Content uploaded by Marinus M. Van Paassen
Author content
All content in this area was uploaded by Marinus M. Van Paassen on Jun 20, 2023
Content may be subject to copyright.
Stereoscopic Depth Cues for Enhancing Pilot Interpretation of
the Artificial Horizon
Dale-Allen Arrundell1, Annemarie Landman1,2, Olaf Stroosma1, René
van Paassen1, Eric Groen2, Max Mulder1
1Delft University Of Technology, Delft, The Netherlands, 2Netherlands Organisation for
Applied Scientific Research, Soesterberg, The Netherlands
Background. Previous studies and accident analyses have shown that pilots
can make roll reversal errors when responding to bank angles shown by the
artificial horizon in the Primary Flight Display (PFD). In the current study,
we tested whether adding stereoscopic depth cues to the artificial horizon
may lead to better bank angle representation due to an improved figure-
ground separation between the symbols. Method. Stereoscopic depth cues
were created by using a half-silvered mirror multi-layer PFD, which pre-
sented the horizon symbol on a lower layer and the aircraft symbol on a
higher layer. A group of 23 non-pilots and 18 general aviation pilots were
shown left or right bank angles on this multi-layer PFD as well as on a nor-
mal single-layer PFD, with the task to roll the wings level using a joystick.
Results. In the pilot group, a similar amount of roll-reversal errors was
made with both displays (median = 3.3%) with no significant difference, p
= 0.635. In the non-pilot group, fewer roll-reversal errors were observed
with the multi-layer display, but this difference did not reach significance
either (median = 3.3% vs. 5.0%, p= 0.182). In both pilots and non-pilots,
the reaction time was longer in the multi-layer display, which reached signifi-
cance in the non-pilots (p= 0.016) but not in pilots (p= 0.215). Participants
noticed that the depth was only visible during the start of the session. Con-
clusions. The results suggest that using stereoscopic depth cues are not a
viable manner to enhance the figure-ground relation in the artificial horizon.
Spatial disorientation is still one of the major causal factors in cases of loss of control in flight
(LOC-I). It was determined to have contributed to 17% of LOC-I accidents in transport
and commuter aircraft between 1981–2016, with no signs of a decreasing trend (Newman &
Rupert, 2020). The Primary Flight Display (PFD) is the main instrument with which pilots
can prevent or counteract spatial disorientation. Several studies have indicated that the
bank indication of the artificial horizon is suboptimal, as it can lead to misinterpretations
of the bank angle direction as well as incorrect roll inputs known as roll-reversal errors
(RREs). In simulator studies, airline pilots were shown to make RREs in 6.9-8.7% of the
cases when being shown a PFD with an unforeseen bank angle and attempting to correct to
wings-level flight (Müller, Sadovitch, & Manzey, 2018; Van den Hoed et al., 2022). When
spatially disorienting roll cues preceded PFD presentation, this percentage increased to
STERESCOPIC DEPTH CUES 2
20% of total cases and 39% of the first encounter (Van den Hoed et al., 2022). Examples of
accident cases that have been associated with spatial disorientation-induced RREs include
Kenya Airways Flight KQA507 (Cameroon Civil Aviation Authority, 2010), Flash airlines
flight 602 (Bureau d’Enquêtes et d’Analyses pour la Sécurité de l’Aviation Civile, 2009),
and Crossair flight 498 (Aircraft Accident Investigation Bureau, 2002).
These misinterpretations of the artificial horizon are thought to be caused by sub-
optimal display design. According to the so-called “figure-ground” principle, we normally
perceive objects moving in the foreground against the fixed horizon in the background. In
standard artificial horizon displays, the horizon symbol is the part that moves, while the
aircraft symbol is fixed, which hampers quick interpretation (Grether, 1947; Johnson &
Roscoe, 1972; Roscoe, Corl, & Jensen, 1981). The format of current displays does allow for
integration with head-up displays. Hence, it is relevant to investigate display adaptations
that may enhance the PFD in its current format.
In the current study, stereoscopic depth cues are investigated as a means to improve
the figure-ground relation in the artificial horizon, in order to prevent RREs. These cues
are presented using a multi-layer display (MLD), presenting the aircraft symbol and horizon
symbol on two different layers. As this produces a different visual image in each eye, the cues
are stereoscopic. MLDs have previously been used to separate categories of information to
improve users’ information uptake and to prevent clutter (Dünser, Billinghurst, & Mancero,
2008; Hayes, Moore, & Wong, 2006). They also have been applied to make information more
salient to improve search performance (Wong, Joyekurun, Nees, Amaldi, & Villanueva,
2005). However, to our knowledge, MLDs have not yet been used to improve figure-ground
separation in cockpit displays.
Method
Design
The experimental tasks were first performed by a group of non-pilots to obtain more in-
formation on optimal depth between two layers of the MLD. These non-pilots performed
the tasks with a single layer display (SLD) PFD (baseline condition) and with a MLD with
either low or high depth between the two layers (two groups, randomly assigned). The
effect of MLD and that of MLD depth were tested using a mixed-model design.
Using these outcomes, one layer of depth was chosen to further test with a group of
private pilots. This group performed the same tasks in the baseline condition and with the
MLD with the depth that had the most effect in the non-pilots. The effect of the MLD was
tested using a within-subject comparison.
Participants
Of the non-pilot participants (n = 23, 18/5 male/female, mean age = 24.8, SD = 6.9), 5
had never seen nor used an artificial horizon, 4 knew what it is but had never used it, 10
had some experience with flight simulators at different fidelity levels, and 8 had experience
with glider flying. None had experience controlling powered aircraft of any category.
The participating pilots (n = 18) held a private pilot license (PPL, all male, mean
age = 43.5, SD = 16.1). Flight experience was on average 324 flight hours, SD = 245 hours,
and 9.8 years of being active as a private pilot, SD = 7.4.
STERESCOPIC DEPTH CUES 3
Figure 1 . Left: the experimental setup with the display and joystick. Right: The configu-
ration of the MLD with two displays and mirror (the black diagonal line).
All participants were right-handed and had normal or corrected-to-normal eyesight.
This experiment was approved by the ethical committee of the Delft University of Technol-
ogy, and all participants provided informed consent.
Apparatus
A desktop setup was used with no outside visuals (see Figure 1, left). A half-silvered mirror-
type display was used for the MLD. A schematic of the inner workings of this display can
be seen in Figure 1 (right). For each display layer, a standard LCD-type display was used
with a resolution of 1152 x 864. These displays were placed perpendicular to each other,
and a half-silvered mirror was placed at a 45°between the displays. This creates a virtual
display (dashed blue line in Figure 2) at some distance x in front of display 1. Distance x
could be set by moving display 2 to increase distance a. For the low depth condition, X
was 1.6 cm, and for the high depth condition 2.1 cm. For the single-layer display (SDL;
baseline) condition, display 2 was turned off and all information was displayed on display
1. A standard Logitech Extreme 3D pro joystick was used as input device. The maximum
angular deflection on the roll axis of this joystick was measured to be 20°.
A Boeing 747-based PFD was used (see Figure 2). In the MLD, the aircraft symbol,
sky pointer and bank angle scale are presented on the upper layer, as well as the speed
and altitude tapes. The horizon and pitch ladder were presented on the lower layer. A
simplified aerodynamic model was used. The model had a fixed speed of 120 knots, altitude
always indicated 10,000 ft., and the attitude was controllable in the pitch and roll axis with
sensitivity and dampening resembling that of a small single-engine piston aircraft.
Procedure and tasks
After a briefing and an intake questionnaire, participants were first showed the SLD and
MLD version of the PFD. Non-pilots were explained the symbols and how the display is used
to control an airplane. Participants then familiarized themselves with the flight dynamics
for 5 minutes by flying several turns and level changes. The familiarization was followed
by 10 practice runs of the experimental task with the SLD and 10 with the MLD. Each
run started with a black screen displayed for 5 seconds. Then, the display showed a PFD
STERESCOPIC DEPTH CUES 4
Figure 2 . Left: The image presented on the upper layer of the MLD. Right: the image
presented on the lower layer of the MLD.
indicating a bank angle at 30°to the left or to the right. When the display appeared, the
participant was to respond immediately by rolling the wings level using the joystick. Ten
seconds after the appearance of the PFD, the run ended. This was immediately followed
by the black screen at the start of the next run. The participant’s hand was always placed
on the joystick. Following practice, this task was performed in two blocks of 30 runs with
the SLD and with the MLD, with a short break in between. The order of conditions was
counterbalanced between participants.
Dependent measures
Data on the roll angle and roll control input were logged and analyzed with MATLAB.
From this, the following dependent measures were obtained:
•Error rate. A roll reversal error is recorded when the first roll control input was
towards the same side of the displayed bank angle. The threshold of input detection
was set at 1.5°stick deflection. The error rate is the percentage of runs with an error
over the total number of runs.
•Reaction time. This is the time between PFD presentation and the start of the first
roll control input.
Lower error rates and faster reaction times were expected with the MLD compared to
the baseline condition, as the MLD was expected to facilitate quicker and more accurate
recognition of the figure-ground relationship.
Statistics
In both groups, the effect of MLD was tested using a paired-samples t-test between MLD
and baseline. If not normally distributed, the Wilcoxon signed rank test was used.
In non-pilots, the effect of the MLD depth between layers was additionally tested
using a mixed-model ANOVA with display type (baseline, MLD) as the within-subject
factor, and MLD depth (low depth group, high depth group) as the between-subject factor.
STERESCOPIC DEPTH CUES 5
Results
Data collection
A number of runs (1% of total) were excluded from analysis due to incorrect detection of an
input immediately following PFD presentation due to the stick position not being centered
within the limits of 1.5°stick deflection.
Non-pilots
The Wilcoxon signed rank test showed that there was no significant difference in error rates
between the baseline (median = 5% errors) and MLD condition (median = 3.3%, Z(22) = -
1.33, p= 0.182). The mixed-model ANOVA showed no significant difference in improvement
between the two MLD depths used (F(1,21) = 1.059, p= 0.315). Nevertheless, as the low
depth showed the largest improvement, we decided to use this depth for the pilot group.
The paired-samples t-test indicated a significant difference in reaction time between the
baseline condition (mean = 591 ms) and the MLD condition (mean = 605 ms, t(22) =
-2.608, p= 0.016), a difference which was opposite to the expected direction.
Pilots
A Wilcoxon Signed Rank test showed that there was no significant difference in error rates
between the baseline (median = 3.3% errors) and MLD condition (median = 3.3%, Z(18) =
-0.475, p= 0.635). A Wilcoxon Signed Rank test t-test indicated no significant difference
in reaction time between the baseline condition (median = 625 ms) and the MLD condition
(median = 644 ms, Z(18) = -1.24, p= 0.215).
Discussion
The results did not indicate a significant improvement in performance when using the MLD
compared with the SLD. In contrast, reaction times in non-pilots were longer when using the
MLD than the SLD, suggesting that it was more difficult to read the bank angle quickly with
the MLD. Participants reported that they indeed perceived depth in the MLD, although this
depth perception was mostly present at the start of the experiment or any time they moved
their head. The requirement of head motions indicates that the MLD was unsuccessful in
presenting stereoscopic cues, as no head motions should be required for such cues. This also
makes it impractical for use in the cockpit.
The required distance between layers in the MLD to obtain a stereoscopic effect
would cause the pitch and roll indications to become inaccurate, as the position of the
symbols relative to each other would then shift greatly depending on head position. Several
participants mentioned that they thought the aircraft model in the MLD condition was
slower to react than that in the SLD condition, which was not the case. We do not know
what may have caused this perception. The results of the experiment lead us to conclude
that stereoscopic depth cues achieved thought MLDs are not suitable for enhancing the
figure-ground representation in the attitude indicator.
Subsequent research into optimizing the PFD for attitude representation could focus
instead on monoscopic cueing. There are several types of monoscopic cues that could
be implemented without making serious changes to the PFD designs currently in use in
STERESCOPIC DEPTH CUES 6
commercial aviation. Examples of this are the use of a horizontal color gradient to simulate
“aerial perspective” (Gibson, 1950), linear perspective lines, ground texture, extending the
horizon behind the speed and altitude tapes, or adding a dark line under the aircraft symbol
to simulate shadow. With each of these additions, it is important to ensure that the salience
of the horizon, aircraft, and pitch and roll ladders remains intact. Additions to the sky and
ground should not contrast too much with the colors of these surfaces, and thickness of
added lines should be minimal. Empirical studies are needed to evaluate the effectiveness
of these design changes, and possibly to fine-tune the optimal use of added symbols.
References
Aircraft Accident Investigation Bureau. (2002). Final Report of the Aircraft Accident
Investigation Bureau on the Accident to the Saab 340B Aircraft, Registration HB-AKK
of Crossair Flight CRX 498 on 10 January 2000 Near Nassenwil/ZH. https://www.
sust.admin.ch/inhalte/AV-berichte/1781_e.pdf.
Bureau d’Enquêtes et d’Analyses pour la Sécurité de l’Aviation Civile. (2009). Final report
on the accident on 1st June 2009 to the Airbus A330-203, registered F-GZCP, operated
by Air France, Flight AF 447 Rio de Janeiro–Paris.
Cameroon Civil Aviation Authority. (2010). Technical Investigation into the Accident of
the B737-800 Registration 5y-Kya Operated by Kenya Airways that Occurred on the
5th of May 2007 in Douala. https://reports.aviation-safety.net/2007/ 20070505-0_
B738_5Y-KYA.pdf.
Dünser, A., Billinghurst, M., & Mancero, G. (2008). Evaluating visual search performance
with a multi layer display. In Proceedings of the 20th Australasian Conference on
Computer-Human Interaction: Designing for Habitus and Habitat (pp. 307–310).
Gibson, J. J. (1950). The Perception of the Visual World. Boston (MA): Houghton Mifflin.
Grether, W. F. (1947). Discussion of pictorial versus symbolic aircraft instrument displays.
USAAF AMC, Engineering Division, Aero Med Lab., TS EAA.
Hayes, J., Moore, A., & Wong, B. L. (2006). Information layering to de-clutter displays
for emergency ambulance dispatch. In Proc. 13th European conference on Cognitive
ergonomics: trust and control in complex socio-technical systems (pp. 10–16).
Johnson, S. L., & Roscoe, S. N. (1972). What Moves, the Airplane or the World? Human
Factors: The Journal of the Human Factors and Ergonomics Society,14 (2), 107-129.
Müller, S., Sadovitch, V., & Manzey, D. (2018). Attitude indicator design in primary flight
display: Revisiting an old issue with current technology. The International Journal
of Aerospace Psychology,28 (1-2), 46–61.
Newman, R. L., & Rupert, A. H. (2020). The magnitude of the spatial disorientation
problem in transport airplanes. Aerosp. medicine and hum. perf.,91 (2), 65–70.
Roscoe, S. N., Corl, L., & Jensen, R. S. (1981). Flight Display Dynamics Revisited. Human
Factors,23 (3), 341–353.
Van den Hoed, A., Landman, A., Van Baelen, D., Stroosma, O., van Paassen, M. M., Groen,
E. L., & Mulder, M. (2022). Leans illusion in hexapod simulator facilitates erroneous
responses to artificial horizon in airline pilots. Human Factors,64 (6), 962–972.
Wong, B. W., Joyekurun, R., Nees, A., Amaldi, P., & Villanueva, R. (2005). Information
layering, depth and transparency effects on multi-layered displays for command and
control. In Proc. Human Factors and Ergonomics Society (Vol. 49, pp. 352–356).
