267
1 INTRODUCTION
Autonomous and remotely controlled vehicles have
beensuccessfullyimplementedinvariousindustries,
e.g. automotive, subsea and airborne as well as in
military applications. Also the safety issues of those
have been addressed by various authors (Gerigk
2015) (Özgüner et al. 2007) (Stokey et al. 1999).
However,sinceunmannedmerchantvesselsarest
ill
in conceptual and design phase (Burmeister et al.
2014), any elaboration on their safety levels must
inevitably be incomplete by simply not including
historical and empirical data. Operational issues of
unmanned ships are elaborated in (Rødseth et al.
2013),andsummarizedinFigure1.In(Kretschmann
&Rødset
h,etal.2015)(Kretschmann&Mcdowell, et
al. 2015) authors optimistically claim that
autonomousshipproposedbythemwillbeingeneral
safer to operate than conventional vessels operating
nowadays.In(Kretschmann,Mcdowell,etal.2015),it
is said that the results of risk analysis are
‘encouraging’ despite acknowledging that there are
someuncertaintiesregardingfut
uresystems’design.
Authors claim that human errors’ contribution to
maritimeaccidents’occurencewillbe tolargeextent
limited. However, the study does not seem to take
intoaccountthat,shouldtheaccidentoccur, therewill
be nobody on board the unmanned vessel to
undertake immediate corrective act
ion in
unfavourable circumstances and restrict the
consequencesofhypotheticalmaritimedisaster.Thus,
vessel’s survivability of the accident will depend
solelyonherdesigners’abilitytoanticipatepotential
accidents’ scenarios and possibly on the operational
performance of any remotely controlled system
installed on board and operated by qualified crew
ba
sedashore.
Towards the Development of a Risk Model for
Unmanned Vessels Design and Operations
K.Wróbel&P.Krata
GdyniaMaritimeUniversity,Gdynia,Poland
J
.Montewka
GdyniaMaritimeUniversity,Gdynia,Poland
ResearchGrouponMaritimeRiskandSafety,AaltoUniversity,Finland
FinnishGeospatialResearchInstitute,Finland
T.Hinz
WaterborneTransportInnovationFoundation,Poland
ABSTRACT:Anunmannedmerchantvesselseemstobeescapingfromthestageofideaexploration.Oncethe
conceptproofsitssafety,itmaybecomeapartofmaritimereality.Althoughthesafetyaspectofsuchashiphas
beenaddressedbyahandfulofscholars,theproblemremainsopen.Thisismainlyduetolackofknowledge
regardingact
ualoperationalcircumstancesanddesignofunmannedships,whichareyettobedeveloped.In
theattemptofbridging this gap, theriskanalysis associated with unmannedshipsneedsto be carried out,
whereallrelevanthazardsandconsequencesareassessedandquant
ifiedinsystematicmanner.Inthispaper
wepresenttheresultsofafirststepofsuchanalysis,namelythehazardanalysisassociatedwiththeunmanned
ships. The list of hazards covers various aspects of unmanned shipping originating from both design and
operationalphasesofvessel’slife.Subsequentlythehazardsandrelatedconsequencesareorganizedinacasual
manner,result
inginthedevelopmentofastructureofariskmodel.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 10
Number 2
June 2016
DOI:10.12716/1001.10.02.09
268
Figure1.Unmannedshipsoperationalcontextrelationshipdiagram(Rødsethetal.2013)
Meanwhile, in (Rødseth et al. 2013), authors
presentedidentificationofmajorrisksassociatedwith
introductionofunmannedvesselsandthenassigned
them with estimated frequency and severity indices
divided into three categories to indicate that a
particularhazardcanbeofeitherhuman,materialor
environmentalnature(Kretschmann&Rødsethet
al.
2015).However,eachandeveryaccidentiscausedby
a series of casual factors, which can be assigned to
various categories.Moreover, malfunctions tend to
propagateand,forinstance,whatinitiallyseemstobe
harmlessmechanicalbreakdowncaneasilyturn into
fire and, eventually, ship’s sinking. Those issues do
notappeartobeaddressedtodate.
Thus,inourstudywemadeanattempttocombine
thepastexperience,encapsulatedinthevastliterature
onmaritimeaccidentcauses(Allianz 2015)(Mazaheri
etal.2015),withtheinsightintothefuture,aselicited
fromtheexperts,inordertodevelopalist
ofpotential
hazardsthatpertaintothedesignandoperationofa
unmanned ships (Gerigk and Skorupski 2012).
Subsequentlythoseareorganizedinacasualmanner
and linked with their potential consequences,
providingastructureofariskmodel.Thelatter,when
fully developed can be seen as a useful
tool
measuringtheeffectofvarious factorsonthesafetyof
unmannedships.
Due to high uncertainty of the potential risk
model, a suitable modelling techniques shall be
adopted, allowing for the uncertainty assessment,
representation and quantification, if possible, if not
some qualitative evaluation needs to be performed,
(Flage & Aven
2009). Therefore, the structure of the
model presented here is developed with the use of
Bayesian Networks, which seems to be suitable for
the given purpose (Weber et al. 2012). At this stage
the network lacks the parameters (probabilities, in
otherwords), sincethe intentionofthe authorsis to
demonstratethestructureonly.Theparametersshall
beelaboratedinfutureworkstocome.
The remainder of the paper is organized as
follows: firstly, data sources are presented together
withmethodology.Then,Bayesiannetworkstructure
aselaboratedisgivenanditsdetailsdescribed.Brief
discussion and suggestion for further study is
followedbyconclusions.
2 MATERIALSANDMETHODS
2.1 Materials
Inorder toinvestigatefactorsconcerning unmanned
vessel’s safety, we used the available description of
system concepts as presented in (Burmeister et al.
2014)(Rødseth and Burmeister 2015). Therein the
authorssuggest thatfutureshipsshallbe capable of
operatingineither
fullyautonomousmode(in open
sea),remotely‐controlled(e.g.incasetheyencountera
situationthatcontrolsystemcannothandleforsome
reason) or manned by so‐called flying or conning
crews(inapproachestoportssimilarlytopilotservice
nowadays and during mooring operations for
instance). To provide them with
sufficient level of
operational safety, they will be equipped with
redundant propulsion systems, secure
communication links and additional environmental
sensors among other purpose‐designed features
(Burmeisteretal.2014).
Similarapproachistakenbyconsortiummembers
involved in currently pursued projects Advanced
Autonomous Waterborne Applications Initiative
(AAWA)(Jokioinen 2016), and US‐
based Defense
AdvancedResearchProjectsAgency’s(DARPA)Anti‐
Submarine Warfare Continuous Trail Unmanned
Vessel (Rogoway 2015) project. As can be seen,
despitethefactthatunmannedmerchantvesselsare
yet to be introduced to international shipping, their
future anticipated design and performance are
decently described in many sources, both scientific
and
popular (Burmeister et al. 2014)(Rødseth and
Burmeister 2015)(Man et al. 2014). We have used
those to create a model describing their expected
safetyfeatures.
269
2.2 Methods
Havingexperienceinbothacademiaandindustryin
various disciplines relevant to unmanned shipping
(navalarchitecture,safetyscience,marinenavigation,
ship stability etc.) we applied brainstorming as a
method of eliciting and systematizing professional
knowledgeregardingexpectedoutcomeofprocesses.
Brainstorming is widely accepted in numerous
disciplines and
valued inter alia for its simplicity,
tendencytomagnify one’s creativity andsortof fun
that comes with it (Isaksen 1998). We used basic
brainstorming guidelines as given in (Rossiter &
Lilien1994),namely:
1 Createstrictbrainstormingrulesandfollowthem
Brainstormingbestpractices have beenused,just
to
mention free‐wheeling and refraining from
criticism;
2 Setaspecificanddifficulttarget
Ourgoalwastocreateastructureofariskmodel
describing best the anticipated relationships
between safety features of unmanned merchant
vessel;
3 Initial ideas should be created by individuals
ratherthangroups
Ideasonincluding
particularriskanditsrelations
toremainingoneswerepresentedbyindividuals;
4 Use group interaction to amalgamate and refine
ideas
If found appropriate or necessary, those were
clarifiedordevelopedbyentiregroup;
5 Useindividualvotingtoselectfinalideas
Results of brainstorming were put under
democraticvote;
6 Keeptimingasshortaspossible
Wehaveapplied a limit ofonehourfor creating
ideasaboutthenetworkandsomeadditionaltime
forresolvingarisingproblems.
The technique was applied among the group of
experts, comprising of ship designers and naval
architects, master mariner, ship stability expert, risk
analyst, all of them having research and industrial
background. The group is deliberately diversified,
containingsixpeoplefromtwocountriesrepresenting
variousfieldofshipping.
2.3 Results
Asaresultofthebrainstormingsessions,wecreateda
structure of Bayesian network describing
relationships between safety issues pertaining to
unmannedvessels.
Wedebatedoncausesandeffects
of some unfortunate events affecting ships’ safety
without considering their actual probabilities and
consequencesasinaclassicdefinitionofriskbeinga
product of those. Instead, we focused on accidents’
potentialcausesandfailures’developmentwithinthe
system. Nevertheless, importance of estimating
likelihood and
consequences of such events is
acknowledgedandwillbeaddressedinfuturework.
In a course of brai nstorming, it has been noted that
seemingly very different types of accidents tend to
originate from similar root causes. Those can be
attributed to malfunctioned sensors or improper
maintenance regime for instance. For
presentations’
simplicity reasons, those were grouped together as
most of them could potentially lead to numerous
typesofaccidents.Theobtainedstructureoftherisk
modelispresentedinFigure2.Wedividedthesafety
features of unmanned vessels into three levels as
follows:
Level 3 focuses on potential root causes of
accidentsasgiveninLevel2;
Level2describespotentialaccidentstowhichthe
unmannedshipcanbesusceptibleandthewaythe
damage might escalate thus causing secondary
accidents;
Level 1 constitutes a set of unwelcome events,
theirdevelopmentpathsandlikelihoods(thelatter
to
be addressedinfuturework)asconsideredin
hereby paper that put a vessel into a risk of
foundering or otherwise directly endangering
maritime safety including the natural
environment.
2.3.1 Level3
Eachofpotentialaccidentshasitsdirectandroot
causes.ThosehavebeenlistedinLevel 3ofanalysis
since they have an influence on most of Level 2
elements:
maintenance regime‐basis on which various
ship’s systems and mechanisms would be
supervised will have a great influence on their
reliability;
sensors’ performance‐quantity, quality and
arrangement of e.g. environmental or machinery
sensors will affect vessel’s ability
to detect
navigational hazards or abnormal operational
conditions;
control algorithms‐vessel’s response to
environmental and navigational conditions she
wouldmeetwilltogreatextentdependonquality
of the software and its developers’ ability to
predict all circumstances the ship can potentially
encounter;
alerting‐command to call for shore
operator’s
assistance must be implemented in software in
proper place for such operator to take over the
vesselinampletime,beforemalfunctionsorother
conditionsdeveloptoapointafterwhichnothing
can be done remotely‐such points shall be
defined;
external information quality‐unmanned vessel’s
safety will
also depend on quality of data
provided by external actors including weather
forecasts,passageplans,stowageplansand,which
is particularly important, shore‐based operator’s
situationalawarenessandabilitytoreactproperly
whenprompted;
operationalregime‐whethershipisoperatingin
fully‐manned,remoteorautonomousmodewould
havegreat
consequences foritssafety,particularly
foritscapabilityofdealingwithuncertaintiesand
magnitudeofhumanerrorlikelihood;
area of operation‐it is widely accepted among
shippingindustrythatsomeregionsoftheworld
aremoredangerous for vessels due toe.g.dense
traffic, long periods of heavy weather
or other
factors.
2.3.2 Level2
Unfortunate events as listed at Level 3 can
potentially initiate a chain of events that can
270
eventuallyleadtoanaccident.Wedividedaccidents,
towhichunmannedvesselscanbeexposedintofour
maincategories:
navigation‐related;
From navigational point of view, there are two
major hazards, namely collision and grounding
whichcanbothresultfromthethirdone,whichis
lossofpositionfix.
Shouldthelatteroccurdueto
e.g.satellitenavigationsystembeingnotavailable,
asecondarymodeofestablishingvessel’sposition
can be applied – some other radio navigation
systemorcontemporarydeadreckoningbasedon
accelerometers and gyrocompasses. Long‐term
accuracyofthelatter,duetoinherentbias,canbe,
however,questionablewhichmayeventuallylead
to a vessel running aground or colliding with
anotherobject(notnecessarilyavessel),especially
wheninlittoralwaters.
engineering‐related;
Problems with proper functioning of variety of
enginesonboard theshipcanbeboth causeand
result of navigational accident. Grounding, for
instance, can be caused by abnormalities in
functioningofsteeringdevices.Ontheotherhand,
if a vessel runs aground it can cause a serious
damagetotherudder.Similarrelationshipscanbe
attributed to propulsion and collision.
Furthermore, loss of electric power can disable
most of ship’s systems including
propulsion,
steering,communicationandballasting.
originatingfromstabilityorbuoyancyissues;
Stability and buoyancy issues can greatly impact
ship’soverall safety.Consequencesofstabilityloss
canbeparticularlydevastatingforshipaswellas
for her crew and cargo. On the other hand the
excessive stability of a ship causes
the undue
rolling resulting in high values of accelerations
acting on ship equipment and cargo. Especially,
the dynamic phenomena that do not cause the
vesseltosinkcandamagethecargo.Goodscarried
themselvescanalsobeareasonforwhichtheloss
ofstabilitymay occur,e.g.in case
ofliquefaction
or containment loss. Meeting inta ct stability
criteriawillalsodependonproperfunctioningof
ballast system since ballast operation will most
likely need to be carried out at sea in order to
compensateforfuelspentandfollowballastwater
managementregulations.
others
Lastbutnotleast,there
aremanymorehazardsto
includeinsafetyanalysis,whichdonotfitintoany
specific category. Unlawful acts can potentially
leadtomanykindsofaccidentsincludingfireand
explosion. Those in turn (regardless their origin)
can badly influence other subsystems, just to
mention propulsion, cargo or structural integrity.
Vessel’s ability to communicate with shore‐based
operatorcanalsobereducedwhichwouldinturn
have critical consequences to further ship’s
functions,e.g.collisionavoidanceorabilitytotake
part in Search and Rescue operation. This aspect
shall be addressed with particular attention as
unmanned vessels might at some point
find
themselves somehow involved in a situation,
which threatens other ship’s crew’s life. Distress
signal might be received by unmanned vessel or
shecanbetheonlyshipinarea capableofpicking
up survivors (she would therefore need to be
equipped with proper appliances and provisions
to accommodate those
unfortunate seafarers). On
the other hand, sufficient solutions aiming in
establishingthatSearchandRescueistakingplace
shall be provided so that the unmanned ship
proactively participates in it or at least does not
interruptit.Thisaspectofdecrewingofshipping
industry does not influence safety of unmanned
vessels
themselves, but can have a significant
impact on safety of marine transportation as a
whole.
Anotherissuethataffectsunmannedships’safety
to a limited extent but can potentially have great
influence on their perception by shippers and P&I
clubs and‐eventually‐on their economic results is
safety of cargo,
hereby referred also to as ‘cargo
damage’. Shipper or cargo owner expects his/her
goods to be delivered in ample time and good
condition. Any deviation from those contractual
conditions is unwelcome. Situation in which cargo
getswetorodorizedhasnoeffectonship’ssafetybut
a great one on shipowner’s
wallet size. Some cases,
however, include hazards to both of them, like self‐
heatingorself‐ignitionofcargoforinstance.
EachandeveryaccidentaslistedinLevel2isby
itselfanunwelcomeeventthatcanpotentiallycausea
majormaritimedisasterandcreateeitherfinancialor
environmental
losses.However,itcanalsoconstitute
justanotherlinkinthechainofeventsleadingtoeven
moredevastatingconsequences.
Ontheotherhand,someminorunwelcomeevents
constituting Levels 3 and 2 do not necessarily
endanger the ship’s safety as a whole, e.g. sensor’s
failure can be of little importance
should proper
redundancyarrangementbeinplace.
2.3.3 Level1
Assessing and ensuring safety of unmanned
vessels is a difficult task but must be carried out in
ordertoprovethatthosecancreatesomeaddedvalue
to global community. As can be deduced from
various papers on this subject
(Burmeister et al.
2014)(Rødseth & Burmeister 2015)(Man et al. 2014),
shipsinquestionwillbecomplexandtosomepoint
revolutionary. Regardless their design, they shall be
exposed to most of the same hazards as today’s
conventional ships, and many more‐most likely.
ThosearegiveninLevel2.
Another issue
in designing safety is its influence
on cost‐effectiveness of the system in focus. In
unmanned vessels’ case, that would express in both
investment and operational costs. For instance,
avoiding heavy weather areas can improve vessel’s
safety,butwillincreasetimerequiredforseapassage
and by that‐fuel costs
and likelihood of being off
schedule. Passing through such area might be risky
fortheshipnotonlybecauseweatherdamagecanbe
expected, but also for the likely necessity to reduce
speed and create even greater delays and fuel
consumption.Ontheotherhand,weatherconditions
intheareacan
provebetterthanexpectedandtherisk
taken might prove beneficial (Krata & Szłapczyńska
2012). Sati sfactory compromise must therefore be
found between vessel’s overall safety and cost‐
effectiveness(Rødseth&Burmeister2015).
271
Figure2.Modelstructuredepictingrelationshipsbetweenthreelevelsofunmannedmerchantvessel’ssafety.
Therefore,bystudyingthewaysandlikelihoodsof
unfavorable situation’s development from e.g.
sensor’smalfunctionintoaseriousmaritimedisaster,
one shallbecapableofestimating risks to which an
unmanned merchant vessel shall be exposed.
Utilizing risks estimation as a measure of assessing
object’s safety conforms to International Maritime
Organization’s
guidelines for Formal Safety
Assessment (IMO 2002). As a result, output of a
model (Level 1) as given in Figure 2 consists of a
likelihood of specific maritime accident (fire, loss of
stability, grounding etc.) to which some less
consequence‐abundant events could have
contributed.Notably,themodelitselfdoes
notspecify
which of events affects accidents’ consequences the
most–suchfeaturecanonlybededucedbystudying
achainofeventsforeachindividualcaseofaccident’s
scenario. We therefore deliberately omit considering
disaster’s consequences and focus on its likelihood
instead.Bylearningthelatter,weshallbecapable
of
estimatinghowsafethevesselinquestionis.
2.3.4 Modelvalidation
Inordertoverifycorrectnessofmodel’sstructure,
we applied it to a maritime accident of fire which
occurredonboardafully‐mannedm/v‘MaerskDoha’
on2
nd
October2006inChesapeakeBay.
Its root cause is attributed to a malfunction of
boiler’s automatic controls which led to operation
withlowwaterlevelanddepositionofsootinexhaust
gas economizer, which was not removed during
routine maintenance (Level 3). Although the engine
crewwasawareofthe
problem,vesselwasallowedto
proceed with reduced speed. Rapid rise in boiler’s
temperature caused nearby equipment to ignite
(Level 2). Carbon dioxide extinguishing system did
272
not function properly due to one of pilot cylinders
being inoperative and fire has been extinguished by
crew using fire hoses despite problems with
emergency power generator and emergency fire
pump. Those were also found inoperative. Fire had
alsocausedissueswithpowergeneratorsstartingair
supplies which in turn
made the generators
themselves impossible to start. Major temporary
repairs had to be completed including main engine
turbochargers’ cleaning before vessel could be
declared ready to proceed for permanent
repairs(MAIB2007).Factthattheaccidentoccurredin
restrictedwatersofChesapeakeBaycausedthevessel
likely to run aground should
the main engine fail –
fortunately,thatwasnotthecase,seeFigure3(Level
1hasnotbeenreached).
As per validation performed basing on real
accident, we were able to confirm that a model’s
structureisvalidfordescribingrisksassociatedwith
today’s vessels’ operation. Future unmanned ships
shall be subject to majority of hazards that are
encounteredbytoday’smerchantvessels.Thatmakes
the herein developed model suitable for describing
theirsafety.
3 DISCUSSION
In the course of presented research we aimed at a
general overview of relationships between safety
featuresofunmannedvessels.Ascanbeseen,
major
subsystems of such vessels are dependent on each
other. For instance, performance and reliability of
main and auxiliary engines will affect vessel’s
navigational capabilities, but will be depend on
performance of fire protection systems and
equipment(Gawdzińskaetal.2015).
Figure3.Validationofmodel’sstructureasappliedtom/v‘MaerskDoha’accident
273
Mostofsafetyfeatureswillalsorelyonsecondary
issuesjusttonamemaintenanceregime,sensorsand
externalinformation.
Uncertaintiesofthemodelaretheresultoffuture
unmanned vessels’ design, which remains to large
extentunknown.Ontheotherhand,thisverypaperis
aimedinassistingfuturedesigners
ofsuchsystemsin
their work by allowing them to trace some basic
relationshipsbetweensafetyfeatures.
Further uncertainties can be related to
imperfections of brainstorming itself as a scientific
method.Thosearelistedin(Isaksen1998)andinclude
issuesregardingchoiceofexperts,theirproductivity
andpotentiallackof
understanding.
4 FURTHERSTEPS
Future work should include eliciting Bayesian
network’s parameters (probabilities). This will be
achieved by involving wider group of stakeholders
takingpartintheprocessofplanningunmannedship
design and operations. This will help to get insight
into the design of a ship, allowing for the
quantification
of the technical reliability of the ship,
alongwiththeprobabilitiesdescribingtheoccurrence
and the negative effects of the anticipated hazards.
Duetolackofhistoricaldatapertainingtounmanned
ship operations, the experts’ knowledge elicitation
techniques will most likely be the only solution for
theriskmodelparameterization.
Another issue relates to the definition of the
acceptablerisklevel,whichwillbeusedascriteriafor
the safety assessment of the unmanned ships.
However this issue remains in the responsibility of
relevantmaritimeauthoritiesratherthanresearchers,
(Vanem2012)(Rødseth&Burmeister2015).
5 CONCLUSIONS
In this paper we present
the results of the hazard
analysis associated with the unmanned ships
adopting brainstorming as a method of compiling
experts’opinionsontheirsafetyfeatures.
Hazardstowhichtheywill beexposedhad been
listedandstructureofriskmodeldevelopedincourse
ofthestudy.Thelatter’sparameters(probabilitiesor
frequencies)areyettobe determined andshallbe a
subjectoffuturework.
Results of our work suggest that safety of an
unmanned ship as a system is made up of several
features, most of which must not be considered
separately from others. Basic incidents like sensors’
malfunctions or flaws
in mechanisms’ maintenance
regimecannotonlycauseoneofships’subsystemsto
fail but can also propagate and trigger a chain of
events that could eventually make a whole system
collapse with catastrophic consequences. This
underlines importance of not only using best‐suited
materials and equipment in construction of
such
vessels but also paying special attention to how
subsystems’ operation will affect others, particularly
inextraordinarycircumstances.
Ourmodelisintendedtobeappliedinunmanned
vessels’designprocessfromitsearlystages. Afterits
parametersaredetermined,itcanserveasusefultool
for naval architects in identifying potential
risks to
unmanned ships and analyzing cost‐effectiveness of
solutionsconsidered.Weexpectittobeimprovedin
years tocomewhenvesselsunderconsiderationare
finally implemented thus making empirical data
available.
REFERENCES
Allianz.2015.Safetyandshippingreview2015.Munich.
Burmeister,H.C.&Bruhn,W.C.&Rødseth,Ø.J.&Porathe,
T. 2014. Can unmanned ships improve navigational
safety
?InProceedingsoftheTransportResearchArena,14‐
17April2014.Paris.
Flage, R. & Aven, T. 2009. On treatment of uncertainty in
systemplanning. Reliability Engineering & System Safety
94(4):884–90.
Gawdzińska, K. & Kwiecińska, B. & Przetakiewicz, W. &
Pelczar, M. 2015. Przyczyny wypadków i
pożarów na
statkachmorskich.ZeszytyNaukoweAkademiiMorskiejw
Gdyni(91):21–29.
Gerigk, M. 2015. Innowacyjne rozwiązania w zakresie
okrętówiobiektówoływających.Logistyka3:1431–1438.
Gerigk, M. & Skorupski, J. 2012. Safety management of
complex airborne and seaborne technical objects.
ArchivesofTransport24(3):285–296.
IMO.2002.GuidelinesforFormalSafetyAssessment(FSA)
foruseintheIMOrule‐makingprocess.
Isaksen,S.G.1998.Areviewofbrainstormingresearch: six
criticalissuesforinquiry.
Jokioinen,E.2016.Remoteandautonomousships‐thenext
steps.
Krata, P. & Szłapczyńska, J. 2012. Weather hazard
avoidance
in modelingsafety of motor‐driven ship for
multicriteriaweatherrouting.TransNav6(1):71–78.
Kretschmann, L. & Rødseth, Ø. et al. 2015. Maritime
UnmannedNavigationthrough IntelligenceinNetworksfinal
report‐QualitativeAssessment.
Kretschmann, L. & Mcdowell, H. et al. 2015. Maritime
UnmannedNavigationthrough IntelligenceinNetworksfinal
report‐
QuantitativeAssessment.
Yemao, M. & Lundh, M. & Porathe, T. 2014. Seeking
harmonyinshore‐basedunmannedshiphandling‐from
theperspectiveofhumanfactors,whatisthedifference
we need to focus on from being onboard to onshore?
AdvancesinHumanAspects ofTransportation.PartI:231‐
239.
Marine
AccidentInvestigationBranch(MAIB).2007.Report
on the investigation of the machinery breakdown and
subsequentfireonboardMaerskDoha.Southampton.
Mazaheri,A.&Montewka,J.&Nisula,J.&Kujala,P.2015.
Usabilityofaccidentandincidentreportsforevidence‐
based riskmodeling‐a case study on shipgrounding
reports.
SafetyScience76:202–14.
Özgüner,Ü.&Stiller,C.&Redmill,K.2007.Systemsfor
safety and autonomous behavior in cars: The DARPA
grandchallengeexperience.ProceedingsoftheIEEE95(2):
397–412.
Rødseth,Ø.J.&Burmeister,H.C.2015.Riskassessmentfor
anunmannedmerchantship.TransNav9(3):357–64.
Rødseth, Ø.J. &
Tjora, Å. & Baltzersen, P. 2013. Maritime
UnmannedNavigationthrough IntelligenceinNetworksfinal
report‐ArchitectureSpecification.
Rogoway,T. 2015.DARPA’s unmannedsubmarine stalker
could change naval warfare forever.
http://foxtrotalpha.jalopnik.com/darpas‐unmanned‐
submarine‐stalker‐could‐change‐naval‐wa‐1695566032
(retrievedJuly1,2016).
274
Rossiter, J.R. & Lilien, G.L. 1994. New ‘brainstorming’
principles.AustralianJournalofManagement.19(1):61‐72.
Stokey,R.etal. 1999.AUVbloopersorwhyMurphymust
have been an optimist: a practical look at achieving
missionlevelreliabilityinanAutonomousUnderwater
Vehicle. 11th International Symposium on Unmanned,
Untethered,
SubmersibleTechnology(UUST’99):32–40.
Vanem,E. 2012.Principlesforsettingriskacceptancecriteria
forsafetycriticalactivities.In:AdvancesinSafetyandRisk
Management:1741–1751.London.
Weber, P.G.& Medina‐Oliva, G. & Simon,C. &Iung, B.
2012. Overview on Bayesian networks applications for
dependability, risk analysis and
maintenance areas.
EngineeringApplicationsofArtificialIntelligence25(4):671‐
682.
