USABILITY EVALUATION OF A FLIGHT-DECK AIRFLOW HAZARDVISUALIZATION SYSTEM

Cecilia R. Aragon, University of California, Berkeley, California and NASA Ames Research Center, Moffett Field, California

AbstractMany aircraft accidents each year are caused

by encounters with unseen airflow hazards near theground, such as vortices, downdrafts, low levelwind shear, microbursts, or turbulence fromsurrounding vegetation or structures near thelanding site. These hazards can be dangerous evento airliners; there have been hundreds of fatalities inthe United States in the last two decades attributableto airliner encounters with microbursts and lowlevel wind shear alone. However, helicopters areespecially vulnerable to airflow hazards becausethey often have to operate in confined spaces andunder operationally stressful conditions (such asemergency search and rescue, military or shipboardoperations).

Providing helicopter pilots with an augmented-reality display visualizing local airflow hazards maybe of significant benefit. However, the form such avisualization might take, and whether it does indeedprovide a benefit, had not been studied before ourexperiment.

We recruited experienced military and civilianhelicopter pilots for a preliminary usability study toevaluate a prototype augmented-realityvisualization system. The study had two goals: first,to assess the efficacy of presenting airflow data inflight; and second, to obtain expert feedback onsample presentations of hazard indicators to refineour design choices.

The study addressed the optimal way toprovide critical safety information to the pilot, whatlevel of detail to provide, whether to displayspecific aerodynamic causes or potential effectsonly, and how to safely and effectively shift thelocus of attention during a high-workload task.Three-dimensional visual cues, with varying shape,color, transparency, texture, depth cueing, and useof motion, depicting regions of hazardous airflow,were developed and presented to the pilots.

The study results indicated that such avisualization system could be of significant value inimproving safety during critical takeoff and landingoperations, and also gave clear indications of thebest design choices in producing the hazard visualcues.

IntroductionTurbulence and other wind-related conditions

were implicated in 2,098 out of 21,380 aircraftaccidents in the NTSB accident database from1989-99 [1]. Addressing the controllabilityproblems created by airflow disturbances has thusbeen a major issue in aviation safety. Disturbancesin airflow, including weather-related hazards (e.g.thunderstorms, low level wind shear ormicrobursts), and locally-generated airwake hazards(such as downdrafts, hot exhaust plumes, wakevortices from other aircraft, turbulence and vorticesfrom surrounding vegetation or structures near thelanding site), have all been documented to behazardous to aircraft of all categories and classes.The weather-related airflow disturbances have beenrelatively well studied and there is commercialhardware for detection of adverse weather ormicrobursts. Our research focuses on the smaller-scale airwake hazards, of which there has been lessstudy (although they are more common), andspecifically on the display of such airflow hazardsto the pilot during approach to landing.

Airflow hazards are hard to detect simplybecause air is invisible. Its flow pattern isundetectable by pilots on a landing approach unlessthe air happens to pick up dust, smoke or otheraerosols that are visible to the human eye. Beingthus unable to directly see a factor of potentiallygreat importance to them, pilots learn to use theirintuition concerning airflow over obstacles near thetakeoff or landing site, and they learn to pick upvisual cues from the surrounding area. Thesemethods are inadequate, however, as airflow-relatedaccidents still occur.

Hazard detection architectureA complete onboard airflow hazard detection

system would consist of three major components:sensors; classification and analysis; and display(human interface). Our research addresses thedisplay stage, but we describe the others here toillustrate the problem in context.

Sensors/detection. Recent technologicaladvances in sensor technology, especially Dopplerlidar [2], PIV (Particle Image Velocimetry) [3], andforward-looking microwave radar, offer thepotential for aircraft-based sensors which can gatherlarge amounts of airflow data in real-time. Thereare currently available commercial systems utilizingthis technology to detect moderate-scale airflowdisturbances such as microbursts and windshear [4].Current research into airflow detection techniquessuch as lidar is promising; it is believed that withina few years hardware capable of being mounted onan aircraft will be able to reliably scan the area afew hundred feet ahead of the aircraft and sampleair particle vector velocities at one-foot intervals orless [5]. With the development of such devices,onboard detection systems that can convey detailed,specific information about airflow hazards to pilotsin real-time become a possibility.

Classification/analysis. This area concernsthe development of algorithms to input the particlepositions and vector velocities, other variables suchas density altitude, aircraft gross weight, and poweravailable; to compute the locations of the areas offlow which may produce a hazard to this particularaircraft on this particular day; and to output thethree-dimensional coordinates of the hazardlocation in real-time.

Display to the pilot (user interface). Giventhe airflow data and the known hazard areas, theproblem then becomes to organize this vast amountof data, describing millions of particles swirling indifferent directions, and present it to the pilot in amanner that does not interfere with the primary taskof operating the aircraft safely. An interface isrequired that can present potentially large amountsof data to the pilot in a non-intrusive yetcomprehensive manner in real-time.

Motivation for visual interface usability studySince an airflow hazard detection system

generates a large amount of disparate data that mustbe organized and presented to a human operating acomplex machine in a high-workload environment,an efficient method of human-machinecommunication is required. The human visualsystem has the highest bandwidth of all the senses.It can process gigabytes of data in real-time andorganize it into patterns that the brain can use todraw conclusions and act very quickly. Beyond thisgeneral observation about the efficacy of visualinput, we also note that the operation of landing anaircraft is an essentially visual operation, even if theflight itself is made under instrument weatherconditions; the pilot looks at least at theinstruments, and finally always at the landing site.It therefore makes sense to organize the airflow datainto some type of visual display.

As with any type of user interface, usabilityevaluation is important to ensure that the displaymost efficiently supports the human operator'sperformance. Given the demanding environmentand the relatively small population of highly trainedpilots, it is especially critical to conduct a usabilitystudy before designing an airflow hazard displaysystem.

Shipboard rotorcraft operationsAlthough the need to detect airflow hazards

exists for all pilots in all aircraft, for our researchwe chose to focus on helicopter operations, andspecifically on Navy shipboard rotorcraftoperations, to which the Navy refers as the"Dynamic Interface." There were several reasonsfor this choice.

Helicopters are especially vulnerable to airflowdisturbances; first, by the nature of the aerodynamicforces involved, and second, because helicopters areoften called upon to operate into and out ofconfined areas or areas that naturally have disturbedairflow. For example, emergency search and rescuemay have to operate in mountainous areas and smallclearings surrounded by vegetation and cliffs wherethe winds are always high. Helicopters also mustland on urban rooftops, offshore oil platforms, or onthe decks of ships. A device for detecting airflow

hazards therefore has a special utility for helicopteroperations.

Operating a helicopter off a moving aircraftcarrier is one of the most demanding tasks ahelicopter pilot can face [6]. Because the ship ismoving, its superstructure will always generate anairwake consisting of vortices and other unseenhazards. In addition, high sea states may causeextreme ship motion, and low visibility maydegrade visual cues. The pilot must maneuver thehelicopter within very tight tolerances underadverse conditions. It is a task that demands theutmost concentration and skill from the pilot[Figure 1]. A system that can deliver even anincremental amount of assistance to the pilot in thishigh-demand environment could have a significantimpact on safety.

Figure 1. Shipboard helicopter landing

Helicopter accidents and incidents that occureach year range from fatal accidents to incidentssuch as "tunnel strikes" (when a rotor blade strikesthe fuselage of the helicopter). There have beenapproximately 120 tunnel strikes since 1960,causing damage ranging from $50-$75K to over$1M [5]. When analysis of these accidents andincidents is performed, the conclusion is frequentlythat they were due to unseen airflow hazards such

as vortices, downdrafts, hot exhaust plumes, orwind shear, where the pilot and ground crew wereinitially unaware of the danger and the pilot wasunable to react in time. Presenting the appropriateinformation to the pilot or flight deck air boss inadvance of the hazard encounter could reduce orprevent these types of accidents in the future.

Finally, because shipboard rotorcraftoperations are such a demanding environment, thearea is very well studied. The Navy has compiledsignificant amounts of data from shipboard flighttests, wind tunnel tests, and computational fluiddynamics computations studying the airflow aroundmoving ships of all types, and how the airwakechanges when helicopters of different makes andmodels land on the ships. Utilizing the data fromthese extensive tests and computational studies, theNavy develops operational envelopes listingallowable wind conditions for many ship-rotorcraftcombinations [7]. However, the envelopes are ofnecessity (for safety reasons) relatively narrow, andconvey fairly limited information, basically ago/no-go decision. The envelopes do not statewhich safety considerations caused a particularoperational limit, thus limiting the informationavailable to the pilot. On the other hand, accidentsand incidents occur during operations within theenvelope every year. On occasion, during the post-accident analysis, the flight test engineers can pointto existing airwake data to show that the accidentwas caused by disturbed airflow over a portion ofthe deck. In other words, the information that couldhave prevented the accident was known, but it hadnot been communicated to the pilot. Thus as Navyflight test engineers seek ways to increase fleetsafety, this problem is ripe for solution.

In this paper, we begin by discussing previousresearch relevant to developing a visual hazarddisplay to solve the airflow hazard problem. Thenwe describe our experimental procedure, discuss theresults obtained, and finally give conclusions anddirections for further work.

Previous ResearchThere are three major areas where there are

significant bodies of research that are relevant toour current efforts: flow visualization, aviationdisplays, and Navy shipboard rotorcraft operations.

Flow visualizationThere is a large body of research on flow

visualization, which often consists of detailedimagery of two- and three-dimensional airflowpatterns, both static and dynamic, steady andunsteady, all designed to help scientists or engineersunderstand—and analyze at length—a particularinstance of a fluid flow. Examples includestreamlines and contour lines for the case ofinstantaneous flow [8], [9] and streaklines,timelines [10], flow volumes [11], spot noise [12],or moving textures for unsteady flow. The imageryis often quite complex and not suitable for rapidglances during time-critical tasks. Otherpotentially applicable work includes automateddetection of swirling flow [13] and terrain andturbulence visualization [14].

Not being designed for viewing in the cockpitor under time pressure, flow visualization methodsemphasize the informational richness of theimagery rather than the user interface. Thesemethods have seldom if ever been subjected tousability studies, and are not in any case expected tobe viewed and acted upon in real-time.

Aviation displaysIt has long been recognized that applying

developing technology to improve aviation displaysmight enhance aviation safety. Significant work inthis area includes synthetic vision and augmented-reality displays (terrain in low-visibilityenvironments, navigation aids) [15], [16], [17],[18], weather visualization including NASA’sAWIN, TPAWS and AWE [19], [20], [21], [22],and turbulence detection/prediction [23]. Holforty'sresearch on wake vortex visualization [24] is one ofthe first to provide three-dimensional hazard cueswith the intent of eventually layering the cues on anaugmented-reality display. There is also a largebody of relevant work concerning human factors inthe cockpit, including the study of attention andcockpit visual displays [25], [26], [27], [28].

Also, it has been known since the eighties thatweather-related airflow phenomena such asmicrobursts and wind shear have been responsiblefor the loss of hundreds of lives in airliner accidents[29]. As a result, a great deal of work has beendone to detect, predict, and display this type of

information to the pilot [30]. There arecommercially available aircraft-based, forward-looking microwave radar and lidar systems that candetect microbursts and windshear. However, theemphasis was placed on integrating the informationinto existing cockpit displays, in order to reducetime to commercial deployment. Accordingly, nousability studies were focused specifically on thedisplay itself or on whether a three-dimensionalhead-up display would be helpful in presentinghazard information to the pilot.

Navy shipboard rotorcraft operationsBecause landing a helicopter on a moving ship

deck is a very hazardous environment [6], the Navyhas long operated a program to perform flighttesting of these conditions [7]. They have run asignificant amount of flight testing with the statedgoal of improving flight safety. The shipsuperstructures always produce airwake. Inaddition, aircraft landing on shipboard are plaguedby hot exhaust plumes, very powerful shipboardradar that interferes with aircraft systems,inaccurate anemometers, and of course theinevitable problems associated with high sea statessuch as extreme values of ship pitch, heave, androll, and high, turbulent winds, as well as therequirement to land in a very confined area.

Because understanding the airwake over theship is so critical, the Navy uses techniquesincluding shipboard flight testing, wind tunnel tests,computational fluid dynamics, and sampling of theairflow vector velocities at various points in theflow field behind the superstructure in thehelicopter landing zones with handheldanemometers. Lidar detectors are being developedand improved and research is ongoing in this area.

The current method of communicating thisinformation to the pilots consists of operationalenvelopes for each ship-rotorcraft combination[Figure 2].

The envelope depicts the allowable windspeeds and directions that a given helicopter isallowed to land on a particular ship. It isnecessarily conservative, as the envelope has toinclude all flight conditions and all fleet pilot skilllevels. The envelopes limit allowable landingconditions significantly; however, even with this

Figure 2. Ship-rotorcraft operational envelope

cautious approach, accidents due to airflow hazardsstill occur. In one recent example, a helicopter wasdamaged when starting up on shipboard, eventhough the winds were within the allowable startingenvelope. Another helicopter was operating upwindof the first, and this configuration caused hazardousairflow to be present at the downwind spot [5].There was knowledge of this problem from theNavy flight testing program; however, theenvelopes cannot portray every combination ofaircraft location on a ship that may have manyhelicopters and aircraft in operation at the sametime.

Experiment DesignDuring potentially hazardous conditions, high

winds, low visibility, or extreme ship motion, thepilot’s attention will naturally be focused outsideduring the critical landing moments; he or she willnot want to look down at a cockpit instrumentdisplay. In designing our experiment, we assumedpilots would prefer an augmented-reality hazardvisualization display (as was verified during theusability study). However, the head-up displaymust be carefully designed not to distract from the

key shipboard visual cues, which may be degradedduring a challenging nighttime or poor-weatherlanding on a ship. Studies have shown that head-updisplays with superimposed symbology may onoccasion cause performance problems due toattentional capture by the perceptual grouping ofthe superimposed symbols [31]. “Scene-linked”head-up displays, or displays where there is nodifferential motion between the superimposedsymbology and the outside scene, can avoid thistype of distraction. For this reason we decided todevelop a head-up display where the hazardindicator is three-dimensional and appears to bephysically part of the world.

Rapid Prototype PhaseWe first constructed a horizontal prototype (a

relatively full-featured simulation of the interfacewith no underlying functionality) [32] of anaugmented-reality hazard visualization system thatincluded many different types of hazard indicators.The usability study on the horizontal prototype hadtwo main goals: first, to determine whetherpresenting airflow hazard data to helicopter pilotswould be helpful to them; and second, to obtainexpert feedback on the presentation of samplehazard indicators, from which we could refine ourdesign choices.

We decided to perform interactive prototyping[32], a technique where the prototype is altered onthe fly as the test user comments on itseffectiveness. This enabled us to rapidly modifythe design and obtain feedback on multiplevariations in a single session.

Selection of Platform for Rapid PrototypeThe next task was to identify the best tool for

creating a relatively realistic, three-dimensionalvisual simulation of the helicopter pilot's view outthe cockpit windscreen during the final approach toa shipboard landing. The tool was required tosupport rapid prototyping, 3D modeling, and simpleanimation. It was especially important that we beable to create new hazard visualizations withinminutes, as we were hoping to get feedback fromthe study participants and implement theirsuggestions during the session so as to tighten thefeedback loop.

Consultation with Navy flight test engineersprovided detailed descriptions of what a landingapproach should look like. Additionally, we wereprovided with an extremely detailed 3D CADmodel of a Spruance-class destroyer (DD 963). Anideal prototype platform would be able to use thisdata to render a realistic approach.

Three approaches were considered for theprototype software platform: a low cost, off-the-shelf flight simulator; a 3D animation system; and a3D CAD tool.

The Microsoft Flight Simulator was consideredbecause it offered the possibility of the pilots beingable to use a joystick, and potentially theopportunity to alter the visual hazard displaywithout affecting the flight simulation. However,there was no convenient interface for importing theexisting ship model into MS Flight Simulator.

We also investigated various 3D animationsystems such as WildTangent, VRML, and Flash.However, although these systems could handle theanimation well, the overhead for changing hazardindicators was considerable, essentially comparableto working in a programming language. (Actualprogramming languages, such as Java, were ruledout for the same reason.)

The CAD modeling tool we selected,Rhino3D, offered rapid construction and alterationof the prototype scenarios and easy access to theship model data. It was very easy and quick tocreate many different types of hazard indicators andmodify their shape, location, color, texture, andtransparency. Although not a flight simulator, theCAD program allowed us to simulate the finalapproach to landing by rotating and zooming themodel of the ship with the hazard indicatordisplayed above it.

MethodologyWe recruited three highly experienced (>1700

hours) helicopter pilots and flight test engineers, allwith shipboard landing experience. Each sessionwith a participant pilot consisted of a 1 ½-hourinterview with the pilot in front of a projectionscreen. All sessions were videotaped. Twoexperimenters conducted the session, one operatingthe computer and the other interviewing the pilotand taking notes.

The operator-experimenter used the Rhino3DCAD program to display on the projection screen amodel of the ship (DD 963), with a hazard indicatordisplayed on the ship's deck where hazardousairwake might be found. The operator manuallysimulated a helicopter's view of an approach tolanding on shipboard as the pilot watched andcommented. A wide selection of different types ofhazard indicators were stored in layers in Rhino3D,so that features could be selectively turned on andoff by the operator. The features that were varied inthe hazard indicator included shape, location, color,texture, transparency, depth cueing, and motion.

Figure 3. Hazard indicators in Rhino3D

Feedback was solicited from the pilot. If thepilot suggested a change, the operator implementedit on the fly and the pilot was asked to judgewhether the change was an improvement. Theexperimenter asked both specific and open-ended

questions throughout the interview designed toelicit the pilots' expertise.

Using the pilots' responses, we attempted toassess the efficacy of presenting airflow data inflight, and to select the most efficacious (as judgedby the pilots) visual presentation for the hazardindicator.

ParticipantsIn choosing participants, we sought pilots with

a great deal of helicopter experience and, ideally,experience with shipboard landings of large militaryhelicopters. Finding pilots with the requisitedomain-specific knowledge was challenging. Thefinal test group for the prototype consisted of twomilitary pilots and one experienced civilianhelicopter pilot:

Participant 1: Navy helicopter test pilot, 2000hours of flight time, 17 years experience.

Participant 2: Navy helicopter flight testengineer, 4000 hours of helicopter simulator time,100 hours of flight time, 17 years experience withshipboard helicopter flight tests.

Participant 3: Civilian helicopter flightinstructor, 1740 hours of flight time, 3 yearsexperience.

ResultsResults of the usability study on the rapid

prototype were encouraging, but in some respectssurprising as to the types of display features pilotsfound helpful. All participants said they liked thesystem and would use it if it were installed on theiraircraft. As they viewed the interface, the pilotsrepeatedly stated that they wanted such a hazardvisualization tool.

As to the type of visualization, the strongestoverriding principle that emerged from thisexperiment is that helicopter pilots are using alltheir attention to focus on the extremely demandingtask of landing on a moving ship deck, perhapsunder low visibility conditions or at night, and thehazard indicator must not distract from that focus.To that end, the participants favored much simplerimagery than we would have expected.

The pilots strongly rejected the use of flowvisualization indicators, and especially of motion toindicate flow. Given the manner in which fixed-wing pilots look for natural flow indicators such asdust devils near the runway, smoke plumes, wind-blown vegetation etc., we had anticipated thathelicopter pilots would prefer a dynamic flowvisualization, capable of indicating the directionand velocity of particles in the hazardous region.However, the participants exhibited resistance tosuch a design. All participants, even whilereiterating their desire for 3D hazard visualization,stated that motion was distracting during theapproach and particularly during the criticalmoments near touchdown. A static visualization,even supplying less information, was stronglypreferred over a dynamic hazard indicator. That is,the participants sought a real-time decision supporttool, not an airflow analysis tool.

Below we describe the hazard displayparameters that were varied in the prototypeusability test, and the results obtained for each.

ColorWe showed hazard indicators in single and in

multiple hues, using colors spanning the spectrum.All pilots preferred single-color hazard indicators,and indeed, preferred only two colors for the finalsystem: yellow for caution and red for danger.Yellow, according to the participants, shouldindicate an airflow hazard that could necessitatestrong pilot input to stay safe, but where the aircraftshould maintain controllability. Red should indicatedanger, an airflow hazard that would likely bebeyond the limits of the aircraft and would put itscontrollability in question.

We were surprised to find the pilots unanimouson the point that a hazard indicator should berendered in a single color (either red or yellow).Multiple-color hazard indicators were considereddistracting and confusing. When the experimenterspointed out that a vortex core could have verystrong winds but the outer portion of the vortexmight not be as hazardous, so that a two-colorindicator with a red core and a yellow mantle mightbe useful, the pilots all disagreed, saying the redvortex core would be difficult to see or to locatecorrectly in a three-dimensional object. In additionto the overall view that the display would be

confusing, a concern was also expressed that a two-color indicator could tempt a pilot to venture intothe yellow mantle while attempting to skirt the redcore. That is, the two-color indicator was thoughtto potentially support an incorrect decision to landin dangerous conditions.

TransparencyWhile holding other variables constant, we

varied the transparency of the displayed hazardindicator from 20% to 80% (according to the Rhinosoftware controls). This test was repeated for arange of objects. The pilots preferred an averagetransparency near 70%. While desiring a hazardindicator sufficiently opaque to come to the pilot'sattention, participants noted the critical need for thepilot to be able to see visual cues on the ship behindthe hazard indicator.

Depth CuesWe displayed hazard indicators that hovered

above the deck and cast no shadow, and others thathad a colored shadow projected onto the deckdirectly below the indicator. Of those withshadows, some had a connecting vertical line fromthe indicator to the deck shadow. All of the pilotspreferred shadows below objects, stating that theyhelped the pilot to localize the 3D indicator inspace. Pilot #1 said shadows alone might besufficient for a shipboard hazard warning system:"just paint the deck red if I need to wave off." Pilot#2 liked the idea of a connecting line between thehazard indicator and the deck. No participantwanted tick marks or numeric information floatingwith the hazard indicators. Again, they preferred tokeep it simple; the purpose is to let the pilot see thelocation and approximate severity of a hazard, notto help them measure or analyze it.

TextureWe displayed hazard indicators having a series

of arrows textured onto their partially transparentsurface, to indicate the direction of airflow in thathazardous area, and asked pilots to compare them toindicators without the texture. Pilots #1 and #2 didnot want the extra detail, saying it could beconfusing or distracting. Pilot #3 suggested stripingas a possible symbology, reminiscent of the yellow

and black caution tape that is a common symbol tomost Americans.

ShapeWe asked the pilots to comment on the effect

of varying the shape of the hazard indicator, such asrectilinear transparent boxes, cloud shapes withrounded corners, spirals, rings both round andrectangular. The rectilinear and cloud shapes werefavored over all others. Again, a preference forsimplicity was displayed. One of the pilots pointedout that the floating rings looked a little bit like theHUD symbology for the “highway in the sky,”perhaps beckoning the pilot to fly into the rings, theexact opposite of the intended action! Thiscomment made clear the need to research all HUDsymbology so as to avoid conflicts with existingsymbology or commonly accepted designs.

MotionThere was a strong consensus that motion in

the display, particularly fast motion, wasdistracting. Pilot #1 (the participant with the mostexperience landing on shipboard in actualhazardous conditions) said the visual indicatorsshould absolutely not use motion at all. It wasdistracting, and in the worst case could inducevertigo, especially at night or in low-visibilitysituations. The pilot stated that if the indicators hadto change their position in real-time to indicate achange in the location of the hazard, they shouldmove smoothly, and attention should be paid to theedges to make sure no flashing or other videoartifacts appear that might distract the pilot from thetask of landing. This pilot also stated that theindicators should fade in and out gradually inresponse to changing hazard conditions (unless thepilot turned them on or off.) A sudden appearanceof a hazard indicator, where there had been none,could be startling and potentially dangerous.Likewise any rapid motion or disappearance out ofthe corner of the pilot's eye during the landing couldbe distracting and potentially dangerous. Pilot #2concurred that there should be no motion in thehazard indicators. Pilot #3, the civilian pilot, statedthat slow motion on the surface of the indicatorcould conceivably be helpful to give an indicationof which way the airflow was moving within, but

that in general, fast motion could be distracting anddangerous.

AudioSome existing hazard warning systems for

commercial aircraft use audible warnings, e.g. a bellor voice. Participants in our study were askedwhether they would judge an audio indicator to behelpful or distracting. The consensus was againstusing audio. Pilots #1 and #2 were clear that theydid not want the hazard indicator to have any audiocomponent. Pilot #3 conjectured that a limitedaudio, such as a soothing female voice, might behelpful under certain limited conditions.

General considerationsOther comments the pilots made were that the

indicator should appear at the 180-degree point, thepoint in the approach where the pilot is abeam theintended landing spot facing downwind. Theindicators should then either turn off as the wheelscross the deck, or remain on throughout the landing.For yellow (caution indicated, but controllable)conditions, it was thought potentially helpful toleave the hazard indicator on display, as the pilotmight choose to fly into the indicated area (the"curtain"). Numeric indicators representing airflowspeed were not preferred; the pilots stated that theywouldn't have the time to read numbers as theyapproached the landing spot. All of the pilotspreferred an idealized representation rather thanexact visualization of airflow, again in the interestof keeping the display simple. One pilot suggestedjust painting the deck or the landing spot red oryellow. It was also suggested that more detailedoptions might be useful at the start of the approach.Perhaps a helicopter silhouette on the deck, or windarrows or airflow lines, could be selected by thepilot at that point, fading to a simpler version as thepilot flew closer. It was also pointed out that it wasimportant for the system to be credible, with nofalse positives or negatives. Finally, it was criticalthat the pilot be able to turn the system on and off,and that a vernier control be present to adjust thebrightness of the display based on the ambient light.

Conclusions and Future WorkA preliminary usability study of an airflow

hazard visualization system for helicopter pilotslanding on board a moving ship indicated that pilotswould use such a system if it were available on theiraircraft. They expressed a need to know moreabout airwake hazards and a desire to have theinformation presented to them in the cockpit as theywere landing. The preference was for a head-updisplay with “scene-linked” indicators vs. aninstrument panel display.

The pilots indicated that any airflow hazardsymbology should present the minimum criticalinformation such as location of the hazard andwhether it was a warning (yellow) or danger (red).There was no desire for detailed quantitativeinformation or even qualitative information such astype of hazard such as vortex, downdraft,turbulence, wind shear, etc. In other words, whatthe pilots are looking for is a decision supportsystem, not a scientific visualization system, andany future work in this area should be done withthis kept in mind. They want to be shown theeffects – e.g. hazards to aircraft – and not causes –e.g. this is a vortex caused by the wind curling upand over the deck edge with downdrafts of up to400 ft/minute. Extensive detail, motion, complexshapes, too many colors, were all considered toodistracting and possibly dangerous in the high-demand environment of shipboard helicopteroperations. Preference was strongly given to staticrather than dynamic indicators. Concerns wereexpressed over distractions such as motion inducingvertigo, confusing symbology causing doubt in thepilot’s mind, etc. Nevertheless, there was a cleardesire to have such a system in the cockpit.

“The Holy Grail”We close with a quote from one of the pilots in

our usability study, asked if he thought a system ofairflow hazard visualization might have thepotential to improve helicopter flight safety:

“...[This system] offers … a chance to avoidmishaps that have happened before, combined withthe opportunity to provide a greatly expandedoperating envelope…

“That’s the holy grail… to be able to bothincrease safety and increase operational capability

at the same time. Usually you don’t find somethingwith the potential to do both. Usually you eitherhave something that makes it a lot safer but tends toimpose certain operational restrictions…or youhave something that gives you greater operationalcapability but there’s risks associated withemploying that additional capability... In this caseyou actually have a concept that could potentiallygive you both.”

Further work is indicated and currently theauthor is conducting a flight simulation study usingthe preferred set of hazard indicators. Airflow dataand ship and helicopter aerodynamic models havebeen loaded into a high-fidelity rotorcraft flightsimulator [33] and scenarios have been createdwhere airflow hazards are known to be present nearthe landing sites on shipboard. A visual hazardindicator system has been developed andimplemented, and integrated into the display systemof the simulator. Experienced helicopter pilots withshipboard landing experience have been recruited tofly multiple approaches to a moving aircraft carrierunder extreme wind and turbulence conditions, bothwith and without visual hazard indicators. Data isbeing gathered both subjectively by having thepilots fill out questionnaires about the hazardvisualization system, and objectively by measuringflight path deviations, control surface motion andpilot workload, landing dispersion, and verticalspeed at touchdown.

AcknowledgmentsThis work was funded by the NASA Ames

Full-Time Graduate Research Program. The authorwishes to thank her adviser at UC Berkeley, Prof.Marti Hearst, for guidance and suggestions. Thiswork could not have taken place without the activesupport and advice of Kurtis R. Long of the NavyDynamic Interface Flight Test Group and FluidMechanics Laboratory at NASA Ames whogenerously shared his extensive knowledge.Thanks are also due to Advanced RotorcraftTechnology, Inc. for the use of their high-fidelityhelicopter flight simulator. The author deeplyappreciates their interest in and support of herresearch.

References[1] FAA National Aviation Safety Analysis Center,https://www.nasdac.faa.gov/.

[2] NASA Marshall Space Flight Center LidarTutorial,http://www.ghcc.msfc.nasa.gov/sparcle/sparcle_tutorial.html.

[3] Particle Image Velocimetry at NASA Glenn,http://www.grc.nasa.gov/WWW/OptInstr/piv/background.htm.

[4] Chambers, J., 2003, Concept to Reality, NASASP-2003-4529,http://oea.larc.nasa.gov/PAIS/Concept2Reality/wind_shear.html.

[5] Long, Kurtis R., 2003, Navy flight test engineer,personal communication.

[6] Wilkinson, C. H., S. J. Zan, N. E. Gilbert, J. D.Funk, 1998, "Modelling and Simulation of Ship AirWakes for Helicopter Operations."

[7] Williams, Suni L., Kurtis R. Long, 1997,“Dynamic Interface Flight Tests and the PilotRating Scale,” American Helicopter Society 53rd

Annual Forum, Virginia Beach, VA.

[8] Buning, P., 1989, “Numerical algorithms inCFD post-processing,” Computer Graphics andFlow Visualization in Computational FluidDynamics, von Karman Institute for FluidDynamics Lecture Series 1989-07.

[9] Strid, T., A. Rizzi, J. Oppelstrup, 1989,“Development and use of some flow visualizationalgorithms,” Computer Graphics and FlowVisualization in Computational Fluid Dynamics,von Karman Institute for Fluid Dynamics LectureSeries 1989-07.

[10] Lane, David A., 1996, “Visualizing Time-Varying Phenomena In Numerical Simulations OfUnsteady Flows,” NASA NAS Technical ReportNAS-96-001.

[11] Max, N., B. Becker, R. Crawfis, 1993, “FlowVolumes for Interactive Vector FieldVisualization,” Proceedings of Visualization ’93,IEEE Computer Society Press, Los Alamitos, CA,pp. 19-23.

[12] Post, F., J. van Wijk, 1993, “Visualrepresentation of vector fields: recent developments

and research directions,” Scientific VisualizationAdvances and Challenges, Academic Press, SanDiego, pp. 367-390.

[13] Sujudi, D., R. Haimes, 1995, “Identification ofSwirling Flow in 3D Vector Fields,” AIAA Paper95-1715, San Diego, CA.

[14] Turner, B., M.Y. Leclerc, M. Gauthier, K.Moore, and D. Fitzjarrald, 1994, “Identification ofTurbulence Structures above a Forest Canopy usingthe Wavelet Transform,” J. Geophys. Res. 99 D1,pp. 1919-1926.

[15] Uenking, M. D., M. F. Hughes, 2002, “TheEfficacy of Using Synthetic Vision Terrain-Textured Images to Improve Pilot SituationAwareness,” SAE World Aviation Congress.

[16] Bailey, R.E., R. V. Parrish, L. J. Kramer, S. D.Harrah, J. J. Arthur III, 2003, “TechnicalChallenges In the Development of a NASASynthetic Vision System Concept,” NATO RTA.

[17] Spitzer, C. R., R. V. Parrish, D. G. Baize, M.S.Lewis, 2001, The Avionics Handbook: SyntheticVision, NASA Aviation Safety Program.

[18] Alter, K. W., A. K. Barrows, C. Jennings, J. D.Powell, August 2000, “3-D Perspective PrimaryFlight Displays for Aircraft,” Proceedings ofHuman Factors and Ergonomics Society, SanDiego, CA.

[19] Latorella, Kara, Jim Chamberlain, 2002,“Graphical Weather Information SystemEvaluation: Usability, Perceived Utility, andPreferences from General Aviation Pilots,” NASAAWIN technical report.

[20] Britt, Charles L., Carol W. Kelly, 2002,“User's Guide For An Airborne Doppler WeatherRadar Simulation (ADWRS),” NASA TPAWS.

[21] Shaw, C. D., F.T. Jiang, R. M. Parry, B. Plale,A. Wasilewski, W. Ribarsky, N. L. Faust, 2001,“Real-Time Weather Data on Terrain,” NASAAWIN, SPIE Vol. #4368, Radar Sensor Technologyand Data Visualization.

[22] Spirkovska, L., S. K. Lodha, 2002. AWE:Aviation weather data visualization environment,Computers and Graphics, 26, pp. 169-191.

[23] Switzer, G., C. Britt, 1996, “Performance ofthe NASA Airborne Radar with the Windshear

Database for Forward-Looking Systems,” NASACR 201607.

[24] Holforty, Wendy, 2003, “Flight-Deck Displayof Neighboring Aircraft Wake Vortices,” Ph.D.dissertation, Stanford University.

[25] Roscoe, S. N., 1968, “Airborne displays forflight and navigation,” Human Factors, 10, pp. 321-322.

[26] Wickens, C. D., R. Carbonari, D. Merwin, E.Morphew, J. O’Brien, 1997, “Cockpit Displays toSupport Hazard Awareness in Free Flight,” NASATechnical Report ARL-97-7/NASA-97-4.

[27] Endsley, Mica R., 2001, “Designing forSituation Awareness in Complex Systems,”Proceedings of the Second International Workshopon symbiosis of humans, artifacts and environment,Kyoto, Japan.

[28] Lee, Alfred T., 1991, “Aircrew Decision-Making Behavior in Hazardous WeatherAvoidance,” Aviation, Space, and EnvironmentalMedicine.

[29] United States Department of Commerce, 1986,The Crash of Delta Flight 191 at Dallas-Fort WorthInternational Airport on 2 August 1985: MultiscaleAnalysis of Weather Conditions, National Oceanicand Atmospheric Administration, Boulder, CO.

[30] Wallace, L., 1993, Airborne Trailblazer,NASA Langley Wind Shear Program,http://oea.larc.nasa.gov/trailblazer/SP-4216/chapter5/ch5.html.

[31] McCann, R. S., D. C. Foyle, 1994,“Superimposed symbology: Attentional problemsand design solutions,” SAE Transactions: Journalof Aerospace, 103, 2009-2016.

[32] Nielsen, Jakob, 1993, Usability Engineering,Morgan Kaufmann, San Francisco, CA, pp. 93-98.

[33] Advanced Rotorcraft Technology, Inc., 1685Plymouth St., Suite 250, Mountain View, CA94043, http://www.flightlab.com.

KeywordsAugmented reality, flight-deck displays,

usability, human-computer interface, airflowhazards, data visualization, helicopters, humanfactors in aviation.

Author’s BiographyCecilia Aragon is a computer scientist at

NASA Ames Research Center and a Ph.D.candidate in the Computer Science Division at theUniversity of California, Berkeley. She earned herM.S. degree in computer science from theUniversity of California, Berkeley, and her B.S. inmathematics from the California Institute ofTechnology. She is also a former airshow and testpilot with over 5,000 cockpit hours.