Cockpit Display of Traffic Information: The Effects of Traffic Load, Dimensionality, and Vertical Profile Orientation

Transcription

1 Aviation Research Lab Institute of Aviation ARL University of Illinois at Urbana-Champaign 1 Airport Road Savoy, Illinois Cockpit Display of Traffic Information: The Effects of Traffic Load, Dimensionality, and Vertical Profile Orientation Amy L. Alexander and Christopher D. Wickens Technical Report ARL-01-17/NASA-01-8 December 2001 Prepared for NASA Ames Research Center Moffett Field, CA Contract NASA NAG

2 Abstract Eighteen certified flight instructors from the University of Illinois Institute of Aviation participated in an experiment exploring the design of the Cockpit Display of Traffic Information for free flight. Pilots flew a sequence of flight scenarios to compare the effects of traffic load, dimensionality, and orientation of the vertical profile view on a coplanar display on maneuver frequency, safety, and maneuver efficiency. Climbs and descents were found to be more frequent than other maneuvers. Climbs were specifically more common with the 2D coplanar displays while descents were more common with the 3D display. Both the rear-view and 3D displays were safer in terms of predicted conflict avoidance than was the side-view display. Climbing maneuvers were more safely implemented with the rear-view and 3D displays than with the sideview display. Altitude maintenance of lateral/vertical maneuvers was best with the side-view, then the rear-view, and then poorest with the 3D display. Airspeed efficiency of airspeed maneuvers was best with the rear-view display. Subjective workload was found to be highest with the 3D display, significantly lower with the side-view display, and lowest with the rearview display. Collectively, these findings suggest that a rear-view coplanar CDTI may be preferable to either a side-view coplanar or 3D display. 1

3 Introduction Free Flight Current air traffic control (ATC) and instrument flight procedures create delays and disruptions within the aviation industry despite the availability of airspace or unused runway capacity (Johnson, Battiste, & Bochow, 1999). The FAA and NASA have recently undertaken a research effort to examine specific ways to improve the efficiency of the National Airspace System. This program has been referred to as Free Flight and is officially defined as a safe and efficient flight operating capability under instrument flight rules in which the operators have the freedom to select their path and speed in real time. Air traffic restrictions are only imposed to ensure separation, to preclude exceeding airport capacity, to prevent unauthorized flight through special use airspace, and to ensure safety of flight. Restrictions are limited in extent and duration to correct the identified problem. (RTCA, 1995). Free flight has in fact been described as a system in which VFR flexibility is provided under IFR protection (Garland, Wise, Abbott, Benel, Hopkin, & Smith, 1995). The evolving air traffic management (ATM) system involves the current ATC goal of ensuring separation as well as the traffic flow management (TFM) system which strives to maximize the use of airspace and airports in the most efficient manner (Planzer & Jenny, 1995). New systems are currently being developed to provide safe separation of traffic while supporting more flexible flight paths. A potential result of free flight is that ATC will have less control over traffic in the en route phase of flight than it does today. Because of this, pilots may be expected to take a greater role in monitoring their own separation from traffic; or more likely, both pilots and controllers will be monitoring the automated systems that are providing separation. Free flight includes a transition to such highly automation-based procedures as well as to satellite technologies, digital communications, and an increase in collaboration between pilots and controllers (Blattner, 1997). While some people fear that free flight will reduce the need for ATC, it has been argued that controllers will still be necessary to handle system constraints such as weather, equipment failures, or any other unforeseen circumstances requiring human problem-solving capabilities. New operational procedures will be needed in order to integrate the advanced avionics and ground automation tools accompanying the free flight concept (Kerns & Small, 1995). These procedures will need to revolve around the use of such technologies as advanced navigation capabilities, automated position reporting, data link for communications, automated conflict detection aids, and, the focus of the current paper, the cockpit display of traffic information (CDTI; Abbott, Moen, Person, Keyser, Yenni, & Garren, 1980; Kreifeldt, 1980), a display which must represent the position of all other aircraft in the 3D space around a given aircraft. The CDTI is an expected evolutionary aspect of free flight (Shelden & Belcher, 1999; Johnson, Battiste, & Bochow, 1999), but at this time, there is no firm consensus on its appropriate format. The integration of the CDTI and other technologies should rely on previous research within each individual realm as well as their functioning within the free flight system as a whole. This research will then help guide the development of the operational procedures that will be necessary for the implementation of free flight. 2

4 The focus of the current research is on how to represent the three dimensionality of the airspace on a CDTI using either 3D (perspective) displays or 2D (coplanar) displays. The following review will first discuss the issues inherent with display dimensionality in general. Dimensionality issues will then be specifically related to the CDTI in terms of conflict avoidance performance under varying traffic densities. Orientation changes of the vertical profile component of the coplanar display will also be discussed, as well as the influence of maneuver characteristics on performance with the CDTI, and conversely, the influence of the CDTI format on the choice of those maneuvers. Display Dimensionality The displays used to support free flight, including the CDTI, must be carefully examined so that their dimensionality is appropriate for the task at hand. Three-dimensional displays are defined as using monocular and/or binocular depth or distance cues to create a 3D image (Wickens, 2000). Several different types of 3D displays have been proposed for aviation use, including panoramic displays (Busquets, Parrish, Williams, & Nold, 1994), pathway in the sky displays (Barfield, Rosenberg, & Furness, 1995; Fadden & Wickens, 1997; Flohr & Huisman, 1997; Haskell & Wickens, 1993; Nordwall, 1997; Reising, Barthelemy, & Hartsock, 1989; Theunissen, 1995), virtual cockpits (Haas, 1992), automatic status displays (Owen & Suiter, 1997), and electronic maps (Olmos, Liang, & Wickens, 1997; Wickens, Liang, Prevett, & Olmos, 1996; Wickens & Prevett, 1995). On the surface, it may seem that using a 3D display would benefit performance when representing a 3D environment. It has been found, however, that 3D displays may actually inhibit performance on certain tasks, and that different types of 3D displays may be better suited for certain tasks (Wickens, 1999). Three-dimensional displays can be distinguished by their viewpoint. Egocentric, or immersed, viewpoints (Figure 1a) depict the world as it would look from the pilot s position. Exocentric viewpoints (Figure 1b) show the aircraft within the world from outside, usually from above and behind, as if a camera were tethered to the back of the aircraft. Comparing egocentric and exocentric 3D displays to a 2D coplanar display (Figure 1c) reveals differences in information processing mechanisms, as well as task-viewpoint interactions (Wickens, 2000). The 2D coplanar display, for instance, imposes a visual scanning cost due to the presentation of lateral and vertical information on two different display panels. This scanning cost is further amplified to the extent that information must be integrated across the two panels depending on the task at hand, as implied by the proximity compatibility principle (Haskell & Wickens, 1993; Wickens & Carswell, 1995). 3

5 (a) (b) (c) Figure 1. Three frames of reference for aviation displays showing at the top, the viewpoint of the image generator and at the bottom, the image viewed by the pilot: (a) 3D egocentric or immersed, (b) 3D exocentric or tethered, (c) 2D coplanar. Adapted from Wickens (2000). Three-dimensional displays, on the other hand, provide a more realistic, integrated picture of the 3D world, an advantage addressed by the principle of pictorial realism (Roscoe, 1968). The benefits of this integration are reversed, however, for tasks that require more focused attention to one plane of space (Haskell & Wickens, 1993). Regardless of the task, a cost to almost all 3D displays is the geometric foreshortening of vectors manifested as the slant underestimation effect, the compression effect, or the line-of-sight ambiguity (Gregory, 1977; McGreevy & Ellis, 1996; Boeckman & Wickens, 2001). Slant underestimation is revealed by the tendency for people to perceive a slanted surface within a 3D display to have a smaller angle (relative to the vertical) than it actually does, or to be shorter and more parallel to the viewing screen than it really is (Perrone & Wenderoth, 1993). Adding depth cues to the display, however, will help decrease the effect of this bias. The compression effect is a spawned cost of portraying a 3D world on a 2D screen (Boeckman & Wickens, 2001). At least two of the three axes visible in a 3D display must be compressed in order for the 3D world to be viewed on a 2D screen. An increase in the amount of compression of an axis is coupled with a reduction in its resolution as the viewing plane or vector approaches an angle parallel to the line of sight. A bias in distance estimation often follows this reduced resolution as it is more difficult to examine information along a compressed 4

6 axis represented by a small number of pixels as well as a small visual angle (Boyer & Wickens, 1994). More cognitive effort may be required then, to accurately estimate distances and positions along this compressed axis, possibly through mentally stretching it to its uncompressed state (Barfield, Hendrix, & Bjorneseth, 1995). Of course, this mental stretching will only lead to accurate estimation to the extent that the transformation of the compressed axis to its uncompressed state is done perfectly. The line-of-sight ambiguity, which affects the determination of object location within a 3D volume (Gregory, 1977; McGreevy & Ellis, 1986), specifically reduces the amount of linear information available within a visible vector as that vector approaches the line of sight. In a 3D display, the line-of-sight ambiguity degrades knowledge of how far away an object is from ownship as well as its absolute distance orthogonal to the viewing plane (Wickens, Vincow, & Yeh, in press). This ambiguity is greater with the exocentric than the egocentric display (Figure 1) because there is ambiguity associated with both where the viewer perceives his or her own position to be as well as that of other objects (Wickens, 2000). The egocentric display, however, imposes an additional cost of only providing a keyhole view of the world, which requires panning of the visual scene to perceive the entire surrounding airspace. Task-specific effects. Comparing the above displays within task-specific contexts, reveals that no single display is best for all possible aviation tasks (Wickens, 2000). Flight path guidance, for example, is best supported by an egocentric display, such as a tunnel in the sky display (Figure 2). This task requires the integration of lateral, vertical, and longitudinal information regarding the flight path, current, and future position of the aircraft. The 2D coplanar display obviously does not provide such integration. Although providing integration, the exocentric display adds the additional cost of ambiguity, discussed previously, concerning the location of ownship relative to the desired flight path. Command Path Boxes Predictor Situational Awareness Support Wire Command Speed Bars Situational Awareness Support Wire Situational Awareness Support Vector Current Time Symbol Figure 2. Tunnel in the sky display. 5

7 Navigational judgments of the similarity between an image presented on a display and an object in the real world, as well as searching through an environment, are best performed using the exocentric 3D display (Olmos, Liang, & Wickens, 1997; Hickox & Wickens, 1999; Schreiber, Wickens, Renner, Alton, & Hickox, 1998; St. John, Cowen, Smallman, & Oonk, 2001). Integration requirements and realism deem the exocentric display to be better than the 2D coplanar display in this situation (Wickens, 2000). The keyhole effect, imposed by the egocentric display, may hide important features necessary to make comparisons, while the exocentric display always provides a larger view needed for such comparisons. A study investigating performance on a target acquisition task compared a 3D exocentric display to a 2D display and found that participants could perform the task much faster using the 3D display (Van Breda & Veltman, 1998). Tracking performance, in terms of roll-angle error, also revealed a 3D advantage over the 2D display. Furthermore, subjective mental workload ratings showed that participants felt less effort was required to perform the task using the 3D display. More precise spatial judgments are best made using a 2D coplanar display (Wickens, 2000). In fact, several studies have found poor performance on position-estimation tasks involving one or more spatial axes with an exocentric 3D display (McGreevy & Ellis, 1986; Tharp & Ellis, 1990; Kim, Ellis, Tyler, Hannaford, & Stark, 1987; Barfield & Rosenberg, 1995; Wickens, 1995; Wickens & May, 1994; Yeh & Silverstein, 1992; Boyer, Campbell, May, Merwin, & Wickens, 1995). As discussed earlier, both 3D displays impose line-sight-ambiguity problems, and the exocentric display, in particular, imposes a double ambiguity when judging the spatial relation between ownship and other elements in the environment. This ambiguity can be detrimental to making judgments of the location of traffic, weather, and terrain hazards, as well as to determining the future position of traffic with respect to ownship. For instance, a study by Jasek, Pioch, and Zelter (1995) found that a coplanar format supported superior performance over a variety of 3D formats for predicting collisions in a simplified ATC task. The coplanar display was also found to support better collision prediction than 3D displays in a follow-up experiment using more complex scenarios. A 2D advantage has also been found in relative position judgment tasks (St. John & Cowen, 1999; St. John, Oonk, & Cowen, 2000; St. John et al., 2001). St. John et al. (2001), for example, investigated the performance effects of 2D and 3D views on determining the relative position of two objects. Participants were asked to complete a series of tasks involving a block consisting of cubes with a ball placed above one of the cubes. All tasks required precise judgment of the location of the ball relative to the cubes on the three axes of space. It was found that participants were generally faster and more accurate using the 2D compared to the 3D views on all tasks. The line-of-sight ambiguity associated with 3D displays degraded performance in these relative position tasks by distorting the height and distance of the ball. Wickens and Prevett (1995) found that pilots flying simulated approaches to landing were more accurate in making judgments regarding the relative location of objects in space when using the 2D coplanar format than when using the 3D displays varying in egocentricity level. Coupling display dimensionality with map rotation in a similar paradigm revealed that pilots judged the relative bearing of terrain features faster with the 3D displays, yet were more accurate with the 2D coplanar format, indicating a speed-accuracy tradeoff and replicating both the scanning time costs of the coplanar display and the line-of-sight ambiguity effect of the 3D display (Wickens, Liang, Prevett, & Olmos, 1996). Olmos, Liang, and Wickens (1997) further 6

8 modified the above paradigm by portraying terrain on a wide screen display simulating the forward-field-of-view (FFOV) from the cockpit. A benefit for the 2D display was found in terms of accurately making vertical judgments in comparing the map display and the FFOV and also in making faster vertical judgments of surrounding terrain relative to ownship. The distinct advantages of 2D displays in making precise spatial judgments and of 3D displays in shape understanding tasks were further explored in a study which required participants to place antennas along a terrain map to establish line-of-sight communications (St. John, Smallman, Bank, & Cowen, 2001). The antennas had to be placed throughout the terrain so that they were in line of sight of each other while hidden from enemy units. The results showed that the 3D view produced slower solution times than the 2D view, revealing that antenna placement was performed better with the 2D than 3D view. This was inferred to be true because the accurate representations of distances and elevations inherent to the 2D view was more helpful to the precise understanding of the terrain than the more realistic 3D view that inevitably invited line-of-sight ambiguity costs. In contrast, when participants needed to choose a promising route through the general terrain layout, results revealed that the 3D view enabled participants to form a general understanding of the terrain more quickly than the 2D view. The authors suggest that these findings support a display design paradigm called Orient and Operate in which 3D views would be best for conveying general shape understanding and then 2D views would be used for solving tactical problems involving precise 3D judgments. 3D views are sometimes found to be more computationally complex and demanding than 2D views. As an example, 2D and 3D views have been compared in fmri studies examining the ability to maintain objects in spatial working memory as the number of objects moving through either view increased (Diwadkar, Carpenter, & Just, 2000). The results showed that increased fmri-measured activation, as a measure of cognitive workload, was associated with the largest number of objects, as well as with the 3D view. Furthermore, the percentage of activation over baseline was greater with the 3D view than with the 2D, and was greatest overall with the largest number of objects in the 3D view. The results also showed that participants took longer to respond to 3D than 2D probes. The problem of ambiguity associated with 3D displays may be addressed through the use of drop-lines (Ellis, 1993; Smallman, Schiller, & Cowen, 2000; St. John et al, 2001). Drop-lines, or altitude posts, unambiguously specify an aircraft s 2D location (X and Y on the ground plane) and also provide altitude (Z) information from the length of the drop-line. While the use of droplines may reduce or eliminate the problem of ambiguity, their use invites increased clutter through the addition of unnatural augmentations that may be detrimental to performance (Yeh & Silverstein, 1992). It is therefore important to determine the appropriate tradeoff between ambiguity and clutter. Smallman et al. (2000) found that the addition of drop-lines significantly improved localization performance on a task where participants reconstructed a 3D view on a test sheet. In comparing 3D and 2D views, the 2D view showed better ground localization performance only when the 3D view was not augmented with drop-lines. CDTI Dimensionality Dimensionality is an important design issue in representing the traffic around ownship on a CDTI. A 3D perspective CDTI (Ellis, McGreevy, & Hitchcock, 1987; Merwin, Wickens, & 7

9 O Brien, 1997) provides an advantage of integrating all three axes of the airspace in a single panel. However, as previously discussed, a 3D view can be somewhat ambiguous in representing the precise location and trajectory of aircraft, (Wickens, 2000), a troubling concern given that the CDTI may need to support precise conflict-avoidance maneuvering. Coplanar displays, which present each aircraft on a map view and a vertical view, on the other hand, allow for ease and precision in making both horizontal and vertical judgments. Potential disadvantages of the coplanar view in this context, however, include the cost of visually scanning between the two panels, the greater demand placed on working memory in retaining values from each panel to be compared or integrated with the other panel (e.g., What is the relative altitude of these two planes that appear to be converging in XY space? ), and the cognitive demands required to integrate or reconstruct the 3D environment from the 2D displays (Wickens, Merwin, & Lin, 1994). Conflict avoidance, as a measure of safety, is of utmost importance in determining the appropriate format of the CDTI. Ellis, McGreevy, and Hitchcock (1987) were the first to explore conflict avoidance in terms of display dimensionality. Pilots were required to monitor a traffic display to determine if an avoidance maneuver was necessary. The results showed that pilots took more time to decide if an avoidance maneuver was needed with a conventional, planar display than with a 3D perspective display. Furthermore, maneuvers chosen while using the 3D display were more likely to result in adequate separation from conflicts than those chosen with the 2D display. It is important to note here that the 2D display used in this experiment was uniplanar, not coplanar, and altitude was therefore represented with an alphanumeric, digital data tag as opposed to the spatial analog code used in coplanar and 3D displays. Bemis, Leeds, and Winer (1988) investigated operator performance on a conventional 2D versus a perspective 3D tactical display. Naval operators were required to detect threats and choose the closest interceptor for each detected threat. The 3D display was found to reduce errors in terms of false detections and omissions for detecting threats and in terms of false interceptions and omissions for selecting the closest interceptors. Errors were expected to be lowered with the 3D display in comparison to the 2D display because relative altitudes of interceptors and threats were presented on the 3D display. Response time was also decreased for choosing interceptors with the 3D display. This was expected because operators had to hook aircraft symbols in order to obtain altitude information while altitude was graphically represented on the 3D display without any need for hooking unless an exact numerical value was desired. The authors pointed out the need, however, to explore these results in terms of display density and complexity. Uniplanar and 3D perspective displays have also been compared in an air traffic control context. Burnett and Barfield (1991) required controllers to resolve impending air traffic conflicts, identify aircraft callsigns at extreme altitudes, and reconstruct air traffic scenarios for two levels of traffic density (7 or 17 planes). It was found that the perspective display, with droplines accurately representing altitude, was fastest in resolving conflicts at the lower traffic density, but performance was the same across displays at the higher density level. Identifying callsigns was faster for the perspective display than the plan view display at both traffic levels. Performance across displays and traffic density levels was the same for the reconstruction task. Overall, these results suggest an advantage for the 3D perspective display, but again, as with Bemis et al., this was only in comparison to a uniplanar display with an alphanumeric, digital data tag. 8

10 In comparing traffic detection and avoidance performance between the 3D display and the 2D coplanar display, such as those shown in Figure 3a and 3b, Merwin et al. (1997) have found that overall performance advantages in avoiding conflicts with a protected zone around each aircraft were supported by the coplanar format. More specifically, fewer predicted and actual conflicts resulted with the coplanar than with the 3D CDTI. Furthermore, the coplanar format fostered more flexible maneuvering strategies, and provided greater separation between ownship and other traffic. A follow-up experiment again examined display dimensionality in terms of traffic avoidance, but did so while adding a weather hazard that also needed to be avoided (Merwin, Wickens, & O Brien, 1997; O Brien & Wickens, 1997). Conflicts were again best avoided with the coplanar display as reflected by the reduction of weather conflicts and predicted and actual traffic conflicts. This advantage was especially prevalent when traffic was ascending or descending (e.g., the 3D displays were particularly problematic when the traffic was flying in three dimensions, rather than along a flat plane). The authors also investigated the effects of data base integration, by presenting both traffic and weather in a single display, and found that such integration supported superior performance compared to separated displays for both coplanar and perspective formats. This integration advantage was enhanced for the coplanar display, however, for two reasons. First, separating the data base across four panels in the coplanar condition led to an excessive amount of visual scanning. Second, integrating the data base onto a single 3D display invoked excessive clutter. Therefore, the coplanar integrated format was found to support the best conflict avoidance performance. In the above cases, integrating across the two panels of the coplanar display was less detrimental to performance than was the ambiguity of the 3D view. However, only a limited number of aircraft were used throughout those experiments. CDTI s may be expected, however, to portray a much greater geographical range of airspace, and therefore depict a much greater number of aircraft than in the simplified paradigms examined in the above studies (Johnson, Battiste, & Bochow, 1999). Increasing the number of aircraft, resulting in a more complex, cluttered traffic environment will require more mental integration and visual scanning between the two panels of the coplanar display, therefore possibly reducing, and even reversing, the advantage seen by Merwin et al., for the coplanar over the 3D format. Such a constraint on the coplanar benefit has not yet been explored by systematically increasing traffic load in a 2D/3D comparison; therefore, one of the issues to be addressed in the current experiment is how traffic load may modulate the effects of dimensionality in affecting pilot performance. 9

11 33 0 A: Vertical extent of ownship s protected zone Traffic 3 6 Current Future Traffic (a) Ownship 0 Traffic 3 6 Traffic Ownship A. Current vertical boundaries of protected zone Traffic Ownship Traffic (b) (c) Ownship Traffic B. Future vertical boundaries of protected zone Figure 3. Three display versions of the CDTI: (a) 3D, (b) 2D coplanar rear-view, (c) 2D coplanar side-view. The rendering of the side view profile in (c) is not what participants actually saw, but is a schematic designed to highlight the different orientation from figure (b). 10

12 Coplanar Viewpoint Orientation: The Profile View In addition to the influence of traffic density, and display dimensionality, a third issue needing to be addressed concerns the orientation of the vertical view in the coplanar display, if such is adapted. Figure 3b shows the behind vertical view, used by Merwin et al., in which both the plan view and the profile view represent left of the aircraft to the left of the display. Thus, each traffic symbol in its vertical depiction on the bottom, is directly positioned under its counterpart on the lateral display. In contrast, the side view in Figure 3c shows a longitudinal depiction of the flight path. The side-view display is examined because such a view appears to be relatively popular, as aviation designers are considering the implementation of vertical situation displays in the cockpit (e.g., DeJonge, 2000; Oman, Kendra, Hayashi, Stearns, & Burki-Cohen, 2001). Visual search issues become critical when dealing with coplanar displays because the number of aircraft icons increases twofold as traffic increases. The side-view representation would appear to be disadvantageous for understanding the 3D locations of multiple traffic aircraft because it amplifies the visual search/scanning cost associated with relating the map view and profile view representation of an aircraft in the 2D coplanar display, as shown in Figure 4. However, since side-view displays are being developed to represent distance from terrain, profile descents, and already exists in static form on approach plates, it is logical to explore the implications of representing traffic on these displays as well. Specifically, viewing ownship from behind on the vertical profile display, rather than from the side, allowed for vertical alignment (represented schematically by the vertical dotted lines in Figure 4a) between representations of the same aircraft on the two different panels. This vertical alignment, based on the principles of visual momentum (Woods, 1984), provides a cognitive link between the two panels such that fixation is directly reoriented upon scanning without the addition of a visual search cost associated with uncertain, inconsistent locations, as was seen with the side-view display. That is, in Figure 4b, each aircraft s vertical depiction is no longer directly located under its lateral depiction with the side-view display. This idea that visual scanning costs are affected by the consistency of aircraft representations across panels can be further explored within the context of the proximity compatibility principle (Wickens & Carswell, 1995). Essentially, the visual search/scanning process involved with relating the vertical and lateral depictions of traffic aircraft will be aided to the extent that there is greater proximity, and therefore less ambiguity, between the upper and lower representations of each aircraft, and hurt to the extent that there is less proximity. In Figure 4, this closer proximity can be seen with the rear-view display, while the side-view display invokes less proximity. 11

13 a) b) Figure 4. (a) Rear-view vs. (b) side-view representation of aircraft across panels. The dotted lines were not visible to pilots, but show how each representation in the top (map) display could be linked to its counterpart in the bottom (profile) display. Another reason the rear-view display is expected to support superior performance compared to the side-view display is that pilots have redundant heading and lateral separation information between the two panels in the rear-view display, while only the top-down panel contains heading information for the side-view display. This means that pilots have to scan between panels to gauge heading and altitude concurrently while using the side-view display. Such a difference might be expected to effect the quality of lateral conflict avoidance maneuvers, that depend heavily on extracting heading information. At the same time, the side-view display contains redundant longitudinal (along-track) information in both panels. For this reason, it may prove superior for judgments on maneuvers involving airspeed, which affect the aircraft s movement along the desired flight path. No study has directly compared these two views in a single experiment; however, a study examining how the amount of computational demand is related to fmri-measured activation compared grid-scanning and mental rotation tasks is somewhat relevant (Carpenter, Just, Keller, Eddy, & Thulborn, 1999). These two tasks are somewhat analogous to the demands imposed by rear and side-view vertical profile displays, respectively. In general, the grid-scanning task was found to invoke a smaller level of activation than the mental rotation task. This is consistent with the idea that less computational demand is associated with scanning objects in fixed positions as opposed to mentally rotating objects to compute and compare coordinates. These results can be generalized to the matching of aircraft between panels of a coplanar display in that the rear-view display allows for the scanning of objects directly in line with one another, while the side-view display requires the pilot to mentally rotate his/her view to determine which planes match up 12

14 across the two panels. It follows that pilot performance is expected to be superior with the rearview display than with the side-view. A pair of more operationally relevant experiments examining simulated combat aircraft maneuvering also support this conclusion by revealing the apparent advantage of the behind view (Olmos, Wickens, & Chudy, 2000). The results of one experiment which used the rear-view coplanar display revealed that visually linking the two panels facilitated navigational awareness in estimating 3D flight paths in comparison to the other experiment which employed the sideview coplanar display. Improvements with the rear-view display were also seen in reducing the total amount of time avoiding hazards, the response times to detect threats, and the errors in altitude judgments of threats. This research, however, did not involve traffic maneuvering. Furthermore, there were a number of other differences between displays in the two experiments that could also have contributed to the performance differences, so the extent to which the improved performance was explicitly related to display orientation could not be established. Maneuver Characteristics The finding that display representations influence the types of maneuvers chosen by pilots is particularly relevant to the current study (Ellis, McGreevy, & Hitchcock, 1987; Merwin & Wickens, 1996; and Pritchett & Hansman, 1997). A dimensionality effect on maneuver choice was first noted by Ellis, McGreevy, and Hitchcock (1987). Specifically, pilots showed a greater preference for making vertical maneuvers when using the 3D perspective display than when using the 2D uniplanar display. The authors concluded that the more natural presentation of vertical separation on the perspective display accounted for the increased use of vertical maneuvers. Merwin and Wickens (1996), however, found that comparing a 3D perspective CDTI to a 2D coplanar CDTI reversed the above trend so that vertical maneuvers were more commonly chosen with the 2D coplanar than with the 3D perspective view. A possible explanation for the increased use of vertical maneuvers with the coplanar CDTI is that the planar CDTI, used by Ellis et al., represented altitude using alphanumeric digital codes while the coplanar CDTI, used by Merwin et al., employed a spatial analog code to represent altitude which was more compatible with the spatial representation of traffic in the top-down view. This better representation of altitude in the coplanar view invited more vertical maneuvering. Merwin and Wickens also found that pilots using the coplanar display tended to maneuver vertically in the opposite direction of the conflict aircraft (e.g., climbed when the conflict aircraft descended), while pilots flying with the 3D display showed a tendency to maneuver in the same vertical direction as conflict aircraft. In summary, the 2D planar display used by Ellis et al. represented altitude in an incompatible, digital format which was not conducive to the implementation of vertical maneuvers. The 3D perspective display used by both Ellis et al. and Merwin et al. employed an analog representation of altitude which provided a more natural setting for vertical separation comparisons, but was somewhat ambiguous in that representation given the inherent ambiguity of 3D displays in depicting altitude. The 2D coplanar view used by Merwin et al. represented altitude in the most precise, unambiguous, analog fashion, which invited vertical maneuvering. This analysis shows that the better the vertical representation, the more often vertical maneuvers 13

15 are chosen. These studies, however, did not involve more than one or two other aircraft besides ownship. In general, initial testing of maneuvering in a free flight environment using a planar CDTI revealed a tendency for pilots to prefer resolving conflicts by maneuvering vertically (Abbott et al., 1980). Maneuvering within the vertical plane, however, conflicts with how pilots are instructed to maneuver under visual flight conditions according to current Federal Aviation Regulations (FAR, ). The FAR instructs pilots to make lateral maneuvers in avoiding traffic due to an inherent inability to determine the altitude of other aircraft. Altitude information provided by a CDTI, however, allows pilots to identify the initial altitude of conflicting aircraft and choose a vertical avoidance maneuver accordingly. The finding of a vertical preference in conflict avoidance maneuvers has been replicated several times since the Abbott et al. study (Palmer, 1983; Merwin & Wickens, 1996; O Brien & Wickens, 1997; Wickens & Morphew, 1997; Gempler & Wickens, 1998; Wickens & Helleberg, 1999; and Helleberg, Wickens, & Xu, 2000). Helleberg, Wickens, and Xu (2000) summarize three related explanations for why vertical maneuvers are chosen over prescribed lateral maneuvers. First, vertical maneuvers result in smaller deviations from the intended flight path and are therefore considered more efficient in terms of returning to the intended path after the conflict has been resolved (Krozel & Peters, 1997). Second, lateral maneuvers are third order control tasks while vertical maneuvers are only second order; hence, they are of a less complex cognitive nature (Wickens, 1986; Wickens & Hollands, 2000). Finally, vertical maneuvers take less time to implement and are therefore more effective under high time pressure. Helleberg, Wickens, and Xu (2000) also found that pilots tended to avoid adjusting airspeed to avoid a conflict despite the fact that airspeed maneuvers were found to be safest in terms of the amount of time spent in a predicted conflict (e.g., a state in which a loss of separation would occur within 45 seconds if no maneuver were undertaken). One possible explanation for this finding is that airspeed maneuvers are more cognitively complex to the extent that the relative lengths of the predictor lines had to be examined in order to determine the effects of airspeed changes. Furthermore, the time dimension, critical for the pilot s understanding of the implications of an airspeed maneuver, is neither graphically nor perceptually represented on the CDTI, while both lateral and vertical dimensions are portrayed quite clearly. Since display representations influence the types of maneuvers chosen by pilots, as previously noted, perhaps airspeed maneuvers could be induced by making the time dimension more salient. Efficiency Another key component in assessing performance effects of different display types deals with the types of maneuvers invited by such display types and their relative efficiency. May, Campbell, and Wickens (1995) explored the efficiency of 2D uniplanar and 3D displays across two experiments in an ATC paradigm. The first experiment required participants to issue a safe set of vectors to an aircraft requesting a flight path change to a new location among five other aircraft and terrain. Due to the constraints of the other aircraft and the terrain, participants had to issue at least one altitude and one heading change in order to keep the target aircraft from penetrating the minimum safe distances of those hazards. Accuracy, in terms of distance from the 14

16 final fix, was slightly better with the uniplanar format; however, this accuracy was achieved at the cost of efficiency in terms of the number of inputs made to revector the aircraft. Essentially, an accuracy/efficiency tradeoff was found for the 2D and 3D display, respectively. In the second experiment conducted by May et al. (1995), participants were first required to determine whether an aircraft would penetrate 1-4 weather volumes, and then issue a single vector to avoid the hazard. Participants stated the approximate vertical and horizontal distance of the aircraft from the weather at the estimated closest point. The findings revealed that clearances allowing the aircraft to pass closer to the weather were issued with the 2D compared to the 3D display. Again, this efficient vectoring allowing a closer passage was made at the cost of a greater number of heading changes being issued. The second phase of this experiment gave participants the responsibility of revectoring an aircraft to a desired new location. It was again found that the 2D display supported more efficient performance in terms of the close passage of the aircraft to the weather element. Merwin, Wickens, and O Brien (1997) explored format-induced maneuver efficiency across 2D coplanar and 3D perspective displays with either a 30 or 60 elevation angle. Pilots using the 2D coplanar display were found to exhibit more flexible maneuvering strategies, regardless of the vertical behavior of the conflict aircraft, than pilots using the 3D displays. This was evidenced by there being no bias for either vertical or lateral maneuvers with the 2D coplanar display when the conflict aircraft was approaching ownship at the same level, for example. There was a bias toward vertical turns in the face of level conflict aircraft, however, when using the 3D displays. In general, the lower efficiency associated with the 3D display was inferred to be due to the inherent ambiguity along the line-of-sight. Basically, the relative lateral and vertical positions of the aircraft to other aircraft, terrain, or weather was less certain with the 3D than the 2D display. This led pilots to be more conservative or cautious in creating paths (and in taking time to create paths) that were far enough away from the hazard to compensate for the distorted separation perception in the 3D display (Boyer, Campbell, May, Merwin, & Wickens, 1995). This hindrance to perception was amplified in the case of complex maneuver vectoring (Wickens et al., 1996). Summary and Goals of the Present Research Display dimensionality has been investigated in several contexts, including a free flight paradigm with the CDTI (Ellis et al., 1987; Merwin et al., 1997). While a 3D perspective CDTI integrates all three axes of airspace in a single panel, this view is ambiguous in representing the precise location and trajectory of aircraft. Coplanar displays, on the other hand, represent horizontal and vertical axes faithfully, allowing for ease and precision in making judgments of traffic location. There are potential costs, however, in the visual demands of scanning between the two panels and cognitive demands associated with reconstructing the 3D environment from the 2D display (Wickens et al., 1994). Ellis et al. (1987) originally explored the effects of dimensionality with a 2D UNIplanar display and 3D perspective display, revealing superior conflict avoidance performance with the 3D display. Comparing the 3D perspective display to a 2D COplanar display, however, as 15

17 Merwin et al. (1997) did, reversed previous findings in that the 2D coplanar display was then found to be more successful at avoiding conflicts than the 3D display. More flexible maneuver strategies were also fostered by the coplanar format, indicative of better maneuver efficiency. It is important to note here that both studies used only a limited number of aircraft (one or two) that may not have fully exploited the effects of workload on dimensionality. The present research increases the number of aircraft from two to six to ten in order to investigate the impact of increasing workload on conflict avoidance and maneuver efficiency. As side-view displays are becoming increasingly popular within the aviation community for the representation of terrain and profile descents, for example, it is important to determine the effects of also representing traffic. As discussed previously, representing traffic on a side-view display would appear to be disadvantageous for understanding the 3D locations of multiple aircraft because it amplifies the visual search/scanning cost associated with relating the map view and profile view representations of aircraft in the 2D coplanar displays. It is the goal of the current study to compare performance with the rear and side profile-view representations of aircraft, as this has not yet been done. Previous research has revealed that display representations influence the types of maneuvers chosen by pilots (Ellis et al., 1987; Merwin et al., 1996; and Pritchett & Hansman, 1997). Merwin et al., in particular, found that vertical maneuvers were more commonly chosen with the 2D coplanar than with the 3D perspective CDTI. A vertical preference for conflict maneuvers in general has been replicated several times (Abbott et al., 1980; Palmer, 1983; Merwin & Wickens, 1996; O Brien & Wickens, 1997; Wickens & Morphew, 1997; Gempler & Wickens, 1998; Wickens & Helleberg, 1999; and Helleberg, Wickens, & Xu, 2000). Again, this finding has not been explored under increasing workload levels as will be done in the present study. The purpose of the current study was to examine the effects of dimensionality, profile view orientation, and traffic density on conflict avoidance (safety), maneuver choice, and efficiency. Pilots were instructed to fly to a predetermined waypoint while avoiding conflicts and minimizing deviations from preset values in altitude, lateral position, and airspeed. Scenarios were designed to include two conflict aircraft which would have to be avoided through maneuvers chosen by the pilots. It was expected, from the previous discussion of Merwin et al. s (1997) findings, that the coplanar display would be superior to the 3D display at low workload due to inherent ambiguities of the 3D display, but the 3D exocentric display (with drop-lines) would support progressively better conflict avoidance performance and maneuver efficiency compared to both of the coplanar displays as traffic density (workload) increased. This was expected due to the increased complexity and demand that would be associated with mentally integrating and visually scanning between the two panels of the coplanar display as the traffic density increased. It was also expected that vertical maneuvers would be more prevalent overall than lateral maneuvers, and would especially be more frequently employed with the 2D coplanar displays than with the perspective display, as shown in previous studies (Merwin et al., 1997; Wickens, Helleberg, & Xu, 2001). 16

18 While no previous research has directly compared the two 2D coplanar displays, it was predicted that the rear-view coplanar display would support superior conflict avoidance and maneuver efficiency performance compared to the side-view coplanar display due to the one-toone, consistent nature of matching aircraft symbols across panels in the rear-view format. The inconsistency involved with representing the two aircraft views in the side-view display was specifically expected to amplify the visual search/scanning cost associated with 2D coplanar displays as described by Wickens (2000). We have described research on the overall safety of displays (2D coplanar safer than 3D) and on the overall choice of maneuvers (vertical preferred) and how this choice is modified by display type (coplanar invites more vertical). However, we have not considered the extent to which different maneuvers with different displays may lead to different levels of safety (conflict avoidance). For example, it is possible (and optimal) that a particular display format may invite more maneuvering of a type for which that particular maneuver-display combination is most safe. This issue will also be examined in the current study. Participants Method Eighteen certified flight instructors (13 male, 5 female) from the University of Illinois Institute of Aviation flew a sequence of flight scenarios designed to compare the three formats of the CDTI. The mean number of flight hours was 1206 hours, with a mean of 237 instrument flight hours. Equipment This experiment was conducted on a low fidelity flight simulator consisting of a Silicon Graphics Iris Octane workstation, a Silicon Graphics 20-in. color display screen, and a flight stick controlling pitch, roll, and throttle. The display screen resolution was 1280 * 1024 pixels and ran at a frequency of 60 Hz. Pitch was controlled by moving the flight stick forward or backward; roll was controlled by moving the flight stick to the right or left. Increased speed was induced by pressing the button on top of the flight stick; decreased speed resulted from pressing the trigger at the front of the flight stick. Speed increased or decreased at a rate of 5 knots/sec of the button being pushed. Traffic Symbology The CDTI presented ownship and traffic, each with 45-second predictor lines. Ownship s predictor line was calculated based on its position, velocity, and rate of turn. Hence, the predictor line would curve in the direction of the bank angle, with a radius based on the degree of aircraft bank. All other predictor lines were based upon the pre-programmed flight path of the individual traffic aircraft. A cylindrical protected zone around ownship, 1000ft. above and below and 3mi. in radius, was also represented. Ownship s aircraft, predictor line, and protected zone were magenta, while all other traffic icons and predictor lines were light gray (see Merwin and Wickens (1996) for a fuller description). 17

19 Predicted conflicts were depicted through the use of orange threat vectors extending from the position on ownship s predictor line indicating the predicted position of ownship at the point of closest pass with the traffic (Figure 3). This threat vector moved along the predictor line toward ownship as the threat moved closer, representing the minimum time to loss of separation. If the threat vector reached another aircraft s predictor line, a state of predicted conflict was defined to occur and the predictor lines of both aircraft would turn white. If sufficient avoidance maneuvers were not taken within 45 seconds after this point, an actual loss of separation would occur and the predictor lines would turn yellow. Display Formats The simulation display presented the CDTI in one of three formats. For all formats, an attitude indicator was placed at the top of the display with airspeed and altitude tapes to the left and right of the display, respectively (Appendix A). A heading indicator was also included as part of the CDTI. All displays consisted of icons representing ownship, traffic, and conflicts as previously described. Coplanar formats. The coplanar display consisted of two windows offering a horizontal, top-down (X-Z axes) view and either a vertical, forward-looking (X-Y axes; Appendix B) or vertical, side-looking (Y-Z axes; Appendix C) view projected orthogonally (without perspective information). The horizontal, top-down display showed the previously described symbology overlaid on a grid of equi-spaced lines representing 5 nautical mile increments. Each line was composed of dots positioned at intervals of 1 nautical mile. This grid rotated with ownship to provide consistent spacing of traffic symbology from ownship s perspective. The lateral scale (number of miles) of the forward-looking view was the same as the top-down display so that the two representations of each aircraft were always matched one-to-one. In other words, each aircraft depicted on the top-down display was always directly above its representation on the forward-looking, vertical display. This was not true for the side-looking view, in that ownship was depicted so that more of the airspace in front of the aircraft (24nmi.) was shown than the airspace behind (6nmi.). Both vertical displays contained two sets of horizontal lines indicating the altitude boundaries of ownship s protected zone. The current 2000ft. altitude region surrounding ownship was depicted by two solid magenta lines. The predicted vertical boundaries of ownship s protected zone was depicted by two dashed yellow lines. 3D format. The 3D display (Figure 3a) depicted an integrated view of the three spatial dimensions viewed from an exocentric position above and behind ownship. The elevation angle of the viewpoint was 45, with an azimuth offset of 10 in the clockwise direction. This azimuth offset was used so that ownship s predictor line would not lie on the line of sight projection. The display showed the previously described symbology along with reference lines to unambiguously show the horizontal positions of the aircraft icons and the ends of the predictive lines. The reference lines extended to the grid and contained yellow regions indicating the vertical boundaries of ownships s protected zone, intending to provide unambiguous relative altitude information. 18

20 Task The task involved flying direct routes to predetermined waypoints while encountering other aircraft. The pilot was to determine and execute what maneuvers would be necessary to avoid conflict with the other aircraft, while minimizing deviations in speed, heading, and altitude from pre-specified target values. After determining that the conflict had been resolved, the pilot was instructed to return to the flight path to intercept the predetermined waypoint. Each trial ended when ownship s aircraft flew across a line running through the waypoint orthogonal to the original path. Two independent variables were manipulated in this experiment. Display type was manipulated by providing the coplanar forward-looking, coplanar side-looking, or 3D display (Figure 3). Traffic level was manipulated by including 2, 6, or 10 aircraft in addition to ownship. This created low, medium, or high levels of traffic, respectively. Each trial, regardless of traffic level, contained one primary (actual) and one secondary (potential) conflict. In other words, an actual loss of separation and a potential loss of separation would occur if the pilot remained on course toward the predetermined waypoint. Potential conflict aircraft were programmed to block one of the likely routes the pilot might choose to avoid the actual conflict aircraft. Conflict aircraft randomly approached from six different horizontal angles: 45, 90, and 135, from both the left and the right. Conflict aircraft also approached at the same altitude, from below (while climbing), or from above (while descending). These traffic geometries are represented in Figure 5. Approaching Crossing Overtaking Figure 5. Traffic geometries. 19

21 Conflict aircraft were programmed to enter ownship s protected zone either in front of or behind ownship. A full representation of the different possible behaviors of intruding aircraft can be seen in Figure 6. Randomly varying the geometry of the traffic encounters both horizontally and vertically without replacement, as indicated above, allowed testing to occur across a variety of possible situations. All other traffic were randomly placed throughout the display on trajectories that would be unlikely to produce conflicts in the face of any maneuver a pilot might choose (i.e., at substantially different altitudes or heading away from ownship s trajectory). Although considered irrelevant to the maneuver choice, this conflict-free property was not known to the pilots beforehand; therefore, the entire display had to be monitored so that avoidance maneuvers would not lead to conflicts with the surrounding aircraft. In particular, the identity of the conflict aircraft could not be assessed immediately by some superficial feature, such as their distance from ownship. Intruder Airspeed > Ownship Intruder Airspeed < Ownship Right Left Overtaking Right Left Crossing Crossing Approaching Approaching Ascending Level Descending Ascending Level Descending Figure 6. Intruder behavior. Design A within-subjects manipulation of display type was used. Pilots flew a series of 18 trials according to instrument flight rules. Pilots were separated into groups of 3, forming a total of 6 cohorts. The presentation of display type was counterbalanced across cohorts so that every possible combinatory order of the displays was used. Low, medium, and high traffic trials were also separated into groups of 3. In other words, there were three main groups of trials, each containing a low, medium, and high traffic condition, for a total of 9 distinct trials. These trials were counterbalanced within each cohort across the 3 pilots so that every display-type order was paired with each main group of trials containing the above traffic conditions (workload). Since traffic geometries and intruder behaviors have been shown to affect the types of avoidance maneuvers used with different display types (Merwin & Wickens, 1996; Wickens, Helleberg, & Xu, 1999), both were counterbalanced so that they were seen with each display type * workload combination by two subjects in every condition. After proceeding through those 9 trials using the 20

22 three different display types, the trials were then repeated using the reverse ordering of the display type. Thus, each pilot experienced a given combination of display * workload twice. Procedure The experiment was conducted in the Beckman Institute at the University of Illinois. When pilots arrived at the laboratory, they completed a statement of consent form (Appendix D) and a brief questionnaire (Appendix E) regarding their previous flight experience. Pilots then read experimental instructions explaining the task (Appendix F) and were shown illustrations (Appendices A, B, C, G) while the experimenter read descriptions of the CDTI symbology (Appendix H). Pilots then flew a series of practice scenarios to familiarize themselves with the task and CDTI symbology, as well as the three different displays described previously. Once practice sessions were successfully completed, pilots began the experimental session as specified above, flying 18 trials. Each trial lasted approximately 2 to 3 minutes. Pilots were instructed to avoid both predicted and actual conflicts while minimizing deviations in speed, heading, and altitude from the target values provided at the beginning of each trial. Subjective mental workload ratings were collected using the NASA TLX scale (Appendix I) after the last three blocks of trials so that the subjective workload experienced with each display type could be compared. Dependent Measures Results Performance measures included: 1) the frequency of maneuvers; 2) the safety in terms of time spent in a state of predicted conflict; 3) the efficiency of avoidance maneuvers defined by the amount of deviation from the target heading, altitude, and airspeed values; and 4) subjective mental workload ratings. Two data points were removed for being more than three standard deviations from the mean (one for combined lateral/vertical maneuvers for altitude efficiency and one for airspeed maneuvers for airspeed efficiency). Maneuver frequencies. The raw data were used to create graphs which were examined to determine the types of maneuvers chosen by pilots for each trial, using similar procedures to those employed by Wickens, Helleberg, and Xu (in press). Maneuvers were categorized as being lateral, vertical descents, vertical climbs, airspeed, or combined lateral/vertical according to the pilot s control inputs (roll, pitch, deceleration/acceleration), flight parameters (altitude, heading, and airspeed deviations), and ownship s relative horizontal position to the original flight path (X deviation). Figures 7 through 11 depict typical trials of respective maneuver types, and illustrate the sort of data that was employed to accomplish the categorization. Figure 7 shows the changes in control inputs and flight parameters associated with executing a lateral maneuver. The altitude deviation and Y stick parameters remained relatively stable, while the airspeed deviation increased as a function of the lateral maneuver. Essentially, the pilot did not try to maintain the prescribed airspeed, as indicated by the lack of button presses. The figure shows a turn to the left in the roll and X stick parameters, followed by a change in heading deviation. A change in X deviation then followed indicating a change in horizontal position away from the original flight 21

23 path. The pilot then turned back to the original flight path. The P01 and P02 alarms refer to the aircraft that ownship came into conflict with. A step up in the line indicates a predicted conflict while a step down shows an actual conflict. It can be seen that this pilot spent three different periods in predicted conflicts. X Deviation P02 Alarm P01 Alarm Button Press Y Stick X Stick Pitch Altitude Deviation Roll Heading Deviation Airspeed Deviation Figure 7. Lateral Maneuver. 22

24 Figure 8 depicts the time history of a vertical descent maneuver. The heading and airspeed deviations remained almost completely stable throughout the trial. The figure shows a pitch down immediately followed by a change in altitude deviation. The pilot successfully avoided all conflicts through the vertical descent maneuver, and also maintained prescribed heading and airspeed values. Button Press Y Stick X Stick X Deviation Pitch Airspeed Deviation Heading Deviation Altitude Deviation Figure 8. Vertical descent maneuver. 23

25 Figure 9 presents a typical vertical climb maneuver. Heading and airspeed deviations remain stable while a pitch up is followed by a change in altitude deviation. The vertical climb maneuver was initiated while the pilot was in predicted conflict. An actual conflict then ensued until the pilot had achieved a high enough altitude above the other aircraft. Altitude Deviation P01 Alarm Button Press X Stick Y Stick Pitch X Deviation Heading Deviation Airspeed Deviation Figure 9. Vertical climb maneuver. 24

26 Figure 10 shows the typical pattern of an airspeed maneuver. The heading and altitude deviations remained stable throughout the trial. Airspeed was accelerated as indicated by a step up in the button press parameter. Such button presses were associated with changes in airspeed deviation. This pilot was not successful in using airspeed to avoid conflicts. P03 Alarm Button Press Y Stick X Stick X Deviation Roll Pitch Heading Deviation Altitude Deviation Airspeed Deviation Figure 10. Airspeed maneuver. 25

27 Figure 11 depicts a combined lateral/vertical descent maneuver. An initial pitch down led to a vertical descent maintained throughout the trial. Upon entering the descent, airspeed deviation increased until the pilot pressed the button to decrease ownship s airspeed to the prescribed 350 knots. The figure shows that the vertical descent maneuver did not allow the pilot to avoid a predicted conflict. The pilot then initiated a left turn as indicated by the roll and X stick parameters, followed by a change in heading deviation. The combined lateral/vertical descent was only successful in avoiding a predicted conflict for a brief time. X Deviation P03 Alarm Button Press Y Stick X Stick Roll Pitch Heading Deviation Airspeed Deviation Altitude Deviation Figure 11. Combined lateral/vertical descent maneuver. Figure 12 presents the distribution of maneuver frequencies as a function of display type. The figure illustrates the clear dominance of vertical over lateral maneuvers χ 2 (1, N = 324) = 97.79, p <.01, with a slight preference for climbs over descents χ 2 (1, N = 251) = 2.9, p <.10. Both of these trends replicate the effects observed by Wickens, Helleberg, and Xu (2001) in a full mission simulation with the rear-view coplanar CDTI and were also representative of the majority of participants (16 out of 18 pilots: 3 at p <.01, 6 at p <.05, 2 at p <.10, and 5 at p >.10). Vertical maneuvers were further explored by examining whether display dimensionality modulated maneuver preferences as had been observed by both Ellis et al. and Merwin et al.. This issue was examined by several contrasts. First, there was no overall difference in vertical maneuver preference for the 3D (27%) versus the coplanar displays (rear-view: 25%, side-view: 26

28 26%). Second, however, within the vertical maneuvers, it appears that dimensionality did influence the tendency to climb versus descend. In this regard, it was revealed that the side-view 2D coplanar display did support more climbing maneuvers than the 3D display χ 2 (1, N = 93) = 6.53, p <.05. The 3D display supported more descending maneuvers than both the side-view χ 2 (1, N = 77) = 9.47, p <.01, and the rear-view χ 2 (1, N = 87) = 3.32, p <.01 displays. This interaction was also examined by comparing climbs and descents within each display type. These comparisons showed that the side-view display elicited significantly more climbing than descending maneuvers χ 2 (1, N = 83) = 13.12, p <.01. The rear-view display showed the same trend, although it was not significant χ 2 (1, N = 81) = 1.49, p >.10. In contrast, the 3D display supported more descending than climbing maneuvers χ 2 (1, N = 87) = 3.32, p <.10. Stated in other terms, the 3D display induced a descent preference and climb aversion that was not only absent with the rear-view display, but reversed with the side-view display. 70 rear-view side-view 3D Frequency lateral descend climb airspeed lat/vert Maneuver Choice Figure 12. Maneuver frequencies by display type. Safety. The mean time spent in a state of predicted conflict (loss of separation within 45 seconds if no maneuver was taken) was the primary measure of safety. The time spent in an actual conflict was not used because such occurrences, while highly correlated with predicted conflicts, occurred very rarely. Figure 13 presents the mean time in predicted conflict as a function of display type and traffic density. These data were analyzed in two ANOVA s 27

29 corresponding to the two separate display issues addressed. One ANOVA examined display dimensionality including the 3D and rear-view display, replicating the study of Merwin et al., and the second examined the effect of vertical view orientation including only the two coplanar displays. A log transformation of the mean seconds per trial spent in predicted conflict was performed to normalize the data. For the dimensionality analysis, a 2(display type: rear-view/3d) x 3(traffic load: 2/6/10) ANOVA revealed no significant differences (all p >.10). For the orientation analysis, a 2(display type: rear-view/side-view) x 3(traffic load: 2/6/10) ANOVA revealed that the rear-view display supported marginally significant superior conflict avoidance performance over the side-view display, F(2, 175) = 2.84, p <.10, a benefit of 4.6 seconds. There was no significant effect of workload for the second ANOVA, F(2,175) = 1.64, p >.10, nor did workload interact with display type, F(2,175) = 0.061, p > Side-View Rear-View 3D Mean Time (Sec) Traffic Density Figure 13. Potential conflict avoidance rate by display type and traffic density. Safety was further explored by examining the predicted conflict time as a consequence of making certain avoidance maneuvers across display type. Figure 13 suggests that the side-view display is less safe than the other two displays. We asked the question why. Is it less safe across all maneuvers, or is it less safe only for certain maneuvers, particularly those that were invited more frequently by this side-view format (e.g., climbing maneuvers; see Figure 12). Correspondingly, this analyses can address whether the equivalence of safety between the 2D rear-view and 3D displays is the result of equivalence across all maneuvers, or results from a tradeoff of the probability of a maneuver and its safety. In order to answer this question, we first assessed the overall safety of each maneuver type, as shown in Figure

30 50 45 Mean Time in Predicted Conflict lateral descend climb airspeed lat/vert Maneuver Choice Figure 14. Mean time in predicted conflict by maneuver choice. Figure 14 presents the distribution of mean times spent in predicted conflicts as a function of maneuver choice, collapsed over display type. This figure illustrates that the safety of maneuvers differed significantly from each other F(7, 266) = 5.7, p <.01, most prominently with both vertical maneuvers being much safer than lateral maneuvers, and with those involving airspeed and combined maneuvers falling in between. It is important to note here that vertical maneuvers were also the most prevalent, as shown in Figure 12. The data were then broken down separately by the safety of each maneuver for a given display. Accordingly, the overall safety of a display, as shown in Figure 13, can be viewed in Table 1 as the average of the safety of the display for a maneuver (top number in each cell) weighted by the frequency with which the display invited that maneuver (bottom number). In the two right columns are the planned contrasts involved in each analysis (dimensionality and profile view orientation). 29

31 Table 1. Safety/probability matrix by display type and maneuver choice. Significant results are in bold type. Display Contrasts Maneuver Rear-view Side-view 3D Rear vs. 3D Rear vs. Side Lateral t = p < 1.0 t = p < 0.9 Descend t = p < 0.5 t = p < 0.7 Climb t = p < 1.0 t = 1.88 p < 0.10 Airspeed t = p <.06 t = p = 0.13 Combined t = p < 0.3 t = p < The side-view display was particularly unsafe because it invited a lot of climbing maneuvers (54%), for which it was less safe than the other two displays (22 vs. 14 and 14 sec, or about 56% more time in predicted conflict, t(103.4) = 1.88, p <.10.) The 3D invitation to descend (seen in Figure 12) could also be viewed as non-optimal because, for this display, climbing maneuvers were consistently more safe (14 sec) than were descending maneuvers (18 sec). Finally, all three displays avoided lateral maneuvers which were the least safe (see Figure 14). Maneuver efficiency. How display format affected maneuver efficiency was examined as a function of the type of maneuver chosen. Efficiency was assumed to be greatest if pilots deviated the smallest amount from the target altitude, airspeed, and heading values. A series of ANOVA s were conducted to determine the effects of these specific maneuvers on efficiency across the three display types. Furthermore, in the following, we only include measures of maneuver efficiency as deviations along, or closely related to the axis chosen for the maneuver (e.g., lateral deviations examined only when lateral or combined lateral/vertical maneuvers were chosen). Lateral efficiency. The overall lateral efficiency of each display type was calculated with an equation where the lateral efficiency (in degrees of heading) of each display given a particular maneuver was weighted by the probability that that maneuver type was chosen. Only trials containing the relevant lateral axis were examined. Table 2 presents a matrix of the efficiency and probability values of these lateral maneuver choices by display type. 30

32 Table 2. Lateral efficiency/probability matrix by display type and maneuver choice. Display Contrasts Maneuver Rear-view Side-view 3D Rear vs. 3D Rear vs. Side Lateral t = p < 0.8 t = p < 0.2 Combined t = p < 0.5 t = p < 0.6 Weighted Avg While the data failed to reveal any significant differences either in overall efficiency of the displays or for displays within the two maneuver types (p <.10), the data in the table do reveal a trend such that the rear-view display exhibited the best lateral efficiency performance on both lateral maneuver types. It is important to note, however, that both lateral and combined lateral/vertical maneuvers were used infrequently with all displays and these findings are therefore based on a relatively small sample size with low power that may prevent what might be an important finding from being significant. Vertical efficiency. Table 3 presents a matrix of vertical efficiency (in feet) and probability values according to display type and maneuver choice, for those maneuvers that either directly involved the vertical dimension (descents, climbs, and combined lateral/vertical) or, in the case of airspeed, a maneuver that often indirectly influences altitude. Table 3. Vertical efficiency/probability matrix by display type and maneuver choice. Significant results are in bold type. Display Contrasts Maneuver Rear-view Side-view 3D Rear vs. 3D Rear vs. Side Descend t = p < 0.4 t = p < 0.5 Climb t = p < 0.9 t = p < 0.4 Airspeed t = p < 0.4 t = p < 0.8 Combined t = p < 0.10 t = p < 0.10 Weighted Avg As with lateral efficiency, the data revealed no overall differences in vertical efficiency as shown across the bottom of the table. Only combined lateral/vertical maneuvers showed a display effect for vertical efficiency. Altitude maintenance of combined lateral/vertical 31

33 maneuvers was best with the side-view display, intermediate with the rear-view, and worst with the 3D display, F(2,11) = 8.76, p <.01. Planned comparisons revealed marginally significant effects for these combined maneuvers when comparing the rear-view and 3D displays, t(6.6) = 1.887, p <.10, and when comparing the rear-view and side-view displays, t(4.9) = , p <.10. It is important to note here that this increased efficiency of the side-view display, which supported the infrequent dual-axis maneuvers, was purchased at a substantial (although nonsignificant) cost of safety (see Table 1). Airspeed efficiency. Table 4 shows airspeed efficiency values (in knots) and corresponding probabilities according to display type and maneuver choice, in the same format as presented for vertical deviations. Table 4. Airspeed efficiency/probability matrix by display type and maneuver choice. Significant results are in bold type. Display Contrasts Maneuver Rear-view Side-view 3D Rear vs. 3D Rear vs. Side Descend t = p < 1.0 t = p < 0.9 Climb t = p < 0.3 t = p < 0.10 Airspeed t = p < 0.10 t = p < 0.10 Combined t = p < 0.5 t = p < 0.6 Weighted Avg The comparisons revealed three marginally significant trends. First, on climbing maneuvers (which were, as we noted, most frequent), the side-view display supported fewer deviations than the rear-view display, t(91.2) = -1.91, p <.10. Second, on the rare occasions when pilots did choose to maneuver directly by adjusting airspeed, efficiency was greater (deviations were smaller) with the rear-view display than with either the side-view, t(7.5) = 2.13, p <.10, or the 3D displays, t(9.6) = -1.89, p <.10. Although workload did not generally interact with displays on any of the efficiency measures, we did observe one marginally significant influence of workload. That is, a 3(display type) x 3(traffic load) ANOVA revealed a marginally significant effect of traffic load on airspeed efficiency, F(2, 315) = 2.82, p <.10. Subjective mental workload. NASA TLX scale data, as shown in Figure 15, were submitted to two planned comparison t-tests. The results of the analysis revealed a significant difference between the rear-view and 3D displays, t(33.1) = 3.12, p <.01. Subjective mental workload was rated highest with the 3D display (11.83), intermediate with the side-view display 32

34 (8.46), and lowest with the rear-view display (6.95). However, the subjective mental workload associated with the rear-view display was not significantly different from that with the side-view display, t(33.7) = 1.11, p > NASA TLX Composite Score D Side-View Rear-View Display Type Figure 15. Mean NASA TLX scores by display type. Discussion The purpose of this study was to examine the effects of the dimensionality and profile view orientation of a CDTI on conflict avoidance performance as assessed by safety (time in conflict), maneuver choice, and maneuver efficiency, in a low fidelity free flight simulation. It was hypothesized that the 3D display would support progressively better conflict avoidance performance and maneuver efficiency compared to both of the coplanar displays as workload levels increased (increased number of aircraft). Although previous research has revealed superior performance on the 2D rear-view display, compared to the 3D display (Merwin & Wickens, 1996; Merwin, Wickens, & O Brien, 1997; O Brien & Wickens, 1997), these studies only used a limited number of aircraft, which did not fully explore the scan/clutter tradeoff that exists between 2D and 3D displays. While the rear-view/side-view profile orientation had not, until now, been directly compared in a CDTI study, it is an important issue considering the ever-growing popularity of vertical situation displays within the aviation community (DeJonge, 2000; Oman et al., 2001). It was hypothesized that the rear-view coplanar display would support better conflict avoidance performance and maneuver efficiency compared to the side-view coplanar display due to the one-to-one matching of aircraft representations across the rear-view panels. 33

35 The current study addressed these issues by varying traffic load across dimensionality and profile view orientation to determine the effects on pilot performance. These effects are discussed first in terms of the effects of dimensionality by contrasting the 3D view with the rearview coplanar display, with which it has previously been compared, then the effects of profile view orientation by comparing the two coplanar views, and finally, in terms of the influence of traffic load itself. Display Dimensionality Safety. The safety of performance with the rear-view coplanar display was not significantly different from that with the 3D display. It was expected that safety would be biased in favor of the 3D display at high levels of workload because the visual search induced by matching an increased number of aircraft representations would be more demanding with the coplanar display. This expected bias between the rear-view 2D coplanar and 3D display was not found, however, indicating that the integrative benefit of the vertical alignment between the two representations of each aircraft in the coplanar view (Figure 4a) was stronger than we had anticipated in offsetting the cost of separating the representations. It is unclear why a 3D cost was not found, even at low workload, as Merwin et al. (1997) found. Maneuver choice. There was a strong preference for vertical over lateral maneuvers for both the rear-view and 3D displays (p <.01). This finding replicates the effect observed by Wickens, Helleberg, and Xu (2001). There are several reasons why this strong vertical preference has been repeatedly found. First, vertical maneuvers are more efficient because they result in smaller deviations from the intended flight path (Krozel & Peters, 1997). Second, vertical maneuvers are less complex cognitively due to their lower control order (second order) compared to lateral maneuvers (third order; Wickens, 1986; Wickens & Hollands, 2000). Finally, vertical maneuvers are more effective under high time pressure because they take less time to implement through greater speed efficiency than lateral maneuvers. The apparent dominance of vertical maneuvers is problematic, however, in that it conflicts with FAR , which instructs pilots to make lateral maneuvers in avoiding traffic. Vertical maneuvers are normally not considered (without the aid of a CDTI) because altitude is inherently difficult to determine through visual inspection. Perhaps this regulation will need to be revised in the case that CDTI s are adopted. The reasoning for requiring lateral avoidance maneuvers does not apply to situations in which a CDTI is employed because the CDTI represents altitude, and, in the case of the coplanar CDTI, represents it in a clear, unambiguous manner. This idea has in fact been utilized in the Traffic Alert and Collision Avoidance System (TCAS), which advises that vertical maneuvers be employed in avoiding traffic for the reasons mentioned above. Display representations have been shown to influence the types of maneuvers chosen by pilots (Ellis et al., 1987; Merwin & Wickens, 1996; and Pritchett & Hansman, 1997). Ellis et al. (1987) specifically noted an effect of dimensionality on maneuver choice in that pilots employed more vertical maneuvers with a perspective display than a uniplanar display. This trend was reversed, however, when comparing a 3D perspective display to a 2D coplanar display due to the superior vertical representation associated with the coplanar display, relative to that with the uniplanar display of Ellis et al., and with the 3D display (Merwin & Wickens, 1996). It was 34

36 therefore predicted that vertical maneuvers would be more prevalent with the rear-view coplanar than with the 3D display. In contrast to these predictions, vertical maneuvers were not selected more often with the rear-view than with the 3D displays for reasons that are not fully explainable. However, there were differences between these displays in the types of vertical maneuvers selected. Essentially, as seen in Figure 12, the 3D display induced a descent preference and climb aversion (p <.10) which was not seen in the rear-view display. In fact, a climb preference was observed with the rear-view display, replicating the effects of Wickens, Helleberg, and Xu (2001). The descent preference with the 3D display seen here partially replicates the findings of Merwin et al. (1997), who found a descent preference, but only when faced with descending traffic. (The vertical approach behavior of potential conflicts was not examined in this study.) Although descending and climbing maneuvers were shown to be equally safe in terms of the amount of time spent in a predicted conflict, the preference for descending maneuvers, found with the 3D display, may invite a greater potential for conflicts than a preference for climbing maneuvers. The reason for this is that when a pilot initiates a descending maneuver, airspeed naturally increases. If a descending maneuver is selected incorrectly, the time until a conflict will be reached will thereby decrease and the pilot will have less time to correct the maneuver. Considering the above speculation, the cause of this 3D invitation for a descent preference becomes quite intriguing. One possibility is that this was due to how descents were illustrated in the rear-view versus the 3D displays. In the rear-view display, the horizontal altitude lines that moved, as a result of vertical manipulations of the control, were quite salient to the pilot (see Figure 3b). In the 3D display, however, altitude changes were more ambiguously represented on the vertical drop-lines due to compression of the vertical axis along the line of sight; therefore, a descent could have potentially been perceived as still being at level flight. Furthermore, the descent bias observed with the 3D display may have been caused by the fact that ascending maneuvers were more likely to obscure or occlude the symbology (Merwin & Wickens, 1996). Specifically, aircraft along the depth axis beyond ownship may have appeared higher than ownship due to the distortion of height associated with 3D displays. An overlap or occlusion of ownship and that aircraft could then occur if ownship were to make a climbing maneuver. A descent, on the other hand, would have allowed ownship to remain visibly below the other aircraft in avoiding a conflict. With this in mind, the preference for climbs over descents with the rear-view coplanar display, observed also by Wickens, Helleberg, and Xu (2001) was perhaps caused by the fact that descents may be viewed as inherently more dangerous than climbs due to the rapid airspeed increase and altitude loss. Altitude representation was more saliently and directly visible to pilots using the rear-view display; therefore, the danger associated with descending maneuvers was also visible and caused somewhat of a descent aversion. This danger was not as evident with the 3D display due to the line-of-sight ambiguity, so descents were not avoided. Maneuver efficiency. It is important to remember that only maneuvers containing the axis relevant to the type of efficiency explored were included in the weighted equations used to determine overall efficiency for each display type. For lateral efficiency, only lateral and combined lateral/vertical maneuvers were considered. The data did not reveal any significant differences in lateral efficiency in comparing the rear-view and 3D displays. An explanation for 35

37 this null effect could be that lateral information was equally unambiguous in both the rear-view and 3D displays. That is, in the 3D display, since the viewpoint was nearly directly behind ownship (10 degree azimuth rotation), there was almost no lateral ambiguity, created by lateral compression (this in marked contrast to the vertical ambiguity created by the 45 elevation rotation of the 3D display as discussed below). Vertical efficiency calculations were based upon descending, climbing, airspeed, and combined lateral/vertical maneuvers. While the data did not reveal significant differences in efficiency between the rear-view and 3D displays overall, altitude maintenance during combined lateral/vertical maneuvers was better with the rear-view display than with the 3D display. This difference is assumed to result from the ambiguity of altitude representation in the 3D display, in contrast to the unambiguous lines across the vertical profile of the rear-view display. In the latter display, vertical trend information for ownship was easily perceived from the displacement of the solid and dashed altitude lines, and for other traffic from the slope of their predictor lines with the rear-view display (see Figure 3b). Vertical trend information had to be extracted from the 3D display, however, by comparing both current and future position drop-lines (in particular, the yellow regions which indicated altitude), a more complex perceptual-cognitive operation. The fact that the yellow regions not only had to be compared, but done so quickly and accurately, could have made such vertical judgments more difficult than when made with the rear-view display. Furthermore, there is evidence from other research that the effects of complex maneuvers, such as the combined lateral/vertical, may be more difficult to visualize with the 3D display than with the rear-view, again due to ambiguity associated with the respective axes (Merwin et al., 1997). This finding shows that complex maneuvers, although rare in occurrence, exacerbate the problem of ambiguity associated with 3D displays. The finding that both lateral and vertical efficiency is better (nonsignificantly and significantly, respectively) with the rear-view than the 3D display replicates several previous findings. For instance, Boyer, Campbell, May, Merwin, and Wickens (1995) found that avoidance paths were slightly longer and farther away from hazards with a 3D display compared to a 2D display. May, Campbell, and Wickens (1995) also found that maneuvers were less efficient with the 3D display, but in an air traffic control paradigm. In particular, highlighting the role of complexity in influencing the 3D costs, Wickens, Miller, and Tham (1996) found an efficiency cost in the time for the air traffic controller to reply with a pilot request, which was associated with 3D displays and was unique to combined lateral/vertical maneuvers. Furthermore, in other CDTI studies, 3D displays have been found to be particularly problematic when traffic moves in all three dimensions (e.g., ascending or descending) rather than along a flat plane (Merwin et al., 1997). Airspeed efficiency was calculated using the same maneuver sets (descents, climbs, airspeed, and combined lateral/vertical) that were used to compute vertical efficiency. Again, the weighted averages revealed no statistically significant differences across display type. However, airspeed efficiency was slightly better (a 20% reduction; p <.10) with the rear-view than with the 3D display when the pilots actually manipulated airspeed as an avoidance maneuver. We infer that this improvement was due to the unambiguous depiction of along-track differences, which are affected by airspeed. Essentially, the top-down panel in the rear-view display depicts a linear representation of along-track information, while this information is compressed in the ambiguous axes of the 3D display because of the 45 degree elevation angle at which it was 36

38 depicted (compare Figures 3a and 3b). The better the representation of this along-track information, the better is the associated airspeed efficiency performance. It is important to note that although airspeed maneuvers were used rather infrequently, airspeed efficiency performance was generally poor whenever any avoidance maneuver was implemented. This finding reveals that the task of maintaining the assigned airspeed was to some extent neglected when pilots attempted to avoid conflicts. This avoidance of airspeed control replicates previous findings suggesting that airspeed was not very salient to pilots and therefore less easily visualized on the CDTI (Wickens, Helleberg, & Xu, 2001). Profile View Orientation Safety. The results supported the hypothesis that the rear-view display exhibited superior conflict avoidance performance compared to the side-view display. This was shown by the finding that the mean time spent in a state of predicted conflict was higher by 4.6 seconds (a nearly 20% increase) for the side-view than the rear-view display (p <.10). These results were expected according to the idea that the side-view display did not utilize a simple and consistent one-to-one matching of aircraft across the two display panels, and therefore invoked a more demanding visual scan process. This finding is further consistent with the proximity compatibility principle in that the two representations of the same aircraft between panels were in greater proximity to one another in the rear-view display than in the side-view, therefore aiding the visual search/scanning process associated with integrating the two panels in the rearview display (Wickens & Carswell, 1995). The superiority of conflict avoidance performance with the rear-view display compared to the side-view display appeared to be specific to climbing (p <.10) maneuvers (and not to lateral maneuvers as would be expected due to the redundant heading information available with the rear-view display). It is important to note that climbing maneuvers were not only the most frequently chosen maneuver type overall, chosen over 42% of the time (see Figure 12), but they were also one of the safest maneuvers in terms of the mean time spent in predicted conflict, although this was only true for the rear-view display (see Figure 14). Interestingly, climbing maneuvers were most prevalent with the side-view display. It is unclear why climbing maneuvers were unsafe with the side-view display. An interesting finding to point out concerns the safety of airspeed maneuvers with either of the coplanar displays. It was initially expected that the side-view display might support superior conflict avoidance performance when implementing airspeed maneuvers due to the availability of redundant longitudinal (along-track) information across panels. A nonsignificant trend (p =.13) supports this expectation that airspeed maneuvers are implemented more safely with the side-view than the rear-view display. Perhaps this finding could be exploited by encouraging more use of airspeed control in avoiding conflicts. Maneuver choice. The finding that display representation influences the types of maneuvers utilized by pilots is again relevant when comparing the rear-view and side-view 2D coplanar displays. The dominance of vertical maneuvers over lateral, and more specifically, climbing maneuvers over descents replicates the previous findings of Wickens, Helleberg, and Xu (2001), who examined the coplanar rear-view display. But there was an apparent tendency of 37

39 the side-view display to enhance/amplify this climbing preference, which itself was exclusive to the coplanar views. It seems that this amplified climbing preference with the side-view display may have been due to the fact that both climbs and descents were more easily visualized with the side-view than with the rear-view display. This improved visualization is due to the along-track representation of altitude in the side-view display, through predictor line slopes (see Figure 3), in contrast to the displacement of the solid and dashed lines with the rear-view display which did not reveal any longitudinal information. We hypothesize that this superior visualization with the side-view display enhanced the perceived danger of descents in that the rapid build-up of airspeed in a descending maneuver was more salient to the pilot from the side rather than from behind. Maneuver efficiency. The weighted average of lateral efficiency with the rear-view display was lower (although not significantly so) than with the side-view display. While this nonsignificant trend may not be worthy of interpretation, it is important to note that the difference could have been due to the redundant coding of heading and lateral deviation information across the two panels in the rear-view display, while only the top-down panel contained heading information for the side-view display. This means that pilots had to scan between panels to access heading and altitude information while using the side-view display. Since the visual search/scanning process has the potential of being a major cost in using 2D coplanar displays, the fact that the side-view display may exacerbate this demand is cause for concern. Vertical efficiency was equivalent for the rear-view and side-view displays when comparing the weighted averages. This null effect is assumed to result from the equivalent, unambiguous representation of the vertical dimension in both coplanar displays. Only combined lateral/vertical maneuvers showed a display effect in that vertical efficiency was better with the side-view than the rear-view display (p <.10). Here again, it may be hypothesized that this effect resulted because the side-view display allowed pilots to integrate altitude and position along the flight path on a longitudinal scale, whereas the rear-view display did not depict this time dimension as clearly. The side-view display would naturally support better performance (greater efficiency) than the rear-view display since such integration would be required in visually and cognitively determining the effects of complex lateral/vertical maneuvers. Although vertical efficiency of combined lateral/vertical maneuvers was better with the side-view than with the rear-view display, it is important to interpret this difference in its connection to the relative safety of the two coplanar displays as discussed above (see Table 1). Doing so reveals that conflict avoidance performance on these combined maneuvers with the side-view display was much less safe than with the rear-view display (37 vs. 14 sec time in predicted conflict). This finding indicates that there was a tradeoff between safety and efficiency with the rear-view and side-view displays. Essentially, the side-view display invited cutting corners, or premature level-offs, which both increased efficiency and decreased safety. Why this happened remains unclear. As with lateral and vertical deviations, the data revealed no overall differences in the weighted averages of airspeed deviations associated with the change in viewpoint of the coplanar displays. There was an apparent trend, however, showing that the side-view display supported fewer airspeed deviations on climbing maneuvers than the rear-view display, a nearly 30% 38

40 reduction in deviations (p <.10). It is important to note that climbing maneuvers were also the most prevalent among both display types (see Figure 12). While the rear-view display revealed worse performance than the side-view display on this frequent climbing maneuver, it supported better airspeed efficiency when the very rare airspeed maneuvers were employed (p <.10). These effects apparently balanced each other out when the maneuvers were weighted by their probability (climbing maneuvers had a high probability while airspeed maneuvers had a low probability). Workload Traffic load. The effects of workload induced by traffic density in the current study were surprisingly small. In general, we found few main effects of workload and in particular, we did not find that increasing workload significantly disrupted performance with the coplanar displays relative to the 3D display. It is inferred that this null effect was observed because of the manner in which workload was manipulated in this particular paradigm. Specifically, the number of potential conflicts did not increase with increasing workload (number of aircraft); therefore, it is possible that pilots were able to search rapidly for relevant, potentially threatening aircraft and eliminate all those that were not in the immediate area of concern to ownship. Increasing the number of potential conflicts along with the number of aircraft overall would be a better workload manipulation. With that said, workload (number of aircraft) did exert at least one effect, in terms of airspeed deviation. Increasing the number of aircraft caused the mean absolute airspeed deviation to increase marginally (p <.10) across all displays. This means that the primary tasks of avoiding predicted conflicts and maneuvering efficiently kept pilots from monitoring their airspeed under high workload conditions. Subjective mental workload. Subjective mental workload ratings revealed that participants felt their workload was lowest with the rear-view display, higher with the side-view display (although this difference was not significant), and highest with the 3D display. This pattern has important implications for the acceptability of a display within the pilot community. The experience of lower subjective mental workload along with superior performance creates ideal circumstances through which a display may be successfully introduced into the cockpit. Summary The following table summarizes the overall findings according to display type and the significance of differences between them on each dependent measure. Scores used in the table for safety and workload reflect trends across all maneuver types, while efficiency scores are based on given maneuvers used with the different display types. Therefore, while overall effects of display on efficiency may not have been found, weak effects are indicated to reflect the effects on efficiency of certain maneuver types. 39

41 Table 5. Summary. Display Maneuver Rear-view Side-view 3D Safety Lat. Eff Vert. Eff AirSpd Eff Workload A plus score means that the display supported the best performance on that dependent measure. A score of 0 reveals moderate, or second-best performance, while a minus score means the display supported the worst performance on that measure. With safety, for example, performance was equivalent with the rear-view and 3D displays; therefore, both displays received a plus score and no moderate score of 0 was assigned. A 0+ scores indicates that efficiency was best with that display on a particular maneuver type. For example, the superior lateral efficiency of climbing maneuvers with the rear-view display compared to the side-view, and especially the 3D display, is reflected by a 0+. These weak effects are added into the final total as half points. The findings summarized in Table 5 suggest that a rear-view coplanar CDTI, with a score of 3, is preferable to either a side-view coplanar, with a score of.5, or a 3D display, with a score of -2. In conclusion, differences or changes in display format may have two separate effects on pilot information: (1) They may invite or inhibit certain types of maneuvers by making certain features of the airspace more or less salient (e.g., the side-view inhibition of descending maneuvers). Also, some of the maneuvers invited by a particular display may actually be inherently less safe or less efficient than desired (e.g., climbing maneuvers when using the sideview display). (2) They may influence the quality of the chosen maneuver, by hiding or degrading the nature of certain information (e.g., the line-of-sight ambiguity of the 3D display) or by increasing the information processing demands of using it (e.g., requiring integration across separate panels). Implications As free flight gets closer to being implemented in the National Airspace System, new systems such as the CDTI will need to be evolved enough to provide the safe separation of aircraft while also supporting more flexible flight paths than those utilized today. Since pilots will need to be provided with traffic information in order to select flight paths and airspeeds in real time, research involving the displays which will accomplish these tasks is extremely important to the safety and success of free flight. The current study suggests that a rear-view 2D coplanar CDTI would perhaps support the best pilot performance in terms of safety (conflict avoidance) and maneuver efficiency. 40

42 Constraints This research is constrained to the extent that there are limitations in generalizing these results to real cockpit flight due to the sample consisting only of general aviation pilots and due to the low fidelity of the flight environment. Future Research Directions It would be useful to further explore the tradeoff between scan/clutter costs associated with 2D coplanar and 3D displays in a more high-fidelity flight environment that would allow the incorporation of aircraft being present in the outside world but not being represented on the CDTI. As pointed out by Wickens, Helleberg, and Xu (2001), free flight scenarios employed in comparing CDTI s have only used a limited number of aircraft. Pilot vulnerability to detecting unannounced (transponder-off) traffic should be explored since greater visual and cognitive workload demands would be expected under high traffic densities. To do so would go beyond the desktop simulator used in the current study and employ a flight simulator and outside-world visualization. In this context, it would also be beneficial to use measures of eye movements to compare visual search patterns across displays under varying workload levels. Eye movement data could possibly confirm the idea discussed previously that pilots initially conducted a rapid visual search of the display and then eliminated those aircraft that were of no interest to them. Analyzing eye movements within a high-fidelity simulation would also lend to the assessment of visual workload associated with making free flight maneuvering decisions in a context of full flight monitoring and control responsibilities (Wickens, Helleberg, & Xu, 2001). The null of effect of workload found in this study should be explored through a paradigm which would increase the number of potential conflicts as well as the number of aircraft overall. This type of workload manipulation would prevent pilots from disregarding all but the two conflict aircraft, regardless of the number of aircraft total, as possibly occurred in this study. The next CDTI study planned in this laboratory does in fact consist of manipulating workload by increasing the number of potential conflicts. Acknowledgments The authors wish to acknowledge invaluable contributions of Roger Marsh and Ron Carbonari for simulation development, and of Don Talleur for professional advice. 41

43 References Abbott, T.S., Moen, G.C., Person, L.H., Keyser, G.L., Yenni, K.R., & Garren, J.F. (1980). Early flight test experience with cockpit displayed traffic information (CDTI) (NASA Technical Memorandum 80221/AVRADCOM Technical Report 80-B-2). Hampton, VA: NASA- Langley. Barfield, W., Hendrix, C., & Bjorneseth, (1995). Spatial performance with perspective displays as a function of computer graphics eyepoint elevation and geometric field of view. Applied Ergonomics, 26, Barfield, W. & Rosenberg, C. (1995). Judgments of azimuth and elevation as a function of monoscopic and binocular depth cues using a perspective display. Human Factors, 37(1), Barfield, W., Rosenberg, C., & Furness, T.A.I. (1995). Situation awareness as a function of frame of reference, computer-graphics eyepoint elevation, and geometric field of view. The International Journal of Aviation Psychology, 5, Bemis, S., Leeds, J., & Winer, E. (1988). Operator performance as a function of type of display: Conventional versus perspective, Human Factors, 30(2), Blattner, L. (1997). Figuring out free flight. Air Line Pilot, Boeckman, K.J. & Wickens, C.D. (2001). The resolution and performance effects of threedimensional display rotation on local guidance and spatial awareness measures (Technical Report ARL-01-4/NASA-01-3). Savoy, IL: University of Illinois, Aviation Research Laboratory. Boyer, B. S., Campbell, M., May, P., Merwin, D. H., & Wickens, C. D. (1995). Threedimensional displays for terrain and weather awareness in the national airspace system. Proceedings of the 39th Annual Meeting of the Human Factors and Ergonomics Society. Santa Monica, CA: Human Factors Society. Boyer, B.S. & Wickens, C.D. (1994). 3D weather displays for aircraft cockpits (Technical Report ARL-94-11/NASA-94-4). Savoy, IL: University of Illinois, Aviation Research Laboratory. Burnett, M.S. & Barfield, W. (1991). Perspective versus plan view air traffic control displays: Survey and empirical results. Proceedings of the Human Factors Society 35 th Annual Meeting (pp ). Santa Monica, CA: Human Factors Society. Busquets, A.M., Parrish, R.V., Williams, S.P., & Nold, D.E. (1994). Comparison of pilots acceptance and spatial awareness when using EFIS vs. pictorial display formats for complex, curved landing approaches. In R.D. Gilson, D.J. Garland, & J.M. Koonce (Eds.), Situational awareness in complex systems (pp ). Daytona Beach, FL: Embry-Riddle Press. 42

44 Carpenter, P.A., Just, M.A., Keller, T.A., Eddy, W., & Thulborn, K. (1999). Graded functional activation in the visuospatial system with the amount of task demand. Journal of Cognitive Science, 11, DeJonge, M. (2000). Displaying the FMS vertical profile. In Proceedings of the 2000 World Aviation Conference ( ). Anaheim, CA. Diwadkar, B.A., Carpenter, P.S., & Just, M.A. (2000). Collaborative activity between parietal and dorso-lateral prefrontal cortex in dynamic spatial working memory revealed by fmri. NeuroImage, 12, Ellis, S.R. (Ed.). (1993). Pictorial communication in virtual and real environments (2 nd ed.). London: Taylor & Francis. Ellis, S.R., McGreevy, M.W., & Hitchcock, R.J. (1987). Perspective traffic display format and airline pilot traffic avoidance. Human Factors, 29, Fadden, S. & Wickens, C.D. (1997). Improving traffic awareness with a head-up flight path highway display. In R.S. Jensen (Ed.), Proceedings of the 9 th International Symposium on Aviation Psychology (pp ). Columbus, Ohio State University, Dept. of Aerospace Engineering, Applied Mechanics, and Aviation. Flohr, E. & Huisman, H. (1997). Perspective primary flight displays in the 4D ATM environment. In R.S. Jensen (Ed.), Proceedings of the 9 th International Symposium on Aviation Psychology (pp ). Columbus, Ohio State University, Dept. of Aerospace Engineering, Applied Mechanics, and Aviation. Garland, D.J., Wise, J.A., Abbott, D.W., Benel, R.A., Hopkin, V.D., & Smith, P.J. (2000). Free flight: Human factors implications. In Proceedings of the Human Factors and Ergonomics Society 39 th Annual Meeting (pp ). Santa Monica, CA: HFES. Gempler, K.S. & Wickens, C.D. (1998). Display of predictor reliability on a cockpit display of traffic information. University of Illinois Institute of Aviation Final Technical Report (ARL-98-6/ROCKWELL-98-1). Savoy, IL: Aviation Research Laboratory. Gregory, R.L. (1977). Eye and brain. London: Weidenfeld & Nicolson. Haas, M.W. (1992). Multi-sensory virtual-interface technology. Proceedings of the 25 th Annual NATO Defense Research Group Seminar. London: Imperial College of Science & Technology. Haskell, I.D. & Wickens, C.D. (1993). Two- and three- dimensional displays for aviation: A theoretical and empirical comparison. International Journal of Aviation Psychology, 3(2), Helleberg, J., Wickens, C. D., & Xu, X. (2000). Pilot maneuver choice and safety in a simulated free flight scenario (Technical Report ARL-00-1/FAA-00-1). Savoy, IL: University of Illinois, Aviation Research Laboratory. 43

45 Hickox, J. & Wickens, C.D. (1999). Effects of elevation angle disparity, complexity and feature type on the cognitive mapping task of navigation checking. Journal of Experimental Psychology: Applied, 5, Jasek, M., Pioch, N., & Zeltzer, D. (1995). Enhanced visual displays for air traffic control collision prediction., Johnson, W.W., Battiste, V., & Bochow, S.H. (1999). A cockpit display designed to enable limited flight deck separation responsibility. In Proceedings of the 1999 World Aviation Conference ( ). Anaheim, CA. Kerns, K. & Small, D.W. (1995). Opportunities for rapid integrations of human factors in developing a free flight capability. (MP 9W ). McLean, VA: MITRE. Kim, W.S., Ellis, S.R., Tyler, M.E., Hannaford, B., & Stark, L.W. (1987). Quantitative evaluation of perspective and stereoscopic displays in three-axis manual tracking tasks. IEEE Trans. on System, Man, and Cybernetics, SMC-17, Kreifeldt, J.G. (1980). Cockpit displayed traffic information and distributed management in air traffic control. Human Factors, 22(6), ). Krozel, J. & Peters, M. (1997). Conflict detection and resolution for free flight. The Air Traffic Control Quarterly Journal Special Issue on Free Flight. May, P.A., Campbell, M., & Wickens, C.D. (1995). Perspective displays for air traffic control: Display of terrain and weather. Air Traffic Control Quarterly, 3, 1-17). McGreevy, M.W. & Ellis, S.R. (1986). The effect of perspective geometry on judged direction in spatial information instruments. Human Factors, 28, Merwin, D.H. & Wickens, C.D. (1996). Evaluation of perspective and coplanar cockpit displays of traffic information to support hazard awareness in free flight. University of Illinois Institute of Aviation Technical Report (ARL-96-5/NASA-96-1). Savoy, IL: Aviation Research Laboratory. Merwin, D.H., Wickens, C.D., & O Brien, J.V. (1997). Display format-induced biases in air traffic avoidance behavior. Proceedings of the World Aviation Congress (98WAC-71). Warrendale, PA: Society of Automotive Engineers, Inc. Nordwall, B.D. (1997). Situational awareness beyond the cockpit. Aviation Week and Space Technology, April 28, O Brien, J.V. & Wickens, C.D. (1997). Free flight cockpit displays of traffic and weather: Effects of dimensionality and data base integration. Proceedings of the Human Factors and Ergonomics Society 41 st Annual Meeting (pp ). Santa Monica, CA: Human Factors and Ergonomics Society. 44

46 Olmos, O., Liang, C.D., & Wickens, C.D. (1997). Electronic map evaluation in simulated visual meteorological conditions. International Journal of Aviation Psychology, 7(1), Olmos, O., Wickens, C.D., & Chudy, A. (2000). Tactical displays for combat awareness: An examination of dimensionality and frame of reference concepts and the application of cognitive engineering. The International Journal of Aviation Psychology, 10, Oman, C., Kendra, A., Hayashi, M., Stearns, M., & Burke-Cohen, J. (2001). Vertical navigation displays: Pilot performance and workload during simulated constant angle descent GPS approaches. International Journal of Aviation Psychology, 11, Owen, G.L. & Suiter, J. (1997). FMS 3D map. In R.S. Jensen (Ed.), Proceedings of the 9 th International Symposium on Aviation Psychology. Columbus, OH: The Ohio State University. Palmer, E.A. (1983). Conflict resolution maneuvers during near miss encounters with cockpit traffic displays. Proceedings of the Human Factors Society 27 th Annual Meeting (pp ). Santa Monica, CA: HFES. Perrone, J.A. & Wenderoth, P. (1993). Visual slant underestimation. In S.R. Ellis, M. Kaiser, & A.J. Grunwald (Eds.), Pictorial communication in virtual and real environments. London: Taylor & Francis. Planzer, N. & Jenny, M.T. (1995). Managing the evolution to free flight. Journal of ATC, Pritchett, A.R. & Hansman, R.J. (1997). Experimental studies of pilot performance at collision avoidance during closely spaced parallel approaches. Proceedings of the 9 th International Symposium on Aviation Psychology. Columbus, OH: Department of Aviation. Reising, J., Barthelemy, K., & Hartsock, D. (1989). Pathway-in-the-sky evaluation. In R.S. Jensen (Ed.), Proceedings of the 5 th International Symposium on Aviation Psychology. Columbus, OH: Department of Aviation. Roscoe, S.N. (1968). Airborne displays for flight and navigation. Human Factors, 10(4), RTCA. (1995). Final report of RTCA task force 3, free flight implementation. Washington, DC. Schreiber, B.T., Wickens, C.D., Renner, G.J., Alton, J., & Hickox, J.C. (1998). Navigational checking using 3D maps: The influence of elevation angle, azimuth, and foreshortening. Human Factors, 40(2), Shelden, S. & Belcher, S. (1999). Cockpit traffic displays of tomorrow. Ergonomics in Design, 7, 4-9. Smallman, H.S., Schiller, E., & Cowen, M.B. (2000). Track location enhancements for perspective view displays (Technical Report 1847). San Diego: SPAWAR Systems Center. 45

47 St. John, M. & Cowen, M.B. (1999). Use of perspective view displays for operational tasks (Technical Report 1795). San Diego, CA: SSC. St. John, M., Cowen, M.B., Smallman, H.S., & Oonk, H.M. (2001). The use of 2D and 3D displays for shape-understanding versus relative-position tasks. Human Factors, 43(1), St. John, M., Oonk, H.M., & Cowen, M.B. (2000). Using two-dimensional and perspective views of terrain (Technical Report 1815). San Diego, CA: SSC. St. John, M., Smallman, H.S., Bank, T.E., & Cowen, M.B. (2001). Tactical routing using twodimensional and three-dimensional views of terrain (Technical Report 1849). San Diego, CA: SSC. Tharp, G.K. & Ellis, S.R. (1990). The effects of training on errors of perceived direction in perspective displays (NASA Technical Memorandum ). Moffett Field, CA: NASA Ames Res. Ctr. Theunissen, E. (1995). In-flight application of 3-D guidance displays: Problems and solutions. Proceedings of the International Federation of Automatic Control. Oxford, England: Pergamon. Van Breda, L. & Veltman, H.A. (1998). Perspective information in the cockpit as a target acquisition aid. Journal of Experimental Psychology: Applied, 4(1), Wickens, C.D. (1986). The effects of control dynamics on performance. In K. Boff, L. Kaufman, & J. Thomas (Eds.), Handbook of Perception and Performance (vol. II, pp ). New York: Wiley. Wickens, C.D. (1995). Display integration of air traffic control information: 3D displays and proximity compatibility. University of Illinois Institute of Aviation Final Technical Report (ARL-95-2/FAA-95-2). Savoy, IL: Aviation Research Laboratory. Wickens, C.D. (1999). Frames of reference for navigation. In D. Gopher & A. Koriat (Eds.), Attention and performance (Vol. 16). Cambridge, MA: MIT Press. Wickens, C.D. (2000). The when and how of using 2-D and 3-D displays for operational tasks. In Proceedings of the IEA 2000/HFES 2000 Congress. Santa Monica, CA: Human Factors and Ergonomics Society. Wickens, C.D. & Carswell, C.M. (1995). The proximity compatibility principle: Its psychological foundation and its relevance to display design. Human Factors, 37(3), Wickens, C. D., & Helleberg, J. (1999). Interactive perspective displays for airborne hazard awareness (Final Technical Report ARL-99-1/ROCKWELL-99-1). Savoy, IL: University of Illinois, Aviation Res. Lab. 46

48 Wickens, C.D., Helleberg, J., & Xu, X. (1999). Maneuver choice in free flight. Proceedings of the 1999 World Aviation Conference ( ). Anaheim, CA. Wickens, C.D., Helleberg, J., & Xu, X. (2001). Pilot maneuver choice and workload in free flight. To appear in Proceedings of the 11 th International Symposium on Aviation Psychology. Columbus, OH: The Ohio State University. Wickens, C. D., Helleberg, J., & Xu, X. (in press). Pilot maneuver choice and workload in free flight. Human Factors. Wickens, C.D. & Hollands, J. (2000). Engineering psychology and human performance (3 rd ed.). New York: Prentice Hall. Wickens, C.D., Liang, C.C., Prevett, T., & Olmos, O. (1996). Electronic maps for terminal area navigation: Effects of frame of reference and dimensionality. The International Journal of Aviation Psychology, 6, Wickens, C.D., & May, T. (1994). Terrain representation for air traffic control: A comparison of perspective with plan view displays (Technical Report ARL FAA-94-2). Savoy, IL: University of Illinois, Aviation Res. Lab. Wickens, C.D., Merwin, D.H., & Lin, E. (1994). Implications of graphics enhancements for the visualization of scientific data: Dimensional integrality, stereopsis, motion, and mesh. Human Factors, 36, Wickens, C.D., Miller, S., & Tham, M. (1996). The implications of data link for representing pilot request information on 2D and 3D air traffic control displays. The International Journal of Industrial Ergonomics, 18, Wickens, C.D. & Morphew, E. (1997). Predictive features of a cockpit traffic display: A workload assessment. University of Illinois Institute of Aviation Technical Report (ARL- 97-6/NASA-97-3). Savoy, IL: Aviation Research Laboratory. Wickens, C.D. & Prevett, T. (1995). Exploring the dimensions of egocentricity in aircraft navigation displays. Journal of Experimental Psychology: Applied, 1, Wickens, C.D., Vincow, M., & Yeh, M. (in press). Design applications of visual spatial thinking: The importance of frame of reference. Woods, D.D. (1984). Visual momentum: A concept to improve the cognitive coupling of person and computer. International Journal of Man-Machine Studies, 21, Yeh, Y.Y. & Silverstein, L.D. (1992). Spatial judgments with monoscopic and stereoscopic presentation of perspective displays. Human Factors, 34(5),

49 Appendix A CDTI Format 48

50 Appendix B CDTI Symbology Illustration 49

51 Appendix C Side-View CDTI 50

52 Appendix D Seamless Display for Aircraft Landing Task Performance and Display Transitions : 3D CDTI Spring 2001 NASA-NAG Research supported by the NASA Ames Research Center Conducted by Amy Alexander Advisor: Dr. Christopher Wickens Institute of Aviation Aviation Research Lab Willard Airport #1 Airport Road Savoy, IL The purpose of this experiment is to examine the technology that will support your awareness of airborne traffic hazards in the advanced general aviation cockpit. You will be asked to fly a navigational path on a low-fidelity flight simulator; and be aware of, and maneuver to avoid, any traffic that is represented on your traffic display. Your total participation will be less than 2 hours. There are no known risks or physical discomforts associated with this experiment beyond those of ordinary life, and you will be paid at the rate of $7/hr. You may terminate your participation at any time, and you will still be paid for the number of hours that you have completed. Your participation in this research will provide for the development of better aircraft display designs, as well as for a better understanding of how pilot performance can be benefited through the use of different displays. We thank you for your involvement. If you have any further questions, please let the experimenter know at any time throughout the experiment, call Amy at , or call Dr. Wickens at Statement of Consent I acknowledge that my participation in this experiment is entirely voluntary and that I am free to withdraw at any time. I have been informed of the general scientific purposes of this experiment and I know that I will be compensated at a rate of $7.00/hour for my participation. If I withdraw from the experiment before its termination, I will receive my total fee earned to that time. I understand that my data will be maintained in confidence, and that I may have a copy of this consent form. Signature of participant: Date : Signature of experimenter: Date: 51

53 Appendix E Pilot Questionnaire 1. Total flight hours (approximate) 2. Total instrument flight hours (sim & actual) 3. Do you use specific strategies to avoid traffic assuming the other pilot IS NOT aware of your presence? YES NO If YES, please describe them below: 52

54 Appendix F Experimental Instructions Introduction The FAA and NASA have recently undertaken a research effort to examine specific ways to improve the efficiency of the National Airspace System. This program has been referred to as Free Flight, and involves providing airspace users with increased flexibility in selecting routes to their destinations. New systems are currently being developed to provide safe separation of traffic while supporting more flexible flight paths. Free Flight has in fact been described as a system in which VFR flexibility is provided under IFR protection. A potential result of Free Flight is that ATC will have less control over traffic in the en route phase of flight than it does today. Because of this, pilots may be expected to take a greater role in monitoring their own separation from traffic; or more likely, monitoring the automated system that is providing separation. The present study examines several issues involved in presenting traffic information in the cockpit. Task Overview In this study you will be asked to fly a series of short trials using a desktop IFR flight simulator which contains an experimental traffic display. The traffic display provides information about the relative position, heading, and altitude of nearby air traffic. During each trial you will be asked to fly a predetermined route to a navigational waypoint while monitoring the traffic display for potential conflicts, which are defined as penetrations of the protected zone around your own aircraft. This protected zone is 1500 ft above and below ownship and 3 miles in radius, for a total of 3000 ft vertically and 6 miles horizontally. The primary goal of your task is to reach the waypoint as efficiently and rapidly as possible while maintaining safe separation from traffic. You will receive several practice trials before beginning the experiment, during which time the experimenter will answer any questions you may have. The process of completing a trial is outlined below: Fly the predetermined route toward the navigational waypoint. Monitor the traffic display for conflicts. The experimenter will explain the symbology on the traffic display to assist you in detecting conflicts after you read these instructions. The other traffic in the display will always maintain their heading and velocity, and will not react to your own aircraft s behavior. When a conflict occurs, avoidance maneuvers should be as efficient as possible, without compromising separation. That is, you should try to deviate as little as possible from your original heading (indicated by a blue dot on the directional gyro), altitude (12000 ft), and airspeed (350 knots) while safely maneuvering around the conflict. You may use any type of maneuver that you feel is appropriate for the situation. The trial will end when you reach the waypoint or a line orthogonal to the original navigational waypoint. That is, if you have maneuvered around traffic, the trial will end as 53

55 you cross any point along that line at any altitude. Nevertheless, after your maneuver is complete, you should try to come back toward your original target heading, altitude, and airspeed. You will first fly a series of 12 practice trials at varying traffic levels (2, 6, or 10 aircraft besides ownship) on the three different displays that will be described to you after you read these instructions. You will then fly a series of 18 experimental trials. The experiment will pause briefly at the end of the last three blocks of trials so that you may rate your mental workload. Maneuvering Avoidance maneuvers will be performed using a flight stick controlling pitch, roll, and throttle. Pitch will be controlled by moving the flight stick forward or backward; roll will be controlled by moving the flight stick to the right or left. Speed will be increased by pressing the button on top of the flight stick; it will be decreased by pressing the trigger at the front of the flight stick. Maximum and minimum airspeeds are set at 400 and 250 knots, respectively. 54

56 Appendix G 3D CDTI 55