10th Portuguese Conference on Automatic Control

16-18 July 2012

CONTROLO2012
Funchal, Portugal

A Linear Design Model for Wave Filtering

and Dynamic Positioning
Vahid Hassani Asgeir J. Srensen Ant
onio M. Pascoal

A. Pedro Aguiar

Laboratory of Robotics and Systems in Engineering and Science

(LARSyS), Instituto Superior Tecnico (IST), Tech. Univ. Lisbon,
Portugal (e-mail: {vahid,antonio,pedro}@isr.ist.utl.pt).

Centre for Ships and Ocean Structures (CeSOS) and Dept. of

Marine Technology, Norwegian Univ. of Science and Technology,
Trondheim, Norway (e-mail: asgeir.sorensen@ntnu.no)
Abstract: This paper describes a procedure to obtain an improved linear design model for
Wave Filtering (WF) and Dynamic Positioning (DP) of ships and offshore rigs subjected to the
influence of sea waves. The proposed model captures the main physics of the problem and allows
for the design of advanced control and estimation algorithms to solve the WF and DP problems.
Numerical simulations, carried out using a high fidelity nonlinear DP simulator, illustrate the
performance improvement in wave filtering as a result of the use of the proposed modified model.
c
CONTROLO2012
Keywords: Vessel Modeling; Dynamic Positioning; Wave Filtering.
1. INTRODUCTION
Dynamic positioning (DP) systems made their appearance in the 1960s for offshore drilling applications, due
to the need to drill in deep waters and the realization
that Jack-up barges and anchoring systems could not
be used economically at the depths required. Early dynamic positioning systems were implemented using PID
controllers. In order to restrain thruster trembling caused
by the wave-induced motion components, notch filters in
cascade with low pass filters were used with the controllers.
However, notch filters restrict the performance of closedloop systems because they introduce phase lag around the
crossover frequency, which in turn tends to decrease phase
margin. An improvement in performance was achieved by
exploiting more advanced control techniques based on optimal control and Kalman filter theory, see [Balchen et al.,
1976]. These techniques were later modified and extended
in [Balchen et al., 1980a,b, Grimble et al., 1980a,b, Slid
et al., 1983, Fossen et al., 1996, Srensen et al., 1996,
Grvlen and Fossen, 1996, Strand, 1999, Fossen, 2000] and
[Torsetnes et al., 2004]. For a survey of dynamic positioning control systems, see [Srensen, 2011b] and the references therein. One of the most fruitful concepts introduced
in the course of the body of work referred above was that of
wave filtering, together with the strategy of modeling the
total vessel motion as the superposition of low-frequency
(LF) vessel motion and wave-frequency (WF) motions. It
was further recognized that in order to reduce the mechanical wear and tear of the propulsion system components,
in small to high see states, the estimates entering the DP
This work was supported in part by projects MORPH (EU
FP7 under grant agreement No. 288704) and CONAV/FCTPT (PTDC/EEA-CRO/113820/2009), and the FCT [PEstOE/EEI/LA0009/2011] and was carried out in cooperation with
the Centre for Ships and Ocean Structures (CeSOS); the Norwegian
research council is acknowledged as the main sponsor of CeSOS.
The first author benefited from grant SFRH/BD/45775/2008 of the
Foundation for Science and Technology (FCT), Portugal.

298

control feedback loop should be filtered by using a so-called

wave filtering technique so as to prevent excessive control
activity in response to WF components. In practice, position and heading measurements are corrupted not only
with sensor noise but also by equivalent colored noise that
captures the influence of wind, waves, and ocean currents;
thus the need for an observer to achieve wave filtering and
separate the LF and WF position and heading estimates
(see [Fossen, 2011] for details). In extreme seas or swell
with very long wave periods, wave filtering is turned off
as described in [Srensen, 2011b, Nguyen et al., 2007] and
[Hassani et al., 2012d].
In [Srensen et al., 1996], WF filtering was done by exploiting the use of Kalman filter theory under the assumption
that the kinematic equations of the ships motion can be
linearized about a set of predefined constant yaw angles
(36 operating points in steps of 10 degrees, covering the
whole heading envelope); this is necessary when applying
linear Kalman filter theory and gain scheduling techniques.
However, global exponential stability (GES) of the complete system cannot be guaranteed. In [Fossen and Strand,
1999], a nonlinear observer with wave filtering capabilities
and bias estimation was designed using passivity theory.
An extension of this observer with adaptive wave filtering was described in [Strand and Fossen, 1999]. Gainscheduled wave filtering was introduced in [Torsetnes et al.,
2004].
To the best of our knowledge, in the wave filtering techniques described in the above mentioned references the
wave-frequency components of marine vessel motion are
computed in the earth-fixed frame (also denoted as NorthEast-Down - NED frame), while the hydrodynamic forces
and torques on a marine vessel are naturally given (via
the hydrodynamic coefficients) in terms of variables that
are best measured in body-axis (because they capture the
influence between the vessel and the environment locally,
no matter what the attitude of the vessel is, in inertial
coordinates). Assuming a stationary wave pattern, this

suggests that the wave-frequency motions be described in

a hydrodynamic frame (or body frame) rather than in a
North-East-Down (NED) frame.
In this paper, inspired by previous pioneering work on
DP, a modified design model of the vessel is proposed.
The model is developed based on practical assumptions
and can be used as a tool to design control and estimation algorithms for dynamic positioning and wave filtering
of the offshore vessels in the presence of sea waves and
disturbances. The main emphasis of the paper is on the
new adopted model; however, two wave filters are designed (based on two different model of the vessel) and
their performance is compared. Moreover, for the sake of
completeness, in the numerical simulations a multivariable
PID is used to control the position of the vessel.
The structure of the paper is as follows. Section 2 is a
brief introduction to important issues that arise in DP.
A modified representative vessel model is also proposed.
An example is described in Section 3 that illustrates the
methodology proposed, via computer simulations. Conclusions and suggestions for future research are summarized
in Section 4.
2. DYNAMIC POSITIONING, WAVE FILTERING
AND SHIP MODELING
In DP systems, the key objective is to maintain the vessels
heading and position within desired limits. Central to their
implementation is the availability of good heading and
position estimates, provided by properly designed filters.
In general, measurements of the vessel velocities are not
available and measurements of position and heading are
corrupted with different noises. Consequently, estimates of
the velocities must be computed from corrupted measurements of position and heading through a state observer.
Furthermore, only the slowly-varying disturbances should
be counterbalanced by the propulsion system, whereas the
oscillatory motion induced by the waves (1st-order wave
induced loads) should not enter the feedback control loop.
To this effect, the DP control systems should be designed
so as to only react to the low frequency forces on the vessel.
As mentioned, wave filtering techniques are exploited to
separate the position and heading measurements into lowfrequency (LF) and wave-frequency (WF) position and
heading estimates. Fig. 1 illustrates this concept graphically. It was this interesting circle of ideas that motivated
the work reported in the present paper on developing a
linear design model for wave filtering and dynamic positioning. In what follows, the vessel model, that is by
now standard, is presented. See [Fossen and Strand, 1999,
Torsetnes et al., 2004, Srensen, 2011a]. The model admits
the realization
= A (0 ) + E w
(1)
= C
(2)
b = T 1 b + Eb wb
(3)
L = R(L )

(4)

M + D = + R (tot )b
(5)
tot = L +
(6)
y = tot + v
(7)
where (1) and (2) capture the 1st-order wave induced
motion in surge, sway, and yaw; equation (3) represents a
1st-order Markov process approximating the unmodelled
dynamics and the slowly varying bias forces (in surge and
sway) and torques (in yaw) due to waves (2nd order wave
induced loads), wind, and currents, where the latter are

299

Total Motion
(WF+LF)

LowFrequency Motion
(LF)

WaveFrequency Motion
(WF)

50

100
Time (s)

150

200

Fig. 1. The total motion of a ship is modeled as a LF

response with the WF motion added as an output
disturbance (adopted from [Fossen and Strand, 1999,
Srensen, 2011a]).
given in earth fixed coordinates but expressed in bodyaxis. The vector R3 is the vessels WF motion due
to 1st-order wave-induced disturbances, consisting of WF
position (xW , yW ) and WF heading W of the vessel;
w R3 and wb R3 are zero mean Gaussian white noise
vectors, and




0
0
I33
,
E = I31 ,
A = 33
33

33

31

C = [033 I33 ] ,

with
2
2
2
= diag{01
, 02
, 03
},
= diag{21 01 , 22 02 , 23 03 },

where 0i and i are the Dominant Wave Frequency

(DWF) and relative damping ratio, respectively. Matrix
T = diag(Tx , Ty , T ) is a diagonal matrix of positive bias
time constants and Eb R33 is a diagonal scaling matrix.
The vector L R3 consists of low frequency (LF), earthfixed position (xL , yL ) and LF heading L of the vessel
relative to an earth-fixed frame, R3 represents the
vessels velocities decomposed in a vessel-fixed reference,
and R(L ) is the standard orthogonal yaw angle rotation
matrix (see [Fossen, 2011] for more details). Equation (5)
describes the vesselss LF motion at low speed (see [Fossen,
2011]), where M R33 is the generalized system inertia
matrix including zero frequency added mass components,
D R33 is the linear damping matrix, and R3 is
a control vector of generalized forces generated by the
propulsion system, that is, by the main propellers aft of
the ship and thrusters which can produce surge and sway
forces as well as a yaw moment. The vector tot R3 is the
vessels total motion, consisting of total position (xtot , ytot )
and total heading tot of the vessel. Finally, (7) represents
the position and heading measurement equation, where
v R3 is zero-mean Gaussian white measurement noise.
Clearly, in the model described in (1)-(7) the evolution
of the WF components of the motion, (1) and (2), are
modeled as 2nd-order linear time invariant (LTI) systems,
driven by Gaussian white noise, in earth-fixed frame.
It is commonly accepted in station keeping operations,
assuming small motions about the coordinates d (xd , yd ,
and d ), that the coupled equations of WF motions can

be formulated in the hydrodynamic frame 1 with

M (w)
R +Dp (w)R = wave1
= R(d ) R

(8)
(9)

where 2 R R3 is the WF motion vector in the hydrodynamic frame, R3 is the WF motion vector
in the Earth-fixed frame, and wave1 R3 is the first
order wave excitation vector, which is dependent on the
vessel heading relative to the incident wave direction. In
the above, M (w) R33 is the system inertia matrix
containing frequency dependent added mass coefficients in
addition to the vessels mass and moment of inertia and
Dp (w) R33 is the wave radiation (potential) damping
matrix; see [Srensen, 2011a] for more details. Here, it is
assumed that the mooring lines, if any, will not affect the
WF motion [Triantafyllou, 1994].
The above formulation indicates that the WF motion
should be computed in the hydrodynamic frame 3 . We
now recall that the problem of modeling the hydrodynamic
forces applied to a vessel in regular waves is solved as
two sub-problems: wave reaction and wave excitation;
the forces calculated in each of these sub-problems can
be added together to give the total hydrodynamic forces
[Faltinsen, 1990]. Potential theory is assumed, neglecting
the viscosity of the fluid. The following effects are important:
Wave Reaction: Forces and moments on the vessel when
the vessel is forced to oscillate with the wave excitation
frequency. The hydrodynamic loads are identified as added
mass and wave radiation damping terms.
Wave Excitation: Forces and moments on the vessel
when the vessel is restrained from oscillating and there
are incident waves. This gives the wave excitation loads
which are composed of so-called Froude-Kriloff (forces
and moments due to the undisturbed pressure field as
if the vessel were not present) and diffraction forces and
moments (forces and moments because the presence of the
vessel changes the pressure field).
Results from model tests and numerical computations for
vessel response analysis often come in the form of transfer
functions or tables of coefficients. This can be applied
to linear wave-induced motions, 2nd-order wave drift and
slowly varying motions. To a large extent, linear theory is
sufficient for describing wave-induced motions and loads
on vessels. This is especially true for small to moderate
sea states. To this effect, a frequency spectrum S() is
selected to describe the energy distribution of the wind
generated sea waves and swell over different frequencies,
with the integral over all frequencies representing the total
energy of the sea state. Common frequency spectra are
the Pierson-Moskowitz (PM) spectrum, the ITTC/ISSC
spectrum, the JONSWAP spectrum, and the more recent
1 In sea keeping analysis (vessel motions in waves) the hydrodynamic
frame is generally moving along the path of the vessel with the x-axis
positive forwards to the desired d , y-axis positive to the starboard,
and z-axis positive downwards. The XY-plane (in hydrodynamic
frame) is assumed fixed and parallel to the mean water surface. The
vessel is assumed to oscillate with small amplitudes about this frame
such that linear theory can be used to model perturbations.
2 In 6DOF dynamics, the equation (8) changes to M (w)
R +
Dp (w)R + GR = wave1 , where G R66 is the linearized
restoring coefficient matrix due to the gravity and buoyancy affecting
heave, roll and pitch only (see [Srensen, 2011a] for more details).
Throughout this paper a 3DOF dynamics is used for the purpose of
design while a 6DOF dynamics is used in the simulation.
3 Assuming that | | 0, then the transformation from the
d
hydrodynamic frame to the body frame, in 3DOF formulation, is the
identity matrix.

300

doubly peaked spectrum by Torsethaugen (for more details

see [Fossen, 2011, Srensen, 2011a] and references therein).
Linear approximations of the wave spectra are studied in
the literature; in particular, 2nd-order wave transfer function approximations have attracted considerable attention,
see [Balchen et al., 1980a,b, Grimble et al., 1980a,b, Slid
et al., 1983, Fossen et al., 1996, Srensen et al., 1996,
Grvlen and Fossen, 1996, Strand, 1999, Fossen, 2000,
Torsetnes et al., 2004], and [Nguyen et al., 2007].
In this paper we also consider a 2nd-order wave transfer
function approximation and we propose a modified model
for the WF components of motion as follows:
= A (0 ) + E w
(10)
= R(L )C
b = T 1b + Eb wb

L = R(L )

(11)
(12)
(13)

M + D = + RT (tot )b
(14)
tot = L +
(15)
y = tot + v
(16)
where all the variables are as defined in (1)-(7). At this
point we would like to highlight the difference between
(2) and (11). As mentioned before, the evolution of the
WF components of the motion in (1) and (2), are modeled
as 2nd-order linear time invariant (LTI) system, driven
by Gaussian white noise, in earth-fixed frame, while (8)
suggests to model the WF motions in body frame. From a
physical point of view, it is obvious that the WF motions
depend on the angle between the heading of the vessel and
the direction of the wave. Assuming stationary waves, 4
one can assume that a linear approximation can be used
to described wave-induced motions in the body frame. This
justifies the modification applied in (11). Modeling the WF
motions in earth-fixed frame means that every time a new
command for a desired heading is issued, the WF motion
dynamic should be updated. By modeling the WF motion
dynamic in body frame such a problem will not occur.
2.1 Observer Design
In the observer design process for wave filtering, the following assumptions are usually made (these assumptions are
widely used in the literature, see [Strand, 1999, Strand and
Fossen, 1999, Fossen, 2000, Loria et al., 2000, Torsetnes
et al., 2004]):
Assumption 1 Position and heading sensor noise are
neglected, that is v = 0.
Assumption 2 The amplitude of the wave-induced yaw
motion is assumed to be small, that is, less than 23 degrees during normal operation of the vessel and less
than 5 degrees in extreme weather conditions. Hence,
R(L ) R(L + W ). From Assumption 1 it follows
that R(L ) R(y ), where y
= L + W denotes the
measured heading.
Assumption 3 Low speed assumption, where the timederivative of the total heading tot is small, bounded, and
close to zero.
We will also exploit the model property that the bias time
constant in the x and y directions are equal, i.e. Tx = Ty .
For observer design, consider a new coordinate system
denoted as vessel parallel coordinates as introduced in [Fossen, 2011, Srensen, 2011a] and [Fossen and Perez, 2009a].
4

In long-crested irregular sea, the sea elevation can be assumed

statistically stable. See [Srensen, 2011a] for details and differences
of Long- and Short-Crested Seas.

Vessel parallel coordinates are defined in a reference frame

fixed to the vessel, with axes parallel to the earth-fixed
p
p
frame. Vector L
R3 consists of the LF position (xpL , yL
)
p
and LF heading L of the vessel expressed in body coordinates, defined as
p
L
= RT (tot )L .
(17)
Computing its derivative with respect to time yields
p
L
= R T (tot )L + RT (tot )L
= R T (tot )R(tot ) p + RT (tot )R(L )
(18)

where b are WF components of motion on bodycoordinate axis and wbf and yf are a new modified disturbance and a modified measurement defined by wbf =
RT (y )Eb wb and yf = RT (y )y , respectively. 5

Using a Taylor series to expand RT (tot ) about L and

neglecting the higher order terms, it follows that
RT (tot )R(L )
(19)
= I + W S,
where
#
"
0 10
S = 1 0 0 .
0 00
Using simple algebra we obtain
R T (tot )R(tot ) = tot S.
(20)
From (18), (19) and (20) we conclude that
p
p
L
tot SL
+ + W S.
(21)

In what follows we test our proposed design model using the Marine Cybernetics Simulator (MCSim), later on
upgraded to the Marine System Simulator (MSS). The
MCSim is a modular multi-disciplinary simulator based
on Matlab/Simulink. It was developed at the CeSOS.
The MCSim incorporates high fidelity models, denoted
as process plant model or simulation model in [Srensen,
2011a], at all levels (plants and actuators). It is composed
of different modules that include the following:
1) Environmental module, containing different wave
models, surface current models, and wind models.
2) Vessel dynamics module, consisting of a LF and a
WF model. The LF model is based on the standard 6DOF
vessel dynamics, whose inputs are the environmental loads
and the interaction forces from thrusters and the external
connected systems.
3) Thruster and shaft module, containing thrust allocation routines for non-rotating thrusters, thruster dynamics, and local thruster control. It may also include
advanced thrust loss models for extreme seas, in which
case detailed information about waves, current, and vessel
motion is required. The shaft is modeled as a rotational
mass, with propeller speed given from motor torque and
propeller load torque.
4) Vessel control module, consisting of of different controllers, namely, nonlinear multivariable PID controllers
for DP.
For more details on the MCSim see [Srensen et al., 2003,
Perez et al., 2005, 2006], and [Fossen and Perez, 2009b].

3. OBSERVER DESIGN AND MONTE CARLO

NUMERICAL SIMULATIONS
3.1 Overview of the Simulator

We now study the time evolution of the slowly varying bias

forces, b, expressed in the vessel parallel coordinates, bp ,
as follows:
bp = RT (tot )b.
(22)
Clearly,
b = R(tot )bp ,
(23)
and differentiating from both sides yields
tot )bp + R(tot )b p .
b = R(
(24)
Recalling (12), (23) and (24) we now have
tot )bp + R(tot )b p = T 1R(tot )bp + Eb wb . (25)
R(

Reordering (25) and multiplying both sides by RT (tot )

gives
tot )bp
b p = RT (tot )T 1 R(tot )bp RT (tot )R(
+ RT (tot )Eb wb .
(26)
Using the assumption that Tx = Ty , it can be checked
that RT (tot )T = T RT (tot ); simple algebra also shows
tot ) = tot S.
that RT (tot )R(
Equation (26) can be expressed as
b p = T 1 bp + tot Sbp + RT (tot )Eb wb .
(27)
Summarizing the equations above yields
= A (0 ) + E w
= R(L )C
b p = T 1 bp + tot Sbp + RT (tot )Eb wb
p = tot S p + + W S

(28)
(29)

3.2 Effect of the Modified Model on the Performance of

the Observers

(30)

In this section, two nonlinear passive observers based on

the two models described by (1)-(7) and (10)-(15) are designed. The passive observers were introduced in [Strand,
1999, Fossen and Strand, 1999, Strand and Fossen, 1999].
The structure of a passive observers for a DP vessel model
described by (1)-(7) is given as

= A (0 ) + K1 y
(39)

(40)
= C

(31)
L
L
p
M + D = + b
(32)
Moreover, using assumptions 1, 2 and 3 a linear model is
obtained that is given by
= A (0 ) + E w
(33)
b = C
b p = T 1 bp + wbf
p
L
=
M + D = + bp
p
yf = L
+ b

In what follows, Monte-Carlo simulations involving two

observers based on two models described by (1)-(7) and
(10)-(15) are designed. The performance of the two observers is compared with the help of numerical simulations
in different environment conditions from calm to high seas.
In these simulations, the different environment conditions
from calm to high seas are simulated using the spectrum
of the Joint North Sea Wave Project (JONSWAP) [Hasselmann et al., 1973].

(34)
(35)
(36)
(37)
(38)

301

When designing the observers for wave filtering in dynamic positioning, and because the controller regulates the heading of the
vessel, the designer can assign a new intensity to wbf ; however,
assigning the intensity of the noise in practice requires considerable
expertise.

L = R(y )
+ K3 y
T
M + D
= + R (y )b + RT (y )K4 y

(41)

15

(42)

10

Table 1. Definition of Sea States from [Price

and Bishop, 1974]

yL
yL by passive observer

yL by modified passive observer

10

15

20

25

30
0

50

100

150

200

250

Time (s)

Fig. 2. Total motion, low frequency component of a (typical 100 meter long) DP vessel (only sway), and its
estimates.
1

yL

0.8

yL by passive observer
0.6

yL by modified passive observer

0.4
0.2
East (m)

(43)
y = L + .
(44)
For details on nonlinear passive observers and how to select
the observer gains, i.e. Ki , i = 1 . . . 4, the reader is referred
to [Strand, 1999, Fossen and Strand, 1999, Strand and
Fossen, 1999, Fossen, 2000, Loria et al., 2000, Torsetnes
et al., 2004, Srensen, 2011a] and [Fossen, 2011].
The structure of the second observer for a DP vessel model
described by (10)-(15) (i.e. observer based on the new
proposed DP model) is adopted from the one in (39)-(44)
except for the WF motion which is transformed to body
coordinate as
(45)
= R(y )C .
Throughout this article we have used same set of observer
gains to study the effect of our model in the performance
of the observers. 6 Three different environment conditions
from calm to high seas are considered, and for each one a
separate observer is designed. Table 1 shows the definition
of the sea condition associated with a particular model of
supply vessel that is used in the MCSim.

ytot = yL + yW

East (m)

b = T 1b + K2 y

0
0.2
0.4

Sea States
Calm Seas
Moderate Seas
High Seas

DWF
0 (rad/s)
> 0.79
[0.67 0.79]
[0.45 0.67]

Significant Wave Height

Hs (m)
< 1.25
[1.25 2.69]
[2.69 9.71]

0.6
0.8
1
0

At this point in the simulation, the nominal values for

the dominant wave frequency (DWF) are selected as
{0.56, 0.73, 0.80} (rad/s). Figs. 2 and 3 show the time
evolution of total motion and the low frequency component
of the motion and its estimates by two passive observers
(one based on the model (1)-(7) and the other based on the
modified model (10)-(15)) in high sea state. The simulation
is done in the station keeping scenario where a nonlinear
multivariable PID controller is keeping the position and
heading of the vessel at zero. It is seen that after the
transient period the DP controller maintains the position
well about the desired set-point equal to zero. It is seen
also that even with wave filtering, some of the 1st-order
wave frequency components are seen in the LF estimates. 7
However, the proposed observer in this paper provides a
better estimation performance compared with that of the
passive observer in [Fossen and Strand, 1999]. To quantify
the performance improvement of our modified observer
over the passive observer of [Fossen and Strand, 1999], we
introduce the following percentage comparison
6 All the gains are optimally tuned for the observer designed based
on the model described in (1)-(7). Such a selection favors the passive
observer; however, the comparison results of the simulations shows
better performance of the observer designed based on the new
proposed model.
7 We should emphasize that the observers are designed according to
the simple model of (10)-(15) (and (1)-(7)) while they are tested in
the MCSim with a high fidelity model that captures hydrodynamic
effects, generalized Coriolis and centripetal forces, nonlinear damping
and current forces, and generalized restoring forces. Moreover, in
the MCSim the JONSWAP wave spectrum is used to simulate the
waves while the observers are designed with linear second order
approximation of the spectral density.

302

50

100

150
Time (s)

200

250

300

Fig. 3. Low frequency component of a (typical 100 meter

long) DP vessel (only sway), and its estimates.
%E =

|
|VARpold | |VARpnew
L
L

|
|VARpnew
L

100

(46)

where VARp is the variance of p; pold

new
are the
L and p
L
estimation error of pL with the passive observer of [Fossen
and Strand, 1999] and our modified observer, respectively;
finally, pL is the LF component of the motion in surge,
sway or yaw.
Table 2 summarizes the result of Monte-Carlo simulations
with different environment conditions from calm to high
seas where a nonlinear multivariable PID controller regulates the position of the ship at origin. We have computed
the performance improvement of our modified observer
over the passive observer of [Fossen and Strand, 1999].
Table 3 shows similar results when the vessel position
was commanded to change 300 (m) foreward in surge and
and sway while keeping the heading at zero; this simple
maneuver was executed with the nonlinear multivariable
PID control law referred to above. As can be seen in
Table 2. Performance improvements of new
observer (Station Keeping)
%E in surge
%E in sway
%E in yaw

Calm Seas
16 %
22 %
-9 %

Moderate Seas
18 %
25 %
51 %

High Seas
78 %
115 %
93 %

Tables 2 and 3, as the sea condition changes from calm

to high seas, there is significant improvement with the use
of the new proposed observer, when compared with that
of the passive observer of [Fossen and Strand, 1999]. The

Table 3. Performance improvements of new

observer (Manoeuvering)
%E in surge
%E in sway
%E in yaw

Calm Seas
17 %
24 %
-8 %

Moderate Seas
18.5 %
27 %
51 %

High Seas
80 %
115 %
94 %

reason for this behavior is the violation of Assumption 2

with the passive observer. In fact, when we move from
calm sea to moderate and high sea, the amplitude of the
wave-induced yaw motion becomes larger and neglecting
this term causes performance degradation. This problem is
alleviated in our proposed observer (by modeling the WF
motion in the body frame).
4. CONCLUSIONS
This paper proposed a new design plant model for offshore
vessels influenced by sea waves, with application to wave
filtering and dynamic positioning. Its key contribution was
the development of a modified model that captures better
the physics of the vessel. Numerical simulations, carried
out in a high fidelity nonlinear DP simulator, showed the
performance improvement in wave filtering due to the use
of modified model. The proposed model is expected to play
an important role in the development of new control and
estimation techniques for DP; the model was already used
in [Hassani et al., 2012d,a,b,c] for simulation and model
testing.
ACKNOWLEDGEMENTS
We thank our colleagues and Thor I. Fossen for many discussions on
wave filtering and adaptive estimation.
REFERENCES
J. Balchen, N. A. Jenssen, and S. Slid. Dynamic positioning using
Kalman filtering and optimal control theory. In the IFAC/IFIP
Symp. On Automation in Offshore Oil Field Operation, pages
183186, Bergen, Norway, 1976.
J. Balchen, N. A. Jenssen, and S. Slid. Dynamic positioning of
floating vessels based on Kalman filtering and optimal control.
In Proc. of the 19th IEEE Conference on Decision and Control,
pages 852864, New York, NY, 1980a.
J. Balchen, N. A. Jenssen, and S. Slid. A dynamic positioning
system based on Kalman filtering and optimal control. Modeling,
Identification and Control (MIC), 1(3):135163, 1980b.
O. M. Faltinsen. Sea Loads on Ships and Offshore Structures.
Cambridge University Press, UK, 1990.
T. I. Fossen. Nonlinear passive control and observer design for ships.
Modeling, Identification and Control (MIC), 21(3):129184, 2000.
T. I. Fossen. Handbook of Marine Craft Hydrodynamics and Motion
Control. John Wiley & Sons. Ltd, Chichester, UK, 2011. ISBN
978-1-1199-9149-6.
T. I. Fossen and T. Perez. Kalman filtering for positioning and
heading control of ships and offshore rigs. IEEE Control Systems
Magazine, 29(6):3246, December 2009a.
T. I. Fossen and T. Perez. Marine systems simulator (MSS).
www.marinecontrol.org, 2009b.
T. I. Fossen and J. P. Strand. Passive nonlinear observer design
for ships using lyapunov methods: Full-scale experiments with a
supply vessel. Automatica, 35:316, 1999.
T. I. Fossen, S. I. Sagatun, and A. J. Srensen. Identification of
dynamically positioned ships. Journal of Control Engineering
Practice, 4(3):369376, 1996.
M. J. Grimble, R. J. Patton, and D. A. Wise. The design of dynamic
ship positioning control systems using stochastic optimal control
theory. Optimal Control Applications and Methods, 1:167202,
1980a.
M. J. Grimble, R. J. Patton, and D. A. Wise. The design of dynamic
ship positioning control systems using stochastic optimal control
theory. IEE Proceedings, 127(3):93102, 1980b.

A Grvlen and T. I. Fossen. Nonlinear control of dynamic positioned

ships using only position feedback: An observer backstepping

303

approach. In Proc. IEEE Conference on Decision and Control

(CDC96), Kobe, Japan, 1996.
Vahid Hassani, A. J. Srensen, and Antonio M. Pascoal. Robust
dynamic positioning of offshore vessels using mixed- synthesis,
part I: Designing process. In Proc. ACOOG12 - IFAC Workshop
on Automatic Control in Offshore Oil and Gas Production,
Trondheim, Norway, 2012a.
Vahid Hassani, A. J. Srensen, and Antonio M. Pascoal. Robust
dynamic positioning of offshore vessels using mixed- synthesis,
part II: Simulation and experimental results. In Proc. ACOOG12
- IFAC Workshop on Automatic Control in Offshore Oil and Gas
Production, Trondheim, Norway, 2012b.
Vahid Hassani, A. J. Srensen, and Antonio M. Pascoal. Evaluation of three dynamic ship positioning controllers: from calm to
extreme conditions. In Proc. NGCUV12 - IFAC Workshop on
Navigation, Guidance and Control of Underwater Vehicles, Porto,
Portugal, 2012c.
Vahid Hassani, A. J. Srensen, Antonio M. Pascoal, and A. Pedro
Aguiar. Multiple model adaptive wave filtering for dynamic
positioning of marine vessels. In Proc. ACC12 - American
Control Conference, Montreal, Canada, 2012d.
K. Hasselmann, T. P. Barnett, E. Bouws, H. Carlson, D.E.
Cartwright, K. Enke, J. A. Ewing, H. Gienapp, D. E. Hasselmann,
P. Kruseman, A. Meerburg, P. M
uller, D. J. Olbers, K. Richter,
W. Sell, and H. Walden. Measurements of wind-wave growth and
swell decay during the joint north sea wave project (JONSWAP).
Ergnzungsheft zur Deutschen Hydrographischen Zeitschrift Reihe,
8(12):195, 1973.
A. Loria, T. I. Fossen, and E. Panteley. A separation principle for
dynamic positioning of ships: theoretical and experimental results.
IEEE Trans. on Contr. Syst. and Tech., 8(2):332343, 2000.
T. D. Nguyen, A. J. Srensen, and Ser Tong Quek. Design of
hybrid controller for dynamic positioning from calm to extreme
sea conditions. Automatica, 43(5):768785, 2007.
T. Perez, . N. Smogeli, T. I. Fossen, and A. J. Srensen. An
overview of marine systems simulator (MSS): A simulink toolbox
for marine control systems. In Proc. of Scandinavian Conference
on Simulation and Modeling (SIMS05), Trondheim, Norway,
2005.
T. Perez, A. J. Srensen, and M. Blanke. Marine vessel models in
changing operational conditions a tutorial. In 14th IFAC Symp.
on System Ident. (SYSID06), Newcastle, Australia, 2006.
W. G. Price and R. E. D. Bishop. Probabilistic Theory of Ship
Dynamics. Chapman and Hall, London, UK, 1974.
S. Slid, N. A. Jenssen, and J. Balchen. Design and analysis of
a dynamic positioning system based on Kalman filtering and
optimal control. IEEE Transactions on Automatic Control, 28
(3):331339, 1983.
A. J. Srensen. Lecture notes on marine control systems. Technical
Report UK-11-76, Norwegian University of Science and Technology, 2011a.
A. J. Srensen. A survey of dynamic positioning control systems.
Annual Reviews in Control, 35:123136, 2011b.
A. J. Srensen, S. I. Sagatun, and T. I. Fossen. Design of a dynamic
positioning system using model-based control. Journal of Control
Engineering Practice, 4(3):359368, 1996.
A. J. Srensen, E. Pedersen, and . Smogeli. Simulation-based
design and testing of dynamically positioned marine vessels. In
Proc. of International Conference on Marine Simulation and Ship
Maneuverability (MARSIM03), Kanazawa, Japan, 2003.
J. P. Strand. Nonlinear Position Control Systems Design for Marine
Vessels. PhD thesis, Dept. of Eng. Cybernetics, Norwegian
University of Science and Technology, Trondheim, Norway, 1999.
J. P. Strand and T. I. Fossen. Nonlinear passive observer for
ships with adaptive wave filtering. New Directions in Nonlinear
Observer Design (H. Nijmeijer and T. I. Fossen, Eds.), SpringerVerlag London Ltd., pages 113134, 1999.
G. Torsetnes, J. Jouffroy, and T. I. Fossen. Nonlinear dynamic
positioning of ships with gain-scheduled wave filtering. In Proc.
IEEE Conference on Decision and Control (CDC04), Paradise
Iceland, Bahamas, 2004.
M. S. Triantafyllou. Cable mechanics for moored floating systems.
In Proc. of 7th International Conference on the Behaviour of
Offshore Structures at Sea (BOSS94), pages 5777, Cambridge,
MA, 1994.