You are on page 1of 6

FUZZY AUTOPILOT FOR SHIPS EXPERIENCING SHALLOW WATER EFFECT

IN MANOEUVERING

Zoran Vuki *, Edin Omerdi and Ljubomir Kulja a

University of Zagreb, Faculty of Electrical Engineering and Computing


Department of Control and Computer Engineering in Automation
Unska 3, HR-10000 Zagreb, Croatia

*
(e-mail: zoran.vukic@fer.hr)

Abstract: A fuzzy autopilot for ship path control is proposed. Nonlinear model of a ship and a steering subsystem is used.
The autopilot uses heading signal and yaw rate signal to produce a command rudder angle. The autopilot does not use
lateral offset from the nominal track. Input variable fuzzyfication, fuzzy associative memory rules and output set
defuzzyfication are described. The influence of the shallow water effect during larger maneuver is analyzed.

Keywords: Disturbance rejection, Fuzzy control, Nonlinear control systems, Ship control

1. INTRODUCTION ship types and ship velocities. Notch filter was used
to minimize the effects of wave disturbance.
The conventional autopilot for ship's course keeping Shallow water effects during manoeuvering were
involves the heading angle feedback. However, by analyzed for the nonlinear model of ESSO 190000
including an additional position feedback, a ship dwt Tanker.
guidance system can be designed (Fossen, 1994).
The desired route is most easily specified by way This paper is organized as follows: Section 2
points. Modern sea going vessels have a range of introduces mathematical models of the ship, the
navigation aids including globa positioning system steering machine and the disturbances. Section 3
(GPS) receivers, Doppler sonar, gyrocompass etc. describes the course controller designed with fuzzy
These devices provide information required to logic. Section 4 introduces the turning concept and
implement track guidance. Nonlinearity of the ship the implementation of course controller in track-
model and the steering gear subsystem during keeping systems. Section 5 presents the simulation
course-changing and the lack of a simple results. Finally, Section 6 summarizes the
mathematical model makes it appropriate to design concluding remarks.
the controller with fuzzy logic instead of the
conventional approach. An accurate following of 2. MATHEMATICAL MODELS
the desired track is of great importance here.
Although strictly speaking normal navigation is also Although the design of a fuzzy controller does not
a track-keeping problem, this paper particularly depend on a mathematical model of the process,
discusses an autopilot for accurate track-keeping in such a model is necessary to simulate various
manoeuver. The fuzzy autopilot proposed here was motions of ship.
designed and tested by simulation in MATLAB
using SIMULINK with Fuzzy Logic Toolbox. The
autopilot is designed to be used for a wide range of
2.1 Ship dynamics In this paper we consider wave disturbance during
course-keeping and change of depth under keel
In order to verify controller behavior in different during track-keeping.
conditions, it is useful to simulate the control law
against a realistic model of the vessel. In this paper Wind-Generated Waves
we shall deal with two nonlinear ship models: We shall deal with the second-order wave transfer
Mariner Class Vessel (Length between function approximation (Fossen, 1994). This model
perpendiculars is Lpp=160.93 m) and ESSO 190000 is written as
dwt Tanker (Lpp=304.8 m). The nonlinear Kω s
mathematical models which describe the dynamic h ( s) = (1)
between the rudder angle δ and the yaw motion for s + 2ξw0 s + w02
2

these ships are given in (Fossen, 1994, Appendix A linear state-space model can be obtained by
E). Ship's models were transformed to S-functions transforming this expression to the time-domain by:
and adapted for the on-line simulation. d2y dy dω
2 + 2ξω 0 + ω 02 y = Kω (2)
dt dt dt
2.2 Steering Gear Subsystem dx h1
Defining = x h 2 and x h 2 = y h , the state-
dt
The steering gear subsystem considered is the "two- space model can be written as:
loop" electrohydraulic steering subsystem common dxh1
on many ships. The nonlinear steering gear model is
dt = 0 1 xh1
+
0
ω (3)
shown in Fig. 1. (Vuki , 1989). dxh 2 − ω02 − 2ξω0 xh 2 Kω h
dt
where ω h is a zero-mean white noise process. Due
to its simplicity, this model is useful for control
systems design.

Depth under keel


The shallow water effect for the ESSO Tanker is
given by a water depth parameter (Fossen, 1994):
T
ζ= (4)
h−T
where T (m) is the ship draft (T=18.46 m) and h > T
Fig. 1. Nonlinear steering gear model. is the water depth. If states ζ ≥ 0.8 (i.e.
h ≤ 41535
. m ), additional terms appear in the
Table 1 Parameter values for steering subsystem nonlinear equations of motion and changes
dynamics of the ship.
Telemotor Position Rudder Servo
Servo Actuator
3. FUZZY LOGIC COURSE AUTOPILOT
K = 4 [°/s] N = 5 [°/s]
2D = 0.4 [°] PB = 7 [°]
The fuzzy autopilot for course keeping uses two
H = 0.8 [°]
control inputs: heading error e = ψ d − ψ and yaw
rate r = dψ / dt . The control action generated by
2.3 Disturbances the autopilot is the command rudder angle δC . Fig.
2. shows a simple block diagram of the autopilot
There are several disturbances with various effects (Tovornik, 1995).
on the system to be taken into account (Amerongen,
1979). Three classes of disturbances can be e error
Conditioning
distinguished: y
Conditioning
δC
r errordot
• "disturbances" which affect the dynamics of the Conditioning

system (e.g. the depth of water), Fuzzy Logic


• disturbances which cause additional signals in Course Controller

the system (e.g. waves),


• disturbances that corrupt the measurements (e.g. Fig. 2. Block diagram of the fuzzy logic course
noise on the position measurements). autopilot.
Block "Conditioning" serves for scaling and
conditioning controller input and output variables.
The membership functions of the fuzzy sets are In Fig. 5. the block diagram of the control system
labeled as follows: for the course-keeping with fuzzy autopilot is given.

Table 2 Labels for the membership functions Disturbances

Ψd Ψ
e
Fuzzy Logic Course δC δ

NB negative big PS positive small Ψ


r Controller Steering Subsystem Ship

NM negative medium PM positive medium


NS negative small PB positive big
Notch Filter
ZE zero
Fig. 5. Block diagram of the control system for the
course-keeping.

4. FUZZY LOGIC TRACK-KEEPING


AUTOPILOT

A track-keeping autopilot (given in Fig. 6.) can be


obtained by introducing an additional position
feedback in the control system shown in Fig. 6. A
Fig. 3. Membership functions of fuzzy sets for ship position (X(t), Y(t)) is calculated from
error and errordot. kinematics equations. In a real system it can be
obtained from GPS (Global Positioning System).

Waypoints(Xd, Yd)

Disturbances
Xd Yd

Calculatingdesiredyaw Ψd e X
FuzzyLogicCourse δC δ
angle
r Controller SteeringSubsystem Ship Y
Ψ
X Y
Ψ

Fig. 4. Membership functions of fuzzy sets for y. Fig. 6. Block diagram of control system for track-
keeping.
Fig. 3. gives the membership functions of fuzzy sets
used for input variables error and errordot The desired route is most easily specified using way
( onlagi , 1996). In Fig. 4. membership functions points (P1, P2, ..., Pn) with coordinates Pi=(Xi, Yi).
of fuzzy sets for output variable y are given. We shall use a turning concept shown in Fig. 7.,
Different shapes of membership functions were where it is supposed that ship moves in a straight
analyzed, but forms of membership functions shown line between way points. A track changing
in Fig. 3. and Fig. 4. with Mamdani inference maneuver is performed in such a way that the ship
mechanism gave the best results (Polkinghorne, moves in a circle arc.
1995).
Pi+1

Table 3 Rulebase of the fuzzy course autopilot Pi-1

error Ship
e NB NM NS ZE PS PM PB ρ0
WOP ρ0 = d (WOP, Pi )
r NB NB NB NB NB NM NS ZE WOP*
ϕ
Pi
r NM NB NB NB NM NS ZE PS
o NS NB NB NM NS ZE PS PM Fig. 7. Turning concept for track-keeping.
r ZE NB NM NS ZE PS PM PB
d PS NM NS ZE PS PM PB PB The wheel over point (WOP*) is the point where a
o PM NS ZE PS PM PB PB PB
t PB ZE PS PM PB PB PB PB ship leaves a straight line motion and enters the
circle arc and vice versa (Holzhüter, 1995). The
Each control input has seven fuzzy sets so that there WOP* will not be a starting point of the turning
is a maximum of 49 fuzzy rules. Table 3 shows the manoeuver, because it is impossible to change the
complete rulebase for the controller. turn rate r of the ship instantaneously. The model-
based wheel-over point WOP which indicates the
start of the manoeuver lies about one ship length
before WOP*. The position of WOP and value of
radius R depend on the angle ϕ=∠Pi-1PiPi+1. Fig. 8.
shows dependence of the position of the WOP ( ρ0 )
on the angle of the ship's course change ( ϕ ) for a
Mariner Class Vessel. This dependence is built in a
simulation algorithm and used for automatic
calculation of WOP. Dependence of the radius of
the circle (R) upon the angle ϕ can be found in a
similar way.

ρ0 = ρ0 (ϕ ) Mariner Class Vessel


Fig. 9. Course changing manoeuver without wave
disturbance.

dy

Fig. 8. Dependence of the position of the WOP


upon the angle of the ship's course change
for the Mariner Class Vessel.

5. SIMULATION RESULTS Fig. 10. Time history and power spectral density of
wave disturbance.
Fig. 9. shows simulation results for the fuzzy course
keeping autopilot. It presents time responses of
heading, yaw rate, command rudder angle and
rudder angle and controller inputs and output during
a course changing maneuver without disturbances.
Heading time response is without overshoot and
oscillation during transient response.

Fig. 10. and 11. shows the influence of wave


disturbance on the course changing performance.
Fig. 10 shows the time history and power spectral
density of wave disturbance. Since most of the
energy in the wave spectrum is located around the
modal frequency of the wave spectrum, notch filter
with the natural frequency equal to the encounter
frequency is used for wave filtering. A small
overshoot appears in the heading time response due Fig. 11. Course changing manoeuver with wave
to an additional phase lag introduced by a notch disturbance.
filter.

Fig. 12. presents way points guidance with the fuzzy


autopilot for a Mariner ship. The desired route is
given with way points P1, P2, ..., P6. Fuzzy autopilot
uses only two inputs (heading and yaw rate) and has
a good tracking perfomance.
P3
P2 P4

------- Trajectory of the ship for water depth


h=30 m

 Trajectory of the ship for water depth


h=250 m

P1 P5

Beginning of the manoeuver


P6 (415, 0)

Fig. 12. Way point guidance by Line of Sight Fig. 14. Manoeuver in a different depths of water.
(LOS). Way points are denoted with P1, Dynamics of the ESSO Tanker differs at
P2, ..., P6. Ship moves in a straight line deep water (h=250 m) and shallow water
between way points. (h=30 m).

Pfinal

Shallow water
h=30 m

R=420 m (580, 420)


shallow water (h=30 m, zeta=1.5997)
introduces additional hydrodynamic terms
Deep water
break point (h=41.535 m, zeta=0.8) h=250 m
Shallow water
B effect
(h=250 m, zeta=0.0797)

P A Pstart

Fig. 13. Dependence of a water depth parameter ς Fig. 15. Shallow water effect in manoeuvering1. At
upon water depth h (see (4)). In a shallow t=130 s depth of water has changed
water (ς≥0.8, h≤41.535 m) additional suddenly from 250 m to 30 m.
hydrodynamic terms are introduced into
mathematical model of ESSO Tanker.

The dependence of a water depth parameter ζ upon


water depth h is given in Fig. 13. Two points are
considered on this graph: point A (deep water,
h=250 m, ζ=0.0797) and point B (shallow water,
h=30 m, ζ=1.5997). Changes in ship dynamics for
different depths of water can be seen from Fig. 14.,
where it is supposed that manoeuver begins at the
same point. The only difference between two
simulations was water depth: in the first case it was
250 m and in the second 30 m.

Now we consider the influence of the shallow water


effect in manoeuvering. A ship moves in a straight
line with forward speed of aprox. 16 knots. Water
Fig. 16. Time responses of variables during
depth is 250 m. The ship begins manoeuver at the
manoeuvering with shallow water effect at
point (415, 0) and moves in a circle arc
t=130 s.
(approximately, see Fig. 15.). At t=130 (s) water
depth suddenly changes to 30 m.

1
The circle in this picture looks like an ellipse
because aspect ratios for x and y axes are not equal
Dynamics of the ship have changed, as can be seen Fossen, T.I. (1994). Guidance and Control of
from Fig. 16. A small picture gives an enlarged part Ocean Vehicles, John Wiley&Sons.,
of time responses for forces and moment around Chichester.
point where depth of water was changed abruptly. Holzhüter, T. and Schultze R. (1995). Operating
Fuzzy autopilot brings a control action in such a Experience With a High Precision Track
way that ship stops moving in a circle arc and enters Controller for Commercial Ships,
the desired direction easily. A fuzzy autopilot reacts Proceedings of the 3rd IFAC Workshop on
appropriately to the shallow water effect by forcing Control Applications in Marine Systems,
the ship to enter the desired heading without Trondheim, pp. 270-277
unacceptable errors. However, the tracking error is Parsons, M.G., A.C. Chubb and Y. Cao (1995). An
higher, as would be expected because of the change Assessment of Fuzzy Logic Vessel Path
in ship dynamics. Control, IEEE Journal of Oceanic
Engineering, Vol. 20, No. 4, pp. 276-284
Polkinghorne, M.N., G.N. Roberts, R.S. Burns and
6. CONCLUSION D. Winwood (1995). The Implementation of
Fixed Rulebase Fuzzy Logic to the Control
This paper presents the development of a fuzzy of Small Surface Ships, Control Eng.
autopilot for course and track-keeping of ships. The Practice, Vol. 3, No. 3, pp. 321-328
research was conducted in two parts: course Tovornik, B. (1995). Fuzzy control of ship's control
keeping and track keeping. The initial research systems, Proceedings of Electronics in
concentrated on the design of a simple course Marine, ELMAR '95, Zadar, Croatia, pp.
keeping system. The next stage of the research used 213-216
this system as a starting point for building accurate Vuki , Z. (1989). Design of Adaptive Guidance
track keeping system. The results of the simulation System for Cargo Ships (in Croatian),
show that it is possible to design an autopilot for Doctoral disertation, University of Zagreb,
track-keeping of ships using fuzzy logic course Faculty of Electrical Engineering, Zagreb.
keeping autopilot.

In a large manoeuver (where nonlinear dynamics of


ship and steering gear play an important role) with
shallow water effect producing change in ship
dynamics, the fuzzy autopilot performs well. There
are two possibilities for improving the perfomance
in manoeuvering:
1. The fuzzy autopilot stays unchanged, but scaling
factors on controller inputs and output are
dynamically modified. This leads to an adaptive
fuzzy autopilot.
2. The fuzzy autopilot is augmented with an
additional input - lateral offset from the nominal
path.
Research in both of these directions is continuing.

The effective use of the same fuzzy autopilot design


with two different ships (Mariner and ESSO 190000
dwt Tanker) with a ratio of lengths near 2 illustrates
the versatility of this type of control. The simplicity
of the proposed autopilot and the possibility of
using it for different ships without retuning its
parameters was one of our goals from the
beginning. This autopilot achieved the goal.

REFERENCES

Amerongen, J.V. (1979). An adaptive autopilot for


trackkeeping, Ship Operation Automation,
III. Proceedings of the 3th IFIP/IFAC
Symp., Tokyo, pp. 105-114
onlagi , D. (1996). Design of Fuzzy Control
Systems (in Croatian), KOREMA, Zagreb.

You might also like