Cognitive Demands of Collision Avoidance in Simulated

8/13/2019 Cognitive Demands of Collision Avoidance in Simulated

1/14

INTRODUCTION

There is now a substantial body of researchon patterns of workload and decrement in com-plex real-life tasks, such as driving, aviation,process control, and other advanced industrialsystems (see Wickens & Hollands, 1999). How-ever, with a few exceptions (e.g., Schuffel, Boer,& van Breda, 1989), no systematic studies havebeen carried out within the maritime domain.

This is particularly relevant at the present time,in view of both the considerable increase in thesize and density of maritime traffic and devel-opments toward the integration of ship controlfunctions into a highly automated ships oper-ation center. With the growth of traffic conges-tion, navigational demands have necessarilyincreased, as has the need for effective naviga-tional decision aids (Grabowski, 1990). Success-ful implementation of such systems will dependon knowledge of how navigational demandsaffect workload and performance.

Mariners are exposed to an increasing num-ber and diversity of supervisory and decisiontasks (Dyer-Smith, 1992; Grocott, 1992; Lee& Sanquist, 1996), needing to divide attentionbetween primary navigation displays and sec-ondary tasks such as engine and cargo functions.Efficient scheduling of these tasks may be criticalfor situation appraisal and reducing the poten-tial for accidents (Sanderson, 1989). Accidentstatistics and the analysis of critical incidents

and near misses in the maritime field providestrong evidence for the role of high workload.Although several simulation studies of naviga-tion workload have been carried out, few haveconsidered performance under stressful, uncer-tain, or emergency conditions. One study using afull mission simulator (Sablowski, 1989) foundno effects on navigation of various emergencies(either a collision threat or rudder failure dur-ing the final phase of a 2-hr scenario), althoughratings of workload were higher for emergencyscenarios.

Cognitive Demands of Collision Avoidance in SimulatedShip Control

G. Robert J. Hockey, University of Leeds, Leeds, U.K., Alex Healey and Martin Craw-shaw, University of Hull, Hull, U.K., David G. Wastell, University of Manchester Instituteof Science and Technology, Manchester, U.K., and Jrgen Sauer, Darmstadt University ofTechnology, Darmstadt, Germany

The study examines the cognitive demands of collision avoidance under a rangeof maritime scenarios. Operators used a PC-based radar simulator to navigate setcourses over 100 6-min trials varying in collision threat and traffic density. Cor-rective maneuvers were made through the application of standard navigation rules

and by using two decision aids (target acquisition and test maneuver). Resultsshowed widespread effects of collision threat in terms of decision aid use, subjec-tive workload, and secondary task performance. Most notably, demand increasedmarkedly over the course of emergency trials, in which collision threat resulted fromrule violation by target vessels. The findings are discussed in terms of the com-parison between predictable demands (requiring standard course changes) andthose involving uncertainty about the others intentions (involving more intensivemonitoring and forced delays in corrective action). The study has relevance for thedesign of collision avoidance systems, specifically for the use of ecological displays.

Address correspondence to G. R. J. Hockey, School of Psychology, University of Leeds, LS2 9JT, UK; [email protected]. HUMAN FACTORS, Vol. 45, No. 2, Summer 2003, pp. 252265. Copyright 2003, Human Factors andErgonomics Society. All rights reserved.


2/14

COLLISION AVOIDANCE 253

It has been reported that approximately 90%of all marine accidents occur in confined waterssuch as channels, fairways, and inshore trafficzones (Cockroft, 1984). These are complex nav-

igational environments in which threats and un-certainty are high and options for action severelyconstrained. One issue contributing to suchproblems is likely to be poor technologicaldesign. Vicente and Rasmussen (1992) notedthat process operators sometimes act as if dis-plays completely describe underlying processes,rather than providing only a partial description,and that this sometimes leads to unanticipatedvariations in the workload and behavior of hu-

man operators. Although periods of low work-load may be less demanding, high-workloadactivities may be more difficult to manage (Raby& Lee, 2001). Lee and Sanquist (1996) ob-served that although a collision avoidance sys-tem was able to monitor increased numbers ofvessels and reduce computational load on theoperator, it also increased the need for interpre-tative skills and knowledge of various predictorfunctions.

To prepare for the study, we conducted aseries of interviews with experienced mariners,using the critical incident technique (Kirwan &Ainsworth, 1992) to inform our understandingof the sources of cognitive demand in collisionavoidance (CA) situations. This showed thatproblems in interpreting intentions, or predictingthe actions, of other vessels was one of the mostcommon sources of near-miss incidents. This iswell established as a problem for mariners inpotential collision encounters (Young & Bell,

1993) and is recognized as major source ofdemand in driving (Zeitlin, 1995). In both cases,effects of uncertainty are exacerbated by theoccurrence of rule violations by others usingthe shared traffic space. Perrow (1984) report-ed data showing that 56% of major maritimecollisions include violations of rules of theroad as a contributory cause. Access to infor-mation about the intentions of other vessels maybe the most important requirement for an oper-

ators effective appraisal of situations (Hammer& Hara, 1990) and remains a major obstacleto the introduction of expert systems (Hobday,Rhoden, & Jones, 1993).

In nautical environments, uncertainty is af-fected by both the physical environment (wind,

currents, sandbanks) and the navigation be-havior of other vessels. Collision encountersare a particularly interesting source of informa-tion about maritime demands. Although deci-

sions about what action to take are based onstandard, internationally agreed rules of theroad (International Maritime Organisation,1972), the dynamic and interpretative natureof the task means that uncertainty and unpre-dictability are common features in the erroneousor risky judgments made in many shipping acci-dents (James, 1994; Perrow, 1984; Wagenaar& Groenweg, 1987).

Our use of uncertainty within CA contexts

is closely related to Woodss (1988) analysis.Woods defined uncertainty as an intrinsic featureof complex systems, associated with unavail-ability or ambiguity of data and reduced pre-dictability of future states. In the application of

Woodss analysis to rule violation by a targetvessel, we argue that cognitive demands are like-ly to be increased because of the loss of reliableinformation for implementing evasive actionsas the scenario develops. In Hutchinss (1995)account of navigational computation, uncer-tainty would therefore be expected to prolongand disturb the familiar fix cycle (p. 133).This will have several effects on navigators,given that they (a) need to consider a greaterrange of potential interpretations of the situa-tion; (b) cannot complete necessary information-processing activities at appropriate times; and(c) are unable to implement routine proceduresand familiar responses when actions do needto be carried out.

There is also a clear link with Rasmussens(1983) analysis, because uncertainty can be con-sidered to push the operator toward the use ofknowledge-based processing, as opposed to therule-based behavior associated with standardCA encounters. Tattersall and Hockey (1995)have shown that this mode of problem solvingmade greater demands on in-air engineers, asindexed by suppression of the 0.10-Hz compo-nent of heart rate variability. In addition, the

higher level of mental load must be sustained forprolonged periods because actions are delayedby the inability to resolve the uncertainty.

The present study was designed to assess thecognitive demands of CA at sea using standardworkload assessment methodology. A specific


3/14

254 Summer 2003 Human Factors

aim was to compare the two major sources ofCA demand, together with those of normal(noncollision-course) encounters. Routine CAencounters have predictable course-change

requirements, following rules-of-the-road pro-cedures. By contrast, emergency encounters,which are brought about by rule violation bythe target vessel, increase uncertainty and makecourse changes unpredictable. We compareddemands under uncertainty with those imposedby predictable course change requirements byexamining the time course of demand over tri-als. Action implications of emergency scenarioswere not resolved until late in the trial, so sec-

ondary task decrement should increase over thecourse of the trial, as compared with routinetrials. The primary task requirement to makecorrective maneuvers under threat conditionsshould be reflected in the increased use of nav-igation aiding devices, particularly test maneu-vers. Because of the possibility of making early,controlled course corrections, routine scenariosshould show sustained use throughout the trial,whereas emergency scenarios, in which reliableinformation was not available until later, wereexpected to show increasing use as the trialprogressed.

METHOD

Participants and Design

The 12 participants (10 men, 2 women) wereselected from a panel recruited for a continuingprogram on performance in complex systems.Their ages ranged from 20 to 31 (median = 24)

years. All were either graduates or students inscience with extensive experience with com-puting and Windows-based applications. Nonehad previous maritime experience of control-

ling large vessels. They were paid 100 ($150U.S.) for their participation in the experiment.

Participants were tested under all conditionsof a 3 2 3 3 repeated-measures design, in

which CA demand was manipulated by threeseparate factors. Collision threat (three levels)was defined as increasing risk of collision withan approaching (target) vessel if no action weretaken. Target behavior(two levels) referred towhether target vessels remained on a fixedcourse or altered during the trial. The thirdfactor, traffic(three levels), was defined by thenumber of distracters (vessels other than ownship or target) present on the display (zero, one,

or three). A fourth factor,phase (three levels),was defined for analytic purposes by dividingthe trial period into three successive phases.

Navigational Scenarios

A set of six generic scenarios was defined bycrossing collision threat and target behavior(Table 1). Three levels of collision threat wereincluded: normal encounters, which did notrequire any action by own ship (OS), and en-

counters requiring either a routine or an emer-gency response. Target behavior involved targetseither remaining on a fixed course throughouta trial or altering at some point, changing thecharacteristics of the developing situation. Innormal encounters the target was shown to bepassing at a safe distance of at least 1 nauticalmile (nm), equivalent to 1852 m. Routine en-counters displayed the target approaching OSfrom the right (starboard) on a direct collision

course. Because such targets have right of wayunder maritime regulations, it was the respon-sibility of OS to take evasive action by imple-menting appropriate course changes.

TABLE 1: The Six Generic Scenarios

Collision Target No. of Encounter Target Approach from Threat Behavior Trialsa

Normal/fixed Port or starboard Normal Fixed 36Normal/altering Port (gives way) Normal Alters 18Routine/fixed Starboard (stands on) Routine Fixed 18Routine/altering Starboard Routine Alters 6Emergency/fixed Port (stands on) Emergency Fixed 6Emergency/altering Port or starboard Emergency Alters 12

a Trials add up to 96 rather than 100. The remaining 4 trials involved ultrasafe scenarios, in which the target passed well ahead. Thesewere not included in the analysis.


4/14


In routine/altering, targets made an (unnec-essary) alteration, as if to give way, althoughOS was still strictly required to make a routineCA maneuver. Emergency scenarios were ex-

pected to increase monitoring demands be-cause of uncertainty about intentions of targetvessels. In these, the target violated rules of theroad, either failing to give way as required (emer-gency/fixed) or altering onto a collision course(emergency/altering), in both cases unexpectedlygenerating a potential collision situation. Thechange from a seemingly safe situation (equiva-lent to normal scenarios) to one in which anemergency response was required was initiated

during the 3rd min of a 6-min trial and took 30to 60 s to become evident. The resulting emer-gency collision threat meant that participantswere put under time pressure to take evasiveaction, but only after determining that the targetwould not revert to normal rule use. Each ofthese six generic scenarios was presented atthree levels of traffic (zero, one, or three dis-tracters) to give a set of 18 distinctive scenariotypes.

In order to explore the time course of effectsof navigational demands, the 6-min period wasanalyzed in three successive phases. Transienteffects associated with the beginning and endof trials were avoided by omitting the initialand final 30 s (when few relevant actionsoccurred), giving three 100-s phases: 30 to130 s, 130 to 230 s, and 230 to 330 s.

Navigational and Collision Avoidance Goals

Participants were required to avoid colli-

sions and, if possible, to stay on a plannedcourse that was always indicated as due northon the radar. The CA goal required operatorsto pass targets and distracters at a safe dis-tance, which was based on approximations tostandards maintained by actual mariners andassessed by measuring the closest point ofapproach (CPA) during a trial between OS andtarget or distracter (whichever was closer). ACPA of 1 nm was defined as a minimum for

good practice. A collision was defined opera-tionally as CPA < 0.5 nm, and a near miss wasdefined as CPA < 0.8 nm. The track-keeping(TK) objective was to remain for as long aspossible within 1 nm of the planned track.Avoiding a collision was emphasized as having

priority over track keeping in those cases wherethe two goals conflicted. However, the TK goalmeant that operators should avoid making un-necessarily large course alterations to avoid

other vessels.

Design of Trials

The duration of each scenario was short(6 min) so as to allow for multiple replicationsof trial types and to increase measurement relia-bility for the different encounters. It also allowedus to provide participants with a more repre-sentative sampling of the relative incidence ofrule-compliance and violation actions. Shorter

trials meant that the starting position and move-ment of both targets and other vessels had to betightly controlled to provide time for collisionsituations to develop. To facilitate this, vesselspeeds were set at a level about 30% higherthan those typically found at sea. (This was estab-lished in pilot tests as not changing behaviorsignificantly.)

The number of trials representing each ofthe 18 scenarios was designed to reflect actualmaritime experience, support expectations ofevent probabilities, and generate sufficient reli-able data. For example, to be consistent with theexceptional nature of transgressions of naviga-tion regulations, the probability of a violationby target vessels was kept fairly low (p = .18).This represents a compromise between the evenlower level of violations in real-life encountersand the minimum number of observations need-ed to obtain reliable measurements on the leastfrequent scenarios (routine/altering and emer-

gency/fixed, both 6 trials out of 100). The prob-abilities and frequencies of trials representingthe six generic encounter types are given inTable 1, with trials in all cases equally dividedamong the three levels of traffic. In addition tothese 18 scenarios, an ultrasafe scenario wasincluded (target passing well ahead,p = .04) inwhich there was no possibility of collision; thisallowed us to check baseline performance, oncases in which collision threat was absent. Data

from these trials showed no errors and very lowlevels of workload and are not included in theanalysis.

Simulation Display

The simulation of a simplified navigation


5/14


control station, judged to be realistic by qualifiednavigation officers, was based on the AdvancedRadar Plotting Aid (ARPA), which is commonlyused on merchant and other vessels. It was pro-

grammed in Visual Basic 3.0 and designed torun on a PC with a minimum specification of

Windows 3.1, Intel 80486 processor, and 17-inch (43.2-cm) VGA color monitor. Scenarioswere presented on a highly simplified relative-motion radarscope, with OS represented at thecenter of the screen. All encounters took placeunder open sea conditions, with no navigationalconstraints apart from other vessels, which weredisplayed as vectors showing relative direction

and speed. On a relative-motion display, anyother vessel that displays a vector pointingdirectly at OS is on a collision course (Figure 1).The planned track was superimposed on theradarscope as a dotted line and was orienteddirectly north. Range rings were set at 1-nmintervals, providing a rough guide to distance.In the scenario represented in Figure 1, the tar-get is shown approaching from port at a rangeof 3 nm and on a collision course (as in normal/altering or emergency/fixed). A distracter isalso represented, approaching from starboard at

3.6 nm and predicted to pass ahead at a rangeof approximately 1 nm.

Both targets and distracters were program-med to appear on the radarscope and behave

as vessels at sea. Their initial position was al-ways ahead (north) and to the side of OS, andthey would move across the screen at a prepro-grammed speed and heading. Starting parame-ters for position, speed, and heading dependedon encounter type but were otherwise variedrandomly across trials. They were calculated toput the target either on a direct collision course(


6/14

remaining to this position (TCPA), could bedisplayed on a panel above the radarscope byacquiring (clicking on) a chosen vessel. InFigure 1, the vessel approaching from port has

been acquired. The predicted CPA enables anoperator immediately to assess the risk of colli-sion, whereas predicted TCPA indicates howmuch time is available to make any necessarycourse change. In addition, a predictor facilitymode (test maneuver) allowed the operator tomodel the consequences of a planned maneu-ver before making it. When the test maneuvermode was toggled on, the OS vector could beset to the heading to be tested, and the predict-

ed relative heading vectors of all other vesselswould be displayed.

Secondary Task

In addition to carrying out the primary navi-gation/collision avoidance task, operators wererequired to monitor a separate display in orderto maintain engine oil temperature within toler-ance limits. The requirement to switch betweendisplays allowed us to infer shifts of attentionbetween primary and secondary tasks. The tem-perature variable would fluctuate slightly arounda set value (50 5 units) when in its normalstate. At random intervals generated by the com-puter (mean 46 s) it would enter a drift state,gradually increasing or decreasing in valuetoward the displayed limits. Operators wereinstructed to reset the system whenever theydiagnosed a drift state. To ensure that partici-pants actively monitored the display for driftsand did not reset it automatically, a cost was in-

corporated in which false positives always trig-gered the onset of a new drift state. If thevariable was allowed to exceed tolerance limits,an omission error was recorded. Oil tempera-ture was reset to normal for each new trial.

Subjective Mental Workload

At the end of each 6-min trial, operatorscompleted a subjective rating of mental work-load (MWL). Because of the length of the ses-

sions (2 hr) and the need to reduce furtherdemands on participants, this took the form of asingle-item visual analog scale. Participantsmoved a slider along a 100-mm line on thescreen (endpoints labeled very low and very highmental workload) and pressed an accept but-

ton when they were happy with their rating.This was scored in 2-mm divisions to give anMWL rating in the range of 1 to 50. Simpleglobal (univariate) scales have been shown to

provide sensitive and reliable indices of subjec-tive workload in similar work situations (Hendy,Hamilton, & Landry, 1993), and the procedureused here has been found to be minimally in-trusive in other long-duration, complex tasks(Hockey, Wastell, & Sauer, 1998; Sauer et al.,2002).

Procedure

Preliminary training. Because participants

had no nautical experience prior to taking partin the experiment, they were given extensiveformal training on the task (a total of around12 hr). Before being confronted with the radarsimulation, they were given instruction in ele-mentary navigational skills (navigational exer-cises and relevant collision regulations for thesimulated encounters). Before continuing withtraining on the simulation task, participants wererequired to practice these component skills intheir own time and to obtain perfect or near-perfect scores on formal tests. (Only one had tobe eliminated for failing to acquire the neces-sary skills.)

Navigation exercises were selected from astandard radar-plotting handbook used for train-ing nautical cadets. These helped participantsto develop an understanding of relative motiondisplays, vessel movements, the geometry ofcollision encounters, and calculation of coursechanges from basic radar-plotting techniques.

Instructions were also provided on the essentialnavigational goals of keeping a safe distancefrom other vessels and maintaining track. Parti-cipants were instructed to make collision avoid-ance the top priority and to attempt to maintainthe good-practice criterion of 1 nm CPAbut to return to the default track line once thecollision threat had passed. An illustrated lec-ture was given, explaining the collision regula-tions relevant for the task. These were selected

from the standard set of maritime regulationsfor the rules of the road at sea (InternationalMaritime Organisation, 1972) and providedclear guidelines for what action should be takenwhenever a collision threat was identified. Es-sentially, these state that vessels should normally



7/14


(a) give way to other vessels approaching ontheir starboard (right) side; (b) make any nec-essary course alterations in a starboard (clock-wise) direction; and (c) stand on (hold present

course) when other vessels approach from theport (left) side (because these should give way).

Simulator training. Participants were askedto familiarize themselves with the user manualprovided, which explained how to use each ofthe simulator controls and interpret the display.They were able to refer to this documentationthroughout a simulator training session of 20practice trials. This allowed participants tofamiliarize themselves with the simulator and

to practice making CA maneuvers. Practice tri-als were a random subset of the full set of 100experimental trials, although they did includean emergency encounter. During early trials,participants were encouraged to clarify theirunderstanding of the interface by asking forclarification if necessary. Performance on prac-tice trials was assessed in terms of success inmeeting primary task goals (CA and TK) andwas augmented by a further set of 10 trials ifnecessary.

Task sessions. Participants carried out thetask in groups of two or three individuals, sep-arated by screens. Five blocks of 20 trials wererun over a period of 3 to 5 days. Experimentalblocks took a little over 2 hr to complete andwere run in either late morning (10:00 a.m.12:00 p.m.) or early afternoon (2:004:00 p.m.).The sequence of trials in each session was self-paced; participants controlled the initiation ofeach trial by clicking a start button, which al-

lowed them to pause between trials if necessary.These breaks were encouraged to prevent fatigueor other effects of continuous testing, but theytypically lasted only 5 to 15 s.

Analysis of Data

Data capture. Automatic data capture en-sured that a detailed record was kept of all sig-nificant events and their time of occurrence(acquisition of other vessels, use of the test

maneuver toggle, heading or speed changes,CPA, and position of OS at 10-s intervals). Forthe secondary task, the system recorded resetactions and switches between radar and oiltemperature displays. The level of the oil tem-perature variable was recorded at 10-s intervals,

allowing the occurrence of drift and error statesto be computed. Data from 1179 of the total1200 trials (12 participants 100) were suc-cessfully collected for analysis, resulting in a

small number of missing observations, whichwere not replaced in the analysis.

Statistical treatment. For most analyses, a 3 2 3 3 repeated-measures analysis of variancewas carried out, with collision threat, target be-havior, traffic, and phase as factors. For someanalyses only whole-trial measures were avail-able, so phase was omitted. Data transforma-tions were carried out if necessary in order tosatisfy assumptions of normality, homogeneity,

and sphericity. (This was necessary becausemeans were based on different numbers ofevents.) A log transform was used for percent-age of time on the secondary task. Square roottransformations were carried out on secondarytask sampling time, duration, frequency, andreset latency. Because no suitable transforma-tion could be carried out on the data for trackkeeping and collision avoidance, we used logis-tic regression based on dichotomized (above orbelow 1 nm) minimum and maximum distanceoff track.

RESULTS

Primary Task

The various overall measures of navigationand collision avoidance behavior for the six dif-ferent scenario types are summarized in Table 2.They include performance measures (rule-following behavior, collisions, track keeping),

course changes, and use of the two navigationaids (target acquisitions and test maneuvers).

Collision avoidance and track keeping. In theabsence of collision threats, overall TK perfor-mance (remaining within the good-practicelimits of 1 nm of the planned course) was gen-erally good. Performance for normal encounterswas near perfect (


8/14


(mean 213 s). Because of this, overall TK errorwas, by default, lower (emergency/fixed 4%,

emergency/altering 2% off track). CA was mea-sured in terms of CPA, using two criteria: colli-sions only (CPA < 0.5 nm) and collisions plusnear misses (CPA < 0.8 nm). Performance wasagain good; overall collision rate was low forboth criteria (2% and 6%) and occurred almostexclusively under emergency conditions.

Course changes and rule following. Parti-cipants generally took appropriate action toachieve the primary task goals. As expected, al-most no course changes were made in normal/fixed encounters, although they occurred on44% of normal/altering trials (in which the tar-get started on a collision course but altered tosafe), with participants following approved pro-cedure (to starboard) on 99% of occasions. Asrequired, maneuvers were made in 99% of rou-tine encounters, with 95% of the first alter-ations again following recommended procedure(to starboard). Finally, required maneuvers weremade in almost all emergency encounters. De-

spite the uncertainty associated with these, rulefollowing was high (91% and 81% for emer-gency/fixed and emergency/altering, respec-tively). The Headings Entered column in Table2 shows that the number of course changesentered was much higher for routine encounters(7.71 and 3.94) than for either normal or emer-gency encounters (around 12 per trial).

Use of decision aids. Table 2 suggests thatboth types of aid were used most often during

routine encounters, in which CA threat waspresent throughout the trial. For target acquisi-tions, however (Figure 2), the only marked ef-fect was that the aid was used more under hightraffic, F(2, 22) = 5.99,p < .01. There was ageneral reduction over the course of the trial,

F(2, 22) = 4.83,p < .05, but no Traffic Phaseinteraction, F(4, 44) = 1.00,p > .05, or any

other effect.Use of the test maneuver facility showedmore widespread differences among encountertypes. Figure 3 shows the percentage of timethe aid was selected. As predicted, this remainedhigh throughout routine threat (around 50%)and increased dramatically over the trial foremergency threat. There were strong effects ofboth threat, F(2, 22) = 42.80,p < .001, andtarget behavior, F(1, 11) = 12.31,p < .001, andtheir interaction, F(2, 22) = 32.75,p < .001.

TABLE 2: Summary of Primary Task (Ship Control) Measures

Collision Deviation Course Rule Target Test Bearings Headings[+ near miss] from Track Changes Following Acquisitions Maneuver Taken Entered

Encounter (% trials) (% time) (% trials) (% trials) (no./trial) (% time) (no./trial) (no./trial)

Normal/fixed 0.0 [0.0]0 0.0 1.5 100.0 1.05 12.8 0.17 2.51Normal/altering 1.4 [6.6]0 0.6 44.1 99.0 0.88 41.1 2.01 2.15Routine/fixed 0.9 [6.0]0 14.7 98.7 96.7 1.20 65.1 5.12 7.71Routine/altering 1.3 [4.0]0 12.6 98.5 93.3 1.07 65.8 5.01 3.94Emergency/fixed 6.7 [21.3]0 3.7 99.7 90.7 0.62 59.6 2.51 1.93Emergency/altering 10.6 [33.3]0 2.4 97.0 81.3 0.88 46.7 1.26 0.92

Figure 2. Mean number of target acquisitions as afunction of phase and traffic density (distracters).


9/14


Unlike target acquisition, this aid had no effectsof traffic. The main effect of phase, F(2, 22) =49.68,p < .001, was associated primarily withthe increase over time under emergency threat.There was an interaction of phase with boththreat, F(4, 44) = 27.34,p < .001, and targetbehavior, F(2, 22) = 8.47,p < .01, as well as a

three-way interaction, F(4, 44) = 9.97,p < .001.For emergency encounters, increase in the useof test maneuvers was greater when targetsaltered course. Figure 3 shows that the use ofthe test maneuver during emergency/alteringchanges from the level of the least-demandingscenario (normal/fixed) during Phase 1 to thatof routine by the end of the trial.

Subjective Mental Workload

Figure 4 summarizes effects on subjectivemental workload ratings (MWLs). These varyover most of the range (442 on the 150scale), reflecting widespread differences amongconditions. There were main effects of threat,F(2, 22) = 98.60,p < .001, target behavior,

F(2, 22) = 46.97,p < .001, and traffic, F(2,22) = 24.91,p < .001, as well as interactionsof threat with both target behavior, F(2, 22) =32.51,p < .001, and traffic, F(2, 22) = 2.66,p

< .05. Effects of target behavior occur onlywith normal encounters, and those of trafficoccur only with routine encounters.

Secondary Task

Performance measures. Table 3 summarizesthe secondary task performance data for thesix encounter types (with MWL means forcomparison purposes). The only sensitive mea-sure was reset latency (time from the onset of a

fault to a reset response). Omission errors wererare, averaging much less than one per trial,and reset false positives were even rarer, occur-ring on only 4% of trials. For reset latency,there were no main effects of encounter vari-ables, but there was an interaction betweenthreat and target behavior, F(2, 22) = 3.36,p .05,although there was a complex interaction,F(1, 22) = 6.56,p < .01. When targets alteredcourse, sampling time was reduced for normal,unchanged for routine, and increased for emer-gency scenarios. There was also a small main ef-fect of traffic density, F(2, 22) = 5.21,p < .05.

Of greater interest are changes in secondarytask sampling over the duration of the trial.Figure 5 shows that the effect of phase, F(2,22) = 10.80,p < .001, is attributable mainly to

TABLE 3: Summary of Secondary Task Measures

MWL Omissions Reset Sampling Sampling SamplingEncounter Rating (no./min) Latency (s) Time (%) Rate (no./min) Duration (s)

Normal/fixed 6.7 0.18 35.2 40.8 2.17 15.07Normal/altering 17.0 0.26 36.6 29.9 2.13 9.33Routine/fixed 27.3 0.29 37.0 19.9 2.13 4.81Routine/altering 28.0 0.31 37.6 18.5 2.03 3.83Emergency/fixed 36.3 0.31 36.8 20.9 2.12 3.58Emergency/altering 40.0 0.56 40.5 27.8 1.11 8.98

Figure 5. Secondary task monitoring (% time selected) as a function of collision threat (separate panels),phase, target behavior (F = fixed, A = altering), and traffic density (0, 1, 3 distracters).


11/14


changes under emergency scenarios, especiallyemergency/altering. This is supported by thePhase Threat interaction, F(4, 44) = 11.63,p < .001, and the three-way interaction among

phase, threat, and target behavior, F(4, 44) =4.22,p < .01. Under emergency/altering, theapparent threat of collision increases fromminimal to maximal over the course of the trial.The pattern of findings here can be seen as amirror image of the increased use of test maneu-vers in Figure 3.

To examine the nature of these effects, wedecomposed sampling time into its two com-ponents, frequency (how often the oil tempera-

ture display was sampled) and duration (lengthof sampling periods). Sampling frequency wasfound to be near constant across all conditions(0.30.4/min). The dramatic reductions in sam-pling time under increasing threat are almostentirely the result of operators making briefer(rather than less-frequent) inspections. However,sampling frequency was higher in trials whereno errors occurred (mean = 0.37/min) than inthose incurring either one (mean = 0.27/min)or two (mean = 0.22/min) errors. Effects onsampling duration were similar to those foroverall sampling time and are not reportedseparately here.

DISCUSSION

Overall, the results confirm that secondarytask techniques can be used effectively to assesscognitive demands in a simulated maritime taskenvironment. Higher levels of collision threat

were associated with markedly increased rat-ings of MWL and with impaired performanceon the secondary oil pressure monitoring task,in the form of increased reset latency, mediatedprimarily by a reduction in the average sam-pling duration. The reduction in secondary taskmonitoring under increased collision threat wasmirrored by an increase in the use of the testmaneuver tool, which strongly reflected CA de-mands. Use of the target acquisition tool proved

sensitive only to traffic density and fell off overthe course of the trial.

Of more specific interest for the purposes ofthe study is that these effects are found not onlywith objective workload factors (traffic densityand marked course change requirements) but

also with uncertainty. All these effects were ob-served most strongly under the two emergencyencounters (emergency/fixed and emergency/altering), in which uncertainty was highest.

Demands associated with traffic density orstandard maneuvering requirements can beassessed quickly and acted upon effectively(using the target acquisition and test maneuverfacilities). By contrast, demands arising fromrule violations build up over the trial and are notfully resolved even when the emergency maneu-ver is carried out (because of the increased riskof collision with the late maneuvers). It is nocoincidence that the problem of uncertainty

about others intended actions has been identi-fied as one of the most serious impediments tothe development of effective collision avoidancesystems (Hobday et. al., 1993). Mariners in ourcritical incident survey reported a number ofincidents in which near collisions involved ruleviolations of the kind simulated in the study.They reported high levels of concern and frus-tration in (a) not knowing what would happenand (b) having to delay course changes untilthe last minute, when judgments of safety mar-gins were difficult.

To understand what is happening in the twoemergency scenarios, recall that they both startout as formally equivalent to normal encoun-ters (emergency/altering with normal/fixed,and emergency/fixed with normal/altering), inwhich no collision threat is indicated. Emer-gencies are generated by the target vessel vio-lating the rules of the road (in the first case byaltering onto a collision course and in the sec-

ond case by failing to alter course). Unlike theevasive response in routine encounters, in thesecases the navigator must continue to monitorthe progress of the target vessel and is discour-aged from making early evasive action by bothnavigation/CA rules and track-keeping require-ments. Although maneuvers still need to beplanned in advance, they have to be constantlyupdated as the threat is sustained and unre-solved. In addition (as in real maritime contexts,

supported by our interviews), maneuvers aretypically knowledge based (Rasmussen, 1983)and can rarely be drawn from the standardrepertoire of evasive actions.

This interpretation is supported by the data,which illustrate the growth of uncertainty for


12/14


emergency scenarios over the course of thetrial as a transition from the levels of workloadassociated with normal scenarios to those forroutine scenarios. Figure 3 shows that time spent

making test maneuvers early in the trial wassmall under emergency scenarios, as for theequivalent normal encounters, confirming thatparticipants did not predict the later misbehav-ior of the target vessel. As uncertainty develops,there is a marked increase in test maneuvers(by a factor of 10 in the case of emergency/alter-ing), until it matches the level for routineencoun-ters, and a corresponding reduction in secondarytask sampling (Figure 5). This all happens dur-

ing a brief (6-min) trial, in which task eventsoccur relatively slowly and infrequently.An alternative explanation for the pattern of

results should, however, be considered. It ispossible that the dramatic increase in demandsover time during emergency scenarios is attrib-utable not to uncertainty but to the demands onevasive action. This is unlikely, given that bothheadings and bearings were much less frequentunder emergency scenarios than under routineones (Table 2) and the time of the first coursechange was much later (mean 213 vs. 80 s,respectively). More specifically, the evasive-action-demands explanation would predict thatthe increasing secondary task neglect (e.g., sam-pling time) under emergency conditions coin-cides with increased course change activity. Bycontrast, an uncertainty explanation predicts agrowing strain on mental resources underemergency conditions, disrupting performanceeven before maneuvers are made.

This cannot be tested easily, because bothmaneuvers and secondary task decrements oc-cur mainly during the middle phase of the trial;however, a detailed analysis of sampling time(using 60-s phases) showed that decrement didnot occur at all for routine scenarios. A morespecific analysis was carried out for all trials inwhich a course change was made. The periodbefore the first maneuver was divided into earlyand late phases, allowing decrements in sam-

pling time before making maneuvers to be ob-served as a reduction in sampling time. Thisrevealed significant decrements for both emer-gency/fixed and emergency/altering, in bothcases for 10 of the 12 participants (sign test;p

Cognitive Demands of Collision Avoidance in Simulated

Documents