163
1 INTRODUCTION
Poor lookouts, i.e. one of errors in situation
awareness, are pointed out as the major cause of
collisionsofships,throughinvestigationsofcollision
accidents. Investigations have long been carried out
on navigators’ errors that contributed to ship
collisions(Romer2009,Takemoto2007,Corovic2013).
Especially there are ma
ny investigation based on
accidents’reports.
In the second generation human reliability
analysis such as “Cognitive reliability and error
analysis method (CREAM)”, accidents caused by
humanerrorsaredefinedasfinalresultsandthereare
many background factors, so called “contexts”, in
human errors (Hollnagel 1998). In order to evaluate
safety mea
sures for preventing collisions caused by
poor lookouts, it is necessary to understand the
contextsofnavigators’errorsinsituationawareness.
However, the contexts of errors in situation
awareness during ships’ navigation have not been
fullyinvestigated.Inordertounderstandthecontexts
in detail, it is not sufficient to invest
igate only
accident reports and errors not resulted in collision
accidents should be also investigated. For the
investigation of such contexts, bridge simulator
experimentsare effective,for thereason thathuman
activitiescanbeobservedindetail.
The purpose of this study is to point out the
possible significant contexts of errors in situation
awareness.Forthi
spurpose,weanalyzedthehuman
activities obtained by bridge simulator experiments
based on “error mode”, i.e. classes of human
activities.Eightbasicerrormodeshavebeenalready
definedbyHollnagel(Hollnagel1998).Inthisstudy,
navigators’ errors in situation awareness are
identifiedthroughtheanalysis.Thecontextsoferrors
in situation awareness are esti
mated based on the
resultsofdetailedobservationofhumanactivitiesin
particularoftheidentifiederrorcases.
Inthepreviousstudy,oneoftheauthorsproposed
amethod foridentifying navigators’errors in watch
keeping basedon the behavior analysis of results of
bridge simulator experiments, and a situation
Error Detection in the Navigational Watch Based on
the Behavior Analysis of Navigators
C.Nishizaki
NationalMaritimeResearchInstitute,Tokyo,Japan
T.Takemoto
TokyoUniversityofMarineScienceandTechnology,Tokyo,Japan
ABSTRACT: Poor lookouts, i.e. one of errors in situation awareness, are pointed out as the major cause of
collisions of ships, through investigations of collision accidents. In order to evaluate safety measures for
preventing collisions caused by poor lookouts, it is necessary to understand background fact
ors, so called
“contexts”,oferrorsinsituationawarenessregardlessofoccurrenceofcollisions.Thepurposeofthisstudyis
topointoutthepossiblesignificantcontexts,usinganavigator’ssituationawarenessmodel.Asaresult,we
pointoutthatoneofthepossiblesignificantcontextsisaproblemonjudgmentofprioritylevelsofotherships
withregardtoattention.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 9
Number 2
June 2015
DOI:10.12716/1001.09.02.01
164
awareness model was developed to describe the
contexts of those errors (Nishizaki 2010, Yoshimura
2012).Fortheidentification,navigators’errorsshould
be determined, and for the determination of such
errors, it is essential to select ships having risks of
collision correctly. In the previous study (Nishizaki
2010), the risks of
collision had been determined
basedonlyon“DistanceofClosestPointofApproach
(DCPA)”. However, it had become evident that the
selection of ships having risks of collision was not
soundinthepreviousstudy.Thenweanalyzederrors
insituationawarenessagain,usingthedataobtained
in the previous
study, with a new method for
evaluatingrisksofcollision.
Inthisstudy,takingintoaccountthedeficiencyof
the pervious study, ships having high risks of
collision are redefined based on a “subjective risk
assessment index” (Hara 1990, Nakamura 1996) and
navigators’errorsinwatchkeepingareidentified.
2 PROCEDURE
FORANALYSIS
Theprocedureforanalysisusedinthisstudyconsists
of six steps as shown in Figure 1. In the first step,
experimentscenariosaredevelopedforobservationof
navigators’ error in situation awareness. For this
purpose, scenarios under heavy workload are used
fortheexperiments.
Figure1.Procedureforanalysis.
Inthe nextstep, simulator experiments using the
scenario are performed to observe navigators’
activities. Simulator experiments and observation of
navigators’activitiesaredescribedindetailinchapter
4.
In the third step, information on navigators’
situation awareness is collected by the behavior
analysis of navigators’ activities during simulator
experiments.
As
the fourth step, the caution ships and the
attention ships, which are determined in chapter 3,
are judged based on tracking data in simulator
experiments.Here,shipshavinghighrisksofcollision
areredefinedbasedona“subjectiveriskassessment
index”inthisstudy.
In the fifth step, navigators’ errors
in watch
keeping are identified based on the redefined ships
havinghighrisksofcollision.Anevaluationmethod
of navigators’ situation awareness is described in
detailinchapter3.
As the final step, the contexts of these errors are
investigatedusingthenavigators’situationawareness
model.Thissituationawarenessmodelis
describedin
detailinchapter5.
3 EVALUATIONMETHODOFNAVIGATORS’
SITUATIONAWARENESS
Indexesofcollisionrisksinshipnavigationhavebeen
proposedbysomeresearches.Inourpreviousstudy,
we used DCPA as an index of risks of collision to
define whether a ship had high risk of collision,
for
thereasonthatDCPAwasoneofsimpleindexesthat
couldbecalculatedeasily(Nishizaki2010).However,
itisnotnecessarytorequirekeepingconstantwatch
onashipofwhichtheestimatedDCPAbecomeszero
atacertaintime.Therefore,inthisstudy,weuseda
subjective risk assessment
index (Hara 1990,
Nakamura1996)focusing onrateof relative bearing
changeofatargetshiptodefinewhethertheshiphad
riskofcollision.Thesubjectiveriskassessmentindex
was designed based on encounter situations of
multiple ships, while the other indexes, such as the
Collision Judgment threshold value
(CJ‐value)
(Kobayashi1976),theSubjectiveJudgmentvalue(SJ‐
value)(Hara1995),andtheObstacleZonebyTargets
(OZT)(Imazu2002),weredesignedbasedonone‐to‐
oneencountersituations.
Whenarelativebearingofashipdoesnotchange,
navigatorsidentifytheshipasthetargetshiphaving
highrisk
ofcollision.Therefore,itisrationaltousea
relative bearing as an index to identify the ships
having high risks of collision. Furthermore, it is
rationaltouserelativedistanceastheotherindex.In
this study, “caution ship” and “attention ship” are
identifiedbasedontheseindexes.
The
caution ship is defined as the ship that
satisfies all the following conditions (Hara 1990,
Nakamura1996):
R
(1)
where,
istherateofrelativebearingchange,Ris
relative distance,
and
are weighing
coefficients.Whenthetargetshipwillcrosstheown
shipheadinginthefuture,therelativedistanceisnot
morethan1800m,andwhenthetargetshipalready
Step 1
Step 2
Step 3
Step 5
Identification of navigators’ errors
Develo
p
ment of an ex
p
eriment scenario
Judgment of caution ships
and attention ships
Investigation about background factor of navigators’
errors based on a situation awareness model
Behavior analysis of
Navigators
Simulator Experiments
Observation and Recording
of Navigators’ Activities
Recording of ships’
tracking data
Step 6
Step 4
165
crossedtheownshipheading,therelativedistanceis
notmorethan1500m.
The attention ship is defined as the ship that
satisfiesallthefollowingconditions:
RR
(2)
where,
and
are weighing coefficients. When
thetargetshipwillcrosstheownshipheadinginthe
future, the relative distance is not less than 3400 m,
and when the target ship already crossed the own
ship heading, the relative distance is not more than
3700m.
The values of
,
,
and
are set based
ontheexperimentaldataof30,000pointsasdescribed
below.
7.1,105.4
5
7.1,108.5
5
(3)
whenthetargetshipwillcrosstheownshipheading
inthefuture,
7.1,100.3
5
6.1,109.1
5
(4)
when the target ship already crossed the own ship
heading.
The caution ship and the attention ship are
identified according to the above mentioned criteria
byusingtrackingdatainsimulatorexperimentsevery
moment. Navigators’ situation awareness are
evaluated by whether or not navigators recognized
the caution ships and
the attention ships at the
moment. In the evaluation, error is defined as a
navigator’sactivitythatconstantwatchisnotkepton
all“cautionships”and“attentionships”.
4 SIMULATOREXPERIMENTS
4.1 ExperimentalScenarios
Thepurposeofsimulatorexperiments wastoobtain
data on navigators’ activities in watch keeping in
order to collecting information on situation
awareness.Anopen sea wasselectedastheareafor
theexperimentalscenariostoconcentratenavigators’
activitiesonwatchingforships.Inthescenario,there
are six ships in relatively‐narrow sea area to realize
heavyworkloadofsubjects,i.e.navigators,incharge
of
theexperiments.
Ten scenarios were used in the experiments.
Figure 2 shows experimental scenario situations.
Threeshipsareapproachingfromtheportsideofthe
own ship, and another three ships are approaching
from the starboard side of the own ship. The initial
positionsoftheallsixshipsaredistributed
from4to6
nauticalmiles(NM)fromtheownship.Speedofeach
shipisfixeduntilexperimentalscenariosarefinished.
Eachshipisapproachingincollisioncourses(DCPA=
0)at3NMfromtheownshipandtheshipkeepsthe
collisioncoursefor240second.The
crossinganglesof
other ships at 3 NM from the own ship were 40
degrees(P1),60degrees(P2),and80degrees(P3)in
theportside,30degrees(S1),50degrees(S2),and70
degrees(S3)inthestarboardside.Eachscenariohas
different ship movements such as courses to
approaching 3 NM line form the own ship, speeds,
crossinganglesaftercollisioncourses.
The environmental conditions were defined as
follows: the weather was fine; visibility was 4 NM;
and wind, wave and current were none. Subjects
detected each ship by the radar at the beginning of
each experiment, and
each ship became visible
gradually.
Own
OtherShip
(S1)
OtherShip
(S2)
(S3)
OtherShip
OtherShip
(P2)
OtherShip
(P1)
OtherShip
(P3)
StarboardSide
PortSide
000
090
1NM
2NM
3NM
WP2
WP1
240
[sec.]
Figure2.Experimentalscenariosituations.
4.2 ExperimentalConditions
Simulatorexperimentswereperformedinthebridge
simulator of National Maritime Research Institute
(Yoshimura 2007). This bridge simulator has a
behavior analyzing system with observation and
recordingsystemofsubjects’activitiesandvoices.
Sixteen subjects who had enough experiences on
board were chosen for the experiments. During
experiments,
theyperformednavigationalwatchwith
ahelmsmaninthebridge.Theywerekeepingwatch
with the standard navigational equipments, i.e.,
compass, binoculars, radar, chart, electronic chart
display and information system (ECDIS), and air
whistle. Communications with VHF radio were not
utilized.Thesubjectswereinstructedtoperformthree
tasksas
describedbelow.
1 Try to navigate as usual and maintain a safe
navigation.
2 Declaretherelativebearingof the perceived ship
measured by compass when respective ships are
perceivedatfirstbyvisualcontact.
3 Declare when they notice something on the
respectiveships’states.
Theinitialcourseofthe
ownshipwas northand
theownshipwasputonautopilot.Theinitialspeed
of the own ship was 12 knots. All subjects were
instructed to maintain autopilot as long as possible.
However,whentheydeterminedtheneedforaction
to avoid collisions, they could change the course by
orderingtothehelmsmanandthespeedbyoperating
theenginetelegraphbythemselves.
We analyzed the navigators’ activities while the
own ships was keeping the original courses, for the
reasonthatthecrossinganglesofshipswerechanged
166
when the own ship course was changed by the
subjects.Eachexperimentendedinabout30minutes.
5 RESULTS
Figures3and4showexamplesofresultsofanalyses
of the records of observations by subjects. In each
figure, the ordinate indicates identifier of ships
generated during the experiments and
the abscissa
indicates the time from starting of the experimental
scenarios. Gray closed circles on the lines for
respectiveshipsindicatedbytheordinatedenotethe
timeswhensubjectspayattentiontotheships.From
thisclosedcirclesymbols,itisobviousthatthereare
two watch patterns: subject pays
attention to plural
shipsatonce;andsubjectpaysattentiontooneship.
Cross symbols denote the time when the subjects
perceiverespectiveshipsbyvisualcontact,andblack
opencirclesdenotethetimewhenthesubjectsnotice
achangeofrespectiveships’states.
0
1
2
3
4
5
6
7
0
60
120
180
240
300
360
420
480
540
600
660
720
780
840
900
960
1020
1080
1140
1200
Identification of Ships
Time(sec.)
P3
P2
P1
S1
S2
S3
Subjectpaysattentiontorespectiveships
Cautionshiporattentionship
Subjectperceivesrespectiveshipsforthefirsttimeby
visualcontact
Subjectnoticesachangeofrespectiveships’states
Timeofstoppingtheautopilot
Figure 3. Example of results of analyses of records of
observationsbysubjectsinnoerrorcase.
0
1
2
3
4
5
6
7
0
60
120
180
240
300
360
420
480
540
600
660
720
780
840
900
960
1020
1080
1140
1200
Identification of Ships
Time(sec.)
P3
P2
P1
S1
S2
S3
Figure 4. Example of results of analyses of records of
observationsbysubjectsinerrorcase.
Thecautionshipandattentionship are indicated
by open squares with sold line. In the experiment
reportedinFigure3,othershipS1istheshiphaving
highrisk of collision during750 to 950 seconds and
should be watched constantly during this time.
Dashedlineinthefigureshows
thetimeofstopping
autopilotoperation.
Ships S1 and P1 have become caution ships or
attention ships in many cases. Ships S1 and P1
navigated near the own ship or in front of the own
ship.Thismeansthatthedefinitionsofshipshaving
highrisksofcollisionareconsistent
withtheordinary
sense of navigators, taking into account that
navigatorstendtofocus on shipsnearthe ownship
courselineingeneral.Insimulatorexperiments,most
ofsubjectsmanuallymaneuveredtoavoidshipS1or
P1.Therewasnocollisioninalltheexperiments.
Aclassificationofhumanactivities
thatmaylead
toseriousaccidentsiscalled“errormodes”,andeight
basicerrormodeswerealreadydefinedbyHollnagel
asshowninTable1(Hollnagel1998).
Table1.Basicerrormodes(Hollnagel1998).
_______________________________________________
No. GeneraleffectSpecificeffect
_______________________________________________
1 TimingTooearly
Toolate
Omission
_______________________________________________
2 DurationToolong
Tooshort
_______________________________________________
3 ForceToolittle
Toomuch
_______________________________________________
4 Distance/magnitudeToofar
Tooshort
_______________________________________________
5 SpeedToofast
Tooslow
_______________________________________________
6 DirectionWrongdetection
Wrongmovementtype
_______________________________________________
7 ObjectWrongobject
_______________________________________________
8 SequenceOmission
Jumpforward
Jumpbackwards
Reversal
Wrongaction
_______________________________________________
Table2.Resultsoferrordetections.
_______________________________________________
S.ID Num. Num. Num. Num.
* of of ofCS of
Exp. SCS &AS detectionerrors
** *** **** _____________________
DelayofDiscontinuation
attention ofattention
_______________________________________________
I 13 78402
II 12 72801
III 11 66912
IV 11 661503
V 10 60801
VI 10 601502
VII 7 42401
VIII 7 42611
IX 5 30
200
X 3 18100
______________________________________________
Total 89 534 72213
______________________________________________
* Identificationofscenario
** Numberofexperiments
*** Numberofshipsanalyzedinthisstudy
**** Numberofshipswhichwerecautionshipsor
attentionshipsorboth
167
Astheresultsoftheanalysisonsubjects’situation
awareness,twotypesoferrorswereidentified.These
typesoferrorsareclassifiedinto“timing”inthebasic
errormodes.Oneofthetypesoferrorswas“delayof
attention”,i.e. navigator’s attention to aship having
high risk of collision
was late. Another error was
“discontinuation of attention”, i.e. navigator’s
attention to a ship having high risk of collision was
discontinuedafterthefirstperception.Table2shows
numbersoferrorsofrespectivetypesidentifiedinthe
experimentsforrespectivescenarios.Asshowninthe
table,discontinuationsofattentionwere
identifiedin
the experiments on eight scenarios. On the other
hand, delays of attention were identified in the
experiments on two scenarios. We analyzed
navigators’situationawarenessfor534shipsintotal,
and 72 ships became “caution ship” or “attention
ship” or both. The total number of errors identified
was 15.
In this study, the frequency of errors is
defined as the number of errors identified par total
number of caution ships and attention ships.
Therefore,the frequency oferrors isabout 21 %, i.e.
15/72, in all experiments, where subjects are kept
underheavyworkloadsituation.
6 DISCUSSION
6.1 Navigator’s
situationawarenessmodelinwatch
keeping
It is well known that other ships are prioritized by
navigatorswithregardtoattention.Theprioritylevel
of a ship can be estimated based on number of
attentions to the ship per unit time. For example
based on Figures 3 and 4, the number
of attentions
per unit time is the total number of gray closed
circles, cross symbols and black open circles
correspondingtoeachshipwithintheunittime.
Inthepreviousstudy,wejudgedprioritylevelsof
other ships determined by each subject, based on
results of analyses of the records
of observations by
subjects. A surveillance list is a fictitious list for
expressing the priority levels and update of the
surveillance list corresponds to change of priority
level.
A situation awareness model was developed to
describe contexts of errors in situation awareness
(Nishizaki2010,Yoshimura2012).Figure5showsthe
navigators’
situation awareness model. This model
was developed based on Endsley’s situation
awareness model (Endsley 1995). The navigators’
situation awareness model enables to describe the
situationawarenesswhenanavigatorpaysattention
to plural ships simultaneously. In this model, it is
assumed that the surveillance list is continuously
updated depending on
change of the navigators’
situation awareness, which always results in change
of priority levels with regard to attention. For
example, when a new ship is perceived, the ship is
addedinthesurveillancelist.
Figure5.Navigators’situationawarenessmodel.
Yes
No
Yes
No
Yes
No
No
Yes
Yes
No
Is new target
detected?
Is there
surveillance
list?
Collecting information around own ship
Collecting information
to create surveillance
list
Collecting information to
select a target that needs
constant surveillance
Predict target’s movement
Create or Update surveillance list
Select a target that require
concentration of watch?
Collecting information of selected target
& Comprehension of the target states
Projection of target future states
and event
Decision making
Action to avoid accident
Start of watch
ki
Is there the necessity
of continual watch to
the target?
Is there the necessity
of action to avoid
accidents?
Trash Box
168
In this study, priority levels of target ships with
regard to attention were estimated for many cases
including 72 cases with caution ships or attention
ships.
6.2 Contextoferrorsinsituationawareness
The contexts of errors in situation awareness were
investigated, using the navigators’ situation
awareness model, based on
the 15 errors identified
through the analysis. As the results of the
investigation, it was found that, in error cases,
navigators have not paid attention to target ships
when the ships were having high risks of collision,
thoughtheshipswereperceivedbefore.Forexample
inthecaseexpressedin
Figure4,thoughtheattention
shouldbepaidtoshipS1during630to780seconds,
the navigator had paid attention to ships other than
shipS1duringthatperiod.Thus,itcanbeconcluded
thatoneofthepossiblesignificantcontextsoferrors
in situation awareness is a problem on
judgment of
priority levels of ships with regard to attention. In
other words, incompleteness of updating of the
surveillance lists is one of the possible significant
contextsoferrorsinsituationawareness.
Here, we discuss the causes of the problem on
judgment of priority levels under heavy workload
situation. Taking
into account the navigators’
situation awareness model shown in Figure 5, the
following three issues can be enumerated as the
possiblecausesoftheproblem:
lackofacquisitionofinformation;
misjudgeonprospect;and
misjudgeoncontinuationofattention.
It is further observed that only a part of
the
surveillance list has been updated repeatedly in a
certainnumberofcases.Thistypeofupdatingisone
of the types of incomplete updating of the
surveillancelists.
6.3 Conditionsofoccurrenceoferrors
We investigated the features of ships to which
attentionwaspaidatacertaintime,
otherthanships
having high risks of collision at that time. Then, we
foundthefollowingtwofeaturesofsuchships:
shipswithlowerriskswereclosertotheownship
than the ships having high risks of collision
around0.5to1.5NM;and
approachinganglesofships
withlowerriskswere
35to45degreesasfarasbasedontheexperiments
without over‐taking ships, and encounter
situations of the ships continued for long time,
hereapproachinganglezeromeansthatcoursesof
bothshipsarethesame.
Namely, existence of a ship, with lower risk
of
collision, having one of the above mentioned two
features is a possible condition for occurrence of
errorsinsituationawareness.
7 CONCLUSION
The contexts of errors in situation awareness under
heavy workload were investigated based on the
results of analyses of the records of observations,
using navigators’ situation awareness model.
The
conclusionsinthisstudyareshownbelow.
1 We identified two types of errors in situation
awareness, i.e. “delay of attention” and
“discontinuationofattention”.
2 One of the possible significant contexts is a
problem on judgment of priority levels of other
ships.
REFERENCES
Imazu,H.&Koyama,T.1986.TheValuationofaCriterion
for Collision Avoidance Action. Journal of Japan
InstituteofNavigation.Vol.74:117‐123.(inJapanese)
Imazu,H &Fukuto,J&Numano,M.2002. ObstacleZone
byTargetsandItsDisplay.JournalofJapanInstituteof
Navigation.Vol.107:191‐197.
(inJapanese)
Endsley, M. R. 1995. Toward a Theory of Situation
Awareness in Dynamic Systems. Human Factors: The
Journal of the Human Factors & Ergonomics Society.
Vol.37Issue1:32‐64.
Corovic, B. M. & Djurovic, P. 2013. Research of Marine
Accidents through the PRISM of Human Factors.
PROMET‐Traffic
& Transportation. Vol.25 No.4: 369‐
377.
Hara, K. & Nagasawa, A. et al. 1990. The Subjective Risk
AssessmentofShipsCollision.JournalofJapanInstitute
ofNavigation.Vol.83:71‐80.(inJapanese)
Hara, K. & Nakamura, S. 1995. A comprehensive
assessmentsystemforthemaritimetrafficenvironment.
SafetyScience.Vol.19Issue
2‐3:203‐215.
Hollnagel, E. 1998. Cognitive reliability and error analysis
method.Oxford,UK:ElsevierScience.
Kobayashi, H. & Endoh, M. 1976. Analysis of Collision
Avoiding Action of Ship – From Viewpoint of Man‐
Machine‐SystemAnalysis‐.JournalofJapanInstituteof
Navigation.Vol.56:101‐109.(inJapanese)
Nakamura, S. 1996. Establishment of safety evaluation
technique concerning sea traffic. Dissertation of
HiroshimaUniversity:142‐144.
Nishizaki,C&Itoh,Hetal.2010.AStudyonErrorModes
and their causes of Navigational Watch. Proceeding of
Autumn Conference of Japan Society of Naval
ArchitectsandOceanEngineering.No.2010E‐G6
‐7:353‐
356.(inJapanese)
Romer,H.&Petersen,H.J.S.&Haastrup,P.2009.Marine
Accident Frequencies‐Review and Recent Empirical
Results‐.JournalofNavigation.vol.48:410‐424.
Takemoto,T&Sakamoto,Y.etal.2007.TheCharacteristics
of Navigator’s Error in Ships’ Collision Accidents.
Proceedings
of 2007 CIN‐JIN‐KINPR Joint Symposium
(AsiaNavigationConference2007):213‐221.
Yoshimura, K & Hikida, K. et al. 2007. Evaluating the
Mariner’sWorkloadUsingtheBridgeSimulator.10th
IFAC/IFIP/IFORS/IEASymposiumonAnalysis:S1.1.4.
Yoshimura, K. & Nishizaki, C. et al. 2012. A Method for
Quantifying the Risks
of Human Error from
Experiments with the ship Bridge Simulator.
Proceedings of Fifth International Conference on
EmergingTrendsinEngineeringandTechnology:86‐92.
