International Journal of Innovative Technology and Exploring Engineering (IJITEE)
ISSN: 2278-3075, Volume-8 Issue-12, October 2019
5714
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
Retrieval Number: L40031081219/2019©BEIESP
DOI: 10.35940/ijitee.L4003.1081219
Abstract: An Artificial Neural Network is a well-known AI
technique for replicating human brain and offering suitable
solution for any unpredictable complicated problem. Taking the
advantage of it, this research will analyse the applicability of
Neural Network Controller for ship manoeuvring, such as course
changing. To train the controller, optimized teaching data are
used to keep the consistency in the data as it could enhance the
learning ability of the controller while training. A double layered
feed-forward neural network and back propagation method are
found suitable for this purpose. Later-on, simulations are done to
justify the effectiveness of the trained controller for unknown
situations.
Index Terms: Artificial Neural Network, Intelligent Controller,
Numerical Analysis, Optimisation, Ship Manoeuvring
I. INTRODUCTION
Manoeuvring of a ship greatly depends on the knowledge
of an operator that he gains through years of experience.
Therefore, an inexperienced captain often faces difficulties in
unfavourable situations, such as manoeuvring through
narrow channels or navigating under high wind disturbances.
In such a situation, relying on the automation would be a
great relief for them. However, while using conventional
control, a successful application could only be found within a
well-constrained environment. As a result, researchers are
very keen on developing intelligent controllers, which can
deal with the robustness and propose a suitable solution for
any given problem. One of such intelligent systems is
Artificial Neural network, which is inspired by the central
nervous system of human‟s and can replicate the human‟s
action in solving complex problems with a lot of
uncertainties. Yamato, Uetsuki and Koyama [1] first used
ANN as a controller and he chose it for automatic ship
berthing. Later on, Fuji and Ura [2] ensured that ANN could
be used as both supervised and non-supervised controller.
After him, researchers continued to use ANN for temperature
control, paper mill waste-water treatment control, process
control etc. Hasegawa and Kitera [3] and Im and Hasegawa
[4], [5] had continued their research on applying ANN for
automatic ship berthing. However, the success was not up to
mark as the controller often confused to navigate the ship up
to the pier in wind disturbances. By this time, Ohtsu, Mizuno,
Kuroda and Okazaki [6] proposed a new strategy to do
optimisation using nonlinear programming language (NPL)
method which allowed the user to carry out the optimisation
Revised Manuscript Received on October 05, 2019.
Yaseen Adnan Ahmed, Maritime Engineering Technology Department,
Umiversity Kuala Lumpur, Malaysian Institute of Marine Engineering
Technology, Lumut, Malaysia. yaseen.ahmed@unikl.edu.my
Iwan Zamil Mustaffa Kamal, Maritime Engineering Technology
Department, Umiversity Kuala Lumpur, Malaysian Institute of Marine
Engineering Technology, Lumut, Malaysia. iwanzamil@unikl.edu.my
Mohammad Abdul Hannan, Marine Technology Department,
Newcastle University in Singapore, Singapore.
for any desired set of equality and non-equality constraints.
His strategy helped a number of researchers who were
struggling to train the controller effectively as they need to do
a lot of experiments to get the data, which were no doubt
expensive, but also time-consuming. A paper published by
Ahmed and Hasegawa [7], [8] clearly demonstrates how to
utilize Ohtsu‟s [6] proposed method in creating consistent
teaching data and get a well-trained controller for ship
manoeuvring. This research will use the same strategy to
create the teaching data, which are consistent and use it to
train a neural network for different course changing.
Changing heading is a simple manoeuvre, however, to
change it within a minimum time is a chal-lenge. Therefore,
the research is aimed to propose an ANN controller which is
able to change the ship‟s course in minimum time. An
optimisation function „fmincon‟ is used for creating teaching
data and Levenberg-Marquardt algorithm is used to train the
net on Matlab. Several types of networks are then analysed
and the one with minimum mean squared error (MSE) is
selected for further research. Simulations are also carried out
for unknown cases to demonstrate the effectiveness of the
proposed controller.
II. METHODOLOGY
Ship Model
As for this research, a S60 container ship has been chosen,
which is equipped with a single rudder and a single propeller.
The principle particulars of this ship is given in Table I.
Table. 1 Ship particulars
Parameter Value
L [m] 276.0
B [m] 39.40
d [m] 15.80
0.7
Creation of Teaching Data
In order to train a neural network controller, a consistent
set of teaching data is needed. As the controller is for course
changing manoeuver, and there are a number of possible
ways to turn a certain degree of course, this research
investigates only the minimum time course changing
manoeuvre to make sure the similarity in teaching data.
To do so, optimisation technique is used on Matlab.
Among the different types of optimisation functions available
on Matlab optimisation toolbox, „fmincon‟ is found to be the
best as it uses the nonlinear programming language (NPL)
method [6].
An Artificial Neural Network Controller for
Course Changing Manoeuvring
Yaseen Adnan Ahmed, Iwan Zamil Mustaffa Kamal, Mohammad Abdul Hannan