Sei sulla pagina 1di 8

13th Congress of Intl. Maritime Assoc.

of Mediterranean
IMAM 2009, Istanbul, Turkey, 12-15 Oct. 2009, Page 1121-1128

Decision making system for the collision avoidance of marine vessel


navigation based on COLREGs rules and regulations.
L. P. PERERA
Centre for Marine Technology and Engineering (CENTEC), Technical University of Lisbon,
Instituto Superior Técnico, Lisbon, Portugal
J. P. CARVALHO
INESC-ID, Technical University of Lisbon, Instituto Superior Técnico, Lisbon, Portugal
C. GUEDES SOARES
Centre for Marine Technology and Engineering (CENTEC), Technical University of Lisbon,
Instituto Superior Técnico, Lisbon, Portugal

ABSTRACT: A fuzzy-logic based decision making (DM) system to facilitate the collision avoidance
capabilities for marine vessels during ocean navigation is presented in this paper. The collision avoidance of
the target vessel with respect to the vessel domain of the own vessel has been analyzed and fuzzy
membership functions have been derived in this study. Fuzzy rule based (IF-THEN) decision making system
has been formulated, implemented and results are summarized. Further, decisions on the DM system are
formulated in accordance with the International Maritime Organization (IMO) rules and regulations
(COLREGs) of ocean navigation to avoid conflict situations.

1 INTRODUCTION Regulations for Preventing Collisions at Sea


(COLREGs) (IMO 1972), were formulated by the
Conventional marine vessel navigational systems International Maritime Organization (IMO) that
consist of human guidance, hence 75-96% of marine represents the importance of regulated prevention of
accidents and causalities are affected by some types the collisions in ocean navigation.
of human errors (Rothblum et al. 2006, Antão and Decision making process and strategies in
Guedes Soares 2008). Therefore implementation of interaction situations with respect to the collision
an intelligent DM system during the ocean avoidance conditions were presented by Chauvin et
navigation is a mandatory requirement to achieve al. (2008). The analysis of quantitative data
higher maritime safety standards. This reality has describing the maneuvers undertaken by ferries and
been characterized as e-Navigation (eNAV 2008). cargo-ships and the behavior of the 'give way' and
Furthermore, to have Autonomous Guidance and 'stand on' vessels with respect to the verbal reports
Navigation capabilities in ocean navigation, such as on-board recorded in a car-ferry in the Dover Strait
the system described in Moreira et al (2007), it is were also presented in the same work.
necessary to have a decision making system onboard The two main terminologies that were used in
vessels to avoid collision situations. This application recent literature with respect the collision avoidance
area is bound to become more important in the conditions could be presented as “Own vessel” (the
future ocean navigation (Fossen, 1999) due to its vessel with the DM system) and “Target vessel” (the
cost reduction and requirements of maritime safety. vessel that need to be avoided). Similar definitions
This paper focuses on a fuzzy logic based have been considered during the formulation of
Decision Making (DM) system to implement on collision situations in this study. The detection of a
vessel navigation to improve safety of the vessel by target vessel location and velocity are two important
avoiding the collision situations. The rules and factors in the decision making process of the
regulations with respect to the collision avoidance collision avoidance. The combination of radar and
conditions, i.e., Convention on the International infrared imaging to detect the other vessel conditions
were proposed by Sato et al. (1998) as a part of the ocean navigation were ignored by most of the recent
collision avoidance system. The method of literature. The negligence of the IMO rules may lead
measuring course of the target vessel and the to conflicts during collision situations. Therefore the
evaluation of the risk of the collision situations by methods ignoring COLREGs rules and regulations
image processing are further proposed as should not be implemented.
improvements of collision avoidance conditions in On the other hand there are practical issues in the
the same work. implementation of the COLREGs rules and
The defined region of vessel domain that was regulations during the ocean navigation. As an
bounded for dynamics of the marine vessel is an example, there will be implementation issues of
important factor to define the collision conditions. "give way” and “stand on" vessels of the COLREGs
The vessel domain, in a collision situation, that rules during the vessel navigation in a situation
depends on parameters of the vessel size, vessel where the moving vessel or moving obstacle with
course and heading angles of the encountered very low or very high speeds with respect to the own
vessels has been proposed by Pietrzykowski et. al vessel. Hence smart DM system must be
(2006) and fuzzy logic based domain determination implemented to be consistent with the COLREGs
system has been considered. Similarly neural- rules and regulations and also to eliminate
classifiers have been proposed by Lisowski et. al previously mentioned implementation conflicts.
(2000), as an element that is supporting the Similarly considerable amount of recent research has
navigator in the process of determining the ship's been focused on design and implementation of
domain where the area around the vessel should be optimization algorithms to find the safest path to
free from stationary or moving navigational avoid collision situations. It is observed that some
obstacles. optimization algorithms always find the safest path
The calculations based on the kinematics of two- behind the target vessel which may lead to a conflict
ship collision encounters conjunction with the with the COLREGs rules and regulations.
equations of motion were presented by Kwik (1989). Similarly some algorithms are enforcing the own
The analysis of a collision avoidance situation was vessel to navigate away from the target vessel or
illustrated with respect to the vessel velocity, turning obstacle by repulsive forces from the vessel or the
rate and direction, and the desired passing distance obstacle during the collision situations. This concept
in the same work. The collision avoidance may lead to a conflict situation when the moving
conditions among ships, shore installations and other vessel with very low speed or very high speed with
obstacles were considered by Yavin et al. (1995). In respect to the own vessel. In addition, complex
addition a case study of a ship moving from one orientation of obstacles may lead to unavoidable
point to another in a narrow zigzag channel was collision situations. Further, the concepts of "give
considered and a computational loop command way” and “stand on" vessels that were derived on the
strategy for the rudder control system associated COLREGs rules and regulations during the vessel
with the numerical differential equation solver was navigation will not be honored in by the repulsive
proposed in the same study. forces based algorithms. Hence optimizations
Ship trajectories are normally simulated by algorithms integrated with a smarter DM system
mathematical models based on maneuvering theory should be considered to overcome the above
(e.g. Sutulo et al, 2002) although approaches based problems.
on neural networks have also been proposed The consideration of course changes and/or speed
(Moreira, and Guedes Soares, 2003). Modeling of changes of vessels in ocean navigation must be
ship trajectory in collision situation by an formulated in order to avoid critical collision
evolutionary algorithm was presented by situations. However some of the recent collision
Smierzchalski et. al (2000) and comparison of avoidance applications have been focused on
computational time for trajectory generation with specific controllability of either course change or
respect to the other maneuvering algorithms were speed change. According to the COLREGs rule 8(b).
also illustrated. Static and dynamic constrains were “Any alteration of course and/or speed to avoid
considered for the optimization process of the safe collision shall, if the circumstances of the case
trajectory in the same work. admit, be large enough to be readily apparent to
another vessel observing visually or by radar; a
succession of small alterations of course and/or
2 COLREGs RULES AND REGULATIONS speed should be avoided”.
It is observed that the COLREGs rules and Hence integrated controls of course changes
regulations with respect to collision situations in as well as speed changes should be implemented
during the vessel navigation to avoid collision
situations. Similarly special measures should be
considered during integrations of course and speed
controls due to the fact that the vessels may not
response to required changes in course or speed.

3 COLLISION SITUATION
Two vessels in a collision situation are presented in
Figure 1 and the following description has been
illustrated with respect to the collision conditions.
The own vessel that has implemented the DM
system is located at the point O (Xo, Yo), and the
target vessel, the vessel need to be avoided is located
at the point A (Xa, Ya). The own and target vessels
velocities of Vo, Va and course of ψo, ψa respectively
are also presented in the same figure. The relative
velocity of the target vessel with respect to the own
vessel was defined as Va,o and relative speed |Va,o| →
Va,o and course of ψa,o could be calculated from
Va,o = Va - Vo (1) Figure 2. Vessel relative collision situation.

These regions have been separated by the


discontinue lines that are coincided with the fuzzy
regions as formulated in the fuzzy membership
function in Figure 4. It is assumed that any target
vessels should be located within these eight regions
and the decisions were formulated accordance to
each region. The gray circular region represents the
vessel domain with the radius of Rvd ( Figure 1). The
white circular region represents the critical collision
risk region due to the target vessel orientation.
As presented in Figure 2, the target vessel
position at the region II has been divided into eight
divisions of target vessel orientations with respect to
the relative course regions: II-a, II-b, II-c, II-d, II-e,
II-f, II-g and II-h. These divisions have been
separated by the discontinue lines that are coincided
with the fuzzy collision risk regions presented later.
Figure 1. Vessel collision situation. 4 COLLISION AVOIDANCE METHODOLOGY

In addition the relative distance and angle of the 4.1 Identification of Obstacles
target vessel with respect to the own vessel are The stationary and moving obstacles in the ocean
derived as the |AO| and θo respectively. All angles navigation can be identified by several instruments
have been measured with respect to the positive Y- and systems: Eye / camera, radar / Automatic Radar
axis as presented in Figure 1 and 2. The curve AB Plotting Aid (ARPA), and Automatic Identification
represents the relative path of the target vessel with System (AIS). ARPA provides accurate information
respect to the own vessel and the collision encounter of range and bearing of nearby obstacles and AIS is
angle is presented by θa,o. The vessel relative capable of giving all the information on vessel
collision situation that is similar to a Radar plot is structural data, position, course, and speed
presented in Figure 2. As presented in the figure the (Hasegawa 2009). The collection of radar data has
collision regions with respect to the own vessel have been considered as the method of identifying
been divided into eight regions: I, II, III, IV, V, VI, stationary and moving obstacles during this study.
VII, and VIII.
4.2 Collection of Navigational Information Table 1. Collision Risk Assessment and Decisions
Reg. Division Risk Rule Condition Decision
The navigational information could be categorized I d Mid. 13 NA NA
into three sections: static information, dynamic e High 14 Va,o << 0 δψ > 0
information and voyage related information. Va,o ≈ 0 δψ > 0
Collection of navigational information is an Va,o >> 0 δψ > 0
important part of the decision making process of the f Mid. NA NA
II e Mid 15 NA NA
collision avoidance in ocean applications. However
f High 16 Va,o << 0 NA
it has not been emphasized on the navigational Va,o ≈ 0 δψ > 0
information of the target vessels at this stage of the Va,o >> 0 δψ > 0
study. g Mid. 17 Va,o << 0 NA
Va,o ≈ 0 δψ > 0
4.3 Analysis of Navigational Information Va,o >> 0 δψ > 0
The collected target vessel information should be III f Mid. 15 NA NA
considered for further analysis of navigational g High 16 Va,o << 0 NA
Va,o ≈ 0 δVo<< 0
information. Three distinct situations that are Va,o >> 0 δVo<< 0
involving risk of collision with respect to the ocean h Mid. 17 Va,o << 0 NA
navigation have been recognized by the recent Va,o ≈ 0 δVo < 0
literature (Smeaton, et. al 1987), overtaking, head-on Va,o >> 0 δVo < 0
and crossing, and same situations have been IV a Mid. 17 Va,o << 0 NA
highlighted in this study. However in ocean Va,o ≈ 0 δVo < 0
navigation, complex collision situations involving Va,o >> 0 δVo < 0
combinations of the above situations could occur g Mid. 15 NA NA
h High 16 Va,o << 0 NA
and identification of each situation with respect to Va,o ≈ 0 δVo<< 0
the each collision conditions will be helpful for the Va,o >> 0 δVo<< 0
overall decisions on vessel navigation. V a High 14 NA* NA*
4.4 Assessments on the Collision Risk b Mid. NA* NA*
h Mid. 13 NA* NA*
The analysis of navigational information will be able VI a Mid. 15 NA* NA*
to predict the collision risk assessments. Therefore b High 16 NA* NA*
the assessment of the collision risk should be c Mid. 17 NA* NA*
VII b Mid. 15 NA* NA*
continuous in the navigational system for safety of
c High 16 NA* NA*
the own vessel. As presented in recent literature, the d Mid. 17 NA* NA*
mathematical analysis of collision risk detection VIII c Mid. 15 NA* NA*
could be divided into two categories: 2D methods d High 16 NA* NA*
and 3D methods. Both methods consist of locations e Mid. 17 NA* NA*
of each vessel in 2D coordinate system and the 3D * Target vessel should take actions to avoid the collision situations.
method consists of an addition time axis. However
NA : Not Applicable
analysis of 2D method in real-time is proposed in
this study to capture the time axis effects. rules and regulations are presented in the last
The summarized collision risk assessments column.
including the fuzzy linguistic variables are presented With respect to the COLREGs rules and
in the Table 1. The first column represents the regulations the vessel coming from the starboard
collision regions with respect to the own vessel and side has the priority for the navigation. Hence, as
the second column represents the divisions of the noted from the table, the collision avoidance
target vessel orientations. Furthermore the third appropriate actions from own vessel have been
column represents the risk assessment with respect limited for the regions of I, II, III and IV. As
to each collision region and that have been divided presented in Table 1, δψ > 0 represents the decision
into three sections: Low risk (Low), Medium Risk to increase of own vessel course to the starboard
(Mid.) and High risk (High). The respective side, and δVo < 0 and δVo << 0 represent decrease of
COLREGs rules and regulations are presented in the own vessel different speed levels respectively.
fourth column. Target vessel relative speed 4.5 Decisions on Navigation
conditions are presented in the fifth column. Finally
the decisions that need to be taken to avoid the The decisions of collision avoidance in ocean
collision situations with respect to the COLREGs navigation should be based on several factors: Speed
and Course of each vessel, distance between two
vessels, Distance of the Closest Point of Approach during this study: Collision Regions µA (Figure 3),
(DCPA) RDCPA (see Figure 1), time to DCPA, Relative Speed µVa,o (Figure 4) and Relative Course
neighboring vessels and other environmental µψa,o (Figure 5). Similarly two fuzzy output
conditions. The initial decisions of collision membership functions have been considered: Speed
avoidance can be categorized into three stages for Change µδVo (Figure 6) and Course Change µδψo
each vessel. (Figure 7).
− When both vessels are at non collision risk range,
both vessels have the options to take appropriate
actions to avoid collision situation.
− When both vessels are at collision risk range,
'give way' vessel should be taken appropriate
actions to achieve safe passing distance
accordance with the COLREGs rules and Figure 3. Collision Regions Fuzzy Membership Function.
regulations and 'stand on' vessel should keep the
course and speed.
− When both vessels are at critical collision risk
range, 'give way' vessel should not be taken
appropriate actions to achieve safe passing
distance accordance with the COLREGs rules,
then 'stand on' vessel should be taken appropriate
actions to avoid the collision situation. Figure 4. Rel. Speed Fuzzy Membership Function.
In this study it is assumed that both vessel take
appropriate actions to avoid the collision situations
to with the respect to the COLREGs rules and
regulations.
4.6 Implementation of Decisions on Navigation
Figure 5. Rel. Course Fuzzy Membership Function.
As the final step, it is assumed that the decisions on
vessel navigation will be formulated with respect to
the collision risk assessments. The actions were
taken by own vessel are proportional to target vessel
behavior as well as the COLREGs rules and
regulations. The expected own vessel and target
actions of collision avoidance situations could be
divided into two categories. Figure 6. Speed change Fuzzy Membership Function.
− Own vessel course change and/or speed change.
− Target vessel course change and/or speed change

5 FUZZIFICATION & DEFUZZIFICATION


The Fuzzy-logic based DM system, which is
formulated for human type thinking, has created a
human friendly environment to facilitate the Figure 7. Course Angle change Fuzzy Membership Function.
collision avoidance capabilities for marine vessels.
Furthermore, a proper formulated fuzzy logic based The Core of the fuzzy set A is defined as the set
DM system has an advantage of simulating the of all elements of the universe that are typical to A
actions of an experienced helmsman in ocean that are associated with the membership value of 1,
navigation. that could be written as
Core ( A) = {x ∈ X | A( x ) = 1} (2)
5.1 Fuzzy sets and membership functions
where x is a generalized variable. The Support of the
Fussy sets are described by the membership fuzzy set is defined as the set of all elements of X
functions that are mappings from one given universe that have non-zero membership degree in A, that
of discourse to a unit interval. It is conceptually and could be written as
formally different from the fundamental concept of Supp ( A) = {x ∈ X | A( x ) ≥ 1} (3)
the probability (Pedrycz et. al 2007). Three fuzzy
The Core and Supp values for the three input and
input membership functions have been considered
two output fuzzy membership functions was
formulated using the given variables in the 6 COMPUTATIONAL SIMULATIONS
respective Figures 3 to 7. Mamdani type IF
The implementation of the DM system has been
<Antecedent > THEN <Consequent> rules based
formulated by using the software MATLAB.
system has been considered and inference via
MATLAB supports for the fuzzy logic schemes of
Min/Max norm has been proposed. Finally the
Mamdani and Sugeno Types (Sivanandam et. al
defuzzification has been calculated by the center of
2007). However this work has been implemented on
gravity method.
the GUI for fitting Mamdani based Fuzzy Inference
5.2 Fuzzy inference system System (FIS).
The Mamdani type fuzzy logic inference
The block diagram for fuzzy inference system with
system consists of utilizing membership functions
integration of the instruments as well as the data
for both inputs and outputs. As previously
acquisition system is presented in Figure 8.
mentioned If-Then rules are formed by applying
fuzzy operations into these membership functions
for given inputs: Collision regions, Relative Speed,
and Relative Course (see Figures 3 to 5).
The resulting output membership functions,
vessel speed change and course change (see Figures
6 & 7) have been calculated by the Fuzzy min-max
configurations. The MATLAB simulation results
with respect to the above formulation have been
presented in Figures from 9 to 15. These figures
contain the start and end positions of the both
vessels with respect to the equal time intervals. In
addition navigational trajectories for both vessels are
also presented in the same figures.
The simulations of two vessels crossing situation
with respect to different speeds and course
conditions in the Cartesian coordinate space have
been presented in Figures 9, 10 and 11. As presented
in Figure 9, for the speed condition of Vo ≈ Va, the
DM system made the proper decision and turned the
own vessel into the starboard side. However the
speed conditions of Vo >> Va and Vo << Va , in Figures
10 and 11, the DM system did not make any
decisions as formulated in the design phase. This is
one of the advantages in the fuzzy logic based DM
system that has not been addressed by recent
Figure 8. Rel. Block diagram for fuzzy inference system
research on collision avoidance.
The avoidance of collisions with respect to
The initial step of the fuzzy inference system
overtake and heading situations are presented in
consists of collection of the target vessel position
Figures 12 and 13 respectively. Further angular
and velocity data. The next step is the calculations of
heading and angular overtake situations are
the relative positions, speed and course which have
presented in Figures 14 and 15. The majority of the
previously been formulated. Then, the data is
collision situations have been avoided by changing
fuzzified by membership functions in the Figures 3
the course angle of the own vessel as preferred by
to 5. The fuzzy rules were developed respecting the
the navigators.
COLREGs rules and regulations in Table 1. The
In an angular overtake situation (in Figure 15),
output of the rules based system could be
the initial speeds of both vessels are equal and the
categorized as the collision risk assessment and the
final speed of the own vessel has been reduced by
fuzzy decisions. Finally the fuzzy decision will be
the DM system to avoid the collision situation. This
defuzzified (see Figure 6 and 7) to obtain the control
is one of the special situations where the speed
actions and that will be executed on own vessel
change has been considered and the course changes
navigation.
could not be implemented to avoid collisions.
NM NM

NM NM

Figure 9. Crossing situation with High Collision Risk Figure 12. Overtake situation with High Collision Risk

NM NM

NM NM

Figure 10. Crossing situation with No Collision Risk Figure 13. Heading situation with High Collision Risk

NM
NM

NM NM

Figure 11. Crossing situation with No Collision Risk Figure 14. Ang. heading situation with High Collision Risk
Fossen, T. I., ed., 1999. “Recent development in Ship Control
Systems Design”. World Super Yachat Review. Sterling
Publication Limited, London.
Hasegawa K., 2009, “Advanced marine traffic automation and
management system for congested water and coastal
areas”, Proceedings of International Conference in Ocean
Engineering, ICOE 2009, February, Chennai, India.
NM IMO, 1972. “Convention on the international regulations for
preventing collisions at sea (COLREGs) ”.
http://www.imo.org.
Imazu, H., 2006. “Advanced topics for marine technology and
logistics”. Lecture Notes on Ship collision and integrated
information system.
Kwik, K. H., 1989. “Calculations of ship collision avoidance
manoeuvres : A simplified approach”. Ocean Engineering,
16(5/6), pp. 475-491.
Lisowski, J., Rak, A., and Czechowicz, W., 2000. “Neural
NM
network classifiers for ship domain assessments”.
Figure 15. Ang. overtake situation with High Collision Risk Mathematics and Computers in Simulations, 51, 399-406.
Moreira, L. and Guedes Soares, C. 2003, Dynamic Model of
Manoeuvrability using Recursive Neural Networks, Ocean
Engineering, 30, pp. 1669-1697.
7 CONCLUSION Moreira, L.; Fossen, T. I., and Guedes Soares, C. 2007, Path
Following Control System for a Tanker Ship Model, Ocean
This paper introduces a new DM system for ocean Engineering. 34, pp. 2074-2085.
navigation based on the fuzzy logic and human Pedrycz W. and Gomide F., 2007, “Fuzzy Systems
expert knowledge. Although successful Engineering, Toward Human-Centric Computing”. John
computational results were obtained, it is assumed Wiley & Sons, Inc., Hoboken, New Jersey.
Pietrzykowski, Z., and Uriasz, J., 2006. “Ship domain in
that more complex collision situations can possibly navigational situation assessment in an open sea area”. 5th
occur and unexpected actions of target vessels could International Euro Conference on Computer Applications
be experienced; hence higher capabilities should be and Information Technology in the Maritime Industries.
formulated into the DM system to overcome similar Rothblum, A. M., Wheal, D., Withington, S., Shappell, S. A.,
Wiegmann, D. A., Boehm,W., and Chaderjian, M., 2002.
situations. As mentioned before, the own vessel “Key to successful incident inquiry”. In 2nd International
collision avoidance actions have been limited for the Workshop on Human Factors in Offshore Operations,
Regions of I, II, III and IV to eliminate the HFW2002.
navigational conflicts. However critical collision Sato, Y., and Ishii, H., 1998. “Study of a collision avoidance
system for ships”. Control Engineering Practice, 6, pp.
situations with respect to the other regions should 1141-1149.
considered and further study should be formulated. Sivanandam S.N., Sumathi S. and Deepa S.N., 2007,
“Introduction to Fuzzy Logic using MATLAB”. Springer
8 ACKNOWLEDGEMENT Berlin Heidelberg, New York.
Smeaton, G. P., and Coenen, F. P., 1987. “Developing an
The research of the first author has been supported intelligent marine navigation system”. Computing &
by a Research Fellowship of the Portuguese Control Engineering Journal, March, pp. 95–103.
Smierzchalski, R., and Michalewicz, Z., 2000. “Modeling of
Foundation for Science and Technology (Fundação ship trajectory in collision situations by an evolutionary
para a Ciência e a Tecnologia) under contract algorithm”. IEEE Transactions on Evolutionary
SFRH/BD/46270/2008. This work is done within the Computation, 4(3), pp. 227-241.
project of “Methodology for ships manoeuvrability Sutulo, S.; Moreira, L., and Guedes Soares, C. 2002,
Mathematical Models for Ship Path Prediction in
tests with self-propelled models”, which is being Manoeuvring Simulation Systems, Ocean Engineering,
funded by the Portuguese Foundation for Science 29(1), pp. 1-19.
and Technology, under contract Yavin, Y., Frangos, C., Zilman, G., and Miloh, T., 1995.
PTDC/TRA/74332/2006. “Computation of feasible command strategies for the
navigation of a ship in a narrow zigzag channel”.
Computers Math. Applic., 30(10), pp. 79-101.
REFERENCES
Antão, P. and Guedes Soares, C. 2008; Causal Factors in
Accidents of High Speed Craft and Conventional Ocean
Going Vessels. Reliability Engineering and System Safety,
93:1292-1304.
Berthold M. R., 2005, “Tutorial: Fuzzy Logic”. Lecture notes
on Advanced course on knowledge discovery, June,
Ljubljana.
Chauvin, C., and Lardjane, S., 2008. “Decision making and
strategies in an interaction situation: Collision avoidance
at sea”. Transportation Research , January , pp. 259-262.
eNAV. 2008, enavigation, http://www.enavigation.org.

Potrebbero piacerti anche