Maritime Autonomous Surface Ships (MASS) and the COLREGS: Do We Need Quantified Rules Or Is “the Ordinary Practice of Seamen” Specific Enough?

Abstract

Maritime Autonomous Surface Ships (MASS) is currently on the agenda in several countries and also in the IMO. In Norway a 120 TEU container feeder is being build and will start sailing autonomously in 2022. The challenge is huge. One question is whether or not the present, quantitative, collision regulations needs to be updated to rules where expressions as “early” and “substantial” are quantified? Or if ships can sail autonomously under the present rules? Another question is if MASS should be marked to signal that the ship is in autonomous mode? Or if it is enough that she follows COLREGS? This discussion paper will take a closer look at these questions and advocate automation transparency, meaning that the behavior of an autonomous vessel has to make sense and be understandable to human operators on other manned ships and crafts.

Full text

511
1 INTRODUCTION
Largeautonomousmerchantvesselsarestillonnot
forreal.However,inNorwaythebuildingcontractis
alreadysignedforYARABirkeland,thefirstMari‐time
Autonomous,SurfaceShip(MASS),anun‐manned
containerfeeder,scheduledtostarttestsin2020
(Kongsberg,2019).LackingIMOregulations,thetests
willhavetocommenceinnationalwaters,whichin
thiscasemeanstheGrenlandareaofPorsgunnand
LarvikinsouthernNorwaywithcomplexnarrow,
inshorearchipelagonavigation.Itisabusyindustrial
areawherealargeportionoftheshiptrafficconsists
ofgascarriersandvesselswithhazardouscargoand,
summertime,anabundanceofsmallleisurecraftsand
kayaks.Theseatrafficintheareaismonitoredbythe
BrevikVTSwhichin2015made623“interventions,”
meaningthattheVTSaskedforsomealterationfrom
theplannedsailingroute(StatisticsNorway,2016).
Conductingautonomousnavigationinsuchanareais
ahugechallenge.
Theprojectisambitious.The80meterslong,
unmanned,autonomousvessel,taking120containers
withafullyelectricpropulsionsystem,willreplace
some40,000truckhaulseveryyear.Thusmoving
heavytrafficfromroadtosea,fromfossilfuelto
hydrogeneratedelectricity.Theplaniscurrentlythat
shewillstarttestrunsin2020.Firstwithamanned
bridgeonboard,thenwiththesamebridgeliftedoff
tothequayside,remotelycontrollingthevessel,
beforefinallyattemptingtogoautonomouslyin2022
(Kongsberg,2019).
1.1 Unmanned,automaticandautonomous
Todaysmannedshipsmaybethoughtofas
“manual.”However,thelevelofautomationisin
manyshipsquitehigh.Withanautopilotin“track‐
following”mode,setsothattheshipcanexecute
turnsalongapre‐plannedroutewithout
acknowledgmentfromtheOfficeroftheWatch
(OOW)‐giventhatthevoyageplaniscorrectand
Maritime Autonomous Surface Ships (MASS) and the
COLREGS: Do We Need Quantified Rules Or Is “the
Ordinary Practice of Seamen” Specific Enough?
T.Porathe
NorwegianUniversityofScienceandTechnology,Trondheim,Norway
ABSTRACT:MaritimeAutonomousSurfaceShips(MASS)iscurrentlyontheagendainseveralcountriesand
alsointheIMO.InNorwaya120TEUcontainerfeederisbeingbuildandwillstartsailingautonomouslyin
2022.Thechallengeishuge.Onequestioniswhetherornotthepresent,quantitative,collisionregulations
needstobeupdatedtoruleswhereexpressionsas“early”and“substantial”arequantified?Orifshipscansail
autonomouslyunderthepresentrules?AnotherquestionisifMASSshouldbemarkedtosignalthattheshipis
inautonomousmode?OrifitisenoughthatshefollowsCOLREGS?Thisdiscussionpaperwilltakeacloser
lookatthesequestionsandadvocateautomationtransparency,meaningthatthebehaviorofanautonomous
vesselhastomakesenseandbeunderstandabletohumanoperatorsonothermannedshipsandcrafts.

http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 13
Number 3
September 2019
DOI:10.12716/1001.13.03.04

512
validatedforasetunderkeelclearance.Thisisthe
waytheNorwegiancoastalexpressHurtigruten
navigatesduringmostofitsinshoreroutefrom
BergentoKirkenes(Porathe,pers.comm.).Butthe
OOWstillhastobepresentonthebridgetolookout
forandhandleencounterswithothershipsandcrafts.
Whatisneededtoremovetheoperatorcompletelyis
differentsensorsthatcanseeandidentifymoving,
unchartedobstaclesinthesea,andanautopilot
connectedwithacollisionavoidancemodule
programmedwiththeInternationalRegulationsfor
PreventingCollisionsatSea,COLREGSforshort
(IMO,1972).Withsuchasystemitisspeculatedthata
shipinautonomousmodemaynavigate
automatically.
However,suchan“automaticship”doesnotneed
tobeunmanned.Itmaycontainamaintenancecrew,
orevenareducednumberofnavigatorswhotake
manualwatchesduringdifficultconditions,ormaybe
daytimewatchesingoodconditions,savingthe
automationforthelongboringnightwatchesor
uneventfuloverseepassages.Withsuchapartly
mannedbridgetheshipwouldhavea“periodically
unattendedbridge”accordingtoIMO’slatest
definitions,(IMO,2018).
ThewatchcanalsobehandedovertoaShore
ControlCentre(SCC)thatcanaccesstheshipssensors
andcommunication,readytowakeuptheOOWif
somethingunexpectedhappens(inwhichcasethe
shipis“remotelymonitored”).Or,theSCCcouldbe
grantedaccesstotheautopilot,inwhichcasetheship
willbe“remotecontrolled”.Itisreasonabletothink
thatthiswillbeagradualevolutiontowardshigher
andhigherlevelsofautomation,maybea
combinationofremotemonitoringandcontrol,and
autonomy.
Itcanalsobeusefultoconsidertheconcept
“OperationalDesignDomain”(ODD)usedbythe
self‐drivingcarindustry(Rodseth&Nordahl,2017).
Inthemaritimedomain,itwouldmeanthattherewill
becertainshippinglanesandfairwayswerethe
automationhasbeenspecificallytrainedandwhich
havebeenspecificallyprepared,maybewith
designatedlanes,orbyspecifictechnical
infrastructure.Intheseareas,ashipmaynavigate
autonomously,whiletheshipinotherareasmust
navigatemanuallywithamannedbridgeorremote
controlledfromtheshore.
TheconceptofOODalsohasdeeperimplications
intothecultureofvesseltrafficinspecificareas.More
onthislater.
Forthediscussioninthispaperthefocuswillbe
onshipsin“autonomousmode”,regardlessof
whetheritispermanentoronlyperiodically.With“in
autonomousmode”Imeanthatacomputerpro‐gram
isnavigating,takingdecisionsandexecutingthem,
regardlessofwhetheranOOWisstandingbyonthe
bridge,orthecaptainisinhiscabinonboardorina
remotecentreashore.Thefocushereisonhowthe
shipautomationcanhandlesinteractionwithother
ships,andparticularlyhowitcouldfollowtherules
oftheroad,theCOLREGS.
2 THECOLREGS
Forseveralcenturiesshipscameandwent,sailing
withthesamewindandtideanditwasnotuntilthe
steamshipsturnedupinthebeginningofthe19th
centurythatcollisionregulationsbecamevital
(Crosbie,2006).In1840theLondonTrinityHouse
drewupasetofregulations,oneofwhichrequireda
steamvesselpassinganothervesselinanarrow
channeltoleavetheotheronherownporthand.The
otherregulationrelatingtosteamshipsrequired
steamvesselsondifferentcrossingcourses,soasto
involveriskofcollision,toaltercoursetostarboard
andpassontheportsideofeachother.Thetwo
TrinityHouserulesforsteamvesselswerecombined
intoasingleruleandincludedintheSteam
NavigationActof1846.Duringtheyearsanumberof
iterationsandinternationalizations,throughwhatis
nowtheInternationalMaritimeOrganization(IMO),
ledtothelatestrevisionoftheInternational
RegulationsforPreventingCollisionsatSea
(COLREGS)onaninternationalconferenceconvened
inLondonin1972.
Onemayaskifmaybenewrulesareneededfor
autonomousships?Ormaybethereshouldbema‐
chine‐to‐machinenegotiationsineveryindividual
caseofconflictingcourses?Thefinalanswertothat
questionisunknown,butitismyfirmopinionthatas
longasMASSwillinteractwithhumansonmanned
shipstherehastobealimitednumberofcommon
andeasytounderstandrulesknownto,andobeyed
by,allvesselsatsea.Onecandreamupotherrules,
butwhatwegot,andneedtoadhereto,isthe
COLREGS.Havingsaidthat,onemightconsiderif
extensionsorrevisionsmaybeneeded.
2.1 Qualitativerules
Thecollisionregulationsare,likelegaltextoftenis,
writteninageneralmannersoastobeapplicablein
asmanysituationsaspossible.Theprecise
interpretationhastobemadeinthecontextofthe
actualsituationjudgednotonlyonknowledgeofthe
rules,butalsoonexperienceandculture,whatthe
rulescall“theordinarypracticeofseamen,”asis
statedalreadyinthesecondrule.
ThequalitativenatureofCOLREGSwillbea
problemfortheprogrammerwhoistowritecodefor
thecollisionavoidancealgorithmsofautonomous
navigationmodules.Iwillinthissectionpointto
somethese“soft,”qualitative,clauseswherethese
problemswillbecomeapparent.
2.2 Rule2:theordinarypracticeofseamen
Rule2oftheCOLREGSisaboutresponsibility.Ithas
twosections.Section(a)state“NothingintheseRules
shallexonerateanyvessel,ortheowner,masteror
crewthereof,fromtheconsequencesofanyneglectto
complywiththeseRulesoroftheneglectofany
precautionswhichmayberequiredbytheordinary
practiceofseamen,orbythespecialcircumstancesof
thecase.”
Section(b)ofthesamerulestatesthat“In
construingandcomplyingwiththeseRulesdue

513
regardshallbehadtoalldangersofnavigationand
collisionandtoanyspecialcircumstances,including
thelimitationsofthevesselsinvolved,whichmay
makeadeparturefromtheseRulesnecessarytoavoid
immediatedanger.”
Whatthisrulebasicallysaysisthatyoumustal‐
waysfollowtheserules,butthatyoumustalso
deviatefromtheseruleswhennecessarytoavoidan
accident.Inessence,ifyouhaveanaccidentitisa
goodchancethatyouhaveviolatedoneorbothof
thesesections.Theproblemforthenavigatorishow
long,orcloseintoanencounter,heorsheshould
followtheRulesandwhenitistimetoskiptherules
anddowhateverisnecessarytoavoidacollision.The
answeris:itdependsonthecircumstances.TheRules
givenohintastothenumberofcablesormiles,
minutesorseconds.Itdoesnoteventrytodefinethe
“ordinarypracticeofseamen.”
Similarsoftenumerationsarefoundforinstancein
Rules15,16and17.
2.3 Rule15to17,riskofcollision
Rule15oftheCOLREGStalksabout“crossing
situations”:“Whentwopower‐drivenvesselsare
crossingsoastoinvolveriskofcollision,thevessel
whichhastheotheronherownstarboardsideshall
keepoutofthewayandshall,ifthecircumstancesof
thecaseadmit,avoidcrossingaheadoftheother
vessel.”
Calculatingwhenacrossingsituationmayleadto
acollisionisprettystraitforwardgiventhatpresent
courseandspeedcanbeextrapolated.(Thisis,how‐
ever,inrealitynotalwaysthecaseastheintentionsof
theothershipmaynotbeknown.)Ifthebearingto
theothershipisconstantovertime,itcanbeassumed
thatthereexistsariskofcollision.Rule15alsodefines
whichvesselshouldtakeactiontoavoidcollision.
“Theonewhichhastheotheronherownstarboard
side.”
Thefollowingrulethendefineshowthisaction
shouldbedonebythe“give‐way”vessel(Rule16):
“Everyvesselwhichisdirectedtokeepoutoftheway
ofanothervesselshall,asfaraspossible,takeearly
andsubstantialactiontokeepwellclear.”
Thisactioncouldbeachangeofspeedorachange
ofcourse,butforthesoftwareprogrammerthe
problematickeywordshereare“earlyand
substantial”.Thereisnosuggestioninmilesorclock
minuteswhatconstitutes“early”,neitherhowlarge
coursechangeorspeedchangeconstitutes
“substantial”.
Rule17definestheactionsoftheshipthatisnot
obligedtoyield,“thestand‐on”vessel:“(a),(i)Where
oneoftwovesselsistokeepoutofthewaytheother
shallkeephercourseandspeed.(ii)Thelattervessel
may,however,takeactiontoavoidcollisionbyher
maneuveralone,assoonasitbecomesapparentto
herthatthevesselrequiredtokeepoutofthewayis
nottakingappropriateactionincompliancewith
theseRules.(b)When,fromanycause,thevessel
requiredtokeephercourseandspeedfindsherselfso
closethatcollisioncannotbeavoidedbytheactionof
thegive‐wayvesselalone,sheshalltakesuchaction
aswillbestaidtoavoidcollision.(c)Apower‐driven
vesselwhichtakesactioninacrossingsituationin
accordancewithsubparagraph(a)(ii)ofthisRuleto
avoidcollisionwithanotherpower‐drivenvessel
shall,ifthecircumstancesatthecaseadmit,notalter
coursetoportforavesselonherownportside.(d)
ThisRuledoesnotrelievethegive‐wayvesselofher
obligationtokeepoutoftheway.”
Thisruleaddstothecomplexitybyusingqualitative
definitionslike“assoonasitbecomesapparent,”
“findsherselfsoclosethatcollisioncannotbeavoided
bytheactionofthegive‐wayvesselalone,”“actionas
willbestaidtoavoidcollision”and“ifthe
circumstancesatthecaseadmit.”



Figure1.ThiswasthetrafficsituationatSkagen,thenortherntipofDenmarkat15:00on5November2018.Onemayreflect
onthedifficultiesofCOLREGalgorithmsneededtodocollisionavoidanceinsuchanareawheregivingwaytooneship
mightleadintoanotherconflictsituationinanunpredictable,cascadingmanner(screenshotfromMarineTraffic.com).



514
Foraprogrammerprogrammingthecollision
avoidancemoduleofanautonomousnavigation
softwarethedifficultyisnotonlyinjudgingwhich
action,butalsowhentoexecuteit“early”and
“substantially”.Theanswerwillbethesameasitwas
intheprevioussection:itdependsonthe
circumstances.Arethereonlytwoshipsmeeting
aloneonthehighseasthetaskmightberelatively
simple,butattheotherendofthespectrum,inahigh
complexitysituation,e.g.inaconstrainedand
intenselytraffickedarealiketheStraitsofMalacca
andSingapore,thetaskisofanentirelydifferent
dimension.Notonlydoesthelargenumberofships
inalimitedspacechangethevalueofvariableslike
“early”and“substantial,”butanevasivemaneuver
foroneshipmayleadintoaclosequarterssituation
withanothershipandsoon,inacascading
interactioneffectwithunpredictableresults.Figure1
showsthecomplicatedtrafficsituationaround
SkagenonthenortherntipofDenmark.
Insomeareastherecanalsobeadifferentculture
ofhowthingsaredone(sometimesquitecontraryto
COLREGS).WhenthehighspeedferryStenaCarisma
traffickedtheGothenburg‐Fredrikshavnlinein30+
knots,anofficerIspoketosaid“Wealwayskeepout
ofthewayofeverythingthatmovesbecauseweare
sofastandmaneuverable”(Porathe,pers.comm.).
AlsointheSoundbetweenSwedenandDenmarkthe
Helsingborg‐Helsingorferrieshasacultureof
keepingoutofthewayinmostsituations(Porathe,
pers.comm.).
Apossiblestrategyforaprogrammertryingto
catch“earlyandsubstantial”aswellas“theordinary
practiceofseamen”foraspecificarea(anODD)could
betostudylargeamountsofAIS(Automatic
IdentificationSystem)dataforthespecificareain
questionsandfromthatdatadeducelimitsof“early”
and“substantialaction”.Ausefulconceptcouldthen
beships“safetyzones”whichisthezonearoundones
shipthatnavigatorstendnottoletothershipswithin.
“Azonearoundavesselwithinwhichallother
vesselsshouldremainclearunlessauthorized,”
(IALA,2008).Thiszonetendstobelargerontheopen
seathaninnarrowwatersorinaportandcanbe
studiedusingAISdata.UsingsuchAISstudies,
establishmentofazoneoutsidewhichanactioncan
beconsidered“early”couldbeattempted.Butthe
contextisimportant,notonlythestaticgeographical
context,butalsothetimedependenttrafficdensity
context.
TheNauticalInstitutementionsthat“Asageneral
guideline,attempttoachieveaCPA(closestpointof
approach)of2(nautical)milesintheopenseaand1
mileinrestrictedwaters”(Lee&Parker,2007,p.35).
Ifallshipsinsuchacomplexsituationwhere
autonomousandgovernedbycleveralgorithmsthere
isachancethatsuchacollisionavoidanceapplication
couldbesuccessful,butinamixedsituationwhere
mostormanyoftheshipsarecontrolledbyhumans,
whicharelesspredictable,theriskofabadoutcome
isevident.
2.4 Rule19,restrictedvisibility
ThefinalrulethatIwanttobringuphereisRule19,
“Conductofvesselsinrestrictedvisibility.”Thisisa
quitlengthyrulewhichsays:
“(a)ThisRuleappliestovesselsnotinsightofone
anotherwhennavigatinginornearanareaof
restrictedvisibility.”
Further“(b)Everyvesselshallproceedatasafe
speedadaptedtotheprevailingcircumstancesand
conditionsofrestrictedvisibility.Apower‐driven
vesselshallhaveherenginesreadyforimmediate
maneuver.”
“(c)Everyvesselshallhavedueregardtothe
prevailingcircumstancesandconditionsofrestricted
visibilitywhencomplyingwiththeRulesofSectionI
ofthisPart.”
“(d)Avesselwhichdetectsbyradaralonethe
presenceofanothervesselshalldetermineifaclose‐
quarterssituationisdevelopingand/orriskof
collisionexists.Ifso,sheshalltakeavoidingactionin
ampletime,providedthatwhensuchactionconsists
ofanalterationofcourse,sofaraspossiblethe
followingshallbeavoided:(i)analterationofcourse
toportforavesselforwardsofthebeam,otherthan
foravesselbeingovertaken;(ii)analterationof
coursetowardsavesselabeamorabaftthebeam.”
“(e)Exceptwhereithasbeendeterminedthata
riskofcollisiondoesnotexist,everyvesselwhich
hearsapparentlyforwardsofherbeamthefogsignal
ofanothervessel,orwhichcannotavoidaclose‐
quarterssituationwithanothervesselforwardsofher
beam,shallreduceherspeedtotheminimumat
whichshecanbekeptonhercourse.Sheshallif
necessarytakeallherwayoffandinanyevent
navigatewithextremecautionuntildangerof
collisionisover.”
TheDutchCouncilofTransportationhasaddedan
amplificationtothisruleforDutchmariners:“During
aperiodofreducedvisibilityunexpectedbehaviorof
othervesselsshouldbeanticipated.Thespeedandthe
correlatedstoppingdistancemustcorrespondwith
thissituation,”(vanDokkum,2016).
ThebigdifferencewiththisruleversusRule15
aboveisthatinrestrictedvisibilitybothvesselsare
suddenlygive‐wayvesselsandtheresponsibilityfor
avoidingacollisionisshared.Theproblemsherefora
quantitativeapproachliesinsofttermslike“safe
speed,”“dueregardtotheprevailingcircumstances
andconditionsofrestrictedvisibility”and“take
avoidingactioninampletime.”Butalsointhe
problemofdefining“restrictedvisibility.”Asa
meteorologicalphenomenon“restricted”isnot
defined,noris“safespeed”,althoughanassumption
mightbethatthevesselshouldbeabletostopwithin
thedistancethatcanbeoverlooked.Anassumption
thatcannotalwaysbefollowedasinmanypartsof
theworldshipsregularlynavigateinconditionsof
visibilitywhereeventheownshipsforecastle(front)
cannotbeseenfromthebridge.
Anotherreflectionisthat“restrictedvisibility”
referstohumanvisibilityoftheeye,whichinthe
autonomouscasecanbetranslatedtothevisibilityof
theday‐lightcameras.Section(d)inRule19which
referstowhenshipsaredetected“byradaralone”

515
wasaddedin1960afteranumberof“radarassisted
accidents”(themostwell‐knownwastheStock‐holm‐
AndreaDoriaaccidentin1956).Anautonomous
vesselwillmostprobably,apartfromday‐light
cameras,AISandradar,alsohaveinfraredcameras
andmaybeLIDAR.Butevenifsensorresourcesonan
autonomousshipcouldbejudgedasbeingbetterthan
thehumaneye,thisrulemakesitnecessarytoinclude
visibilitysensorstodecideifRule19,“restricted
visibility,”ortherules11to18,“conductofvesselsin
sightofeachother,”shouldapply.Aconfounding
factorhere,thatneedstobetakenintoconsideration,
isthatfogoftenappearsinpatchesorbanks,soeven
iftheautonomousshipitselfmaybeinanareaof
goodvisibility,theothervesselmightbehiddenina
fogbank,inwhichcaseRule19apply.Apossible
solutionfortheMASSmightbetocompareradarand
cameraimages.
Aphenomenonworthtakenintoconsiderationis
thatwhileanautonomousvesselwillweighits
differentsensorinputsinanobjectivemanner
resultinginasightingwithaprobabilitymeasure,the
humanoperatoronamanualvesselhasacognitive
systemthatprefervisualegocentricinputthroughthe
eyesascomparedtoexocentricimagesfromradar
andelectronicchartsthatneedstobementallyrotated
tobeaddedtotheinnermentalmap,(Porathe,2006).
Anexampleofthisistheallisionofthecontainer
vesselCoscoBusanin2007withtheSanFrancisco
OaklandBayBridgeinheavyfogbutwithfully
workingradarandGNSS/AISsupport(NTSB,2009).
Thehumancognitivesystemhasotherlimitations
suchase.g.“normalitybias”and“confirmationbi‐
as.”(Poratheetal.,2018).Withthis,togetherwith
otherhumanshortcomingslikefatigue,aninclination
towardsshort‐cuts,andsometimessheerviolations,
theriskisthatthelistofpotentialinteraction
problemsbetweenhumanandmachineguided
navigationwillbelong.
3 QUANTITATIVECOLREGS
Thecodeforacollisionavoidancesoftwarethatisto
coverallpossiblesituationswillhavetobeverylong
andhewouldstillnotsuffice.Theunknown
unknowns,blackswans,wouldkeepappearing.
Fromacomputerprogrammer’spointofview,it
mightseemhelpfulifallqualitative,soft,
enumerationsofCOLREGScouldbequantifiedinto
nauticalmiles,degreesofarcandclockminutesonce
andforall.Thiswouldgreatlyfacilitatethe
developmentofthenecessaryalgorithmsthatwill
governfuturecollisionavoidancesystems.However,
suchaquantifiedregulatorytextwould,inthesame
way,havetobeverylengthyanditwouldstillnot
coverallpossiblesituations.InsteadCOLREGS,like
otherlegaltextwillneedtohaveageneralformatthat
isopentointerpretationsinacourtofmaritimelaw,
andtheoppositeof“theordinarypracticeofseamen,”
i.e.“goodseamanship,”includejuridicaloptionssuch
as“negligence”and“grossnegligence”,(van
Dokkum,2016).Shipstechnicalperformanceand
maneuverability,experienceandtrainingofseamen,
allevolvewithtime,sofortherulesoftheroadtobe
validtheymustbewritteninageneralmanner.
Insteaditisthealgorithmsofcollisionavoidance
applicationsthatneedtobepreciseandquantitative.
ByusingAISdataandlargescalesimulations,
applicationscanbemadetolearnthemosteffective
andefficientwayofmaneuveringindifferent
situations,stillfollowingtheCOLREGS.Itwould
probablybebeneficialifsuchmachinelearningwas
ongoing“lifelong”fortheAI(ArtificialIntelligence)
onthebridge,whichthenwouldbecomemoreand
moreexperiencedthroughtheyears.However,itis
unlikelythattheIMOwouldacceptanAIonthe
bridgewhichwasnotcertifiedandwhobehavedina
preciselypredeterminedwayforaspecificsituation
(evenifthiscouldbedefendedbycomparingtheAI
toatrainedandlicensedthirdmateworkinghisway
upthroughtheranksgainingmoreandmore
experience).
Anotherpointtopayattentiontoisthat,aslongas
therearemanualshipsgovernedbyhumansonthe
sea,theactionsofautonomousshipshastobe
predictableforthesehumans.Autonomous
navigation,supportedbyartificialintelligenceonthe
bridge,hasanumberofadvantagescomparedto
human,manualnavigation:improvedvigilance,
improvedsensingandperception,longerendurance,
anabilitytolookfurtherintothefutureandtokeep
morealternativeoptionsopenduringthedecision
makingprocess.Forinstance,bykeepingtrackofall
shipmovementsonaverylongrangeanAImightbe
abletopredictapossibleclosequarterssituation
severalhoursaheadofahumannavigatorbutmay
thereformakemaneuverswhichmightnotmake
sensetoanOOWonamanualshipinthevicinity.
Therefore,itisofoutmostimportancethat
autonomousshipsarepredictableandtransparentto
humans.
4 AUTOMATIONTRANSPARENCY
4.1 Anthropomorphism
Everyoneofusthatarestrugglingwiththe
complexityofdigitaltoolsknowthattheydonot
alwaysdowhatwewantorassumetheywilldo.
They“think”differentlyfromus.Aninnatetendency
ofhumanpsychologyistoattributehumantraits,
emotions,orintentionstonon‐humanentities.Thisis
calledanthropomorphism.Wedosobecauseitgivesus
asimple(butfaulty)methodto“understand”
machines.However,thechanceisthatifweknow
thatMASSalwayswillfollowCOLREGS,wecan
learntoknowtheirbehaviorandinahumanmanner
beabletounderstandtheirworking.Thisin
oppositiontonormal,mannedships,whereyou
alwayshavetobecautiousofmisunderstandingsor
violations.
4.2 Identificationlight
Inmyopinionitisthereforeimportantthatships
navigationinautonomousmodeshowsomekindof
identificationsignal.Itcouldbean“A”addedtotheir
AISiconinECDISorontheradarscreen.Duringdark
alightsignalcouldbeadded(e.g.apurplemast‐head
all‐aroundlight,seeFig.2).

516
Theassumptionaboveisthatifautonomousships
alwaysfollowCOLREGStheirbehaviorwillbea
hundredpercentpredictable.Butaswehaveseen
above,thismightnotbetrueife.g.thespectrometers
onboardtheautonomousshipdoesnotinterpret
“restrictedvisibility”thesamewaywedo(and
thereforeRule19shouldorshouldnotbeused).

Figure2:Shouldshipsnavigatinginautonomousmode
carryaspecialidentificationlight?Thebehaviorofthe
navigationAImaybedifferentfromthebehaviorofnormal,
mannedships.Thelightcouldbepurplewhichisacolor
thatisnotusedforotherpurposes.Thesamediscussion
andcolorchoiceisdebatedintheautonomouscarindustry.
4.3 Intentions
Anotherimportantissueisunderstandingintentions.
Interpretingtheintentionsofothershipscorrectlyis
imperativetorulefollowing.Anoldaccidentinthe
EnglishChannel1972canserveasanexampleof
whatmisinterpretedintentions(andtherefore
applyingthewrongrules)mayleadto:
TheferrySt.Germain,comingfromDunkirkin
FranceanddestinedforDover,wasturningslowlyto
port,awayfromthestraitwesterlycoursetoDo‐ver.
Insteadhercaptainintendedtotakehersouth‐west,
downontheoutsideoftheTrafficSeparationScheme
(TSS),intheInshoreZone,inordertofindaclearer
placetocrosstheTSSata“rightangle”accordingto
Rule10oftheCOLREGS.ThebulkcarrierAdartewas
headingnortheastuptheTSSto‐wardstheNorthSea.
Thepilotonboardrecognizedtheradartargetasthe
Dunkirk‐Doverferryandassumed,quitewrongly,
thatshewouldcrossaheadofhimandthattherenow
existedariskofcollision(Rule15).Adartewouldthen
bethegive‐wayshipandwasobligedtogive‐wayby
turningtostarboard.AtthesametimeSt.Germain
startedherportturn,thepilotonAdartestartedto
madeaseriesofsmallcoursealternationstostarboard
togiveway(quitecontrarytothe“substantialaction”
requiredbyRule16).ButinsteadSt.Germain
continuedherportturnandthetwoshipscollided.St.
Germainsank,killinganumberofpassengers(Lee&
Parker,2007).
Thisaccidentisretoldtoillustratetheneedto
understandintentionsandthisgoesforbothmanned
andunmannedships.Iftheintentionoftheothership
isnotunderstood,theriskisthatCOLREGwillnot
saveasituation.Itisimportantthatautomationshare
informationaboutitsworkings,itssituation
awarenessanditsintentions.Questionslike:What
doestheautonomousshipknowaboutits
surroundings?Whatothervesselshasbeenobserved
byitssensors?Thesequestionscouldbeansweredby
e.g.alivechartscreenaccessibleon‐linethrougha
webportalbyothervessels,VTS,coastguardetc.See
Figure3.
Basedonitssituationawarenesstheautomation
willmakedecisionsonhowitinterpretstherulesof
collisionavoidance.Itwouldbeabenefitifthe
intentionsofshipscouldbecommunicated,asargued
inPorathe&Brodje(2015).Largeshipsobeyunder
IMO’sSOLASconvention.ASOLASship(asdefined
inMaritimeRulePart21)isanyshiptowhichthe
InternationalConventionfortheSafetyofLifeatSea
(SOLAS)1974applies;namely:apassengership
engagedonaninternationalvoyage,oranon‐
passengershipof500tons’grosstonnageormore
engagedonaninternationalvoyage(IMO,1980).
SOLASshipsmusttransmittheirpositionand
someotherinformationusingAIS.Inaddition,
SOLASshipsareusuallybigandmakegoodradar
targets,whichwillprovideasecondsourceof
information.Furthermore,allSOLASshipmustmake
avoyageplanfromporttoport.Severalpassedand
ongoingprojectsaimatcollectingvoyageplansand
coordinatingshiptrafficforreasonsofsafetyand
efficiency(e.g.EfficienSea,ACCSEAS,MONALISA,
SMARTnavigation,SESAME,andtheSTM
Validationprojects).Theseattemptsinrouteexchange
wouldmakeitpossibleforSOLASships–alsoMASS
‐tocoordinatetheirvoyagesandshowintentionswell
aheadoftimetoavoidenteringintoaclosequarters
situationwheretheCOLREGswillapply.


Fig.3:Exampleofautomationtransparency:Anon‐linechartportalshowingthesituationawarenessoftheautonomous
ship(hereAutomatExpress),whereshethinkssheis,whatothershipsandobjectsshehasobserved,andwhatintentionsshe
hasfortheclosefuture.An“A”isaddedtotheAISsymbolfor“Iamnavigatingautonomously”andtheintendedroute
showncouldalsobevisibleinECDISandradarsofshipsinthevicinity.

517
Routeexchangewouldforinstancealloweach
shiptosendanumberofwaypointsaheadofthe
shipspresentpositionthoughAIStoallshipswithin
radiorange.Allshipscanthenseeotherships
intendedroute(asinFig.3).IntheACCSEASproject
2014asimulatorstudywasmadewith11professional
British,SwedishandDanishbridgeofficers,harbor
masters,pilotsandVTSoperatorswithexperience
fromcomplextrafficinthetestareawhichwasthe
HumberEstuary.Thefeedbackfromtheparticipants
onthebenefitsofshowingintentionswereoverall
positive(Porathe&Brodje,2015).
5 CONCLUSIONS
Ihaveinthisdiscussionpointedatsomechallenges
facingdevelopersofcollisionavoidancesoftware.
Muchofthishastodowiththequalitativenatureof
COLREGSvisaviethequantitativeneedsofreallife
situations.
However,alsotheinteractionbetweentraditional
shipsin“manualmode”isfromtimetotime
problematic.Theintroductionofautonomousships
whichintheirnavigationfollowsamachine
interpretationofCOLREGSmightleadtomanymore
problemsifnotimplementedcarefully.
Itisofgreatimportancethatthemaneuversof
autonomousshipsarepredictabletohumanoperators
onmanualships.TheAIonbordhasapotentialto
becomemuch“smarter”thanhumans,andtobeable
toextrapolatefurtherintothefutureandthereby
behaveinawaythatmightsurprisepeople
(“automationsurprise”).Insteadthesoftwareshould
focusonbehavinginahumanlikemanner.
Suchautomationtransparencymightconsistof
MASSshowingitsnavigationmode(thepurplemast‐
headlight=inautonomousmode),thecontentofits
situationawareness(whichvesselsareobserved–and
therebywhicharenotobserved)anditsintentions.
Intentionscanbesharede.g.usingrouteexchange
technologydevelopedinrecente‐Navigationprojects
likeEfficienSea,ACCSEASandMONALISA.
Onlyifothermarinerscanunderstandthe
workingsofMASS,apeacefulcoexistenceispossible.
ACKNOWLEDGEMENTS
Thisdiscussionpaperbuildsonapreviouspaperpublished
intheCOMPIT2019conference.However,thesolutions
andargumentationhasbeenfurtherdeveloped.The
researchisconductedwithintheSAREPTA(Safety,
autonomy,remotecontrolandoperationsoftransport
systems)projectfundedbytheNorwegianResearch
Council,whichisherebygratefullyacknowledged.
REFERENCES
Crosbie,J.W.2006.LookoutVersusLights:SomeSidelights
ontheDarkHistoryofNavigationLights.Journalof
navigation59(1):1‐7.
vanDokkum,K.2016.TheCOLRGSGuide(6thed.).
Vlissingen,TheNetherlands:DokMarMaritime
Publisher.
IALA,theInternationalAssociationforAidstoNavigation
andLighthouseAuthorities.2008.VTSManual.
https://www.iala‐
aism.org/wiki/dictionary/index.php/Ship_Safety_Zone
[Acc.2019‐02‐10]
IMO,theInternationalMaritimeOrganization.1972.
ConventionontheInternationalRegulationsforPreventing
CollisionsatSea,1972,(COLREGs).London:IMO.
IMO.1980.TheInternationalConventionfortheSafetyofLifeat
Sea(SOLAS)1974.London:IMO.
http://www.imo.org/en/About/Conventions/ListOfConv
entions/Pages/International‐Convention‐for‐the‐Safety‐
of‐Life‐at‐Sea‐(SOLAS),‐1974.aspx[Acc.2019‐02‐10]
IMO.2018.RegulatoryscopingexercisefortheuseofMaritime
AutonomousSurfaceShips(MASS).MSC100/5/6,12
October2018.London:IMO.
Kongsberg.2019.Autonomousshipproject,keyfactsabout
YARABirkeland.https://www.km.kongsberg.com/ks/
web/nokbg0240.nsf/AllWeb/4B8113B707A50A4FC12581
1D00407045?OpenDocument[Acc.2019‐02‐10]
Lee,G.W.U.&Parker,J.2007.ManagingCollisionAvoidance
atSea.London:NauticalInstitute.
NTSB,NationalTransportationSafetyBoard.2009.Allision
ofHongKong‐RegisteredContainershipM/VCoscoBusan
withtheDeltaToweroftheSanFrancisco–OaklandBay
BridgeSanFrancisco,CaliforniaNovember7,2007.
WashingtonDC:AccidentReportNTSB/MAR‐09/0,
PB2009‐91640.
Porathe,T.2006.3‐DNauticalChartsandSafeNavigation.
Vasteras:MalardalenUniversityPress.
Porathe,T.&Brodje,A.2015.HumanFactorAspectsinSea
TrafficManagement.InV.Bertram(ed.),Proceedingsof
the14thInter‐nationalConferenceonComputerApplications
andInformationTechnologyintheMaritimeIndustries
(COMPIT‘15).TechnischeUniversitätHamburg‐
Harburg.
Porathe,T.,Hoem,A.,Rodseth,O.,Fjoltoft,K.&Johnsen,
S.O.2018.Atleastassafeasmannedshipping?
Autonomousshipping,safetyand“humanerror.InS.
Haugenetal.(eds),SafetyandReliability–SafeSocietiesin
aChangingWorld.London:Taylor&FrancisGroup.
StatisticsNorway.2016.Variousindicatorsfromtheoperational
areaofVTScentres.https://www.ssb.no/191461/various‐
indicators‐from‐the‐operational‐area‐of‐vts‐centres[Acc.
2019‐02‐10]
RodsethO.J.&NordahlH.2017.Definitionforautonomous
merchantships.Version1.0,October10.2017.Trondheim:
NorwegianForumforAutonomousShips.
http://nfas.autonomous‐ship.org/resources‐en.html[Acc.
2019‐02‐10].