17. Pitch angular displacement of the ship.

0.6

0.2

0.4

0.6

0.8

20

40
60
Time [sec]

80

100

4
0

20

60
40
Time [sec]

80

100

Figure 16. Ship displacement in heave.

Figure 18. Roll angular displacement of the ship.

thickness of 0:01 rad assigned for the boundary layer

surrounding the sliding surface s3 . In addition, for this
particular maneuver, Figures 16 and 17 illustrate that,
the ship undergoes a realistic heave and pitch angular
displacement. The roll motion was one order less in
magnitude than the pitch angular displacement (see
Figures 17 and 18). This is because a following seastate was considered in this study along with a straight
line desired trajectory; thus, resulting in an incident
wave angle of 0o.

existing literature on ship modeling. It accounts for

the eects of inertial forces, wave excitation forces,
retardation forces, nonlinear restoring forces, linear viscous damping terms, wind and current loads. In addition, the physical limitations of the propulsion system
are accounted for in the model formulation of the ship.
A nonlinear robust controller has been designed
based on a reduced-order model that takes into consideration the surge, sway and yaw motions of the ship.
The vessel is considered to be fully actuated and all its
state variables are assumed to be available through
measurements.
The simulation results illustrate the capability of the
nonlinear six degree-of-freedom model in predicting the
ship response during a turning-circle maneuver.
Moreover, they demonstrate a robust tracking
characteristic of the proposed controller on both the

5. Conclusions
A nonlinear six degree-of-freedom dynamic model is
presented herein for the purpose of control of marine
surface vessels. The formulation closely follows the

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

Khaled and Chalhoub

811

reduced-order and the full-order models of the ship, in

spite of signicant modeling dierences between the
two cases. Furthermore, they illustrate the adverse
eects of the physical limitations of the propulsion
system on the controlled response of the ship.
Future work will focus on developing a dynamic
model of the rudder and will employ the line of sight
concept to determine the desired yaw angle of the ship.
Acknowledgments
The authors would like to acknowledge the nancial support
provided by the ONR grant under Award No.: N00014-05-10040. Dr. Kelly B. Cooper is the Program Director.

Notation
AT , AL ,
Arud,Adisk,
Dpr ,
H1=3 ,
k, ,
L, Lpp ,
m, I,
p, q, r,

transverse and longitudinal areas of the

ships superstructure.
area of the rudder and the propeller disk,
respectively.
diameter of the propeller.
average height of the highest one-third
peaks of the wave.
wave number for infinite sea depth and
incident wave angle, respectively.
overall length and length between the
perpendiculars of the ship, respectively.
mass and mass moment of inertia of the
ship.
angular velocity of the ship along the
i , j and k directions, respectively.

o o

T,
u, v, w,
xG , yG , zG ,
water , air ,
, , ,

o

draft of the ship.

translational velocity of the ship along
the i , j and k directions, respectively.
o o
o
coordinates
of the ship mass center with
respect to the body-fixed frame.
density of water and air, respectively.
yaw, pitch and roll angular displacements
of the ship, respectively.

References
Abbott IA and Von Doenhoff AE (1958) Theory of Wing
Sections, including a Summary of Airfoil Data. New
York, USA: Doves.
Barr RA, Miller ER, Ankudinov V and Lee FC (1981)
Technical basis for maneuvering performance standards,
U.S. Coast Guard Report CG-8-81, NTIS ADA 11474.
Berge SP, Ohtsu K and Fossen TI (1998) Nonlinear control
of ships minimizing the position tracking errors, in
Proceedings of the International Federation of Automatic
Control (IFAC) Conference on Control Applications in
Marine Systems (CAMS98). 2730 October 1998,
Japan: Fukuoka, pp. 141147.

Blanke M (1982) Ship Propulsion Losses Related to Automatic

Steering and Prime Mover Control. Copenhagen,
Denmark: Technical University of Denmark.
Faltinsen OM (1990) Sea Loads on Ships and Offshore
Structures. Cambridge, England: Cambridge University
Press.
Fossen TI (1993) High performance ship autopilot with wave
filter, in Proceedings of the 10th Ship Control System
Symposium (SCSS93), 2529 October 1993, Ottawa,
Canada.
Fossen TI (1994) Guidance and Control of Ocean Marine
Vehicles. New York: Wiley.
Fossen TI (2000) A survey on nonlinear ship control: from
theory to practice. Plenary talk, in Proceedings of the Fifth
International Federation of Automatic Control (IFAC)
Conference on Manoeuvring and Control of Marine Craft,
2325 August 2000, Aalborg, Denmark.
Fossen TI (2005) A nonlinear unified state-space model for
ship maneuvering and control in a seaway. International
Journal of Bifurcation and Chaos 15: 27172746.
Fossen TI and Grvlen A (1998) Nonlinear output feedback
control of dynamically positioned ships using vectorial
observer backstepping. IEEE Transactions on Control
Systems Technology 6(1): 121128.
Fossen TI and Strand JP (1999) A tutorial on nonlinear backstepping: applications to ship control. Modelling,
Identification and Control 20(2): 83135.
Fox RW and McDonald AT (1992) Introduction to Fluid
Mechanics. New York: Wiley.
Godhavn JM (1996) Nonlinear tracking of underactuated
surface vessels, in Proceedings of the 35th IEEE
Conference on Decision and Control, 1113 December
1996, Kobe, Japan.
Isherwood RM (1973) Wind resistance of merchant ships. The
Royal Institution of Naval Architects 15: 327338.
Journee JMJ and Massie WW (2001) Offshore
Hydrodynamics, Delft University of Technology, http://
www.shipmotions.nl/LectureNotes.html.
Journee JMJ and Pinkster J (2002) Introduction in
Ship
Hydromechanics,
http://www.shipmotions.nl/
LectureNotes/ShipHydromechanics_Intro.pdf.
Khalil HK (1996) Nonlinear Systems, 2nd edition. Englewood
Cliffs, NJ, USA: Prentice Hall.
Kristiansen E, Hjulstad A and Egeland O (2005) State space
representation of radiation forces in time domain vessel models. Oceanic Engineering 32: 21952216.
Kuiper G (1992) The Wageningen propeller series. MARIN
Research Institute Publication No. 92-001, http://
www.marin.nl
Le MD, Tran QT, Nguyen TN and Gap VD (2004) Control
of large ship motions in harbor maneuvers by applying
sliding mode control, in 8th IEEE International Worshop,
Avanced Motion Control, AMC04. 2528 March 2004,
Kawasaki, Japan, 695700.
Lee CH and Newman JN (2004) Computation of Wave Effects
Using the Panel Method. Southampton, England: WIT
Press.
Lewis EV (1988) Principles of Naval Architecture, 2nd edition.
Jersey City, NJ, USA: Society of Naval Architects and
Marine Engineers (SNAME).

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016

812

Journal of Vibration and Control 17(6)

Moreira L, Fossen TI and Guedes Soares C (2007) Path following control system for a tanker ship model. Ocean
Engineering OE-34: 20742085.
Newman JN (1977) Marine Hydrodynamics. Cambridge,
Massachusetts: MIT Press.
Oil Companies International Marine Forum (OCIMF)
(1994). Prediction of Wind and Current Loads on Very
Large Crude Carriers (VLCCs). London: Witherby & Co.
Ogilvie TF (1964) Recent progress toward the understanding
and prediction of ship motion, in The Office of Naval
Research
(ONR)
5th
Symposium
on
Naval
Hydrodynamics. 1012 September 1964, Bergan, Norway.
Perez T (2005) Ship Motion Control. New York: SpringerVerlag.
Pettersen KY and Nijmeijer H (2001) Underactuated ship
tracking control: theory and experiments. International
Journal of Control 74(14): 14351446.
Pivano L, Johansen TA, Smogeli N and Fossen TI (2007)
Nonlinear thrust controller for marine propellers in

four-quadrant operations.
in American Control
Conference, 1113 July 2007, New York, USA.
Roddy RF, Hess DE and Faller WE (2006) Neural network
predictions of the 4-quadrant wageningen b-screw series,
in Fifth International Conference on Computer and IT
Applications in the Maritime Industries, 811 May 2006,
Netherlands: Leiden.
Slotine JJE and Li W (1991) Applied Nonlinear Control.
Englewood Cliffs, NJ, USA: Prentice-Hall.
Society of Naval Architectures and Marine Engineers
(SNAME) (1950). Nomenclature for Treating the Motion
of Submerged Body through a Fluid. New York, NY, 74
Trinity Place, 10006.
Strand JP, Ezal K, Fossen TI and Kokotovic PV (1998)
Nonlinear control of ships: a locally optimal design, in
Preprints of the IFAC NOLCOS98, Enschede, The
Netherlands, pp.732738.
Ueng SK, Lin D and Liu CH (2008) A ship motion simulation system. Virtual Reality 12: 6576.

Downloaded from jvc.sagepub.com at YildizTeknik Univ on February 22, 2016