31
1 INTRODUCTION
Thecrewonboardofavesselisresponsibletoensure
asafeandefficientvoyagetoday.Incontrasttosucha
conventionally manned vessel, however, an
unmanned and autonomous vessel’s navigation
systemhastodecideindependently onhowto react
tounfavourableweatherconditionsandhowtoav
oid
collisions. A respective concept is developed in the
collaborative research project Maritime Unmanned
Navigation through Intelligence in Networks
(MUNIN) using an integrated approach of weather
routeingandcollisionavoidancetoenableunmanned
andautonomousshippingondeep‐searoutes.
It envisions a dry bulk carrier which shall be
operatingunmannedandaut
onomouslyonthedeep‐
sealegsofaporttoportvoyageonly.Thereasonfor
choosing this specific vessel type as a case for this
projectisthatcomparativelylittlemanualcargocare
isrequiredandthatvoyagesareoftentimeslongwith
less tight schedules. These facts favour the
implementationofslowsteamingconceptswhichadd
totheholist
icsustainabilityapproachoftheMUNIN
concept(Rødseth&Burmeister2013).
According to the concept, an on‐board crew
masters the vessel during berthing as well as while
transiting of dense traffic and coastal areas. At all
times,thevesselisalsomonitoredand assistedbya
shore‐ba
sed supervisory entity. Via satellite links,
information exchange is established and interaction
and intervention mechanisms are ensured. As
depictedinFigure1,thevesselitselfiscontrolledbya
newly designed autonomous ship controller
operatingbridgeandenginesystemswhichisassisted
by an adv
anced sensor module, during times of
unmannedoperation(Burmeisteretal.2014b).
This paper provides an overview about the
development of an autonomous navigation system
which is part of MUNIN’s autonomous ship
controller,andthespecificinteractionsbetweenharsh
weatheroperationandcollisionavoidance.Section2
recapitulatesthecornerstonesofma
ritimenavigation
processes,technologyandhazards.Thedevelopment
approachforanunmannednavigationsystemandits
architecture can be found in the 3
rd
section. The
Interaction of Harsh Weather Operation and
Collision Avoidance in Autonomous Navigation
H.‐C.Burmeister,W.C.Bruhn&L.Walther
FraunhoferCenterforMaritimeLogisticsandServices,Hamburg,Germany
ABSTRACT: Taking into account the autonomous navigation system design and today’s state of the art
navigation with regards to weather and collision avoidance this paper presents the architecture of the
integratedapproach,itslinkstoexistingrulesandregulationsandthetestscenarios.Thesedemonstratehow
safeandefficientnavigationofaut
onomousvesselscanbeachievedbyshowingthemoduleʹsinteractionand
validatingthefeasibilityoftheapproach.Theseanalyseswillbebasedonhistoricaltrafficdatasetsaswellas
simulationresults.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 9
Number 1
March 2015
DOI:10.12716/1001.09.01.04
32
system’smodules for weather routeingand collision
avoidance are described in detail in the subsequent
sections4and5andfollowedbytheconclusioninthe
6
th
section.
Figure1.TheMUNINmodules
2 SHIPNAVIGATION
2.1 Thenavigationalprocess
Navigationcanbedescribedastheprocessoractivity
ofaccuratelyascertainingone’spositionandplanning
andfollowing a route (Oxford Dictionaries 2014). In
order to be able to do so, mankind has developed
ever‐more precise instruments such as sextants,
chronometers, gyro compasses
as well as radio and
satellite navigation devices which, along with many
other inventions, lead the way for modern‐day
integrated bridge systems. Especially the vast
technologicaladvancementswhichcouldbeachieved
within the most recent decades have resulted in a
considerabledeclineinbridgemanningrequirements.
Many tasks are
now carried out more effectively,
precisely and reliably by today’s automated bridge
equipmentthaneverbefore.Thetaskoftheofficerof
thewatch(OOW)onpresent‐daymerchantvesselsis
thus to be characterised by his function as a system
integratorandfinaldecision maker (Hetheringtonet
al.2006).
In
governingtheprocess of navigation,theOOW
must continuously consolidate information from
varioussourcesandevaluatethepresentconditionof
thevesselandtheconditionsofitssurrounding.This
subjective perception then provides the basis for
decision‐making in the context of safe and efficient
shiphandling.
In the course
of developing a concept for an
autonomousnavigationsystemfordeep‐seavoyages
within the MUNIN project, the present navigational
processes have been recorded, mapped and adapted
to meet the requirements imposed by the project’s
overall framework. A thorough analysis has been
conducted with regards to technology availability,
information requirements, legal
framework, work
flowandresponsibilities.Basedonthefundamentals
ofpresent‐daymarinenavigation,specifictaskshave
beenidentifiedandclassifiedtobe redesigned.As a
vessel is a complex system, a large number of
interconnected requirements and dependencies have
tobeaccountedforduringtheprocessredesign.This
approach
isusedduetotheintentionofminimizing
the necessary process‐related and legal adjustments
whenintroducinganautonomousnavigationsystem
(Bruhn2013).
2.2 Currenttechnology
State of the art Integrated Navigation Systems (INS)
whicharepresentlyinstalledonmerchantvesselsare
designedincompliancewiththeresolutionMSC.252
(83) adopted
by the International Maritime
Organization(IMO).
Suchasystem’spurposeistoenhancethesafetyof
navigation by providing combined and augmented
functions to avoid geographic, traffic and
environmentalhazardsbyintegrationofinformation
from various source systems (IMO 2007a), such as
Electronic Chart Display and Information System
(ECDIS), radar,
Automated Radar Plotting Aid
(ARPA), Automatic Identification System (AIS),
GlobalNavigationSatelliteSystem(GNSS),speedlog
andecho sounder. This supportsthe bridge team in
effectivelyconductingnavigationaltasks priortoand
throughoutthevoyage.Multifunctionalworkstations
displayintegratedinformationandprovidefunctions
such as route planning and monitoring as
well as
manualand/orautomaticnavigationcontrolfunctions
andan alert management system. Generally, an INS
does not necessarily provide functions which would
not be available on conventional stand‐alone bridge
equipment. Thus, the actual benefit to the OOW is
mainly limited to a holistic user interface approach
which improves
accessibility of functions and
informationofbridgeequipment(Hetherington2006).
The installation of INS technology on board
typically also comprises advanced track control
devices.Incontrasttoconventionalautopilots,which
provide heading control only, these appliances are
capable of guiding vessels along pre‐planned tracks
over ground (Berking & Huth
2010). Depending on
inputdataregardingthevessel’sgeographicposition,
its heading and speed, track control systems are
designedtoleadvessels oneitherstraightorcurved
routes along a sequence of waypoints. Considering
the limitations of manoeuvrability of the vessel,
course changes are conducted automatically as well
as compensation
for drift and leeway. Doing so,
manually set turning radii and turning rates are
observedwithincertaindegreesoftolerance.Further,
track pilot systems must be capable to adapt to
changingsteeringcharacteristicsofthevessel due to
differentweather,speedandloadingconditions(IMO
1998). In conclusion of this review
of existing
navigation technology it must be stated that for the
purpose of autonomous maritime routeing neither
INSnortrackcontroltechnologyissuitedtoaccount
to an adequate degree for external factors such as
weatherandtraffic.
Additionallytotheabovementionednavigational
appliances, meteorological assistance systems are
employed
onalargenumberofvesselstosupportthe
bridge team in the task of passage planning. These
systemsprovideoptimisedrouteswithregardstofuel
efficiency,voyagedurationandtheprinciplesofsafe
navigation under the forecasted environmental
conditionsalongthetrack.Numericalcalculationsare
based on short‐ and medium
term meteorological
dataofupto10daysingrid‐format.Variousnational
and international meteorological services produce
localand/orglobalmaritimeweatherpredictions,also
including predictions of wave and swell conditions.
These forecasts are composed of observations from
33
respective satellites, stations, buoys as well as of
vessels (Bott 2012). Typically, meteorological
assistance systems use several sources to combine
models and improve route optimisation results. The
offeredservicesensureshore‐basedrouteing,all‐time
availabilityanddigitaldataexchange.Additionallyto
environmental forecasts, the optimisation algorithms
also use specifications
of the vessel as input. This
includes constructional data such as the main
particulars as well as variable data such as the
loading conditions. The quality of the route
optimisationverymuchdependsontheconsideration
of all relevant parameters. The most‐widespread
systemsinuseare:
BonVoyage
byAWT,Inc.
SeawarebyStormGeoAS
VVOSbyJeppesen,Inc.
SPOSbyMeteoGroup,Ltd.
Alloftheexaminedsystemsaredesignedasstand‐
aloneapplicationswithoutanyconnectionstoe.g.the
vessel’scargocomputerorsensorsystem.Inclusionof
variabledata such asdraught, trim, wind and
wave
characteristicsisonlypartlypossibleanddependson
additional sensors or manual input. This data is
imperativelyrequiredforcontinuouslycalculatingthe
vessel’s resonances, roll motion and damping and
assessing its seaworthiness. Further, routeing
restrictions due to traffic regulations and/or
geographic constraints are not included in all route
optimisation
systems. A direct import of optimised
routes into the vessel’s ECDIS isn’t always possible
either(Waltheretal.2014).
This analysis of four leading weather routeing
systems concludes that the available capabilities are
presently insufficient for the purpose of unmanned
andautonomousshipping.
2.3 Mainnavigationalhazards
Shipping accidents in general
pose a threat to
maritimesafety andtothemarineenvironment.Thus,
preventionofaccidentsisimperativetoautonomous
navigation systems as much as it is to conventional
bridge teams. Based on maritime accident statistics
andanalyses,twoessentialtaskshavebeenidentified
as key enabler for autonomous deep‐sea
navigation.
Theseare:
Weatherrouteingand
Collisionavoidance.
Studies imply that the yearly number of vessels
lost remains rather steady in a range between app.
110and170inrecentyears.Almosteverysecond of
theselossesistobeaccountedtofoundering,whichis
definedas:“sinkingdue
toroughweather,leaks,breaking
in two, etc., but not due to other categories such as
collisions […]” (Arendt at al. 2010). This leads to the
conclusion that navigation in challenging sea states
poses a major threat to the safety of deep‐sea
shipping.
Analysingtheprimarycausesofvessel
casualties,
only2%offoundering accidentsoriginateincollision
situations.(Pikeetal.2013)Consequencesareusually
less severe but still can result in a constructive or
economic total loss. Further, collisions are ranked
secondhighestcauseofseriousvesselincidentswhich
are not attributed to harsh weather conditions or
machinery
failures following grounding accidents
(Mandryk2011).Butasgroundingistobeconsidered
an unlikely hazard scenario under deep‐sea
conditions, collision avoidance is identified as the
other main task for the autonomous navigation
system. Such systems are expected to operate
alongside with conventionally manned vessels and
thusmustrespect
theInternational Conventiononthe
InternationalRegulationsforPreventingCollisions at
Sea(COLREG)tothesamefullextend.Furthermore,
researchestimatesthat 65‐95%of shipping accidents
are to be attributed to human error. Low safe
manning standards, high workloads and night
watches lead to crew fatigue, increasing the risk of
collision.(Sanquist1992&Rothblumn.d.).Thus,also
misbehaviourornegligenceofothervessels must be
takenintoaccount.
The total number of global maritime collision
situationscanonlybeassessedroughly.Accordingto
e.g.theGermanFederalBureauofMaritimeCasualty
Investigation, collision accidents sum up to annual
numbers
up to 154 cases under its jurisdiction only
(BSU2013).Thus,thenumberofworldwidecollision
cases must be expected to be considerably higher.
Further, the number of near miss situations are not
documentedbutarealsoestimatedtobesignificant.
3 AUTONOMOUSNAVIGATION
3.1 Processredesign
Shiftingfromconventionalto
autonomousnavigation
must be based on a review of manned bridge
procedures.Parasuramanetal.2000proposesafour‐
stagemodelofhumaninformationprocessingwhich
consistsofthefollowingfunctionclasses:
Informationacquisition,
Informationanalysis,
Decisionandactionselectionaswellas
Actionimplementation.
For
the purpose of designing an autonomous
navigation system, the principle level of autonomy
whichshallbeachievedforthesefunctionclassesby
suchkindofsystemmustbedefined.Itistheaimof
the MUNIN project to design a deep‐sea navigation
system which is able to autonomously navigate a
vessel safely and efficiently along a predefined
voyage plan with respect to weather and traffic
conditions. According to the often cited levels of
automation defined by Sheridan 2011, the
autonomous navigation system must at least reach
level 7, which means that it “executes automatically
[and]thennecessarilyinformsthehuman”
toensurethat
thevesselcanstillbesafelyoperatedevenincaseof
possible communication drop‐outs during critical
situations (Rødseth et al. 2013). While this level is
already achieved for the functional class of action
implementation by state‐of‐the‐art track pilots, it is
notthecasefor
theotherfunctions.Yet,shiftingfrom
mannedtounmannedoperationofmerchantvessels
inherits certain elementary challenges. While the
approach towards more autonomous information
acquisition is described in Bruhn et al. 2014, a
thorough analysis of the available information and
34
decision‐making process is necessary to design the
functional framework of the autonomous navigation
system correctly and to determine which processes
requiresuchhighdegreesofautonomy.
Nowadays, the structures on board as well as
connections to shore‐based stakeholders are tailored
for on‐board crews. As a consequence, redesign
of
bridgeprocesseshasbeenapproachedinsuchaway
to ease compliance with existing standards as far as
possible. Being also dedicated to the proposition to
contributeto the enhancementof safety in shipping,
the initial point for the redesign of the navigation
process is to be found in
today’s conventional and
mannedshipping.
Consecutively,theidentifiedandclassifiedbridge
tasksarethenbeingadoptedtomeettherequirements
ofautonomousnavigationandgroupedintoactivities
whichareeitherrelatedto(Bruhn2013):
Voyageplanning,
Look‐out,
Bridgewatch,
Manoeuvring,
Communication,
Administrationor
Emergencies.
Thesegeneralactivitiesarethenparticularisedinto
individual work processes to be assigned to
individual modules of the vessel’s autonomous
navigation system. While some activities, such as
voyage planning and administration will be
transferred to shore‐based entities, others
imperativelymustremainonboard.Thisconcernsall
activities for
in‐situ information gathering and
processing. Under the proposition that an
autonomous navigation system would have reliable
informationabouttheconditionofthevesselandthe
situationinitssurrounding, theprocessesofdeep‐sea
navigationcanbeperformedbythesystem.
3.2 Autonomousnavigationsystemarchitecture
The autonomous navigation system
ensures that the
vessel“followsapredefinedvoyageplan,butwithacertain
degree of freedom to adjust the route in accordance with
legislationandgoodseamanshipautonomously,e.g.,dueto
an arising collision situation or significant weather
changes”(Burmeisteretal.2014b).Hereby,itrelieson
the data provided
by an INS or an even more
advanced separate sensor module, like MUNIN’s
Advanced Sensor Module (ASM) as described in
Bruhnetal.2014.
Theautonomousnavigationsystem itself consists
oftwomainapplicationmodules,whichensuresafety
withregardstothetwomainhazardsasidentifiedin
paragraph2.3,namely
the:
Weatherrouteingmoduleandthe
Collisionavoidancemodule.
The former module ensures safe operation by
avoiding unfavourable weather conditions (strategic
routeing) and reducing negative impacts of
environmentalforces(Waltheretal.2014).Thelatter
ensures operation in compliance with COLREGs
(Burmeister&Bruhn2014).
Theautonomousnavigation
systemdoesgenerally
not execute commands directly, but updates the
waypointlistandusesstate‐of‐the‐arttrackpilotsfor
controlling rudder and engine. However, in certain
criticalsituations,likeoperationsinharshweatheror
closeencountersituations,thetrackpilotisbypassed
anddirectcommandsaregiventoavoid
unnecessary
delays in resolving the critical situations due to
restrictions of the track pilot’s operational envelope.
Figure 2 provides an overview about the system’s
architectureandmoduleinteraction.
Figure2.Autonomousnavigationsystemarchitecture
However, collision avoidance and weather
routeing can’t be resolved independently of each
otherasthismightresultincontrarycommandsand
thus safety‐critical situations. For this reason, the
individualdecisionsmadebythetwomodulesneed
tobebalancedbytheoverallautonomousnavigation
system, before being passed to the
track pilot. The
way this is achieved is dependent on the prevailing
circumstances of the vessel. Of certain interest are
herebysituations, where collisionrisks occur during
harsh weather operations. In case of an immediate
danger by other vessels, the collision avoidance
service consults the operational weather routeing
service,if
certainlimitsmustberespected,butincase
itcan’tresolvethesituation,the weather restrictions
are neglected, as the risk of collision is higher
prioritized in that case. However, if no immediate
danger is present, the collision avoidance service
operatesundertightrestrictionsfromtheoperational
routeingservicetoavoid
replacingtheriskofcollision
byariskoffoundering(seeTable2).
Table2.Prioritisationofcommands
_______________________________________________
Situation Noharshweather Harshweather
_______________________________________________
NoRiskof StrategicWeather Operational
Weather
Collision RouteingRouteing
RiskofNormalCollisionCollisionAvoidance
Collision
AvoidanceControl Controllimitedby
WeatherRouteing
Restrictions
Immediate Manoeuvre‐of‐the‐ Maneuver‐of‐the‐
Dangerlast‐secondcontrol last‐secondcontrol
consultingweather
routeingrestrictions
_______________________________________________
35
4 WEATHERROUTEINGMODULE
4.1 Requirements
The working hypothesis of the autonomous
navigation system is, that it “can autonomously
navigate a ship safely and efficiently along a
predefinedvoyageplanwithrespecttoweatherand
traffic conditions” (Krüger et al. 2014). On manned
vessels,thebridgeteamisrequiredto
ensureaproper
planning before commencing a voyage and to take
any factor into account which potentially might
compromisethesafetyofthevessel(IMO2007b).This
includesenvironmentalfactorssuchaswind,current
andseastate.Typically,thisvoyageplanisproduced
bythenavigationalofficerandverifiedby
themaster.
Yetstill,thefinalrouteingdecisionsaremadeby
the vessel’s master. To bear in mind this key fact is
evenmorecrucialwhenavesselisactuallynavigating
under harsh weather conditions. High wind loads
and rough sea states from unfavourable angles of
encounter can provoke heavy ship
responses.
Resulting rolling and slamming motions can cause
cargo to shift, endanger safe stability and hull
integrity and even lead to capsizing. To assist
navigators in their routeing decisions, IMO has
published principles which represent the guidelines
forwatch‐keepingofficersintermsofhowtoreactto
adverse weather
situations. This document defines
general minimum safety requirements for avoiding
potentially perilous environmental conditions and is
valid for all types of merchant vessels (IMO 2007c).
Thephenomena,whichtheweatherrouteingmodule
ofadeep‐seanavigationsystemistoobviate,are:
Surf‐riding andbroaching‐to,
Reduction of
intact stability when riding a wave
crestamidships,
Synchronousrollingmotionand
Parametricrollmotion.
Development of an autonomous deep‐sea
navigation system in compliance with the IMO
routeing guidelines (IMO 2007c) requires a weather
routeingmodulewiththetwomainobjectivesof:
optimising the voyage plan
based on the vesselʹs
hydrodynamicswithregardtofuelefficiencyfora
givenweatherforecastand
maintaining the effects of sea state and wind on
ship responses (all 6 degrees of freedom) below
definedsafetylevels.
Theseobjectivesalignwiththetesthypothesesfor
development of prototypes in the
context of the
MUNIN project (Krüger 2014). Further, it is of the
utmost importance for an autonomous weather
routeingmoduletobefittothespecificcharacteristics
of the vessel which it serves. This allows to
adequately respect fixed and variable parameters
such as hull form, propulsion system, displacement,
draught
and trim during the continuous navigation
process.
4.2 Approach
Inthepreviousparagraphithasbeenoutlinedthatit
is important to distinguish between strategic and
operational weather routeing. While the former
intendstoavoidunfavourableweatherconditionsin
the first place, the latter one aims to minimize the
negativeeffects
ofsuch.
Strategic routeing decisions are made based on
detailedweatherforecastsfromshore‐basedservices.
Very much like existing services, whose approaches
canbefoundinvariouspublications(Adegeest2008,
Böttner 2007 & Shao et al. 2014), the autonomous
navigation system optimises the vessel’s route with
regardstopassage
duration,arrivaltimeand/orfuel
consumption. Under close consideration of the
vessel’s safety and given routeing restrictions,
specificationsaboutcourseandspeedofthevesselare
givenforeachindividualrouteleg.Thispath‐finding
problem is solved by employing the A* algorithm
(Waltheretal.2014).
In case navigation in
harsh weather conditions
cannot be avoided, the autonomous navigation
system takes suitable measures to minimise the
negative impacts of environmental forces. This is
necessaryasunfavourablewindloads,wavelengths,
waveheights or anglesof encountercan causelarge
roll angles and accelerations, slamming, loss of
stability, shift of cargo
or even capsizing. For an
autonomous navigation system to prevent such an
occurrence,inputofreliablereal‐timedatafromlocal
meteorological observations, made by the before‐
mentioned advanced sensor system are required. A
variety of on‐board sensors provide information
about prevailing wind characteristics, precipitations,
atmospheric pressure and humidity. Sea
state
characteristicsandoceansurfacecurrentareobtained
by processing radar and camera imagery while
motionandstresssensorsinstalledatcriticalpositions
allow to closely monitor induced responses of the
vessel.Thisdataiscollected,evaluatedandprocessed
to produce a thorough perception of the
environmental conditions in the
proximity of the
vessel. As a result, critical areas are identified
accordingtosafety requirementsand also visualised
in a polar plot. In combination with the vessel’s
specific characteristics and routeing restrictions, the
autonomous navigation system determines an
optimisedroutewithinthosesetboundaries.
The architecture of the autonomous navigation
system’s
weatherrouteingmodulecombinesthetasks
of strategic and operational routeing. Figure 3
illustratestherelevantinputandoutputdata.
Figure3.Overviewoftheweatherrouteingmodule’sinputs
andoutputs
36
Forconductingits continuous route optimisation,
the autonomous navigation system requires static
informationaboutthevessel’scharacteristicsandthe
applicable safety requirements. This, in combination
withvoyagerelated information such asthe vessel’s
loading condition, draught, trim, centre of gravity,
metacentric height, ETD, ETA and routeing
restrictions represent the basis
for computing the
responses of the vessel to severe environmental
conditions.Fromitshullgeometryandactualstability
parameters, a hydrodynamic model is derived. This
accounts for resistance of the vessel and allows
calculating its interaction with the surrounding
environment.
As mentioned before, strategic routeing depends
on meteorological forecasts as
input for route
optimisation while operational routeing uses local
real‐timesensordata.Thisdatastreamcanbedivided
into ship conditions and environmental conditions.
Theformerdefinestheposeofthevessel,namelyits
position and rotation and consists of velocities and
accelerations in all six degrees of freedom. The
environmental conditions, however, cover
meteorologicalas well as oceanographic information
ofwhichwavecharacteristicsarethemostsignificant
one. Further, it must of course be mentioned that
routeingrestrictions are not only basedon guidance
given by e.g. IMO but can also be set by a human
operator,ifrequired.
The results of the weather routeing process
conductedbytheautonomousnavigationsystemare
threefold. Firstly, a documentation records the data
received from all sources as well as the computed
outputs. Theses consist secondly of an optimised
voyage plan, containing route waypoints, speed
profiles and estimated fuel consumption from
strategic routeing,
based on the available forecasts.
Thirdly,theoperationalrouteingprovidesimmedia te
measuresincommandingcertainacertaincourseand
speed for the vessel under given harsh weather
conditions.
4.3 Validation
Theimplementationsofthestrategicandoperational
weatherrouteingmodulearetestedby:
running of different scenarios in a
ship handling
simulationenvironmentand
comparingresultswiththoseoccurredtracksand
othertools.
Testingofthestrategicweatherrouteingmoduleis
conductedbyverificationofcalculationsaswellasby
validationofitsrouteoptimisationresults.Tracksare
being calculated from the western entrance of the
EnglishChannelto
NorthAmericaaswellasanother
oneto South America and viceversa. In eithercase,
typically great circle navigation would be used and
notmanydeviationsmustbemadeduetorestrictions
bye.g.landmasses.Comparisonofcalculationresults
for both directions of the same route differ only
insignificantly.
These variations can be attributed to
theweatherforecastdatawhichprovidesmoredetails
for the first 72 hours and allows for the use of data
resolution of 0.25° compared to 1.00° for the
remaining passage time. These optimisation results
are then compared with recorded tracks from that
region. In both
cases, high wind speeds and high
wave height areas are avoided and tailwinds are
favoured. Greater disparities can be found only in
thoseroutelegswheredifferentdataresolutionshave
been used. Thus, the overall results of the strategic
weather routeing module are satisfactorily while
especially implementation of a smoothing
algorithm
presentspotentialforfurtherimprovement.
Astheoperationalweatherrouteingmoduledeals
with the immediate environmental conditions at the
vessel’s position, a test scenario of a much smaller
area is required compared to testing of the strategic
weather routeing module. An eight‐shaped route is
defined for the vessel to
follow to ascertain that the
vesselwillfacevariousanglesofencounterofrough
sea state which is set from easterly directions (see
figure 4). A pre‐defined cross‐track error as well as
minimum and maximum speed profiles ensure that
deviationsfromthevoyageplanarelimited.Asa
test
result,theoutlinedtrackisfollowedratherpreciseon
most route legs, yet the voyage speed is adjusted.
Especially on long route legs on which waves from
abeam are encountered, the vessel cannot avoid
critical areas for para metric rolling only by speed
reduction. When track deviation reaches the
maximum allowance,
the course is altered, leading
thevesseltocrossagainsttheseaalongthetrack.Due
totheeight‐likeshapeofthepre‐definedtrack,most
encounter situations re‐occur and lead to similar
measuresbytheoperationalweatherrouteingmodule
aswellastocomparablebehaviourofthe
vessel.With
regardstothevalidityoftheresults,itmustofcourse
be mentioned that validation of the harsh weather
control in a ship handling simulation environment
onlyallowstotesttheprinciplefunctionalitiesofthe
approach,butdoesnotreplacetheneedofanin‐situ‐
testascorrect
harshweatherresponsesareonlytoa
certain degree covered by ship handling simulators
available.
Figure4.Validationrunofthe operationalweatherrouteing
37
5 COLLISIONAVOIDANCEMODULE
5.1 Requirements
Asfortheweatherrouteingmodule,theautonomous
navigation system’s overall working hypothesis as
laidout in paragraph 4.1 isalso the baseline for the
collisionavoidancemodule.Fromtheconceptdesign
perspective, it is the function of the collision
avoidance module to ensure
the latter requirement.
The fundamental regulative framework for collision
avoidanceonthehigh‐seasislaiddowninCOLREG,
whichaccordingtorule1is applicable toall vessels
on the high seas. Hereby, the steering and sailing
rules defined in Part B of COLREG are of special
importancefor
theautonomousnavigationsystem,as
it defines the obligations and correct collision
avoidance measures of any vessel (IMO 1972). In
general,theprocessofcollisionavoidancecanbesplit
upintothetwotasks(Froese&Mathes1995):
Analyseactualtrafficsituationand
DetermineCOLREG‐conformcountermeasures.
While
theformer task alsoincludes requirements
fromSTCWandSOLASbesidesCOLREG,especially
tomonitorandevaluateothervessels’behaviour,the
lattermustmainlycomplywithpartBaslaidoutin
Bruhn 2013. Thus, from a legal requirement
perspective, the two main tasks that the collision
avoidance module of
the autonomous navigation
systemmustconductareto:
identify the COLREG obligation of the vessel
towards all objects in the vicinity in unrestricted
watersand
calculate possible, COLREG compliant deviation
measures for a given traffic situation in
unrestricted waters that shall minimise the
necessarytrackdeviation.
5.2 Approach
Automated
collisionavoidanceaccordingtoCOLREG
isnotauniquetopic,buthasbeencoveredbyseveral
authorsbefore.E.g.Kreutzmannetal.2013provided
a formalisation system to enable a machine to
determine COLREG situations correctly and
exemplarily implemented this for rule 12. Other
approacheshavebeenmadebyZeng
2000orPereraet
al. 2009 which used genetic algorithms and fuzzy
logic respectively to determine the give‐way
obligationfortwo‐vesselsituations.Incontrast,Liuet
al. 2006 & Xue et al. 2008 do also provide counter
measures, in the case of Xue et al. 2008 even for
multiple
encounters.However,mostoftheseconcepts
require perfect information and are only covering
certain rules of COLREG. Additionally, further
restrictions e.g. due to the prevailing weather
circumstancesarenottakenintoaccount.
As outlined in paragraph 3.2, the autonomous
navigation system is not gathering traffic or
environmental data itself but
relies on a separate
system to provide this information. Nevertheless,
evenifthislook‐outobligationaccordingtoCOLREG
rule5 isoutsourced to that system, the autonomous
navigationsystemstillneedstodealwithincomplete
data provisions to be applicable in real world
applications. Hereby, the autonomous navigation
system distinguishes
three different kinds of data
qualities as outlined in Table 3 (Burmeister et al.
2014a).
Table3.Availableobjectdatapercategory
_______________________________________________
AvailabledataDetected Classified Identified
_______________________________________________
Positionxxx
Speedovergroundxxx
Courseovergroundxxx
Headingxxx
Bearingxxx
Rateofturnxxx
CPAxxx
TCPAxxx
Objecttype‐xx
MMSI‐number‐‐x
Shiptype‐‐x
Navigationalstatus‐‐x
_______________________________________________
Incaseavesselisappearingandariskofcollision
is developing according to COLREG rule 7 or any
furtherdefinedcriteria,likeathresholdvalueforthe
closest point of approach (CPA), the autonomous
navigationsystemdeterminesitsobligationsbasedon
the relative bearing RB
OSTS from the vessel to the
object, the relative bearing RB
TSOS vice‐versa, the
prevailing visibility and the navigational statuses of
bothvessels
1
.Afterwards,theautonomousnavigation
systemdetermineswhetherCOLREGrule13,14,15,
18 or 19 applies and reasons the own vessels’
obligationswhichmayeitherbe:
COLREGrule16Give‐wayvesseland
COLREGrule17Stand‐onvessel.
Whileitisrequiredforagive‐wayvessel
tomake
a large enough alteration of course and speed in
ample time to ensure a safe passing distance, it is
required for the stand‐on vessel to maintain speed
and course. However, stand‐on obligations are
commonlyonlyfulfilledinatwovesselsituationfor
vessels in sight, as
in case of multiple vessel
situations,itmightotherwiseoccurthatthevesselis
give‐way and stand‐on vessel at the same time and
thustheseopposedrequirementsgenerateadeadlock
situationresultinginacollision(Cockcroftetal.2012).
In case of being a give‐way vessel, the autonomous
navigation system algorithm adds waypoints to the
route plan taking hydrodynamic and environmental
restrictions into account, like e.g. actual turning
diametersandweatherrestrictionsasperIMO2007b.
Those are then forwarded to the track pilot for
execution(seeFigure4).
In case of incomplete data sets, assumptions for
the
missingdataaremade.Thisisespeciallyrelevant
for a missing navigational status, which today is
mainly taken from AIS data even though COLREG
PartConlyrequireslightandshapesignalstodisplay
the navigational status (IMO 1972). However, in
several situations the give‐way obligation is over‐
determinated, meaning
that irrespectively of the
navigational status the actual situation would result
1
Asnarrowchannelsandtrafficseparationarerareduring
deep‐seapassage,thesespecialcasesarenotfurtherde‐
tailedinthispaper
38
in a give‐way obligation of the vessel by a
subordinate rule as well, thus an assumption of the
navigational status is not necessarily increasing the
uncertainty(Burmeister&Bruhn2014).
Figure4.Decisiontreeforthecollisionavoidancemodule
Additionally, the autonomous navigation system
also monitors the immediate danger criteria and
initiatesthemanoeuvreofthelastsecondasrequired
by rule 17. Hereby, the autonomous navigation
systemuseshydrodynamictrackpredictionsbasedon
fasttimesimulationmethodsasdescribedinBenedict
etal.2014todeterminethefinalmoment
untilwhich
thevesselitselfcanresolvethecollisionsituation.Of
course,accordingtoCOLREGrule17themomentthe
give‐way vessel alone can no longer avoid the
collision shall be used as an criteria, but as this is
difficult to predict,one may also use this moment
instead
(Cockcroft et al. 2012). For executing the
manoeuvre, the track pilot is bypassed by the
autonomousnavigationsystemanddirectcommands
are send to engine and rudder control. Hereby, the
two derivatives of Bow Crossing Distance (BC) and
CPAareimportant:
TT
TT
T
d BC α , v
BC α , v
dα
(1)
'
TT
'
TT
T
d CPA α , v
CPA α , v
dα
(2)
with
TascourseovergroundandvTasspeedover
ground. Depending on the type of encounter
situation, the algebraic sign of BC’ or CPA’
determineswhatruddercommandisgiven. Asimilar
approach will be taken regarding the engine
command. However, in case of harsh weather, a
subordinatesmoothingproceduretriestoreducethe
riskof
foundering.Afterinitiation,thedirectcontrol
isexecuteduntilthesituationisresolved.
Alongwiththe normalrisk of collision handling,
the immediate danger handling has been
implemented within the MUNIN ship handling
simulation test‐bed environment. Besides the
described two‐vessel‐situation‐handling, this also
includes additional routines to ensure
multiple‐
encounter‐handling.
5.3 Validation
Validating the feasibility of the collision avoidance
moduleisathreefoldactivity:
Internallegallogictest,
Historicaldatatestand
Real‐timeshiphandlingsimulationtest.
For the logic test, random traffic situations have
been generated and are executed with the
autonomous
navigation system being active, but
usingasimplifiedshipmodel.Figure5displaysone
of the resulting plots, including a simple logbook
when and how the collision avoidance module has
made certain decision with regards to determining
COLREG obligations and counter measures. The
resulting plots and logs are afterwards analysed by
experiencednauticalexpertswithregardstoCOLREG
compliance. The main objective is to test if general
situationsarehandledcorrectly.
Within the historical data test, AIS records are
used to re‐run real situations and compare the
autonomous navigation system’s outcome with the
documented behaviour of conventional vessels.
Again,it
usessimplifiedshipmodelsforthere‐runas
no hydrodynamic ship data is available from AIS
records.WithinMUNIN, a two‐month data‐setfrom
vesseltrafficbetweenBarcelonaandMallorcaisused
for this analysis including 4000 different tracks and
above 100 close quarter situations with a CPA less
than1nauticalmile.Theareaunderinvestigationhas
notrafficrestrictionsandwaterdepthsofmorethan
100 metres so that navigational conditions are
comparable to a trans‐ocean voyage. The main
objective of this test is to directly compare, if the
autonomous navigation system would result in
similar
decisions and tracks like a manned vessel to
see how well it can be integrated into existing
shipping practice. Additionally, handling of multi‐
encountersituationsisevaluated.
Finally, the autonomous navigation system is
appliedwithintheMUNINshiphandlingsimulation‐
based test‐bed. Here, it controls a complete
hydrodynamic ship model
travelling in different
weather situations. This set‐up allows testing of
interaction between weather routeing and collision
avoidance module as well as all hydrodynamic
specificdecisions.
39
Figure5.Testplotofcollisionavoidancemodule
6 CONCLUSION
Thispaperhaspresentedanintegratedconceptfora
maritime autonomous navigation system with
regards to weather routeing and collision avoidance
as well as a validation approach. From the first
internaltestsitcanbeconcludedthatthedeveloped
operational weather routeing module performs
according to its purpose. Nevertheless,
it must be
mentionedthatduetotherathergeneralcharacterof
the IMO guidelines (IMO 2007c) potentially
hazardous situations might develop even under
conditionswhichwouldbeconsideredsafe.Thus,the
developed tool represents a good basis but also
requires further enhancement. With regards to the
collisionavoidancemodule,
thefirsttestsresultedin
acceptable situation handling, given the fact that
COLREG itself allows for a certain degree of
interpretation.
Besides its application on a potential unmanned
ship, the autonomous navigation system also
providesagenericapproachwhichcanbeaugmented
to a nautical assistance or even monitoring tool to
improve situational awareness of the officer of the
watch.Thus,itholdsthepotentialtofurtherdecrease
humanerrorsandtoimprovesafetyatseas.
ACKNOWLEDGEMENTS
The research leading to these results has received
funding from the European Union’s Seventh
Framework Programme under the agreement SCP2‐
GA‐2012‐314286.
REFERENCES
Arendt, F. et al. 2010. Shipping statistics yearbook 2010:
Shippingmarket,shipbuilding,portsandseacanals.Bremen,
ISL
Benedict, K.Kirchhoff, M. Gluch, M. Fischer, S. Schaub,
M. Baldauf, M. Klaes, S. 2014. Simulation Augmented
ManoeuvringDesignandMonitoring –aNewMethod
for Advanced Ship Handling. In TransNav, the
InternationalJournalonMarineNavigationandSafetyofSea
TransportationVolume2(3):131‐141.
Berking, B. Huth, W. (eds) 2010. Handbuch Nautik.
Navigatorische Schiffsführung. Hamburg: DVV Media
Group
Bott A. 2012. Synoptische Meteorologie. Methoden der
Wetteranalyse und –prognose.
Heidelberg/Dordrecht/London/NewYork:Springer
Brown, A.J. 2002. Collision scenarios and probabilistic
collision damage.
In Marine Structures Volume 15: 335‐
364.
Bruhn, W.‐C. (ed.) 2013. MUNIN Deliverable D5.2:Process
Map for autonomous Navigation (Available from
MUNINprojectonrequest).
Bruhn,W.‐C.Burmeister, H.C.Long,M.T. Moræus,J.A.
2014. Conducting look‐out on an unmanned vessel:
Introduction to the advanced sensor module for
MUNIN’sautonomous
drybulkcarrier.InProceedingsof
International Symposium Information on Ships—ISIS 2014:
04‐05.
BSH2011.INT1Symbols,Abbreviations,TermsusedonCharts.
Hamburg/Rostock: Bundesamt für Seeschifffahrt und
Hydrographie.
BSU 2013. Annual Report 2013 (Online) Available:
http://www.bsu‐
bund.de/SharedDocs/pdf/DE/Jahresstatistik/Jahresbericht_201
3.pdf?__blob=publicationFile[Accessed13.01.2015]
Burmeister, H. ‐C. Bruhn, W.C 2014a. Designing
an
autonomous collision avoidance controller respecting
COLREG S. In Ehlers, B.E. Asbjørnslett, Ø.J. Rødseth,
T.E.Berg(eds),Maritime‐PortTechnologyandDevelopment
2015:83‐88,London:Taylor&Francis Group.
Burmeister, H.C. Bruhn, W.C. Rødseth, T.E. Porathe, T.
2014b.AutonomousUnmannedMerchantVesselandits
Contributiontowardsthee‐Navigation
Implementation:
The MUNIN Perspective. In International Journal of e‐
NavigationandMaritimeEconomyVolume1:01‐13.
Cockcroft,A.N.Lameijer,J.N.F.2012.AGuidetotheCollision
Avoidance Rules International Regulations for Preventing
Collisions at Sea (Seventh Edition). Oxford: Butterworth‐
Heinemann
Hetherington, C.Flin, R. Mearns, K. 2006. Safety in
shipping: The human element. InJournal of Safety
ResearchVolume37:401‐411.
IMO1972.ConventionontheInternational Regulations for
PreventingCollisionsatSea(COLREG)
IMO 1998. Resolution MSC.74(69) Annex2. Adoption of new
andamendedperformancestandardsfortrackcontrolsystems.
London:InternationalMaritimeOrganization
IMO 2007a. Resolution MSC.252(83).
Adoption of the revised
performance standards for integrated navigation systems
(INS).London:InternationalMaritimeOrganization
IMO 2007b. SOLAS‐Chapter V‐Safety of Navigation‐
Regulation34–SafeNavigationandAvoidanceofdangerous
situations.London:InternationalMaritimeOrganization
IMO 2007c. MSC.1/Circ.1228 Revised Guidance to the
Master for avoiding dangerous situations in
adverse
weather and sea conditions. London: International
MaritimeOrganization
Kreutzmann,A.Wolter,D.Dylla,F.Lee,J.H.2013.Towards
Safe Navigation Rules. In TransNav, the International
Journal on Marine Navigation and Safety of Sea
TransportationVolume7(2):161‐168
Krüger, C.‐M. (ed.) 2014. MUNIN Deliverable D.8.1: Test
environment set‐
up description (Available from
MUNINprojectonrequest).
40
Liu,Y.Du,X.Yang,S.2006.TheDesignofaFuzzy‐Neural
Network for Ship Collision Avoidance. In Advances in
Machine Learning and Cybernetics Lecture Notes in
Computer Science Volume 3930: 804‐812. Berlin/
Heidelberg:Springer
Mandryk, W. 2011. Marine Casualty Profiles. (Online)
Available:
www.imsf.info/documents/2011/Wally%20Mandryk
%20‐%20Lloyds%20List%20Intelligence%20‐%20
Shipping%20Casualty%20Profiles.pdf
. [Accessed
12.01.2015].
Oxford Dictionaries 2014. [Online]. Available:
http://www.oxforddictionaries.com/definition/english/n
avigation.[Accessed14.10.2014].
Parasuraman, R. Sheridan, T.B. Wickens, C.D. 2000. A
ModelforTypesandLevelsofHumanInteractionwith
Automation. In IEEE Transactions on Systems, Man and
Cybernetics, Part A: Systems and Humans Volume 30(3):
286‐297
Perera, L.P. Carvalho, J.P.
Guedes Soares, C. 2009.
Autonomous guidance and navigation based on the
COLREGsRulesandregulationsofcollisionavoidance.
In Guedes Soares C. & Parunov J. (eds), Proc. Of the
International Workshop “Adavanced Ship Design for
PollutionPrevention”:205‐216.London:Taylor&Francis
Group
Pike,K.etal.2013. 15
YearsofShippingAccidents:Areviewfor
WWF.Southampton:SouthamptonSolentUniversity
Rødseth,Ø.J.Burmeister,H.‐C.2012.Developmentstoward
the unmanned ship, In Proceedings of International
Symposium Information on Ships – ISIS 201. Hamburg,
Germany,August30‐31,2012.
Rødseth, Ø.J. Kvamstad, B. Porathe, T. Burmeister, H.‐C.
2013. Communication Architecture for an Unmanned
Merchant Ship. OCEANS‐Bergen, 2013 MTS/IEEE.
http://dx.doi.org/10.1109/OCEANS‐Bergen.2013.6608075.
[Accessed07.2014].
Rothblum, A.M. (n.d.). Human Error and Marine Safety.
USCG. http://www.bowles‐langley.com/wp‐
content/files_mf/humanerrorandmarinesafety26.pdf
.[Accessed28.02.2013].
Sanquist, T. F. 1992. Human Factors in Maritime
Applications: A New Opportunity for Multi‐Modal
Transportation Research. In Proceedings of
the Human
Factors 36th Annual Meeting‐1992: 1123‐1127. Sage
Publications.
Sheridan, T.B. 2011. Adaptive Automation, Level of
Automation,AllocationAuthority,SupervisoryControl,
and Adaptive Control: Distinctions and Modes of
Adaptation. In IEEE Transactions on Systems, Man and
Cybernetics, Part A: Systems and Humans Volume 41(4):
662–667.
Walther,
L. Burmeister, H.‐C. Bruhn, W. 2014. Safe and
efficient autonomous navigation with regards to
weather.InBertramV.(ed.),13thInternationalConference
onComputerandIT‐ApplicationsintheMaritimeIndustries
Redworth, 12‐14 May 2014: 303‐317. Hamburg:
TechnischeUniversitätHamburg‐Harburg.
Xue, Y. Lee, B.S. Han, D.
2008. Automatic collision
avoidance of ships. In Journal of Engineering for the
MaritimeEnvironmentVolume223(1):33‐46.
Zeng,X.Masanori,I.Shimizu,E.2000.CollisionAvoidance
ofMovingObstaclesforShipwithGeneticAlgorithm.In
6
th
International Workshop on Advanced Motion Control :
513‐518.
