59
1 INTRODUCTION
Thetransportationamountatsea hasbeenincreasing
allovertheworldanditisexpectedthatships’size
become bigger and marine traffic increases.
Therefore,itisessentialtoevaluatesafetyofmarine
trafficfor the purpose of improvementof efficiency
and safety of marine traffic. Traditionally, visua
l or
radarobservationhasbeenconductedtoinvestigate
actualtrafficflowofships.Thentheobserveddatais
used as base data for the evaluation of the
aforementionedmarinetraffic.Ontheotherhand,it
is becoming possible to observe the actual marine
trafficmore easily andfaster than before,according
to rapid spread of Automa
tic Identification System
(AIS).Goerlandt etal. (Goerlandt 2011)presented a
method for obtaining a realistic input data for the
marine traffic simulation through analysis of AIS
data.
Besides, description of collision avoidance
behaviours of ships are indispensable to simulate a
realistic marine traffic. However, the act
ual actions
for collision avoidance depend on circumstances
where ships are sailing, i.e. ship domains which
never let other ships enter are different between in
harboursandinthecoastalseaarea(e.g.Inoue1994,
Miyake 2015). Therefore, it is important to develop
and implement an algorithm of collision avoidance
corresponding to a ta
rget traffic or target area into
themarinetrafficsimulation.
The authors developed an automated marine
trafficsimulationsystemwithAISdata.Inthispaper,
we propose a series of systematic procedures for
marine traffic simulation including analysing for
collisionavoidancebehavioursusingAISdata.
Firstlytheprocedureforma
rinetrafficsimulation
withAISdataareintroduced.Andtheprocedurefor
analysing collision avoidance behaviours using AIS
data is introduced. Specifically, we proposed a
methodforidentifyingatimewhenagive‐wayship
evadeastand‐onship.Finally,theexaminationofthe
methodisshown.
Procedure for Marine Traffic Simulation with AIS
Data
R.Miyake&J.Fukuto
NationalMaritimeResearchInstitute,Tokyo,Japan
K.Hasegawa
OsakaUniversity,Osaka,Japan
ABSTRACT:Itisessentialtoevaluatesafetyofmarinetrafficfortheimprovementofefficiencyandsafetyof
marine traffic. Spreadof AISmakes observation of actual marine traffic more easily and faster than before.
Besides,descriptionofcollisionavoidancebehavioursofshipsareindispensabletosimulatearealist
icmarine
traffic.Itisimportanttodevelopandimplementanalgorithmofcollisionavoidancecorrespondingtoatarget
trafficortargetareaintothemarinetrafficsimulationbecauseactualactionsforcollisionavoidancedependon
circumstanceswhereshipsaresailing.Theauthorsdevelopedanautomatedmarinetrafficsimulationsystem
withAISdata.Andinthi
spaper,weproposedaseriesofsystematicproceduresformarinetrafficsimulation
includinganalysingforcollisionavoidancebehavioursusingAISdata.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 9
Number 1
March 2015
DOI:10.12716/1001.09.01.07
60
2 MARINETRAFFICSIMULATION
2.1 Generalstepsofmarinetrafficsimulationand
requireddata
General procedures of marine traffic simulation are
composed of the following steps: (e.g. Nakamura
2011)
Step1:observationofactualmarinetraffic;
Step2:descriptionoftheobservedmarinetraffic;
Step3:simulationoftheobserved
marinetraffic;and
Step4:evaluationofthesimulatedmarinetraffic.
Atthefirststep,dataofactualmarinetrafficina
targeted area is acquired through an investigation
based on visual or radar observation, or video
recording (e.g. The Japan Association of Marine
Safety1991).Atthesecondstep,
theacquireddatais
analysedtogetattributesofshipsandthetrafficsuch
asshiptypeandsize.Thentrafficvolumeorgeneral
trafficrouteincludingwaypointsareobtained.Atthe
third step, a series of traffic flow data, which
represents traffic conditions of target area, is
generated using these
attributes. Then the target
trafficissimulatedbasedonthetrafficflowdata.In
this step, it is possible to include the future
prediction or new traffic systems such as traffic
separation schemes according to the purpose of
simulation. At the last step, the results of the
simulation are evaluated
with appropriate means
corresponding to the purposes. For instance,
frequency or position of one to one ship encounter
situationsisexaminedforthepurposeofassessment
oftheeffectofmodificationofashapeofchannel.
2.2 RequireddataobtainedfromAIS
Theauthorshaveproposedmodifiedproceduresfor
marine traffic
simulation using AIS data. We
developed tools for obtaining some attributes from
AISdataandanautomateddatagenerationsystems
for these attributes. Table 1 shows analysable
attributes by traditional procedure with radar
observationandbytheproposedprocedurewithAIS
data, which are the required data for the marine
traffic simulation. The attributes denoted by “yes”
and “no” are available and not available,
respectively.Theattributes denoted by “hard”have
seldom been obtained from visual or radar
observationbecauseitrequires unrealisticamountof
worktoanalysethembyhand.Forthisreason,many
researchers have focused on whole marine
traffic
ratherthanthemovementofindividualshipsintheir
analyses.
The procedure with AIS data makes analyses
easier,fasterandmore accurate thanthe traditional
procedure.This procedure alsomakes itpossible to
analysethedetailof the individualshipmovement,
in addition to the analyses of whole traffic
flow.
Here, it should be noted that AIS data does not
includetheinformationofsmallshipswhereAISare
notinstalled,suchasdomesticshipsoflessthan500
GTandinternationalshipsoflessthan300GT(IMO
2003).
Table1.Comparisonofanalysableattributeswhicharethe
requireddataforthemarinetrafficsimulation
_______________________________________________
ObjectRequiredAvailability
___________________
datawithRadar withAIS
_______________________________________________
ODdata(gate) hard yes
Individual ODdata(port) hard no
shipShip’sspeedhard yes
movement Waypointshard no
CA*behaviours hard yes
_______________________________________________
Shipsizeandtype yes yes
Ship’sspeedyes yes
Distribution Trajectoriesyes yes
oftraffic Trafficvolume yes yes
flowTrafficrouteyes yes
ODtableyes yes
_______________________________________________
*collisionavoidance
3 MODIFIEDPROCEDUREFORMARINE
TRAFFICSIMULATIONWITHAISDATA
Figure1showstheproposedmodifiedprocedureof
marine traffic simulation with AIS data. In this
figure, rectangles with grey shading indicate the
unique process by using AIS data. Rectangles
withoutshadingindicatethecommonprocess,which
arealso included in
the traditional procedures with
visual or radar observation. Rectangles drawn by
heavylineindicatetheautomatedprocesses.
Here, we briefly describe the outline of the
procedure with AIS data for marine traffic
simulation.
3.1 Step1‐GatesettingforODsurvey
Preparation of an origin and a destination survey
(OD
survey) in a targeted area is conducted at
Process 1‐1. Gates, which represent a starting point
andanendpointoftrafficflowaswellasanorigin
and a destination of a ship, should be set at
appropriate places, considering the observed traffic
flowofthetargetedareain
ordertocoverallarterial
trafficroutes.
3.2 Step2‐ODsurveyandmodelling
At the next step, OD survey and modelling for the
trafficsimulationareconducted.AtProcess2‐1,OD
survey is conducted based on the set gates. At
Process2‐2,eachdataofeachship
whichpassedthe
gatebasedontheODsurveyisextractedasaseries
ofpositiondata,i.e.trajectory.Thenthetrajectories,
which enable to reproduce the target traffic for
marinetrafficsimulation,areclassifiedaccording to
the routes defined by origin and destination gates.
Irrelevant data for the purpose of
analysis such as
anchoredshipsisfilteredoutinthisprocess.
61
AIS data
Decode
Extract target area
Gate setting for OD survey
Extract ships passed gates based on OD survey
Analysis of ship type and size
Category of ship grouped by ship type and size
Modelling of traffic route
Navigation data
Evaluation
Extract ships passing gate based on OD survey
Synchronized data of state of individual ship
Time series data of state quantity of encountered two ships
Analysis of collision avoidance behaviours
Development of algorithm for collision avoidance
Modelling of
occurrence frequency
of ships
number of ship
OD table
Ship speed
Traffic route
WP
Step 1
Step 3
Step 4
Step 2
Process 2-2
Collision avoidance behaviours
Process 2-3
Process 2-4
Process 1-1
Process
2-5
Process 2-6
Process 5-2
Process 5-3
Process 5-4
Process 5-5
Analysis object
Whole traffic flow
OD survey
Process 2-1
OD survey
Marine traffic simulation
algorithm for collision
avoidance
Step 5
Process 5-1
Process 5-6
Figure1.ModifiedprocedureofmarinetrafficsimulationwithAISdata
Figure 2 shows the trajectories of all extracted
shipsthatpassedtheorigingateandthedestination
gate. Based on attribute data of the extracted ships
from AIS data, at Processes 2‐3 and 2‐4, ships are
grouped according to ships’ types and sizes. Then
categories of ships are created
for the model. They
shouldbeclassifiedaccordingtotheappropriateship
size and type because the attributes of traffic route
dependonshipsizeandtype.
Through Processes 2‐5 and 2‐6, modelled data
such as traffic routes and occurrence frequency of
ships in the simulation is generated
using the
developed tools. Specifically, at Process 2‐5, the
numbers of ships in the respective categories are
counted,andthenODtableisgenerated.AtProcess
2‐6, average speed of ships or waypoints of traffic
flow are observed andthen a traffic routemodel is
generated for respective categories
of ships’ types
andsizes.
62
3.3 Step3‐SimulationandStep4 ‐Evaluation
Atthethirdstep,basedonthemodelleddatasuchas
ODtableandtrafficroutes,navigation dataofships
for simulation is created. The navigation data
describes simulation conditions of individual ship
suchasappearancetimetothesimulation,waypoints
and
speedandsoon.
Using the navigation data, the marine traffic is
simulated at Step 3 and the simulation results are
evaluated according to the purpose of the study at
Step4.
3.4 Step5‐Analysisforcollisionavoidancebehaviours
Description of collision avoidance behaviours of
ships is indispensable to
simulate a realistic marine
traffic. However, the actual actions for collision
avoidancedependoncircumstanceswhereshipsare
sailing,forexample,ships’domainswhereneverlet
other ships enter are different between in harbours
and in coastal sea areas (e.g. Inoue 1994, Miyake
2015). Therefore it is important to develop and
implement an algorithm of collision avoidance
correspondingtothetargettrafficorthetargetarea
intothemarinetrafficsimulation.
At Step 5, collision avoidance behaviours of
individualshipareanalysed,inordertodevelopthe
algorithm.The detailedprocedureof theanalysis is
described in Chapter4. Inour
other paper (Miyake
2015), ships domains where never let other ships
enter in a coastal sea area were modelled based on
theprocedureofanalysispresentedinthispaper.
Figure2. Trajectories of extracted ship passed the gate
basedontheODsurvey
4 PROCEDUREOFANALYSISFORCOLLISION
AVOIDANCEBEHAVIOURS
4.1 ODsurvey
An origin and destination survey (OD survey) in a
targetedareaisalsoconductedfirstly.
The easiest means to analyse the collision
avoidance behaviours is to focus on one‐to‐one
encountersituationbetweentwoships.Therefore,at
Process1
‐1,gatesaresetinordertoextractshipdata
passingintersectedtworoutes.
Specifically, in the case mentioned in Figure 2,
fourgates,i.e.Gates1,3,5and6asillustratedinthe
figure,weresetatProcess1‐1,inordertoanalysethe
collisionavoidancebehaviours
incrossing situation.
AtProcess5‐1,ODsurveyisconductedbasedonthe
set gates. Then, data of ships passing the two
intersectedroutesareextractedfromwholeAISdata
atProcess 5‐2,e.g. inFigure 2, oneroute is Gates1
and5,andtheotherisGates
3and6.
4.2 Calculationofsynchronizedstatedataofships
Beforeanalysisforcollisionavoidancebehaviourson
one‐to‐oneencountersituationatProcess5‐5,stateof
shipsiscalculatedateverysynchronized10seconds
at Process 5‐3, because it is essential to compare
respectivestate quantities
ofships at thesametime
for analysis of collision avoidance behaviours. The
statesofshipsareposition,headingandspeedandso
on.
Thesynchronizedintervalshouldbesetaccording
to the purpose of study. In this study, the interval
wassetat10second,forthereasonthatships’
dataat
this interval was deemed to include an appropriate
amount of information about variations of rate of
turn or acceleration by altering course or reducing
speed for the purpose of the analysis of collision
avoidancebehaviours.
4.3 Calculationofstatequantitiesandanalysisof
collisionavoidancebehaviours
Fromthe
synchronizedstatedataofindividualship,
time series data of state quantities of encountered
two ships, such as DCPA, TCPA and distance of
them, are calculated at Process 5‐4. In the example,
thetime series datawascalculated in thesituations
satisfying all the following conditions: (1) distance
between give
‐way and stand‐on ships was within
18520m(10nauticalmiles);(2)TCPAwaswithin30
minutes; and (3) DCPA was within 3704 m (2
nauticalmiles).
And then, at Process 5‐5, collision avoidance
behaviourofindividualshipisanalysedbasedonthe
time series data. The detailed
procedure of the
analysisisdescribedinChapter5.
63
Which side does
the stand-on ship cross
forward or backward?
Exploration of time when the stand-on ship
crosses the longitudinal line of give-way ship.
Exploration of time when the stand-on ship
crosses the lateral line of give-way ship.
Is it corresponding
to a bottom value of DCPA
where DCPA starts rapid
increase ?
Rate of turn >= 0.10
deg/sec ?
Start
Time series data of state quantities
of encountered two ships
End
Loop from terminal point of time-series data
Smoothing time-series data
No
Yes
Backward
Forward
T2 - T3 < 30sec ?
No
No
Yes
Yes
Exploration for T1
Exploration for T2
Exploration for T3
starting point of time-series data
Identification of Ts
Is it corresponding to a
rapid change of heading of
the give-way ship ?
Yes
Yes
No
Figure3.Methodforidentificationofatimewhengive‐wayshipstartstoevadeitstargetship
4.4 Developmentofalgorithmforcollisionavoidance
AsastepafterProcess5‐5, an algorithm of collision
avoidanceisdevelopedbasedontheanalysis,andbe
implementedtothetrafficsimulator.
Furthermore,itisessentialtocomplementdataof
smallshipswithvisualorradarobservation,because
AISdatahardly
includesthedataofdomesticshipsof
lessthan500GTandinternationalshipsoflessthan
300GT,asmentionedinSection2.2.
5 ANALYSISFORCOLLISIONAVOIDANCE
5.1 Summaryofmethodforanalysisofcollisionavoidance
behaviours
Inthischapter,wedescribethedetailedprocedurefor
analysingcollision
avoidancebehavioursinProcess 5‐
5.Specifically,thepurpose,here,istoidentifyatime
when give‐way ship starts to evade its target ship
becausedeterminingamomentofinitiatingcollision
avoidance action is critical in order to establish a
collision avoidance algorithm. Figure 3 shows the
method for
identification of a time when give‐way
shipstartstoevadeitstargetship.
64
As described in the previous chapter, the time
series data of state quantities of encountered two
ships is used for the analysis. It is also possible to
distinguish an encounter situation whether a give‐
wayshiptakesacollisionavoidanceactionornot.If
the give‐way ship evades its
target ship, the time
when the give‐way ship starts to evade the other is
identified.
Collision avoidance behaviours are, generally,
understoodasactionsforincreasingDCPAorTCPA
ofencounteredtwoships.Namely,theyarechanging
the course or speed of ship resulting in increasing
DCPA or TCPA. Therefore,
in this method, collision
avoidance behaviours are identified by analysing
DCPAandTCPA.
Tobeprecise,inthismethod,collisionavoidance
behavioursarelimitedtothosemeetingthefollowing
conditions: (1)one‐to‐oneencounter situation
between a give‐way ship and a stand‐on ship; (2)
action which a give
‐way ship changed its course to
increaseDCPA;and(3)a stand‐onshipislocatedon
the starboard bow of the give‐way ship at the time
when both ships encountered under the conditions
specifiedinSection4.3.
Inthismethod,threetimesareexplored,i.e.T1,T2
and
T3,asexplainedinSections5.2to5.4,inorderto
identify the time, Ts, which is defined as the time
whenagive‐wayshipstartstoevadetheother.
Here, an example of crossing situation is
illustrated in Figures 4 and 5. Figure 4 shows a
relative
trajectory of a stand‐on ship to a give‐way
ship.Therelativetrajectoryisplottedonabody‐fixed
coordinatesystem,andtheoriginofcoordinatesisthe
centre of the give ‐way ship. Figure 5 shows a time
series of status data of the give‐way ship in the
situationspecifiedinFigure4.
Give-way ship
Stand-on ship
-15 -10 -5 0 5 10 15
-15
-10
-5
0
5
10
15
X (km)
Y (km)
×
×
×
t = T2
t = T1
t = T3
Figure4. Relative trajectory of stand‐on ship to give‐way
ship
81000 81500 82000 82500 83000 83500
UTC time (sec)
230
220
210
200
4
0
8
12
0
250
500
750
1000
0
1000
-1000
1200
1500
500
-500
-2000
2
6
10
14
Heading (deg)speed (deg)DCPA (m)TCPA (sec)
t = T2
t = T1
t = T3
Figure5.Timeseriesdataofgive‐wayship
Underthesituationexpressedinthesefigures,the
stand‐on ship was crossing from ahead of the give‐
wayshipontherightandthegive ‐wayship altered
its course to the right. After the stand‐on ship had
crossed in front of the give‐way ship, the give‐way
shipreturnedtoitsoriginalcourse.
Atimeseriesdatawasfirstlysmoothedbecauseit
included a lot of insignificant small variations. The
smoothing process is indispensable to explore,
automatically, extreme values of each data such as
heading and DCPA. We smoothed it based on the
method of moving average
of one‐minute data.
Through this process, small variations were
eliminated.ThegraphsinFigure5areplottedbased
onthedataaftersmoothing.
5.2 ExplorationforT1whenstand‐onshipcrosses
longitudinalorlaterallineofgive‐wayship
The time T1 is when the stand‐on ship crossed:
the
lateral line of the give‐way ship from the starboard
sideofthegive‐wayship;orthelongitudinallineof
thegive‐wayshipfromthefrontofthegive‐wayship.
Namely,incasewherethestand‐onshippassesahead
ofthegive‐wayship,T1
isdefinedwhenthestand‐on
ship crosses the longitudinal line of the give‐way
ship.Onthe other hand, incase where thestand‐on
ship passes behind the give‐way ship, T1 is defined
whenthestand‐onshipcrossesthelaterallineofthe
give‐way ship. In
the case where the stand‐on ship
crosses the longitudinal line of the give‐way ship
pluraltimes,T1isdefinedasthetimewhenthestand‐
on ship crossesthe closest point on the longitudinal
lineofthegive‐wayship.Similarly,inthecasewhere
thestand‐on
shipcrossesthelaterallineofthegive‐
wayshipplura l times,T1isdefinedasthetimewhen
the stand‐on ship crosses the closest point on the
laterallineofgive‐wayship.
InencounteredsituationillustratedinFigure4,T1
isthetimecorrespondingtothecrossingpoint
ofthe
relativetrajectoryandtheordinate.
65
5.3 ExplorationforT2whenDCPAincreased
The time T2 is corresponding to a bottom value of
DCPAwhereDCPAstartsrapidincreaseowingtothe
changeoftheheadingofthegive‐wayshipattherate
equal or greater than 0.10 deg/sec, as illustrated in
Figure5. The
time T2comes always earlierthan T1,
because T2 corresponds to the collision avoidance
behaviours taken by the give‐way ship against the
stand‐on ship sailing on the starboard bow of the
give‐way ship. Namely, T2 corresponds to the time
before the stand‐on ship crosses the longitudinal
or
laterallineofthegive‐wayship.
In Figure 4, cross mark corresponding to T2
denotes the relative position of the stand‐on ship to
thegive‐wayshipatT2.
5.4 ExplorationforT3whenheadingofgive–wayship
changed
The time T3 is corresponding to a
rapid change of
headingofthegive‐wayshipasillustratedinFigure
5.
In Figure 4, cross mark corresponding to T3
denotes the relative position of the stand‐on ship to
give‐wayshipatT3.
5.5 IdentificationofTswhengive‐wayshipstartedto
evadethestand
‐onship
Finally,thetimeTs, i.e. the timewhenthegive‐way
ship starts to evade the stand‐on ship, is identified
with T1, T2 and T3. The time Ts becomes T3 in the
casewherethefollowing conditionsaresatisfied:(1)
T1isidentified;(2)T3is
beforeorequaltoT2;and(3)
the difference between T2 and T3 is within 60
seconds.Intheothercases,Tscannotbeidentified.In
otherwords,suchcasesaredistinguishedasthatthe
give‐wayshiptakesnocollisionavoidancebehaviour
tothestand‐onship.
In encountered
situation illustrated in Figures 4
and5,TswereT3.
6 EXAMINATIONOFMETHODFORANALYSIS
OFCOLLISIONAVOIDANCEBEHAVIOURS
6.1 Comparisonofthetimesforstartingcollision
avoidancebehaviours
For examination of the method, we compared the
identified times Ts through the analyses based on
data obtained by both
manual exploration and
automatic exploration with the method. The both
explorations have been conducted using the
individual time series data of state quantities of
encounteredtwoshipsin59encountersituations.
Among 59 encounter situations, no collision
avoidance behaviours were distinguished in two
situations. According to our detailed consideration,
these
situationswerecorrespondingtowhereT2were
not identified. Specifically, in these encounter
situations, DCPA decreased after the give‐way ship
changeditsheading.
Table 2 shows the difference between both
manuallyandautomaticallyidentifiedtimesTsin57
situations.Outof57situations,only7situationswere
actual one‐to
‐one encounter, and the others were
situationswhereagive‐wayshipencounteredplural
ships and it took collision avoidance behaviours
continuously.
Table2. Difference between both manual and automatic
identifiedstartingtimeofcollisionavoidance
_______________________________________________
Differencenumberofencountersituation
__________ ___________
One‐to‐one* nonone‐to‐one**
_______________________________________________
10sec04
10sec110
20sec410
30sec17
30sec<diff.<60sec 14
60sec<=diff.<90sec‐3
90sec<=diff.<120sec‐5
120sec<=diff.<150sec‐0
150sec
<=diff.<180sec‐1
180sec<=diff.<210sec‐0
210sec<=diff.<240sec‐1
240sec<=diff.<270sec‐0
270sec<=diff.<300sec‐0
300sec<=diff.‐5
_______________________________________________
Total750
_______________________________________________
*actualone‐to‐oneencountersituation
**nonone‐to‐oneencountersituation
Here, we examine accuracy of automatic
identification using 30 seconds as the threshold,
takingintoaccounttheinaccuracyofstatequantities
of encountered two ships owing to the following
reasons:(1)AISdataincludesalotofsmallvariations
originally; (2) AIS data might include some
anomalous values; and (3)
interpolation of AIS data
wasnecessarytosynchronizealltimeseriesdata.
In 6 situations among 7 actual one‐to‐one
encounter situations, the differences between the
times Ts, which were manually and automatically
identified, were within 30 seconds. In 31 situations
among 50 non one‐to‐one encounter situations,
the
differenceswerealsowithin30seconds.
Forthesereasons,itcanbesaidthatthismethodis
valid for evaluating collision avoidance behaviours
onthewhole,thoughmoreaccurateanalysismaybe
necessary for evaluating encounter situations in
detail.
6.2 Modificationofmethodforaccuracyimprovement
To conduct more accurate
analysis, identification of
collision avoidance behaviours should be improved,
especiallywithregardtothefollowingpoints:
treatment of decreasing DCPA after collision
avoidancebehaviours;
simultaneouscollisionavoidancebehavioursboth
byastand‐onshipandagive‐wayship;and
encountered situations with multiple ships and
continuous
collisionavoidancebehaviours.
66
Thethirdbulletpointiscausedbythesmoothing
ofthetime‐seriesdata.Namely,manysmallextreme
values of statequantitiesare eliminated through the
smoothingprocess.Therefore,weneedtoconsidera
betterwayofsmoothinginordertoimproveaccuracy
oftheanalyses.
7 CONCLUSIONS
The authors
developed an automated marine traffic
simulationsystemwithAISdata.Andweproposeda
series of systematic procedures for marine traffic
simulation including analysing for collision
avoidancebehavioursusingAISdata.
We showed that some attributes, which are
required data for the marine traffic simulation, are
analysable with AIS data
and that analyses become
moreeasilyandfasterthanbasedonvisualorradar
observation.
And then we briefly introduced the modified
procedure for traffic simulation based on analysis
withAISdata,inwhichweproposedfivesteps.
Besides, we described the detailed procedure for
analysing collision avoidance behaviours with AIS
data. Then we examined the method for analysis of
collision avoidance behaviours. Based on the results
ofexaminationofthemethod,itcanbesaidthatthe
method is valid for evaluating collision avoidance
behaviours on the whole, though more accurate
analysis may be necessary for evaluating encounter
situationsin
detail.
REFERENCES
GoerlandtF.&PenttiK.2011.Trafficsimulationbasedship
collision probability modeling. Reliability Engineering &
SystemSafety.vol.96(1):91‐107
IMO, 2003: Guidelines for the installation of a shipborne
automaticidentificationsystem(AIS),SN/Circ.227
Inoue K. & Usami S. & Shibata T. 1994. Modelling of
Mariners’senseson
MinimumPassingDistancebetween
Ships in Harbour. The Journal of Japan Institute of
Navigation90:297‐306(inJapanese)
MiyakeR.&Hasegawa K.& FukutoJ. 2015: Modellingof
observedshipdomainincoastalseaareabasedonAIS
data,ProceedingsofTransNav2015
NakamuraS.2011.EvaluationofNavigation Safety
bythe
Marine Traffic Predictions and Simulations:
NAVIGATION.PublishedJapanInstituteofNavigation
178:18‐22(inJapanese)
TheJapan Associationof Marine Safety,1991: (provisional
title) “ Investigation of methods for quantitative
evaluationofnavigationenvironment” (inJapanese)