347
1 INTRODUCTION
Most ship collisions derive from human error and
have moderate to severe consequences on both
environment and human lives. The International
MaritimeOrganization(IMO)stilldefinesshippingas
a highly dangerous industry, as the number of
accidents is still relatively high compared to other
industries (IMO, 2003). According
to Allianz Marine
Insurance,humanerrorisstilloneofthemajorcauses
of marine accidents and may vary from 80 to 96 %
(AllianzMarineInsurance,2023).
According to the collecting and studying data
regarding navigation accidents and human actions
accordingly, recent investigations showed us that
misunderstanding of COLREGs
is more severe than
thought(DemirelandBayer,2015).CurrentICTand
sensor technologies can significantly reduce the
misunderstandingandthenumberofcollisionsatsea.
In particular decision support system based on the
data coming from the onboard sensors is oneof the
best technology, suitable for several kinds of
ships
independently from the year of construction
(Lazarowska,2017).
Adecisionsupportsystemforshipnavigationisa
computer‐basedsystemthataidsashipʹsnavigatorin
makingdecisionsaboutthevesselʹscourseandspeed.
It uses information from various sources, including
electroniccharts, satellitepositioningsystems,radar,
and other sensors, to generate recommendations or
A COLREGs-Compliant Decision Support Tool to Prevent
Collisions at Sea
M.Martelli
1
,S.Žuškin
2
,R.Zaccone
1
&I.Rudan
2
1
UniversityofGenova,Genova,Italy
2
UniversityofRijeka,Rijeka,Croatia
ABSTRACT:Groundingsandcollisionsstillrepresentthehighestpercentageofmarineaccidentsdespitethe
currentattentiononMaritimeEducationandTrainingandtheimprovementofsensorcapability.Mostofthe
time, a collision is caused by a human error with consequences ranging from moderate to
severe, with a
substantial impact on both environment and life safeguarded at sea. In this paper, a brief statistical data
regarding human element as a root cause of marine incidents together with collision regulations
misunderstandingispresentedasabackgroundchapter.
Furthermore, the present work discusses a decision support system architecture
to suggest an appropriate
actionwhentheriskofapotentialcollisionisdetected.Theproposedarchitecturesystemisbasedonvarious
modulesintegratedwithpropersensorinputdataregardingthesurroundingnavigationarea.
As a result, the tool can support the Officers of Watch in the decision‐making process
providing an early
suggestion in compliance with the COLlision REGulations. The proposed system is intended to be used
onboardindependentlyfromthedegreeofautomationoftheship,anditisbasedonAIS,whichismandatory,
makingitwidelyapplicable.Theproperuseofthesystemcanconsiderablyreducethe
numberofcollisions,as
demonstratedbytheobtainedresults.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 17
Number 2
June 2023
DOI:10.12716/1001.17.02.11
348
alerts for the navigator. Such a system aims to
improve the safety and efficiency of shipnavigation
byprovidingthenavigatorwithtimely andaccurate
information that can help them make informed
decisions (Pietrzykowski et al., 2017). Some decision
supportsystemsforshipnavigationalsoincluderoute
planningandrisk
assessmenttoolsformingacollision
avoidancesystem (COLAV).The COLAV is a safety
system designed to help prevent collisions between
two or more ships and rely on sensors to detect the
presence and location of other objects and then use
thisinformationtocalculatetheriskofacollisionand
theevasivemanoeuvre(Zaccone&Martelli,2018).If
theriskisdeemedtobehigh,thesystemmayissuean
alerttotheoperatorortakeautomatedactiontoavoid
thecollision.SomeexamplesofCOLAVsusedinthe
maritime industry include radar‐based systems that
candetectthepresenceof
othervessels(Wilthiletal.,
2018) and automatic identification systems (AIS)‐
basedsystemsthatcanexchangeinformationabouta
vesselʹs position and course with other ships in the
same area (Jincan & Maoyan, 2015). CASs are an
essentialsafetyfeatureonmanyvehiclesandcanhelp
reducetherisk
ofaccidentsandfatalities.
A COLREGs‐compliant decision support tool for
preventing collisionsat sea, as a consequence of the
maritime advanced technology development in
navigation and the above‐mentioned motivations, is
indeed proposed by the authors for this transitional
period where interaction between human
interpretation and the help of a
decision support
system tool is inevitable and valuable. In particular,
the paper structure is the following: in Section 2
statics on the collision at sea are reported to raise
awarenessofthisspecifictypeofaccident.Insection3
the general architecture of the proposed system is
shownwitha
detaileddescriptionofeachsub‐module
andsomeexamples.IntheConclusion,Section4,the
advantagesandthefuturedevelopmentofthesystem
aredrawn.
2 STATISTICSONCOLLISIONS
According to the Annual Overview of Marine
CasualtiesandIncidents(EMSA,2022),thepercentage
of reported accident events from 2014 to 2022
was
obtainedbycountingeachoccurrenceintheEuropean
MaritimeCasualtyInformationPlatform(EMCIP)for
everyeventtype(Figure1).
Figure1.Percentageofaccidenteventsintheperiod2014‐
2021organizedbyaccidenteventtype(EMSA2022).
During this period, human action accounted for
the highest percentage of accident events at 59.6 %,
followedbySystem/equipmentfailureat24.5%,Other
agentsorvesselsat8.6%,Hazardousmaterialat5.3%,
andUnknownat2.0%.Similartrendswereobserved
across all ship types with minor variations in the
percentages;however,
theHumanfactorremainedthe
mostcommonaccidenteventtypeforallshiptypes.
According to the Safety analysis of EMCIP data
regardingnavigationalaccidents,thestudyexamined
1,637 elements that contributed to navigational
accidents, which were reported in 351 safety
investigationreports(Figure2).
Figure2.Percentageofaccidenteventsfortheperiod2014‐
2021organizedbyaccidenteventtypeAccordingtoSafety
analysisofEMCIPdata‐navigationalaccidents,2022.
These factors were classified into nine safety
issues. According to the findings, most oftheissues
raised were related to work/operation methods,
organizational factors, and risk assessment,
accounting for nearly two‐thirds (66%) of the
contributingfactors.Themaintakeoutoftheanalysis
is that the top three safety issues, Work/operation
methods and Organisational factors, are directly
linked to COLREG rules which make up 55% of all
contributingfactors.
Table1.Humanactiongeneralconditionsand
interpretation.(AccordingtoSafetyanalysisofEMCIPdata
‐navigationalaccidents,2022)
________________________________________________
InterpretationNr. %
________________________________________________
Delayedinterpretation70 27.5%
Localdiagnosis‐Wrongdiagnosis45 17.6%
Decisionerror‐Wrongdecision41 16.1%
Localdiagnosis‐IncompleteDiagnosis 28 11.0%
Localprediction‐Unexpectedstatechange 16 6.3%
Localprediction‐Processspeedmisjudged 12 4.7%
Wrongreasoning‐Wrongpriorities 11 4.3%
Wrongreasoning‐Deductionerror 9 3.5%
Decisionerror‐Decision
paralysis8 3.1%
Wrongreasoning‐Inductionerror 5 2.0%
Decisionerror‐Partialdecision4 1.6%
Localdiagnosis‐Other2 0.8%
Localprediction‐Unexpectedside 2 0.8%
Decisionerror‐Other1 0.4%
localprediction‐Other1 0.4%
________________________________________________
Total255 100.0%
________________________________________________
Accordingtothecollectingandstudyingdatafrom
the abovementioned study, the action analysis
includes255 casesbasedon various types ofhuman
action interpretation errors. More than one‐fourth
(27,5%) of interpretation issues concern delays in
interpretation,followed bywrong diagnosis(17,6%),
349
andwrongdecision(16,1%)thattogethermakesmore
than60%ofallreportedcausesofcognitivefunctions.
Thiscanbeexemplified by theOfficerOftheWatch
(OOW)monitoringthevesselʹsprogress,considering
the interpretation of data displayed from navigation
instruments,alarmstriggeredbyECDIS,ortheshipʹs
motionduringamanoeuvre.
According to EMSA analysis, out of significant
occurrences of 8,800 navigation accidents, the
following 370 cases have undergone safety
investigationswithareporteddatasetsuitableforthe
furtherrequiredanalysis.
Figure3. Navigation accidents are dealt with by safety
investigations.AccordingtoSafetyanalysisofEMCIPdata‐
navigationalaccidents,2022.
Among these, collisions comprised the highest
proportion (almost 44%), followed by groundings
(38%) and contacts (18%). The analysis aims to
contextualize navigation accidents by reviewing
informationfromallrelevantoccurrences.
Inadditiontothenavigationaccidentandhuman
action data collected from the abovementioned
studies, the MAIB investigation showed us that
the
misunderstanding of the COLREGs is more severe
thanthought.AfteranalysingtheMAIB databaseon
collision investigation, it is evident that the highest
number of cases had violated the following
COLREGs’ rules: Rule 5 (Look‐out), Rule 7 (Risk of
Collision), and Rule 8 (Action to avoid collision).
Regarding
Rule 5, it is evident from the MAIB’s
incident report that well‐known, established
international maritime standards are sometimes
neglected.Regardingtheriskofcollisionandactionto
avoid it, the complex COLREG misunderstanding is
arising. Some recognized hazards are based on
improper or delayed Rules interpretation,
misunderstanding of vessel manoeuvring
characteristics, acomplacency of self‐confidence and
overreliance of the navigator, and reluctance to use
shippropulsionincollisionsituations.Currentstateof
arttechnologiescanhelpinreducingnotablehuman
error, for example with Decision Support System
(DSS).
3 DECISIONSUPPORTSYSTEMARCHITECTURE
The general architecture of the DSS
proposed in the
current paper is shown in Figure 4. The system can
receive, as input, the data coming from several
sensors.Havebeenhypothesizedthatthedatacoming
from Automatic Identification System (AIS) and
GlobalPositioningSystems(GPS)willbethestarting
pointsincethesesystemsareinstalledonevery
ship.
From AIS, the data of the vessels navigating in the
surrounding are provided; among the several
messages, the most relevant are targets’ position,
attitude, course over ground (COG), speed over
ground (SOG), and navigation status. Moreover, the
GPS provides the own ship’s current position while
the gyro provides its
orientation. Furthermore, it is
necessary to keep in mind that target acquiring is a
majorfocusforCOLREGRuledeterminationbyusing
water‐stabilized RADARs, which must be used for
DSS systems. Also, the RADAR can be used as
double‐check for consistency. Additionally, ECDIS
with an appropriate Electronic Navigational Chart
(ENC) is needed as available layer information,
especially for the position of safe waters, fixed
obstructions, temporary notices, Traffic Separation
Scheme (TSS), narrow channels, and other relevant
informationneededforsafenavigation.
Data from sensors are the input of the detection
module that constantly checks for collision risk; if a
collision risk is detected, the module triggers the
COLREG Classification Module that identifies the
COLREG scenario and the rules involved. With this
information,theRouteselectionmodulecansuggesta
course change to avoid the collision.Eventually,the
finalcheckoncompliancewithCOLREGisdone.The
specificdetailsonthe
functionalitiesof eachmodule
areprovidedinthefollowingparagraphs.
Eventually, a Graphical User Interface (GUI),
locatedinthebridgeorany relevantspaceonboard,
shows the outcome of the system. The Officer Of
Watch (OOW) can indeed follow the suggestion
providedbytheDSSordecidebythemselfand
then
actontheHumanMachineinterface(HMI)togivethe
commandontherudderhelmorsettheautopilot.In
the case of an autonomous ship, this action can be
completelyautomatic.
Figure4.Proposeddecisionsupportsystemarchitecture.
3.1 Detectionmodule
Thedetectionmodule runswithasamplerategiven
by the slower sensor and runs continuously in the
background.Itisresponsiblefor providinganalarm
incasetheriskofcollisionisdetected.
350
Figure5.Referenceframes.
Considering the scenario reported in Figure 5,
consider the following variables in the earth‐fixed
referenceframe
12
,, nn
:theshipinitialpositionS,
00102
SS S
txtnytn; and the ship current
velocity,
12
SS S
vxnyn.
Similarly, the kinematic variables of the target,
initiallydetectedinH,areassumedtobeknown:
00102
12
HH H
HH H
txtnytn
vxnyn
(1)
The motion laws of both the ship and target are
then known, assuming that the velocity and
orientationofthetwovesselsdonotvaryintime:
0
0
SS S
H
HH
ttvt
ttvt
(2)
ThePathInterceptionPoint(PIP)isthegeometric
intersection of the two trajectories and can be
expressed as
12
PIP PIP PIP
x
nyn
. It can be
addressedbysolvingthefollowingsystem,expressed
inthematrixialform:
SS SSSS
H
HHHHP
yx yxxy
x
yx yxxy
y
(3)
The system admits only one solution if the
followingcondition(trajectoryincidencecondition)is
ensured:
det 0
SS
HH
yx
yx
(4)
The time needed to reach the PIP is denoted as
TPIP.AcollisionoccursifbothshipsreachthePIPat
the same time, or rather, atʺexcessivelyʺ close
instants. As a matter of safety, considering actual
operatingconditions,shipsshouldalwayskeepasafe
distance.
Therefore,oneapproach
tocollisiondetectionisto
require that the difference between ship and target
TPIP is greater than a threshold time Θ. This
parameter can, for example, be calculated as a
function of safety distance by the following
relationship:
min ,
safety
SH
D
vv
(5)
Thiscorrespondstoverifyingthatatthetimethe
shipengagesthePIP,thetargetisatadistancegreater
thanthesafedistance.However,thereisnoguarantee
that the target will not violate the safety distance
before or after that time. For such a reason is
necessary
tointroducetheexpressionofthedistance
inthe timedomain betweenthe shipand thetarget,
D(t), that should be lower than the threshold to
provideanalarm:
S H safety
Dt t t D (6)
Theminimumdistanceovertime(ClosestPointof
Approach, CPA) is so defined, using the previous
expression,as:
00
00
2
min
SH S H
SH SH
t
SH
vv t t
Dt t t v v
vv
(7)
Theevaluatedkinematicvariablesarereportedin
the GUI with an alert to the OOW, as reported in
Figure6.
Figure6. Example of the information coming from the
detectionmodule.
An exampleof the result is reported In Figure6.
The yellow circle represents the time when the
violatingthesafetydistance,D
safety.Thereafter,dueto
relative motions, the vessels will continue to move
closer until they reach the CPA, identifying the
minimumencounterdistancewithareddashedline.
Thenumericalinformationisreportedinthepop‐up
windows.
3.2 COLREGClassificationmodule
As some of the COLREG Rules are applied
consecutively,
it is necessary to find a way how to
determinewhichruleisapplicableatcertaincollision
situations for the DSS proposed. According to
(Sunmer, 2021) an individual approach to each
COLREG Rules assessment is necessary for this
approach.Thekeyinfluencingfactoristheattitudeof
atargetconcerning
ownvessel.Itisessentialtokeep
in mind that there are performance limits with
351
sensory equipment on commercial vessels. For
collision avoidance, it is only possible to ensure
complete autonomy with audio and visual sensory
equipment that can replace the human navigatorʹs
sightandhearing.Consequently,itisstillnecessaryto
involve human navigators in the Decision Support
System(DSS)process.
The key
element in theartificially intelligent and
automatedcollisionavoidancesystemistheCOLREG
ClassificationModule. Thisinitialandcriticalstep is
thecornerstoneofthedecisionprocessonwhichthe
collision avoidance architecture is based. The initial
step of the COLREG Classification Module
development relies on COLREG Rules description
and
elaboration together with the legal framework,
empirical studies, case laws, and scientific survey
analyses.Thispartisparticularlysensitiveduetopast
research that shows numerous accidents and
COLREG misunderstandings. Nowadays, basic
communication between vessels is established by
using VHF which is not recommended according to
the COLREG rules. Furthermore, electronic
direct
communication between other vessels will be
developed in the near future where two ECDIS
systems on board vessels, as primary means of
navigation, will exchange relevant data regarding
navigation safety. Also, a significant number of
commercial ships that sail with navigators not
knowingornotunderstandingtheCOLREGsis
stilla
challenge that requires a global solution. After the
initial step, the COLREG Classification Module
developmentneedsto be setwithoutambiguityand
strongandconcretesafetyparametersthatcanlaterbe
encodedintoCOLREGverificationAlgorithmsforthe
Module.
Considering the previous scenario, Rule 15
(Crossing situation), according to
International
RegulationsforPreventingCollisionsatSea(2020)for
the COLREG Classification Module development, is
elaborated. The crossing Rule states that when two
power‐driven vessels are crossing, and there is a
collisionrisk,theshipwiththeothershiponherown
starboard side shall keep out of the
way and, if
possible,shouldavoidcrossingthebow ofthe other
vessel. Regulators did not restrict manoeuvring to
starboard only. Still, if thinking about crossing from
thestarboardside, avesselshouldaltertostarboard
and pass astern of the crossing vessel if the
circumstancesofthecasepermit.
Rule
15 assigns give‐way and stand‐on
responsibilities among two crossing power‐driven
vessels,butonlywhenthereisariskofcollision.Even
though the own threshold for collision risk might
differ from the target vessel, the own vessel can act
conservatively and risk‐averse, which is always a
good approach to avoid close‐quarter situations. If
thinking about other COLREGs geometries, it is
noticeable that crossing includes any situation not
classified as head‐on or overtaking.Therefore, when
the relative bearing of the target vessel is in the
spectrumof[6°,112.5°]and [247.5°,354°], andifthe
trajectory
isbringingthetargetvesseltotheminimum
CPAradius,thenthenavigatorconfirmsthatthereis
a risk of collision and then decides on the crossing
action. Furthermore, in several Admiralty cases, the
notion that a crossing give‐way vessel should not
crossaheadofthestand‐onvessel
hasbeenconfirmed
(Benjamin et al., 2006); therefore, it is necessary to
ensurethatthecrossingthesternisoptimalbehaviour
when verifying generated trajectories. The following
figuredepictsthecrossingwhererule15isadopted.
Figure7. Rule 15‐Crossing situation. (Courtesy of:
www.ecolregs.com).
InaccordancewithRule15(Crossingsituation),if
the circumstances of the case admit, vessel A shall
avoidcrossingaheadofvesselB.
Furthermore, the COLREGs classification
algorithm for the COLREG Classification Module
focuses on the own vessel’sand target’s attitude, so
course and speed through water are significant to
determine
which Rule will be appropriate for each
collisionsituation.Asstatedbefore,itisnecessaryto
keep in mind that the classification algorithm is
developedincludingtheinteractionwithhuman.On
thecontrary,itshouldbemodifiedforfuturepotential
autonomous navigation applications. For setting the
mathematical formulation for the
COLREG
Classification Algorithm, the following variables are
used:H
o–headingoftheownvesseltakenfromthe
gyro compass, COG – Course Over Ground for the
own vessel taken from the GPS or radar, CTW –
CourseThroughWatertakenfrom the radar, SOG –
Speed Over Ground taken from the GPS or radar,
RPM–Revolutions Per
Minutetakenfromtheengine
speed indicator directly, performance measurement
monitoring, or conning display, STW – Speed
ThroughWatertakenfromthespeedlogorradar,n
OV
–GNSSnorthpositionoftheownvessel,e
OV–GNSS
eastpositionoftheownvessel,h–draughtoftheown
vesseltakenfromtheloadicatorcomputerormanual
inputtoverifysafewaters,D–depthofwater,ECDIS
info – various ECDIS available layer information,
especiallypositionofsafewaters,fixed obstructions,
temporary notices, TSS,
narrow channel, and other
relevantinformationneededforsafenavigation.
When tracking a new target T
n, the following
information is of interest: H
T – heading of a target
takenfromtheradar,COG
T–CourseOverGroundfor
atargettakenfromtheradar,CTW
T–CourseThrough
Water of a target taken from radar, SOG
T – Speed
Over Ground of a target taken from radar, STW
T –
Speed Through Water taken from radar, n
T – GNSS
northpositionofatarget,e
T–GNSSeastpositionofa
target,AIS
T–variousAutomaticIdentificationSystem
information of a target taken from the AIS receiver,
dCPA
T – distance to Closest Point of Approach
(usuallycalledsimplyaCPA)ofatargetinrelationto
the own vessel taken from the Automatic Radar
PlottingAid(ARPA),TCPA
T–TimetotheCPAofa
targetinrelationtotheownvesseltakenfromARPA,
352
R
T– Rangeofatargettakenfromradar,θT–bearing
ofatargettakenfromradar,θ
OV–bearingoftheown
vessel from the perspective of a target (calculated
afteracquiringnewtarget),andBCR–BowCrossing
RangetakenfromARPA,whichcanbepositive(bow
crossing),ornegative(sternpassing).
COLREGClassificationalgorithmfortheCOLREG
ClassificationModuleisseparateandnowrepresents
the
firststepbeforethecollisionavoidancealgorithm
withRouteselectionforDSSsysteminthenextphase.
The main function is utilizing vessels’ water
geometriesand determining theappropriateattitude
foraccurateCOLREGsRuledetermination.
Considering the scenario reported, COLREG
ClassificationAlgorithmforRule15isproposed:
Input:Ho,COG,CTW,SOG,RPM,STW,nOV,eOV,h,D,
ECDISinfo,T
1,2,…,n(HT,COGT,CTWT,SOGT,STWT,nT,eT,
AIS
T,dCPAT,TCPAT,RT,
T,BCR).
Output:DisplayrelevantCOLREGsRules
Every10secondsdo:
Rule15
verifyinformationextractedfromwaterstabilized
RADAR
foreachT
n:
readT
n(HT,CTWT,STWT,nT,eT,dCPAT,TCPAT,RT,
T,BCR)
if247.5°<
Tn<354°,RT<6NM,BCR>0,and
CPACPA
PREF:
display:RULE15–T
nCROSSINGBOWFROM
PORT–STAND‐ON
endif
if247.5°<
Tn<354°,RT<6NM,BCR0,and
CPACPA
PREF:
display:RULE15–T
nCROSSINGSTERN
FROMPORT–STAND‐ON
endif
if006°<
Tn<112.5°,RT<6NM,BCR>0,and
CPACPA
PREF:
display:RULE15–T
nCROSSINGBOWFROM
STARBOARD–GIVEWAY
endif
if006°<
Tn<112.5°,RT<6NM,BCR0,and
CPACPA
PREF:
display:RULE15–T
nCROSSINGSTERN
FROMSTARBOARD–GIVE‐WAY
endif
endfor
end
TheproposedCOLREGClassificationModulewill
give directly four possible classifications related to
Rule15,takingintoconsiderationthevessel’ssensor
parametersoftheownvesselandtargetvessels.These
four possibilities in the proposed Module represent
significant information to determine give‐way or
stand‐on vessel and to determine
bow or stern
crossing situation. Once the appropriate Rules have
been classified, the collision avoidance algorithm
exploits Rules classification as constraints and/or
input parameters, which are then used for system
designanddecisionsinthenextphase.
3.3 Routeselection
The route selection module will be enabled if the
detection module
detects that the safety distance is
exceeded and suggests an evasive manoeuvre. The
algorithmassumesa minimumcoursechange,setin
advance, towards the side established by the
COLREGclassificationmoduleandre‐checkswhether
thesafetyconditionisensured.Ifthesafetycondition
is achieved, the last course value
is recommended;
otherwise, a new course angle is assumed, iterating
until the achievement of a safe solution. When the
safetyconditionisassessedalltheinformationwillbe
displayed on the GUI both in graphical and textual
form,anexampleisreportedinFigure8.
Figure8.Exampleofsuggestionofanewcourseincrossing
scenario.
4 CONCLUSIONS
Groundings and collisions still represent the highest
percentage of marine accidents caused by human
error, with consequences on the environment and
human life. The analysed statistical data regarding
human element as a root cause of marine incidents
and collision regulations misunderstanding show us
thatmisunderstandingCOLREGsismoresevere
than
thought.
The proposed COLREGs‐compliant decision
support tool for the end‐users (OOW and Masters)
providesan earlysuggestionincompliance with the
COLlision REGulations in raising navigation safety.
As suggested, the critical element in an automated
collisionavoidancesystemtechnologyistheCOLREG
ClassificationModu le.Therefore,futurework
willbe
basedonallCOLREGs‐individualRules’assessment
together with COLREGs implementation and
compliance by the legal framework for establishing
COLREG Classification Module. Also, as a result of
future steps regarding individual COLREG Rule
assessment,safetyparametersneedtobeestablished
and elaborated as input parameters for the DSS
system.
Furthermore,adetectionandtrackingmodulefor
a DSS system can also rely on data coming from
sensors not fully exploited for navigation purposes,
suchasLiDARsandcameras.Thisdata,especiallyif
elaborated by a data fusion algorithm, could help
classify the objects when data coming from AIS are
unavailable(e.g.,yacht,AIS switchedoff,etc.)orfor
floating obstacles. Moreover, in the perspective of a
completely autonomous vessel, the proposed
approach can be further developed by providing a
complete evasive manoeuvre that can be directly
actuated by the track keeping and motion control
system,withhumansthatonlysupervise
theprocess;
thiswillbetheobjectoffurtherstudies.
353
Nevertheless, the properly established and
proposed DSS system can considerably reduce the
numberofcollisionsinraisingnavigationsafetyand
environmentalprotection.
ACKNOWLEDGEMENTS
This research was partially funded by European Union’s
HorizonEuropeunder thecallHORIZON‐CL5‐2022‐D6‐01
(Safe, Resilient Transport and Smart Mobility services for
passengers and goods), grant number 101077026, project
nameSafeNav.Viewsandopinionsexpressedarehowever
those of the author(s) only and do not necessarily reflect
thoseoftheEuropeanUnionorExecutiveAgency(CINEA).
Neitherthe EuropeanUnionnorthegrantingauthoritycan
beheldresponsibleforthem.
REFERENCES
[1]Allianz Marine Insurance. Archived on 10‐March‐2023.
Retrieved from https://www.agcs.allianz.com/global‐
offices/united‐states.html#marine
[2]Annual Overview of Marine Casulties and Incidents –
European Maritime Safety Agency (EMSA) 2022.
Archived on 25‐March‐2023. Retrieved from
https://www.emsa.europa.eu/accident‐investigation‐
publications/annual‐overview.html
[3]Benjamin, R.M. Multi‐objective autonomous vehicle
navigation in the presence
of cooperative and
adversarialmovingcontacts.InOCEANS’02MTS/IEEE,
volume3,pp.1878‐1885.IEEE,2002.
[4]Benjamin,R.M.,andCurcioJ.Colregs‐basednavigation
ofautonomousmarinevehicles.IEEE/OESautonomous
underwatervehiclespp.32–39,2004.
[5]Benjamin,R.M.,Curcio,J.A.,Leonard,J.J.,andNewman,
P.M. Navigation of unmanned marine
vehicles in
accordance with the rules of the road. In Proceedings
2006 IEEE International Conference on Robotics and
Automation,2006.ICFA2006.,pp.3581‐3587,May2006.
[6]Demirel, E, and Bayer, D. Further studies on the
COLREGs(collisionregulations).Transnav‐International
Journal on Marine Navigation and Safety on Sea
Transportation
(2015).
[7]Ecolreg. Archived on 03‐March‐2023. Retrieved from
https://ecolregs.com/index.php?lang=en
[8]EMSA,2022 Annualoverviewof marinecasualtiesand
incidents2022.
[9]Lazarowska, A.ʺA new deterministic approach in a
decisionsupportsystemforship’strajectoryplanning.ʺ
ExpertSystemswithApplications71(2017):469‐478.
[10]IMO. HumanElement
Vision, Principlesand Goals of
the Organisation. International Maritime Organisation,
ResolutionA.947(23).Adoptedon27November2003.
[11]International Regulations for Preventing Collisions at
Sea(Consolidatededition,2018).Archivedon03‐March‐
2023. Retrieved from
https://www.samgongustofa.is/media/log‐og‐
reglur/COLREG‐Consolidated‐2018.pdf
[12]Jincan,H.,&Maoyan,F.(2015,June).BasedonECDIS
and AIS ship collision avoidance warning system
research. In 2015 8th International Conference on
Intelligent Computation Technology and Automation
(ICICTA)(pp.242‐245).IEEE.
[13]Pietrzykowski,Z.,Wołejsza,P.,&Borkowski,P.(2017).
Decision support in collision situations at sea. The
JournalofNavigation,70(3),447‐464.
[14]Safety
AnalysisofEMCIPData.AnalysisofNavigation
Accidents‐European Maritime Safety Agency (EMSA)
2022. Archived on 25‐March‐2023. Retrieved from
https://www.emsa.europa.eu/publications/item/4830‐
safety‐analysis‐of‐emcip‐data‐analysis‐of‐navigation‐
accidents.html
[15]Sunmer, M. Dynamic Collision Avoidance for Sea
SurfaceVehicleswithaHiddenMarkovModel.doctoral
dissertation. University of Rijeka,
Faculty of Maritime
Studies,Rijeka2021.
[16]Wilthil,E.F.,Flåten,A.L.,Brekke,E.F.,&Breivik,M.
(2018,March).Radar‐basedmaritimecollisionavoidance
using dynamic window. In 2018 IEEE aerospace
conference(pp.1‐9).IEEE.
[17]Zaccone,R.,&Martelli,M.(2018,October).Arandom
sampling based
algorithm for ship path planning with
obstacles. In Proceedings of the International Ship
ControlSystemsSymposium(iSCSS)(Vol.2,p.4).
