487
1 INTRODUCTION
It is commonly believed that human errors are the
main causation factor for maritime accidents and
incidents.Theterm“humanerror”isabroadcategory
covering a wide variety of unintentional unsafe
behavior.FromAllianzfiguresarangefrom50to80%
are often seen, with 75% being
the figure used by
Allianz (2018). With this background, it could be
arguedthatanunmannedandfullyautonomousship
shouldbemuchsaferthan acorresponding manned
ship. However, there are several parameters which
willdeterminethesafetyofanautonomousshipand
this paper will attempt to present
a more complete
picture.
Section two will define the types of autonomous
ships we believe is the most relevant in the near
future, i.e. next 10 years. Section three will compare
autonomous ships, as understood by the authors,
withmannedshipsandlistthemaindifferencesthat
can also be the
basis for comparison of risk factors.
Section four discusses types of accidents and
causation factors and how this picture will be
modifiedforautonomousships.Sectionsfivetoseven
discuss different classes of accidents and try to
providesomequantitativeexpectationsforhowthese
classes will change when autonomy is
introduced.
Sectioneightwillgiveasummaryandconclusions.
2 WHATISANAUTONOMOUSSHIP?
Autonomy literally means “self‐governing” and
comes in very different forms. Rødseth (2018) dis‐
cusses this topic and provides a characterization
schemeforautonomyinships.MaritimeAutonomous
SurfaceShip(MASS)isbyIMOdefinedas
ashipthat,
to a varying degree, can operate independently of
human interaction. Autonomy is also closely
connected to unmanned operation: Having a
completelyunmanned shipisdesirable as itrealizes
significant gains by removing the hotel section and
Addressing the Accidental Risks of Maritime
Transportation: Could Autonomous Shipping
Technology Improve the Statistics?
Å.S.Hoem
NorwegianUniversityofScienceandTechnology,Trondheim,Norway
K.Fjørtoft&Ø.J.Rødseth
SINTEFOcean,Trondheim,Norway
ABSTRACT:Aparadigmshiftis presently underway in the shippingindustry promising safer,greener and
moreefficientshiptraffic.Inthisarticle,wewilllookatsomeoftheaccidentsfromconventionalshippingand
seeiftheycouldhavebeenavoidedwithautonomousshiptechnology.Ahypothesis
ofincreasedsafetyisoften
broughtforward,andweknowfromvariousstudiesthatthenumberofmaritimeaccidentsthatinvolveswhat
iscalled“humanerror”rangesfromsome60‐90percent.Ifwereplacethehumanwithautomation,canwethen
reducethenumberofaccidents?Ontheother
hand,isthereapossibilityfornewtypesofaccidentstoappear?
Whatabouttheaccidentsthataretodayavertedbythecrew?Thispaperwillpresentamethodtoassessthese
differentaspectsoftheriskscenariosinlightofthespecificcapabilitiesandconstraintsofautonomousships.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 13
Number 3
September 2019
DOI:10.12716/1001.13.03.01
488
associated energy use, removing much safety
equipment and reduces crew costs and by that also
allows easier scaling down of ship sizes (Rødseth
2018b).Centralinthisisalsotheuseofashorecontrol
center(SCC)asdiscussedinManetal(2015).Inthis
context,autonomyisimportant
toenableoperatorsin
thecontrolcentertomonitorandcontrolseveralships
andbythatreducecostsofoperationsintheSCC.
It is theoretically possible to design a fully
autonomousshipwithoutanyhumanoversightatall,
but this is extremely unlikely in all but very special
cases,
due to the resulting extreme demands on the
on‐board technology. Being able to operate with
“constrained autonomy” (Rødseth 2018) and having
humans as back‐up in cases where operational
demandsexceedtheautomationsystem’scapabilities
isamuchmorelikelyalternative.Inaddition,current
public and private law and regulations
associated
with safety of ship operations as well as with the
commercial issues related to shipping is also
dependentonhavingalegallyresponsiblepersonin
charge of the ship. Changing laws and regulations
will take a long time if it is at all possible (Rødseth
2017).
As the technology
improves, the shipping
community gets more experience with the operation
ofautonomousshipsandwhenlawsandregulations
have been updated, it is very likely that fully
autonomousshipswillbelaunched,butthiswilltake
manyyears.Technologywillbeusedforsensing,AI
andIoThavebeenrapidly advanced
and utilized in
variousfields.Automatedoperationsystemsofships
havebeenactivewithaimsoffurthersafenavigation
by preventing human errors, improving working
conditionsofshipʹsetc.(Matsumoto2018).
Inlinewiththeabovediscussion,inthefollowing
wewillassumethatanautonomousshipisa
shipthat
is completely unmanned, but with a shore control
center and limited (constrained) autonomy in the
onboardcontrolsystems.
3 COMPARISONTOMANNEDSHIPS
In the following paragraphs, we will attempt to
identify the main factors that distinguish an
autonomousshipfrom a conventional manned ship,
basedontheassumptions
fromtheprevioussection:
Fully unmanned cargo ship with constrained
shipboardautonomyandashorecontrolcenter(SCC)
tohandleeventsthattheautomationcannothandle.
3.1 Fullyunmanned
The most interesting autonomous ship projects are
associated with fully unmanned operations as
discussedintheprevioussection.Whiletherewill
be
provisions for having people onboard during
maintenanceandportoperations,unmannedvoyages
haveanumberofimportanteffects:
1 Higherdemandonsensors,automationandshore
control as operators in SCC lack some of the
ʺpersonal touchʺ, both on environment, ship and
technicalsystemʹsperformance.
2 Much
lowerexposuretodangerforthecrew.
3 May be unable to inspect equipment or systems
thatreporterrorsorproblems.
4 Lower risk of fires in accommodation, galleys,
laundry and waste systems, which is relatively
highonmannedships.
3.2 Constrainedautonomy
Autonomy will be limited for the onboard systems
andtheshipwillbedependentonoccasionalsupport
from the SCC. To avoid known problems with
human‐automation interfaces (HAI) in the shore
control center, the ship automation will have
ʺconstrained autonomyʺ (Rødseth 2018). The
assumption is that this also helps in testing and
qualifying sensor and automation systems
to
specified performance level. This has a number of
effects:
1 More limited, but also more deterministic action
responsesfromsensorsandautomation.
2 Dependence on shore control operatorsʹ
performanceandsituationalawareness.
3 Dependenceoncommunicationlinktoshore.
4 Dependence on high quality implementation of
fallbacksolutionsand
definitionofminimumrisk
conditionsfortheship.
3.3 Shorecontrolcenter
The shore control center will be manned with
supervision operators as well as specialist
intervention teams that are activated in cases of
special demands from a ship (Man et al. 2015). In
additiontoissuesmentionedintheprevious
sections,
thiswillhavethefollowingeffects:
1 Dependent on good training and cooperation in
theshorecontrolcenter.
2 Intervention crew do not have to worry about
personalriskandadverseconditionsonboard.
3.4 Highertechnicalresilience
Anotherimportantaspectisthereliabilityoftechnical
systems onboard and
increased redundancy in the
same systems. As there is no crew available to
provideasafetybarrierincaseoftechnicalfailures,it
is necessary to add new technical barriers where
necessary, e.g. by using increased redundancy. This
requirement is already included in the guidelines
publishedbyDNVGL(2018).
Today’s
crew use much of their time on
maintenanceoftheshipanditssystems.Thiswillnot
be possible on an unmanned ship and to avoid
increased off‐hire due to more and longer dry‐
dockings, it will be necessary to use systems with
lower maintenance requirements. This can typically
be diesel‐electric energy and propulsion systems, no
useofheavyfuel,improvedcoatingsontheshipand
incargoholdsetc.Effectsare:
1 Moretechnicalbarriersagainsttechnicalfaults.
2 Much improved technical systems with built in
predictivemaintenancefunctionality.
3 Moredependentonmaintenanceatshore.
489
3.5 Improved voyageplanning
Finally, unmanned ships will be used in liner type
operations where they trade between a relatively
limited number of ports where infrastructure and
trained personnel are available to handle the
unmannedship safely and efficiently. In addition to
infrastructure requirements, also the current legal
systems rule
out tramp type shipping where the
unmanned ship calls on arbitrary ports: Until
international regulations have been established,
unmanned operation will need to be based on
bilateral agreements between the involved flag,
coastal and port states. This also means that
operations of unmanned ship will be able to take
advantage
of better cooperation with coastal state
authorities, better described fairways, possibly
additional infrastructure in the fairways and
improvedplanningofthevoyage.Theeffectsofthis
are:
1 Lesschanceofsurprisesduringvoyage.
2 Moresupportfrompublicfunctionsonland.
4 ACCIDENTSCENARIOS
4.1 Today’saccidentpicture
Thereare
anumberofdifferent papers investigating
accident statistics and causation factors in the
available literature. They use different data sources
anddifferentmethodsandresultsvaryquitewidely.
Thepublicly availabledatabasesof marineaccidents
have different database structure and approaches to
analyze the accident causation and consequence
mitigation. There
are many reasons for this, among
them large variations in accidents between
geographicregions,typesofships,ageofships,flags
andinsurance,seee.g.Eleftheriaetal.(2016),Equasis
(2018)andAllianz(2018).
In this paper, we will use statistics from the
EuropeanMaritimeSafetyAgency(EMSA2018)and
mainly
figures from the period 2011 to 2017. This
covers accident reports from EU and associated
countries.
4.2 Occupationalfatalities
Working on a ship is in general considered more
dangerous than similar jobs on land. In the UK, the
fatalityrateatseaisabout12timeshigherthaninthe
general population and in Poland it is eight times
higherthanthatagain(Allianz2012).
From the EMSA statistics it can be seen a split
between occupational fatalities, e.g. slipping, falling
orbeing hit byobjects, and fatalitiescaused by ship
accident.Intheperiod2012to2017suchoccupational
fatalities
amountedto43%ofatotalof683fatalitiesin
theperiod.
Ifashipisoperatedwithoutacrew,itisobvious
thatthiswillbeasignificantcontributiontothesafety
ofthevoyageasseenfromthenowon‐shorecrew.
4.3 Shipaccidents
EMSA uses a
special classification system that is
implemented in EMCIP (European Marine Casualty
InformationPlatform).Amucha bbreviatedversionof
theclassificationsystemisshowninFig.1.
Figure1.EMCIPelements(EMSA2018)
Most casualties should be seen as processes that
involveanumberoferrors,failuresanduncontrolled
environmental impacts, and not just the more
dramaticCasualtyEventitself.Thisgroupof events
willcollectivelybetermedAccidentalEvents(Caridis
1999). The categories of accidental events used by
EMSAarelistedinFigure
2. Contributingfactorsis
somethingthathelpscausearesult.Thelattertwoare
often called causal factors, which in general mean
general actions, omissions, events or conditions,
withoutwhichthemarinecasualtyormarineincident
would probably not have occurred or have been as
serious (IMO 2008). Over the
period 2011 to 2017,
EMSA has analyzed 1645 accidental events with a
distributionasshowninFigure2below.
Figure2.AccidentaleventsfromEMPIC(EMSA2018)
Thispresentsalowerpercentageforhumanerrors
than what has been common in other literature
(Allianz2018,Baker2009),butitisstillasubstantial
contributingfactorwith58%.It is also interestingto
see that equipment failure represents 25% of the
accidentalevents.Wewillcomebacktothis
insection
5.
4.4 Thehumanfactor isstillanissue
Another statistics of interest is how respectively
shipborneoperationsandshoremanagementactsasa
maincontributingfactortothecasualtyevents.Thisis
rendered in Fig. 3, where around 2900 contributing
factorshavebeenanalyzed.
490
Figure3. Relationship between ship and shore as
contributingfactorformarine casualtiesin general(EMSA
2018)
Thismay have animpact on expectations from a
shorecontrolcenterinthecontextofunmannedships.
However,shipboardoperationisamaincontributing
factorto70%ofthecasualtyevents.
This bring us to the human role in MASS
operations. Humans still need to intervene with a
MASS vessel,
however the human element of the
operationsseemoftentobeforgottenwhendesigning
a MASS. The human, i.e. operator, still need to
supervise and analyze the operations done by the
autonomous systems, either from a SCC or when a
MASSismanned.Whenlookingataccidentstatistics
of conventional
shipping, we tend to look at the
negative side of human intervening. In the design
phase of MASS, the human machine interactions
(HMI)shouldbeaddressed.AConceptofOperation
(CONOPS) refer to the awareness of a situation. It
givestheperceptionofaneventwithrespecttotime
and condition,
and the system behavior (actual and
future).ACONOPSwilladdressthehumanfactorsin
the MASS operation aspect. Known relevant human
factorchallengesofremotelyoperatedandautomated
systemsthatshouldbeincluded(Karvonen2018)are:
Situationandautomationawareness
The understanding between automation and
humanrole
Userexperiencesandusabilityofthesolutions
Trustinautomation
Graphicaluserinterfaceandvisualization
4.5 Accidentsinautonomousships
Itisanexpectationthatmoreautomationcanremove
someoftheaccidentstodaycausedbyhumanerror:
Automation address human shortcomings like
fatigue,limitedattentionspan,information
overload,
i.e.limitsofthehumanworking memory,normality
biasetc.Howmuchthatautomationcanimprovethe
accident statistic is still an open question. The full
picture is also more complicated than this, as
illustratedinFig.4.(Poratheetal.2018).
Figure4.Threemaingroupsofaccidentsandincidents
Themiddlecirclesrepresenttodayʹsincidentsand
accidentsinshipping,whichwasdiscussedinsection
4.Therightcirclesrepresenttheaccidentsthattodayʹs
crewareabletoavoidbybeingpresentonboard.The
left circle represents new types of incidents that are
caused by the advanced automation systems
themselves.Thedarkcirclesarethedamagepotential
and the white circles represent actions by the
automation systemstoavoidorminimizetheeffects
of these incidents. This picture needs to include the
effectsoftheSCC.Here,itisimportanttoremember
that our increasing dependence on information
systems,
and increasingly sharing of control of
systemswithautomation,arecreatingaconsiderable
potential for loss of information and control leading
to new types of “human errors” (Leveson 2012).
Whicharecontributingtotheobservedpercentageof
“humanerror”involvedintheaccidentrates.
For the evaluation of the accidentʹs
causes, it is
possible to apply Human Factors Analysis and
ClassificationSystemforMarineAccidents (HFACS‐
MA). In a study by Wróbel et al. from 2017, 100
accidents reports were analyzed by applying this
method and paying particular attention to the
followingtwofollowingaspects:
Ifthe shipwere
unmanned, how would that fact
affectthelikelihoodofparticularaccident?
If the accident occurred anyway, would its
consequencesbemoreorlessseriousiftherewere
nocrewonboard?
According to HFACS‐MA, the accidentʹs causes
are divided into 21 causal categories grouped in 5
levels:
External Factors: Legislation gaps, administration
oversights,anddesignflaws.
Organizational Influences: Fallible decisions of
upper‐level management affecting supervisory
practices as well as the conditions and actions of
theoperator(Scarborough2005).
Unsafe Supervision: Supervisory actions that
influence the conditions of the operator and the
typeofenvironmentinwhichtheyoperate.
Preconditions: latentunsafeconditions for unsafe
acts that exist within a given work system (IMO
1999).
UnsafeActs:errors(slipsandlapse),mistakesand
violationsperformedbytheoperator.
The study concluded that the remoteness of the
humanoperatorsandcrewhasthebenefitofreducing
the risk to the personnel significantly, and reducing
the number of navigation‐related accidents like
collision or groundings (Wróbel et al. 2017, pp 10).
However, the results also showed
that the damage
491
assessment and control is likely to be one of the
biggestdifficultiesfortheunmannedvessel.
Onedrawbackofthestudyisthattheyevaluated
anunmannedvesselasavesselwiththesamedesign
andtechnical systems inplace, onlywith the bridge
and crew being remote. The design
and system
architecture of autonomous systems will be
completely different as discussed in section 3.4.
Another drawback of the study is the subjective
evaluation of the effect of unmanned ships on the
likelihoodoftheaccidentsandthemanyassumptions
about which HFACS‐MA causal category has the
largestimpact
onanaccidentʹsoccurrence.Asoneof
therecommendationsforfurtherresearch theauthor
emphasizetheneedtoidentifyandlistallanticipated
hazardsandtheirevaluatedeffects; only then can the
level of safety associated with the unmanned ships
operationsbeassessed(Wróbeletal.2017,pp.11).
In
this paper, we take a similar approach, but
instead of analyzing accident investigation reports,
we look at the larger picture and qualitatively
evaluate the potential for the causal factors most
commonfortheknownaccidentsandincidentstoday.
4.6 Experiencesfromaccidentsrelatedtosensemaking
andHMI
The past decades
we have seen a decline in marine
accidents leading to loss of property, life and
environmental damage. Particularly after 1980 the
introduction of new technology has been
accompanied by a steady and significant
improvementinshipsafety.Thesefirststepstowards
greater use of automation in machinery spaces
continued with advanced
ships with smaller crews
and increased operating efficiency through new
technologies, particularly with regard to navigation
system(Pomeroy2017,Hetherington,Flinetal.2006).
However, more automation has also been related to
thefollowingissues:diminishedshipsense,mishaps
during changeovers and handoffs, latency and
cognitive horizon, potential skill degradation,
and
resilienceinabnormalsituations.
Oneofthebiggestchallengesinhighlyautomated
systems is the disconnect, suggested as one of the
ironiesofautomation(Bainbridge1983),betweenthe
demandoftheshipsanditssystemandtheskillsand
knowledgeofthepeopleoperatingitbothatseasand
ashore
iscausingnewtypesofincidentsandaccident.
Inareviewof 14 MAIB accidentreports from 2005–
2016, Kilskar and Johnsen (In Press) identified the
followingsafety issues concerningautomationatthe
bridgecontributingtoseveraloftheaccidents,hence
contributingfactors:
Lossofsituationawareness/poorsensemaking
Insufficienttraining
Alarmrelatedissues
Poorsystemdesignordisplaylayout
Poor(safety)management
Poorormissingworkloadassessment
Lackingorinsufficientpassageplanning
Missing,poororunclearregulationsorstandards
Although these safety issues where identified in
accidentinvestigationreports,they
concernHMIand
willapplytooperatorsinaSCC.
5 AQUALITATIVECOMPARISONOF
AUTONOMOUSANDMANNEDSHIPS
We have listed the main factors that distinguish an
autonomousshipfromamannedship,asdiscussedin
paragraph 3.1 – 3.5. With the identified causal and
contributing factors, conditions, activities, systems,
components, etc. that are critical with respect to
accidental risk, presented in section 4 and 5, we
attempt to classify the potential for higher or lower
contributionstotodayʹsincidentsinshipping.
Table 1 lists the characteristics of different
technological solutions and shortly describes their
strengthand/orshortcomings.For
eachcharacteristic,
acolorindicatestowhichdegreethisiscontributing
tothethreerisktypeslistedinthethreelastcolumns:
New accidents caused by new technology, Today’s
known accidents, and accidents adverted by crew)
illustrated in Fig. 4. The factors contribution to the
risk types is indicated by the following colors:
increasedrisk(red‐R),neutralimpact(yellow‐Y),or
lesserimpact/likelihood(green‐G).Notethatforthe
firsttypeofaccidentalrisks,newaccidentscausedby
newtechnology,autonomousshipscanobviouslynot
be
better than today.At best, it is neutral (Y). For a
fullyunmannedship,onedifferentiatingfactorfrom
manned ship is a higher demand and reliance on
sensors,automationandshorecontrol(row1intable
1below).Moreadvancedtechnologymeansahigher
degree of system complexity causing new
technologicalfailureslikeunknownsoftwarefailures
for example. This contributes to a higher likelihood
(risk) of new accidents caused by new technology,
indicatedbyared“R”underthecolumn“New”.For
today’sknownincidentsandaccidentslikecollisions
andallisionscausedbyhumanerroneousactionsdue
to fatigue, new technology
will be able to address
such human shortcomings with collision detection
andavoidancesystems.Hence,agreen“G”indicates
thepositivecontributiononrisk,asknownaccidents
areavoided bynew technology. However, accidents
adverted by crew today should also be possible in
autonomous operations by remote control and
operation
from the SCC. The technology in a fully
unmannedshipandSCCshallbedesignedforremote
operation, and the crew will still have impact, in
order to avoid accidents and incidents. Hence, the
contributionisneutral,indicatedbyayellow“Y”.
6 DETAILEDDISCUSSION
First category, fully unmanned, points to
a higher
risk for software and technical failure. Due to for
example:
Sensor failure/degradationofhardware
Insufficientredundancy
Lossofpropulsionorsteeringcontrol
Cybersecuritybreaches
LossofcommunicationwithSCC
However, unmanned vessels will improve on
some of todayʹs operators’ errors caused
by human
erroneous actions due to fatigue or other harsh
workingconditions.
492
Table1.Qualitativecomparisonofautonomousandmannedshipping
Important factors to address in the design and
development of MASSs is robust sensor quality,
redundancy on key technology, and good education
for land ‐based operators, that builds the situational
awarenessbasedontechnology.Nextfactorthathas
been pointed to is less exposure to danger for the
crew. Statistics tells
that about 40% of deaths at sea
areoccupationalhazards.Anotherelementisthatitis
expectedthatitwillbeslightlylower riskoffiresin
accommodation, galleys, laundry and waste system,
because of no installation of such technology due to
thefactthatthereisnoneedfor
itsincethereareno
people on board. The expectations are fewer
accidents,butwhenanaccidenthappens,itmightbe
moredifficulttocombatwhenpeopleisnotavailable
and the only trust is technology, as addressed by
Wróbeletal.(2017).
For a constrained autonomy vessel, we have
pointed
tobetterhuman‐automationinterfaces,dueto
time to get situational awareness before action. The
designofSCCwilllearnfromaccidentswherealarm
related issues and poor HMI were major causal
factors.Itislikelythatthehumansarenotdirectlyin
the loop (manually steering and navigating the
vessel). To let the SCC take control there are
dependencies to the infrastructure, such as the
communicationinfrastructure,thatwillhaveenough
coverageandbandwidthtobringdatafromthevessel
totheSCCforawarenessbeforedecisionsaretaken.
This also points to more conservative and safer
operationalprocedures,to
bothoperationalpractices
andahighersafetydegree.
493
Shorecontrol center isanother categorythat has
beenpointedto.ThesameappliesforaSCCasona
vessel’s bridge today, a good crew is those who
collaborate and use each other’s expertise in
operations and problem solving. It is even more
importantataSCCsincethe
possibilitytoinspectthe
vesselisnotthesame.Weassumehereanincreased
riskofaccidentsthatistodayadvertedbycrew,aswe
knowtherewillbecontrollabilityissueswitharemote
crew,andahighdependenceontheSCCteam’sskills
and knowledge. At the same time,
the human risk
factorislowersincetheinterventioncrewdonothave
toworryaboutpersonalriskandadverseconditions
on board. Training and resource management are
important.
ThecategoryHighertechnicalresiliencebringsus
back to the technology. It is important to build
technical barriers towards technical failures with
built‐inpredictivemaintenancefunctionality.
Technical resilience is essential for MASS. The
dangeristhatnewunpredictablesituations,thathave
notbeenthoughtof,canoccurduetoahighnumber
oftechnical systems.Componentinteractionaccidents
are becoming more common as the complexity of
systemdesignsincreases(Leveson2012).
Improved voyage planning is a safety‐critical
function for autonomous vessels. Good planning
means to prepare the voyage, the loads, the
maintenanceandallreportingduringavoyage.This
is a significant requirement compared with
conventional vessels, were good planning is crucial
forsuccess,butoftenoverlooked(NTSB2015,DMAIB
2013,
Bell2006).
7 CONCLUSION
This paper provides a more realistic description of
what an autonomous ship will be in the foreseeable
future,i.e.unmanned,havingmonitoringandcontrol
personnelonshore,exhibitingconstrainedautonomy
andhavingbetteroperationalplanningandtechnical
equipmentthanamannedship.
While the overall risk
picture for autonomous
shipsmay lookunpromising (Fig. 4),the differences
in implementation have significant impacts on the
individualrisktypes.Thequalitativeassessmentdone
inTable1indicatesthatthereisindeedasignificant
possibilitytoimproveoverallsafety for autonomous
ships compared to manned, although there are also
areas
thatrequirespecialattention.
Thispaperonlyprovidesacursoryandqualitative
analysisoftheriskissues,butitishopedthatitcan
contribute to a more systematic process for risk
assessment, also more accurately incorporating the
positive technical contributions from autonomous
shipdesigns.
REFERENCES
Allianz Global Corporate and Specialty (2018), Safety and
shippingReview2018–anannualreviewoftrendsand
developments in shipping losses and safety, Munich,
Germany,June2018.
Allianz Global Corporate and Specialty (2012), Safety and
Shipping 1912‐2012: From Titanic to Costa Concordia,
Munich,Germany,March2012.
Baker,C.,McCafferty,
D.(2009).ABSReviewandAnalysis
ofAccidentDatabases.
Bainbridge, L. (1983). Ironies of automation. In Analysis,
Design and Evaluation of Man–Machine Systems 1982
(pp.129‐135).
Bell, J., & Healey, N. (2006). The causes of major hazard
incidents and how to improve risk control and health
and safety management:
A review of the existing
literature.HealthandSafetyLaboratory.
DNV GL (2018), Class Guideline‐Autonomous and
remotely operated ships, DNVGL‐CG‐0264, September
2018.
DMAIB(2013).TheDanishMaritimeAccidentInvestigation
Board:VEGA SAGITTARIUSGroundingon16 August
2012. Marine accident report 2012003009. Issued on 27
March2013.Valdby:DMAIB.
Eleftheria,E.,Apostolos,P.,&Markos,V.(2016).Statistical
analysis of ship accidents and review of safety level.
Safetyscience,85,282‐292.
EMSA (2018), Annual Overview of Marine Casualties and
Incidents2018.EMSA,Lisbon,Portugal,2018.
Equasis (2018), The World Merchant Fleet in 2017 –
Statistics from Equasis, www.equasis.org,
retrieved
January2019.
Caridis,P.(1999).CASMET.Casualtyanalysismethodology
for maritime operations. National Technical University
ofAthens.
Hetherington, C., Flin, R., & Mearns, K. (2006). Safety in
shipping: The human element. Journal of safety
research,37(4),401‐411.
IMO MSC/Circ.102/MEPC/Circ.392. 2002. Guidelines for
Formal Safety Assessment (FSA) for use in
the IMO
Rule‐MakingProcess.Asamended.London:IMO2002.
IMOMSC.255(84).2008.Codeoftheinternationalstandards
and recommended practices for a safety investigation
into a marine casualty or marine incident (casualty
investigation code). Adopted May 16, 2008. London:
IMO2008.
IMOResolutionA.884(21).Amendmentstothecodefor
the
investigation of marine casualties and incidents
(A.849(20)).London:IMO1999.
Karvonen, I. 2018. Human Factors Issues in Maritime
AutonomousSurfaceShipSystemsDevelopment.The1
st
International Conference on Maritime Autonomous
SurfaceShip.
Leveson, N. 2012. Engineering a safer world: applying
systemsthinkingtosafety.
Man, Y., Lundh, M., Porathe, T., & MacKinnon, S. (2015).
From Desk to Field–Human Factor Issues in Remote
MonitoringandControllingofAutonomousUnmanned
Vessels.ProcediaManufacturing,3,2674‐2681.
Matsumoto
T.et.al2018.Guidelinesforconceptdesignof
automated operation/autonomous operation of ships.
International conference on maritime autonomous
surfaceship.
NTSB,NationalTransportSafetyBoard.2015.Groundingof
MobileOffshore DrillingUnitKulluk, near OceanBay,
Sitkalidak Island, Alaska December 31, 2012.
NTSB/MAB‐15/10.
Pomeroy,R.V.,&Earthy,J.
V.(2017).Merchantshipping’s
relianceonlearningfromincidents–Ahabitthatneedsto
changeforachallengingfuture.Safetyscience,99,45‐57.
PoratheT., HoemÅ.,RødsethØ.J.,FjørtoftK.,JohnsenS.O.
(2018), At least as safe as manned shipping?
Autonomousshipping,safetyandʺhumanerrorʺ.XXXX
Rødseth
Ø.J (2018). Defining Ship Autonomy by
Characteristic Factors, Proceedings of ICMASS 2019,
Busan,Korea,ISSN2387‐4287.
Rødseth Ø.J (2018b). Assessing Business Cases for
AutonomousandUnmannedShips.In:Technologyand
494
SciencefortheShipsoftheFuture.ProceedingsofNAV
2018:19thInternationalConferenceonShip&Maritime
Research.IOSPress2018ISBN978‐1‐61499‐870‐9
Rødseth Ø. J. (2017). From concept to reality: Unmanned
merchant ship research in Norway. Proceedings of
UnderwaterTechnology(UT), IEEE,Busan,Korea,ISBN
978‐1‐5090‐5266‐0.
Rødseth Ø.J. & Nordahl H. (eds.). 2017. Definition for
autonomous merchant ships. Version 1.0, October 10.
2017. Norwegian Forum for Autonomous Ships.
http://nfas.autonomous‐ship.org/resources‐en.html.
[Accessed2018‐12‐12].
Scarborough,A.,Bailey,L.,&Pounds,J.(2005).Examining
ATCoperationalerrorsusingthehumanfactors
analysis
and classification system (No. DOT‐FAA‐AM‐05‐25).
Federal Aviation Administration Oklahoma City OK
CIVILAEROMEDICALINST.
Wróbel,K.,Montewka,J.,&Kujala, P.(2017).Towardsthe
assessmentofpotential impactofunmannedvesselson
maritime transportation safety. Reliability Engineering
&SystemSafety,165,155‐169.
