You are on page 1of 6

MODELING, SIMULATION AND FEEDBACK LINEARIZATION

CONTROL OF NONLINEAR SURFACE VESSELS

Mehmet Haklidir, Deniz Aldogan, Isa Tasdelen and Semuel Franko


TUBITAK Marmara Research Centre, Information Technologies Institute, 41470, Gebze-Kocaeli, Turkey
mehmet.haklidir@bte.mam.gov.tr, deniz.aldogan@bte.mam.gov.tr
isa.tasdelen@bte.mam.gov.tr, semuel.franko@bte.mam.gov.tr

Keywords: Surface Vessels, Nonlinear Analysis and Control, Feedback Linearization.

Abstract: Realistic models and robust control are vital to reach a sufficient fidelity in military simulation projects
including surface vessels. In this study, a nonlinear model including sea-state modelling is obtained and
feedback linearization control is implemented in this model. To control the system, nonlinear analysis
techniques are used. The model is integrated into a commercial framework based CGF application within a
high-fidelity military training simulation.The simulation results are presented at the end of the study.

1 INTRODUCTION Table 1: DoF Description and Notation.

The aim of this study is to observe the dynamic


behaviors of the surface vessels under the effect of
hydrodynamic force-moments and environmental
conditions such as waves, current, wind, season that
pertaining to the tactical environment.
The analysis and control of nonlinear motion
model of surface vessels are obtained by using SNAME’s (1950) notation is used in this study.
following techniques: The first three parameters and time derivatives that
• Linearization by Taylor Series are shown on Table define the position and the
• Phase Plane Analysis motion of the platform in x-, y-, z- axes. Last three
o Course Keeping parameters define the orientation and rotary motion
o Zig Zag Maneuver of the platform. After analyzing 6 degrees of
• Lyapunov Stability Theorem freedom motion of surface vessel, it is observed that
• Feedback Linearization 2 axis systems are needed to perform the motion.
Ship dynamics model and disturbance model are Therefore, North – East- Down (NED), is the local
introduced in Section 2; the phase plane analysis and geodetic coordinate system fixed to the Earth, and
lyapunov stability therom in Section 3 and 4, the Body Fixed, is fixed to the hull of ship, coordinate
proposed controller is discussed in Section 5; frames are used. Motion axis system X0 Y0 Z0 has
simulation results are presented in Section 6. been fixed to the platform and called as Body Fixed
axes system. The point O, which is the origin of this
axes system, is always selected as the ship’s centre
of gravity. (Figure 1)
2 THE SURFACE PLATFORM
MOTION MODULE
2.1 Coordinate System and Vector
Notation
The motion of surface vessels has 6 degrees of
freedom. The description and notation of each
degree of freedom has been shown on Table 1. Figure 1: Coordinate System.

92
MODELING, SIMULATION AND FEEDBACK LINEARIZATION CONTROL OF NONLINEAR SURFACE VESSELS

2.2 Surface Platform Motion Equations Surge: X = m[u − rv − xG r 2 + zG rp ] (1)


Sway: Y = m[v + ru − zG p + xG r] (2)
Fossen (1991), by inspiring Craig’s (1989) robot
model, contrary to classical representation, modeled Roll: K = I X p − mzG (u + ru )] (3)
6 degrees of freedom motion of the surface vessel Yaw: N = I Z r + mxG (v + ru ) (4)
vectorially.
η = J ( η) υ 2.2.4 Force and Moments Acting on Surface
Mυ + C (υ)υ + D (υ)υ + g ( η) = τ +g0+w Vessel
Above; M is the moment of inertia including Basically force and moments acting on surface
added mass, C(υ) is Coriolis matrix, D(υ) is vessel can be divided to four as; hydrodynamics
damping matrix, g(η) is gravitational force vector force and moments, external (environmental) loads,
and τ is the vector showing the force and moments control surface forces (rudder, fin..) and propulsion
of the propulsion system that causes motion. This (propeller) forces. Force and moments can be
representation will be used in this study. expressed according to axis system;
2.2.1 Motion Equations Surge: X = XH + XR + XE + T
Sway: Y = YH + YR + YE
Representing the motion equations in the Cartesian
system of coordinates (body-fixed reference frame) Roll: K = KH + KR + KE
and defining xG, yG and zG as the position of the Yaw: N = NH + N R + N E
ship’s CG, the well known motion equations of a Description of indices is; H, Hydrodynamic
rigid body are giving by the following (Fossen, force and moments originating from fluid-structure
1991): interaction, R , forces that affects control surface are,
Surge: E, environmental external loads (Wave, current,
X = m[u + qw − rv + xG (q 2 + r 2 ) + yG ( pq − r) + zG (rp + q )] wind), T, propulsion force.
Sway:
Y = m[v + ru − pw − yG ( r 2 + p 2 ) + zG ( qr − p ) + xG ( qp + r )] Hydrodynamic Forces and Moments
Heave: Integration of the water pressure along the wetted
Z = m[ w + pv − qu − zG ( p 2 + q 2 ) + xG ( rp − q ) + yG ( rq + p )] area of the surface vessel causes hydrodynamic force
and moments within the platform. These force and
Roll :
K = I X p + ( I Z − I y ) qr + m[ yG ( w + pv − qu ) − zG (u + ru − pw)]
moments can be defined, with the velocity and
acceleration terms as a nonlinear axes system, by
Pitch: using Abkowitz method.
M = I y q + ( I x − I Z ) rp + m[ zG (u + qw − rv) − xG ( w + pv − qu )]
Yaw: Most important step on developing maneuver
N = I Z r + ( I y − I x ) pq + m[ xG (v + ru − pw) − yG (u + qw − rv)] model is expanding force and moment terms in
Taylor’s series. This way, nonlinear terms act as
independent variables and form a polynomial
2.2.2 Simplifying Assumptions equation. The function and its derivatives have to be
continuous and finite in the region of values of the
Simplifying assumptions used in this study are
variables to use the Taylor's expansion. Certainty of
following:
the model alters depending on where the expansion
• The rotational velocity and acceleration about
is finished.
the y-axis are zero (q, = 0).
• The translational velocity and acceleration in Force and moments, which were obtained by
the z direction are zero. (w, = 0). expanding Taylor series until third power, are under
• The vertical heave and pitch motions are mentioned (Abkowitz, 1969; Sicuro, 2003)
decoupled from the horizontal plane motions.
• The vertical centre of gravity, (VCG), is on the X hid = X u (u ) + X vr vr + X uu v 2 (5)
centerline and symmetrical (yG=0) Yhid = Yv v + Yr r + Yp p + Yφ uv φ uv + Yφ ur φ ur

2.2.3 Simplified Motion Equations +Yφ uu φ uu + Yu v u v + Yur ur + Yv v v v (6)

Applying simplifying assumptions to the general +Yv r v r + Yr v r v


motion equations, the following simplified equations
of motion are obtained

93
ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

K hid = Kv v + K p p + K u v u v + Kur ur + Kv v v v where, ω : Wave Frequency [rad/sec], V: Wind


+ Kv r v r + K r v r v + K uv φ uv φ Speed (at 19,5 m above sea) [m/s]
(7)
+ K ur φ ur φ + Kuuφ uuφ + K u p u p The Current and Wind Models
+ K p p p p + K p p + Kφφφ φφφ − ΔGz (φ )
Typically wind models only treat the force and
moments that are directly related to surge, sway and
yaw motions. In this study, the wind model is
Obtaining Hydrodynamic Derivatives
obtained by using Isherwood Method.(Isherwood
In order to obtain hydrodynamic derivatives three
1972)
basic methods can be used.
Wind forces and moments acting on a surface
¾ By means of basin test using the realistic
platform are usually defined in terms of relative
model
wind speed VR (knots) and relative angle γR (deg).
¾ By using CFD (Computational Fluid
The wind forces for surge and sway and the wind
Dynamics) software
moment for yaw as is shown.
¾ By using empirical formulae
In this study third method was used. 1
X wr = C X (γ R ) ρ wVR2 AT (9)
Hydrodynamic derivatives have been used by the 2
empirical formulae of the source Inoue et al. (1981). 1
Ywr = CY (γ R ) ρ wVR2 AL (10)
To have an opinion about validity and fidelity of the 2
empirical formulae, parameters of a merchant ship 1
that was chosen from literature was used. By using N wr = C N (γ R ) ρ wVR2 AL L (11)
2
these parameters and related formulae hydrodynamic
derivatives were calculated and compared with the where CX, CY and CN are the force and moment
equivalent in the literature.(Table 2) coefficients, ρw is the density of the air, AT and AL
are the transverse and lateral projected areas and L is
Table 2: Comparing the hydrodynamic derivatives the overall length of the ship. (Isherwood, 1972).
obtained from the model data and empirical formulae. The equations of current forces and moments are
similar with wind forces and moments.

2.3 Nonlinear Equations of Motion


When the simplified 4 degrees of freedom motion
model, which was obtained in previous section, was
associated with hydrodynamic forces and
environmental external loads a nonlinear maneuver
The Environmental Disturbances model can be obtained. To behave like independent
The environmental disturbances acting on the variables and become coefficients of a polynomial
surface vessels can be grouped into two main motion equation, hydrodynamic derivatives are
categories; the wave model, the current and wind derived by another software that makes use of ship
models. geometry. As well, terms of ship motion equations
are normalized relative to the ship velocity. (Fossen,
The Wave Model 1991)
When real data regarding the complicated seas lacks,
idealized mathematical spectrum functions are ′ v ′2
X ′ = X ′(u ′) + (1 − t )T ′( J ) + X vr′ v′r ′ + X vv
(12)
generally used for marine calculations. One of the + X rr′ r ′2 + X φφ
′ φ ′2 + cRX FN′ sin δ ′
easiest and commonly used of these calculations is
the Pierson – Moskowitz spectrum where a wave ′ v ′3 + Yrrr
Y ′ = Yv′v′ + Yr′r ′ + Yp′ p ′ + Yφ′φ ′ + Yvvv ′ r ′3
spectrum formula is provided for winds blowing
over an infinite area and at a constant speed for over ′ v ′2 r ′ + Yvrr
+Yvvr ′ v ′r ′2 + Yvv′ φ v ′2φ ′ + Yv′φφ v ′φ ′2 (13)
a sea of full state. In this study this spectrum is used +Yrr′ φ r ′2φ ′ + Yr′φφ r ′φ ′2 + (1 + aH ) FN′ cos δ ′
while a wave model is created.(Berteaux, 1976)
This spectrum is expressed as follows due to the ′ v′3 + K rrr
K ′ = Kv′ v′ + K r′ r ′ + K ′p p′ + Kφ′ φ ′ + K vvv ′ r ′3
wave frequency and wind speed. ′ v′2 r ′ + K vrr
+ K vvr ′ v′r ′2 + Kvv
′ φ v′2φ ′ + Kv′φφ v′φ ′2 (14)

0.0081g 2 ⎡ ⎛ g ⎞
4⎤
+ K rr′ φ r ′2φ ′ + K r′φφ r ′φ ′2 − (1 + aH ) z R′ FN′ cos δ ′
Sξ = exp ⎢ −0.74 ⎜ ⎟ ⎥ (8)
ω5 ⎢⎣ ⎝ Vω ⎠ ⎥⎦

94
MODELING, SIMULATION AND FEEDBACK LINEARIZATION CONTROL OF NONLINEAR SURFACE VESSELS

′ v′3 + N rrr
N ′ = N v′ v′ + N r′ r ′ + N ′p p′ + Nφ′ φ ′ + N vvv ′ r ′3 20° rudder angle. For a zig-zag maneuver, when the
angular acceleration plotted is against angular
′ v′2 r ′ + Nvrr
+ N vvr ′ v′r ′2 + N vv
′ φ v′2φ ′ + Nv′φφ v′φ ′2 (15)
velocity it shows how non-linear ship response can
+ N rr′ φ r ′2φ ′ + N r′φφ r ′φ ′2 + ( xR′ + aH xH′ ) FN′ cos δ ′ be (Figure 4).

3 PHASE PLANE ANALYSIS


3.1 Phase Portrait of Course Keeping
Phase portraits of surface platform are shown. Yaw
angle (psi) versus its derivative yaw rate (r) in
Figure 2 and Roll angle versus roll rate in Figure 3
are used to obtain the phase portraits. If the real part
of the eigenvalues is positive, then x(t) and x(t) both
diverge to infinity, and the singularity point is called
an unstable focus.

Figure 4: Phase Portrait of Zig Zag Maneuver.

4 LYAPUNOV STABILITY
THEOREM FOR SURFACE
PLATFORM DYNAMIC
A fully actuated surface platform can be described
by
Mυ + C ( υ)υ + D ( υ) υ + g ( η) = Bu = τ
η = J ( η) υ
Figure 2: The Phase Portrait (Yaw vs Yaw rate).
where J(η) is singular for θ = ±90 degrees (Euler
angles), M= MT>0 and D(ν) = DT(ν) > 0. The
The phase portrait in Figure 3 demonstrates that position is controlled by
the unstable free motion of the surface platform.
u = BT ( BBT )−1 ⎡⎣ g (η) − J T (η) K P η⎤⎦

(
where Kp = KTp > 0. Let V = 1 υ T Mυ + η T K Pη
2
)
be a Lyapunov function candidate for the closed-
loop system (4.1), (4.2) and (4.3). We take the time
derivative of the Lyapunov function candidate to
obtain
(
V = υT M υ + J T (η) K P η )
(
= υ Bu − C (υ )υ − D (υ )υ − g (η ) + J T (η ) K Pη
T
)
= υ (− C (υ )υ − D(υ )υ )
T

= −υ T D(υ )υ
Figure 3: The Phase Portrait(Roll angle vs Roll rate).
which is negative semidefinite. Asymptotic
3.2 Phase Portrait of Zig Zag stability can then be established by applying
Maneuver LaSalle’s invariance principle, but the equilibrium
point (η, ν)=(0, 0) is only locally asymptotically
It is intended that the surface platform makes zig- stable since J(η) is singular for θ = ±90 degrees.
zag maneuvers of 45° with a velocity of 8 m/s with

95
ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics

5 FEEDBACK LINEARIZATION 6 EXPERIMENTAL RESULTS


The basic idea with feedback linearization is to The crucial parameters of the surface platform
transform the nonlinear systems dynamics into a chosen for the illustration have been displayed in
linear system (Freund (1973). Conventional control Table 3.
techniques like pole placement and linear quadratic
optimal control theory can then be applied to the Table 3: The main parameters of the surface platform.
linear system. Feedback linearization allows us to
design the controller directly based on a nonlinear
dynamic model that better describes a ship
maneuvering behavior. Consider Norrbin's nonlinear
ship steering equations of motion in the form
(Fossen 1992):
mψ + d1ψ + d 3ψ 3 = δ (16)
here m = T/K, d1 = n1/K and d3 = n3/K. In the sample application, it is intended that the
Taking the control law to be: surface platform makes zig-zag maneuvers of 45°
δ = mˆ aψ + dˆ1ψ + dˆ3ψ 3 (17) with a velocity of 8 m/s. The route information
regarding this task is inputted by the VR-Forces
where the hat denotes the estimates of the graphical user interface (Figure 6).The results below
parameters and a, can be interpreted as the have been produced after running the simulation for
commanded acceleration, yields: 800 seconds.
~ a + d~ ψ + d~ ψ 3
m(ψ − aψ ) = m (18)
ψ 1 3

Here m~ = mˆ − m , d~ = dˆ − d and d~ = dˆ − d
1 1 1 3 3 3
are the parameter errors. Consequently, the error
dynamics can be made globally asymptotically
stable by proper choices of the commanded
acceleration aψ. (Fossen 1992) In the case of no
parametric uncertainties, equation (18) reduces to:
Figure 6: The route defined for the platform.
ψ = aψ which suggests that the commanded
acceleration should be chosen as: In this application, which is known as the zig zag
test of Kempf in the literature (Kempf, 1932), the
aψ = ψd − K dψ~ − K pψ~ (19)
initial speed of the platform has been given as 0. The
where ψ d is the desired heading angle and platform is ordered to move to the specified
ψ~ = ψ − ψ d is the heading error. This in turn yields waypoints one by one by increasing its velocity up
to 8 m/s. It takes the platform 96 seconds to reach to
the error dynamics: the first point. The first loop is accomplished in
ψ + K dψ~ + K pψ~ = 0 (20) approximately 295 seconds. The results are
acceptable for the motion behaviors that are
The block diagram of the control system is supposed to be realized by a large platform and
shown in Figure 5. satisfactory in terms of simulation.

Figure 5: Block Diagram of System.

Figure 7: (a) Change of location.

96
MODELING, SIMULATION AND FEEDBACK LINEARIZATION CONTROL OF NONLINEAR SURFACE VESSELS

7 CONCLUSIONS
In this study, feedback linearization control has been
implemented in a nonlinear surface vessel model
including sea-state modeling (wave, current, wind).
The performance of the maneuver controller has
been illustrated through a simulation study. The
results are acceptable and satisfy for the needs of
military simulation. Although we have designed our
control to cover all influences, a more specified
design can upgrade the performance in each
Figure 7: (b) Change of velocity. different case. In the future work, the performance
of the controller may be compared with an
intelligent control technique.

REFERENCES
Abkowitz, M. A. , 1969. Stability and Motion Control of
Ocean Vehicles, M.I.T. Press, Cambridge,
Massachusetts.
Berteaux, H. O. , 1976. Buoy Engineering, Wiley and
Sons, New York.
Fossen, T. I., 1991. Nonlinear Modeling and Control of
Underwater Vehicles, Dr. Ing. thesis, Dept. of
Engineering Cybernetics, The Norwegian Institute of
Figure 8: Changes in the yaw angle and rudder angle of Technology, Trondheim.
the surface platform. Fossen, T. I. and Paulsen, M. J., 1992. Adaptive Feedback
Linearization Applied to Steering of Ships,
Controller performance can tried by some Proceedings of the 1st IEEE Conference on Control
Applications (CCA'92), Dayton, Ohio, September 13-
different route applications:
16, 1992, pp. 1088-1093.
Freund, E., 1973. Decoupling and Pole Assignment in
Nonlinear Systems. Electronics Letter, No.16.
Inoue, S., Hirano, M., Kijima, K., 1981. Hydrodynamic
derivatives on ship manoeuvring; International Ship
Building Progress, Vol. 28.
Isherwood, R. M. , 1972. Wind Resistance of Merchant
Ships, RINA Trans., Vol. 115,pp. 327-338.
Kempf, G., 1932. Measurements of the Propulsive and
Structural Characteristics of Ships, Transactions of
SNAME, Vol. 40, pp. 42-57.
SNAME, 1950. The Society of Naval Architects and
Marine Engineers. Nomenclature for treating the
motion of submerged body through a fluid, Technical
Research Bulletin No. 1-5

Figure 9: Controller performance in different routes.

97