International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 2

Number 1

March 2008

Intelligent Ship Control System

A. Lebkowski, R. Smierzchalski, W. Gierusz & K. Dziedzicki

Gdynia Maritime University, Gdynia, Poland

ABSTRACT: The article present intelligent control system of ship motion in situations threatening with

collision. The goal of the presented system is to support the navigator in decision making, with possible full

replacement of his work in the future. In this article, it was introduced system joins work of three computer

techniques, evolutionary algorithms to marking of optimum path of passages, fuzzy logic to control ship after

set path of passage and multivariable robust control to precise movement of the vessel with very small

velocity and any drift angle. Introduced system has to assure safe trip of ship in each navigational conditions

with regard of weather conditions and met navigational objects of static or dynamic nature. For testing of the

operation the system, the marine environment simulator was used to present navigational situations in a 3D

graphical mode at the poor hydro- and meteorological conditions.

1 INTRODUCTION

Modern marine transport requires preserving dates

of delivery to harbours located all over the world,

irrespective of weather conditions and volumes of

transported cargo. In case of passenger transport, an

additional requirement should be taken into account

which is providing passengers with adequate level of

comfort. On the other hand, there is a tendency to

reduce ship operation costs, and realisation of this

task may unintentionally involve threats to human

life and natural environment. Losing transported

cargo is also possible. That is why securing safety of

sailing is one of more important issues in present-

-time marine navigation.

Among all causes of sea accidents, navigational

errors compose a relatively big group (Soares,

Teixeira 2001). Out of fifteen biggest ships lost in

years 2003-2004, as many as nine cases referred to

collision or stranding (www.isl.org).

A method leading to the reduction of sea accident

risk may be introducing solutions that support the

navigator in decision making in the situations

threatening with collisions.

For the voyage of the ship carrying passengers

and cargo to be safe, a passing path meeting safety

conditions should be determined for the ship. A

basic safety condition assumes a minimum distance

Db of safe passing of objects. Besides, the

determined passing path should meet economic

criteria, important for the shipowners, which include

the length of the passing path, time needed to cover

it, changes of ship’s speed along particular path

sections and number of manoeuvres to be performed

by the ship.

Taking into account a defined by the operator

range of observation of a navigational situation (3, 8,

12, 24, 48 sea miles) covered by the ARPA

(Automatic Radar Plotting Aids) system, the time

horizon for solving a navigational problem can range

from several minutes to 1-2 hours. The area of

observation may be obscured by navigational

constraints, of static or dynamic nature, which can

considerably affect the process of the passing path

determination.

The task of determining a safe passing path for a

ship at sea is reduced to the selection of an optimum

solution, or a subset of optimum solutions, from a

set of permissible paths. The paths are selected using

an assumed cost criterion, provided safety conditions

are unconditionally met and constraints taken into

account. When determining a safe path for the ship

motion, a compromise solution is searched for. The

compromise is, generally, made between the cost of

trajectory deviation from that assumed, or from the

shortest way leading to the assumed endpoint, and

the safety of passing navigational constraints. In this

situation, steering the ship along the determined path

taking into account parameters of ship dynamics and

meteorological conditions is reduced to determining

a passing trajectory.

In order to determine an optimal passing path for

the ship, a ISCS (Intelligent Ship Control System)

has been developed. The system makes use of united

work of two computer techniques: evolutionary

algorithms (EA) for determining the optimal passing

path, and fuzzy steering for directing the ship along

the assumed path. The trajectory of the ship motion

determined by fuzzy steering along the assumed path

is called a passing trajectory. The information on the

navigational environment is delivered to a moving

ship by a measuring system. The navigational

constraints, both static and dynamic, which ship

meets on its way, compose the navigational

environment and are modelled in the form of

polygons, the shape and dimensions of which

depend on weather conditions, region of navigation,

manoeuvring ability of the ship, its dimensions,

speed, course and bearing line, as well as speeds of

the passed objects. Concluding, the task of the

intelligent ship control system (ISCS) is controlling,

in a fuzzy way, the motion of the ship in the

navigational environment along the passing path

determined in an evolutionary way.

An essential feature of the system is its ability to

control safely and automatically the ship motion in

navigational situations. The use of the system

considerably facilitates operator’s work concerning

calculations performed in order to determine the

passing path for the ship, as well as actions taken to

keep the ship on the already determined passing

path. When performing this function, the system

takes into account all safety related legal regulations.

The proposed solution is expected to contribute in a

considerable way to the reduction in the number of

accidents recorded, and to the increase of the safety

of sea navigation. The article also presents a

simulator used for verifying the operation of

intelligent system controlling ship’s motion at sea.

Analyzed is the operation of the system in the

situation threatening with collision, and in the

presence of unfavorable hydro- and meteorological

conditions. The developed simulator presents

navigational situations using 3D graphics.

2 DESCRIPTION OF THE ENVIRONMENT

AND OBSTACLES

The ship, moving in the sea environment, meets

various navigational constraints, of both static and

dynamic nature. The static constraints include lands,

canals, shallows, straits, and/or areas with legal

traffic restrictions (traffic separation areas, water

lanes, etc.). The static navigational constraints are

approximated by polygons in a similar manner to

that used for creating electronic vector maps. The

dynamic constraints include other ships and moving

objects passed by the own ship. These obstacles are

modelled as moving hexahedrons. The area

surrounding the own ship and all approaching

moving objects is called a domain. The dimension of

the domain depends on the navigational situation

and parameters of motion and positions of the own

ship and approaching objects. The positions, speeds

and bearing lines of the approaching objects are

determined by the ARPA system. Part of the

approaching objects create collision threat for the

motion of the own ship. In the evolutionary task of

avoiding collisions it was assumed that the object is

considered dangerous if it has come into the area of

observation and can cross the course of the own ship

at a dangerously close distance, defined by the

operator depending on weather conditions and the

navigation area.

Initial conditions, assumed when determining the

passing path for the ship, include current position of

the own ship and parameters of motion of the strange

objects, determined at the initial instant by ARPA.

The determined trajectory of the ship motion has a

form of a broken line, consisting of line segments,

linking the starting point with the assumed target

point.

determinig optimum

passing path

environment

(wind, sea currents, waves)

navigational environment

(static and dynamic constraints)

STAGE 2

Trajectory Controller

STAGE 1

Evolutionary Algorithm

Multivariable Robust Controller

very small velocity and drift angle

PHASE 3

Correction of the route

of the passage

STAGE 3

Evolutionary Algorithm

- bearing line to object j

- course of object j

- speed of object j

- assumed course of own ship

- assumed speed of own ship

- speed control signal

- course control signal

Ψ - ship's course

V - ship's speed

Z - disturbances

Course Controller

compensation of weather conditions

PHASE 1

PHASE 2

Speed Controller

steering along assumed trajectory

Fig. 1. The structure of ship control in a collision situation with

use of ISCS

Fig. 2. The control subsystems for the ship motion: a) heading

and/or speed stabilisation, b) stabilisation of the turning radius,

c) roll damping, d) trajectory tracking, e) precise steering with

the low speed, f) dynamic ship positioning (DSP), g) automatic

single-buoy mooring, h) turret-mooring for FSO, FPSO or

FPDSO ships

Taking into account the hierarchical manner of

ship steering, shown in Fig. 1, the optimum passing

path (trajectory of motion) is determined by EA

using a kinetic model of ship motion. Ship’s

dynamics is taken into account, when the fuzzy

controller steers the ship along the selected trajectory

of motion.

The adopted structure of steering secures:

− At the first stage: determining a safe passing path

for the own ship on the basis of an assumed target

point and current navigational situation in the

area of navigation. The trajectory consists of a

sequence of line segments characterized by

constant course and speed depending on the

navigational situation recorded by the ARPA

system.

− At the second stage: steering the own ship along

the selected passing path, time and distance

parameters being preserved and disturbances

acting on the ship taken into account. At the

second stage, steering is determined for a selected

passing path of the ship, taking into account

dynamics of the ship, and hydro- and

meteorological conditions in the navigational

area. This stage consists of three calculation

phases, executed simultaneously: phase one, in

which the deviation from the assumed course is

currently controlled and corrected depending on

the navigational situation and effect of sea

currents, waves and wind (course controller);

phase two, in which the speed of ship motion

along the selected path is controlled (speed

controller); phase three, is realize precise

movement of the vessel with very small velocity

and any drift angle (case in Fig. 2). Such kind of

the motion is used mainly in harbours or other

constrained areas.

− At the third stage: adaptive correction of the

passing path in a navigational situation changing

in an unpredicted manner.

3 EVOLUTIONARY ALGORITHM (EA)

The passing path was determined using an

evolutionary algorithm, presented in detail in

(Łebkowski & Śmierzchalski 2003a, Śmierzchalski

1999, Śmierzchalski 2000). On the basis of EA tests

one can conclude that genetic operators are used

with different frequencies during particular phases of

operation of the algorithm (Łebkowski et al., 2005).

In order to increase EA efficiency, its operation was

divided into two phases.

In the first phase, an area of possible solutions is

searched, which contains the location of a global

optimum. EA procedures used in this phase base on

a population consisting of 50 individuals. The task

performed by EA in this phase includes the

examination of the space of permissible solutions.

The individuals composing this population are

characterized by increased probability of the use of

genetic operators, such as: crossing, soft mutation,

and repair of individual, due to the most frequent use

of those operators in the initial phase of operation of

the algorithm.

In the second phase of EA operation, the area of

solutions obtained in the first phase is exploited in

order to obtain an approximation of the global

optimum. EA procedures used in this phase base on

a population consisting of 10 individuals. The

individuals composing this population would be

characterized by increased probability of use of the

genetic operators of mutation, smoothing and gene

removal.

4 FUZZY CONTROLLER FOR SHIP’S

POSITION STEERING

The ship is kept on the assumed passing trajectory

using the rules of fuzzy inference, first proposed by

Mamdani (Mamdani 1974). The fuzzy controllers

are characterized by lower sensibility to disturbances

than that revealed by conventional controllers widely

used in naval autopilots. Another quality which

makes fuzzy controllers more effective than their

conventional counterparts is possibility to incorporate

expert’s elements and basis of knowledge into the

controller’s basis of knowledge.

As mentioned above, the fuzzy controller of

ship’s motion is divided into two parts: the course

controller and the speed controller, working

simultaneously. These two parts are structurally

identical. The input signals for the both controllers

are: the deviation of the output value from that

assumed, and speed of the deviation changes in time.

For the course controller the assumed value is the

course, while for the speed controller – speed.

The difference between the fuzzy course

controller and the fuzzy speed controller consists in

the application of different bases of rules in each

controller. For the course controller, 9 linguistic

values have been defined at each input and output.

They are: NH, NM, NL, NVL, Z, PVL, PL, PM, and

PH which, respectively, stand for "Negative High",

"Negative Medium", "Negative Low", "Negative

Very Low", "Zero", "Positive Very Low", "Positive

Low", "Positive Medium", and "Positive High".

Membership functions for rule predecessors and

successors are shown in Fig. 3.

0,00- 0,25- 0,50- 1,00

NM NL NVL Z PVL PL PM

0 1 2 3- 1- 2- 3 ∆Ψ

4- 4

- 0,75 0,25 0,50 1,000,75

PHNH

Fig. 3. The shape of the membership function for an 81-rule

fuzzy course controller

For the fuzzy course controller defined in the

above manner 81 possible Mamdani-type rules are

obtained and placed in the controller’s basis of rules.

The output signal from the controller is the signal for

steering the rudder deflection, passed to the steering

engine. A positive/negative value means steering the

ship to the left/right.

For the speed controller 7 linguistic values have

been defined. They are: NH, NM, NL, Z, PL, PM,

PH and stand, respectively, for "Negative High",

"Negative Medium", "Negative Low", "Zero",

"Positive Low", "Positive Medium", and "Positive

High". Membership functions for rule predecessors

and successors are shown in Fig. 4. The fuzzy speed

controller includes 49 Mamdani-type rules placed

in the controller’s basis of rules. The output

signal from the speed controller is the change of

position of the speed adjuster lever on the main

engine speed governor. A positive/negative value

means increase/decrease of rotational speed of the

main engine.

0,00 0,33 0,66 1,00- 0,33- 0,66- 1,00

NH NM NL Z PL PM PH

0 1 2 3- 1- 2- 3 ∆V

Fig. 4. The shape of the membership function for a 49-rule

fuzzy speed controller

The required passing trajectory of the ship,

determined by a two-phase EA and corrected by

the fuzzy controller, should be an optimal trajectory

of the ship motion for given navigational

situation taking into account current hydro- and

meteorological conditions of the own ship

environment. The hydrological conditions, which

include wind, sea currents and waves, considerably

affect the steering generated by the fuzzy controller.

The above weather disturbances are modelled by

equations which defining forces and moments

generated by them (Łebkowski & Śmierzchalski

2003b).

5 MULTIVARIABLE ROBUST CONTROLLER

The main goal of the presented control system is the

precise steering of three ship's velocities: surge,

sway and yaw. The values of velocities during such

manoeuvres are very small (often close to zero) and

therefore standard navigation logs and angular rate

meters are unserviceable due to their poor accuracy.

Kalman filters or observators systems should be used

for estimation. The block diagram of the process is

presented in Fig. 5.

For performing of any manoeuvres the regulator

calculates the two demanded forces τ

and τ

for

longitudinal and lateral directions and one moment

for turning (in the ship-fixed frame).

Fig. 5. The block diagram of the robust control system

These signals should be distributed in optimal

way between propellers installed on the ship - see

Fig. 2.The system for this purpose is commonly

named 'thrust allocation unit' and should be

created individually for every driving system. One

of the possibilities is the algorithm based on the

pseudoinverse matrix operations and logic

inequalities described in (Gierusz & Tomera 2006).

A rational methodology for designing robust regu-

lator for ship velocities can consist of the following

items:

− derivation a linear state model or transient

functions model of the vessel in the chosen range

of velocities,

− estimation of uncertainties in the system,

− defining of performance requirements,

− building open-loop scheme for calculating of

regulator,

− computing of the controller by means D-K

iteration procedure from ”µ - Analysis and

Synthesis Toolbox” (Balas et al., 2001).

Case study

The presented scheme was applied to synthesis of

the robust controller for floating training ship. The

vessel named 'Blue Lady' is used by the Foundation

for Safety of Navigation and Environment Protection

at the Silm lake near Ilawa in Poland for training of

navigators. It is one of the series of 7 various

training ships exploited on the lake. The ship 'Blue

Lady' is an isomorphous model of a VLCC tanker,

built of the epoxied resin laminate in 1:24 scale. It is

equipped with battery-fed electric drives and the

control steering post at the stern for two persons.

Details of the robust controller one can find in

(Gierusz 2006). Exemplary results of the steering are

presented below for two levels of wind. Every

Figure is divided into two parts. The left-hand side

presents the results of steering for weak wind and the

right-hand side presents the same trials performed

with medium level of wind. Presented example is

illustrated by means of 2 Figures:

− the trajectory, drawn by ship's silhouettes every

30 s, (every trajectory starts from point 0,0),

− ship's velocities (reference signals and real

values), supplemented by wind runs recalculated

in Beaufort scale.

As one can expect the tracking of the ship's

velocities is acceptable when wind is small. If it

increasing worse control quality is observed

especially in velocities. To compensate additive

disturbances we used feedforward regulator, which

the scheme was shown on Figure 6.

Fig. 6. The block diagram of the robust control system

Figures 7 and 8 present trajectories and velocities

of the ship when robust regulator is used with

feedforward part.

Fig. 7. The trajectory of the ship when feedforward controller is

implemented in the system, drawn by silhouettes every 60 s.

Initial heading o = 180 deg, the trial period t = 1950 s. The

arrows indicate the average wind direction

Fig. 8. The velocities of the ship - from the top: surge, sway

and yaw. The bottom figures present the wind speed in

Beaufort scale. Solid lines denote real values, dashed lines –

commands

The feedforward regulator used in presented

control system consists of three parts:

− low pass filters for relative wind speed and

direction, measured by means of the ultrasonic

anemometer, installed on board,

− calculation unit for wind forces and moment

estimation,

− simple regulators for three channels (longitudinal,

lateral and angular ones).

6 OPERATION OF THE SIMULATOR AND

SIMULATION TESTS

Parameters of motion of the dynamic objects and the

positions of static objects, including land contours

and shallow water regions, are initialized once when

the program is started. During the operation of the

program the information is cyclically exchanged

between the mathematical model of the ship and the

graphic environment. Changes in ship’s position,

course and/or speed are visualized in the displayed

graphics. The simulator user can control the ship and

particular parameters of its operation. There is also

a ssibility to observe the vicinity of the ship.

The vigating window of the simulator is shown

in Fig. 9.

Fig. 9. Simulator navigating window

In order to model navigational situations, twenty

3D silhouettes were implemented of various types of

vessels (tankers, bulk cargo ships, passenger ferries,

sailing vessels, and yachts) essential from the point

of view of sea low regulations. The simulator user

can observe changes in weather situation and sea

state, presented in 3D graphical technique. The

length and height of waves are changed according to

Pedersen scale, while atmospheric conditions are

determined using the Beaufort scale, for which the

visibility ranges have been determined. The radar

screen with the ARPA system implemented in the

simulator is shown in Fig. 10.

HDG 084.7

SPD 12.3 KTS

TARGET No. 09

T RANGE 12.8 NM

T BEARING 340.2

T COURSE 188.5

T SPEED 12.4 KTS

000

020

030

040

050

060

070

080

090

010

100

110

120

130

140

150

160

170

180

340

330

320

310

300

290

280

270

350

260

250

240

230

220

210

200

190

AIS

Name AQUILA

Radio call 8WQ34

Type BULK

Destinat. SAN DIEGO

Ship size 250 [m] 32 [m]

Draught 10 [m]

T RANGE 7.4 NM

T BEARING 340.2

T COURSE 188.5

T SPEED 12.4 KTS

TARGET No. 08

WAIPOINT No. 08

WPT RNG 12.8 NM

WPT BRG 340.2

DSFT 0.5 NM

CPA 3.9 NM TCPA 8:25 min

BCR 2.8 NM BCT 9:25 min

SCR -.- NM SCT -:-- min

DEPTH 129 m

WIND 2.8 KT 235.4

CURRENT 1.6 KT 230.4

UTC 14:28:31

01 / 06 / 2005

RANGE 12 NM

NORTH UP

N 052

19.3421

E 018

15.3467

T08

T05

W23

W24

Fig. 10. The window of the radar simulator

7 CONCLUSIONS

The presented Intelligent Ship Control System in a

collision situation, making use of computer

techniques: evolutionary algorithms, fuzzy control

and multivariable robust controller. The simulator

models basic dynamic parameters of the marine

environment. Taken into account are phenomena

connected with bad visibility, the effect of shallow

water, and/or the presence of other navigational

objects of static (lands, water lanes, navigational

buoys, restricted traffic areas, lighthouses) and

dynamic nature (other moving ships and areas of

unfavorable weather conditions). The simulator

allows modeling various navigational situations,

thus providing opportunities for verification of

the proposed ship control system. The presented

simulator, operating with the ISCS, may make an

effective tool for learning sea navigation. It can also

be used as the system supporting navigators in

decision making at sea.

REFERENCES

Balas, G.J., J.C. Doyle, K. Glover, A.Packard and R. Smith

2001. fi-Analysis and Sythesis Toolbox. Ver.4. The

Mathworks Inc. Natick, USA.

Gierusz, W. 2006. The steering of the ship motion: a -synthesis

approach. Archives of Control Sciences 16, 5-27.

Gierusz, W. and M. Tomera 2006. Logic thrust allocation

applied to multivariable control of the training ship. Control

Engineer-ing Practice 14, 511-524.

Łebkowski A., Śmierzchalski R., 2003a. Evolutionary-fuzzy

Hybrid System of Steering the Moveable Object in

Dynamic Environment, IFAC Conference Control

Application in Marine Systems, Girona Spain.

Łebkowski A., Śmierzchalski R., 2003b. Hybrid System of Safe

Ship Steering in a Collision Situation at Sea, VI Krajowa

Konferencja Algorytmy Ewolucyjne i Optymalizacja Glo-

balna, Łagów.

Łebkowski A., Dziedzicki K., Tobiasz M., Tomera M.,

Śmierzchalski R., 2005. A Marine Environment Simulator

for Testing Ship Intelligent Control Systems in Dangerous

Situations, Conference Advanced Computer Systems, Ełk,

Poland.

Mamdani B.E., 1974. Application of Fuzzy Algorithms for

Control of Simple Dynamic Plant, proceedings of the IEE,

vol. 121, no.12.

Soares C.G., Teixeira A.P., 2001. Risk Assessment in Maritime

Transportation, Reliability, Engineering and System Safety

74 pp.299-309.

Śmierzchalski R., 1999. The Decision Support System to

Design the Safe Manoeuvre Avoiding Collision at Sea,

3rd World Multiconference on Systemics, Cybernetics and

Informatics, USA, Orlando pp. 1124-1130.

Śmierzchalski R., Michalewicz Z., 2000. Modelling of a Ship

Trajectory in Collision Situations at Sea by Evolutionary

Algorithm, Journal of IEEE Transaction on Evolutionary

Computation No. 3, Vol.4, pp.227-241.

Śmierzchalski R., 2003. Evolutionary algorithm in problem

of avoidance collision at sea. Artificial Intelligence

and Security in Computing Systems, Kluwer Academic

Publishers, pp. 81 – 90.

Śmierzchalski R., Łebkowski A., 2003. Hybrid System of Safe

Ship Steering at Sea, IEEE International Conference on

Methods and Models in Automation and Robotics,

Międzyzdroje Poland.

World Shipbuilding and Maritime Casualties”, www.isl.org.