Content uploaded by Daan Marinus Pool
Author content
All content in this area was uploaded by Daan Marinus Pool on Jul 29, 2015
Content may be subject to copyright.
Investigation into Pilot Perception and Control
During Decrab Maneuvers in Simulated Flight
J.T. Beukers,
∗
O. Stroosma,
†
D.M. Pool,
‡
M. Mulder,
§
and M.M. van Paassen
¶
Delft University of Technology, Delft, The Netherlands
An experimentwasconducted in theSIMONAResearch Simulator atDelftUniversity
of Technology to investigate the influence of sway, roll and yaw motion cues on pilot per-
formance, control and motion perception during decrab maneuvers. In the experiment,
six pilots were instructed to perform manual decrab maneuvers in heavy crosswind con-
ditions with a Cessna Citation 500 model. The contributions of yaw, roll and sway motion
stimuli were varied such that their effects on objective measures and subjective ratings
could be examined. The results of this experiment show that yaw motion had a posi-
tive influence on performance in terms of lateral touchdown distance from the runway
centerline. Roll motion significantly decreased roll rate variations during decrab. High
workload and the relatively low intensity of lateral motion cues led to the fact that pilots
were unable to give consistent fidelity ratings. This result emphasizes the need for an
objective and quantifiable method to determine motion fidelity for such maneuvers. Pilot
models can possibly be used to further investigate the influence of different motion cues
in these transient control tasks.
I. Introduction
Flight simulators are important tools for both pilot training and research into aerospace systems.
Most simulators consist of a cockpit replica, an outside visual system and a motion system that
together provide cues to achieve simulated flight at a desired level of fidelity. Due to the restricted
freedom of movement of a simulator, however, it is virtually impossible to provide exactly the
same motion cues as in real flight. This induces a mismatch in stimuli compared to actual flight
∗
MSc Student, Control and Simulation Division, Faculty of Aerospace Engineering; j.t.beukers@gmail.com.
†
Researcher, Control and Simulation Division; o.stroosma@TUDelft.nl. Senior Member AIAA.
‡
Ph.D. Candidate, Control and Simulation Division; d.m.pool@TUDelft.nl. Student member AIAA.
§
Professor, Control and Simulation Division; m.mulder@TUDelft.nl. Senior Member AIAA.
¶
Associate Professor, Control and Simulation Division; m.m.vanpaassen@TUDelft.nl. Member AIAA.
1 of 35
J. T. Beukers, Olaf Stroosma, Daan M. Pool, Max Mulder and Marinus M. van Paassen,
Investigation into Pilot Perception and Control During Decrab Maneuvers in Simulated
Flight (2010), in: Journal of Guidance, Control, and Dynamics, 33:4(1048-1063)
http://dx.doi.org/10.2514/1.47774
which in turn causes pilots to adapt their control behavior, thereby limiting the simulator’s fidelity
and usefulness. Pilot models play an important role in gaining insight in this process.
1–9
In this light, a five year research project was initiated at Delft University of Technology (TU
Delft) to investigate simulator fidelity in more detail. The goal of this project is to develop a
method to objectively and quantifiably assess the extent to which a flight simulator supports real-
flight pilot behavior.
10
A cybernetic approach
11
is adopted that uses control-theoretical models to
mathematically describe a pilot’s multimodal response to visual and motion cues in real flight and
in the simulator.
12,13
Currently, considerable experience has been gained by applying this approach
to continuous control tasks, such as disturbance-rejection or target-tracking tasks, in which pilots
manually control an aircraft relative to an equilibrium point.
14–18
Maneuvers that are essential ingredients of pilot training, however, are typically of a more
transient nature and thus require transient rather than continuous pilot control inputs. In these
transient control tasks, pilots bring the aircraft from one equilibrium state to another, such as in the
case of an engine-out after take-off, a helicopter deceleration to hover, or a decrab before landing
in heavy crosswind.
Fundamental differences between continuous control tasks and these discrete transient maneu-
vers make the analysis of the latter much more complicated. First, these maneuvers are of relative
short duration compared to tracking tasks, which can last several minutes. As a result, the infor-
mation per measurement is relatively low and maneuvers need to be repeated extensively to obtain
reliable results. Second, pilot control behavior in continuous control tasks can be described with
quasi-linear pilot models which can be identified from measured data using frequency
19,20
and
time-domain
21
identification techniques. Such linear models are not available for transient maneu-
vers. In addition, due to their short duration, frequency domain identification techniques are not
expected to yield satisfactory results when studying pilots’ transient responses.
As a prelude to a cybernetic approach in which pilot control behavior is modeled for transient
tasks, this paper describes a preliminary investigation into the influence of different motion stimuli
on pilot behavior in transient maneuvers. The present study will form a basis for further investi-
gation into the use of pilot models in describing and understanding transient control tasks. In this
paper, a decrab maneuver before landing in heavy crosswind is considered. A decrab maneuver is
almost always executed manually and, as it induces high pilot workload, requires extensive pilot
training in simulators, making it a relevant maneuver to study. In addition, decrab maneuvers usu-
ally come with lateral acceleration profiles that are difficult to replicate with a reasonable level of
fidelity in most hexapod simulators.
To examine the effects of lateral motion cueing components on pilot perception and control
an experiment was performed in the SIMONA Research Simulator (SRS) of TU Delft. The ex-
periment can be considered as a follow-up to the study performed by Smaili et al.,
22
extending
his experiment in a number of ways. First, to better represent the way a typical motion base is
2 of 35
configured and tuned, a different workspace optimization method was chosen, as described in Sec-
tion IV. In this method the lateral motion components were tuned for maximum workspace use in
each experimental condition, rather than using constant filter gains which are tuned for one (full
motion) condition only. Second, to allow for later comparison to real flight with TU Delft’s labora-
tory aircraft, a Cessna Citation 500 aircraft model was used which is considerably smaller than the
Fokker 100 model used by Smaili et al. Third, additional performance and control metrics were
used. Finally, only pilots actively performing a decrab maneuver were considered in this paper,
while in the research of Smaili et al. both pilots-flying and pilots-not-flying were examined.
The paper is structured as follows. In the next section some general aspects of the decrab ma-
neuver are discussed. In Section III, results from previous research regarding the effects of motion
stimuli on pilot perception and control are summarized. Section IV describes the simulator motion
cueing, including workspace optimization, used in the experiment. In Section V the experimental
method is described, followed by Section VI in which the results are presented. The results are
discussed in Section VII and the paper ends with conclusions and recommendations.
II. The Decrab Maneuver
A decrab maneuver is an asymmetric aircraft maneuver executedduring final approach in cross-
wind conditions. Under such conditions, the initial approach is usually performed wings-level with
the nose of the aircraft pointed into the wind such that the aircraft approaches the runway slightly
skewed, or crabbed, with respect to the runway. Figure 1(a) illustrates the aircraft attitude during
initial approach in left crosswind conditions. Here, the aircraft flies with a compensating heading
angle or crab angle, ψ, such that the track angle, χ, (i.e., the heading of the ground speed vector)
is aligned with the runway. Stick and pedals are in neutral position at this moment and wings are
level.
In order to keep the lateral forces acting on the landing gear low when touching down, the pilot
should ultimately align the aircraft with the runway. To achieve this, a decrab maneuver is gener-
ally performed using both aileron and rudder, with the desired outcome illustrated in Figure 1(b).
Now the aircraft is lined up with the runway, inducing a side slip angle β. Subsequently, roll cor-
rection is applied to prevent the aircraft from drifting sideways as a result of the slip angle. Lateral
touchdown position and lateral velocity with respect to the runway are now controlled by carefully
adjusting the aircraft roll attitude.
Decrabbing motion of an aircraft is characterized by roll and yaw rotations. In most aircraft the
pilot station is located above and in front of the aircraft center of gravity and the rotations cause
linear accelerations at the pilot station. The relation between the specific forces in the aircraft
center of gravity,
¯
f
cg
, and the specific forces at the pilot station,
¯
f
ps
, are given by:
3 of 35
V
wind
V
T AS
V
ground
ψ
(a) Before decrab maneuver
V
wind
V
T AS
V
ground
β
(b) After decrab maneuver
Figure 1. Geometry of a left crosswind decrab maneuver.
¯
f
ps
=
¯
f
cg
+
˙
¯ω
b
× ¯r + ¯ω
b
× ¯ω
b
× ¯r , (1)
in which ¯ω
b
is the aircraft body rate vector and ¯r is the spatial offset of the pilot station with
respect to the center of rotation. In the Cessna Citation 500 aircraft, the pilot’s head is located
3.2 m forward and 0.75 m above the center of gravity. As Eq. (1) indicates, the spatial offset can
have a considerable influence on the specific forces during a decrab maneuver. Through dynamic
coupling between both side slip angle β and yaw rate r
b
on the one hand and roll rate p
b
on the other
hand, the aircraft will also tend to roll when rudder inputs are given during a decrab maneuver. A
representative motion profile obtained by using an autopilot in steady wind conditions, is given in
Figure 2.
Here, the lateral motion cues at the pilot head are shown in the time domain as well as in the
frequency domain. In the frequency domain, the cues are compared with the vestibular thresholds
for motion detection, as found by Heerspink et al.
23
Note that the upper and lower thresholds as
defined by Heerspink et al. are sensory thresholds, i.e., thresholds measured in the dark with single
axis motion. During a decrab maneuver, the actual thresholds are expected to be higher, e.g. due to
indifference effects from the combination of motion cues.
24
As such, these thresholds are not used
here as a predictor of the motions’ perceivability, but as a perception-based measure of motion
magnitude.
From Figure 2 it can be observed that the low frequency parts of the sway specific force f
y
and the roll acceleration are above absolute perception thresholds. Yaw cues on the other hand are
found to be subthreshold for most runs. The influence of yaw motion cueing on perception ratings
and performance can therefore be expected to be small, as will be further discussed in Section III.B.
4 of 35
f
y
(m/s
2
)
0 5 10 15 20
-1
-0.5
0
0.5
Experiment Data
Mean threshold line
10
0
10
1
10
−3
10
−2
10
−1
10
0
˙p (deg/s
2
)
0 5 10 15 20
-6
-4
-2
0
2
4
10
0
10
1
10
−2
10
−1
10
0
10
1
Time (seconds)
˙r (deg/s
2
)
0 5 10 15 20
-6
-4
-2
0
2
Frequency (rad/s)
10
0
10
1
10
−2
10
−1
10
0
10
1
Figure 2. Lateral motion cues at pilot head during an automated decrab maneuver compared to perception thresholds.
23
The characteristic frequencies for the aircraft model in the landing configuration (including a
yaw damper) as used in the experiment are given in Table 1. Both sway cues and roll cues at the
Dutch Roll frequency, the dominant lateral mode, of 1.7 rad/s are expected to be perceivable during
decrab maneuvers in the actual aircraft.
Table 1. Cessna Citation eigenmotions for experiment landing configuration.
eigenvalue natural freq. damping period time to damp
λ ω
0
ζ P T
1
2
(-) (rad/s) (-) (s) (s)
Short period -1.30 ± 1.66j 2.11 0.617 2.98 0.533
Phugoid -0.0251 ± 0.221j 0.223 0.113 28.2 27.6
Dutch Roll -1.05 ± 1.32j 1.69
0.623 3.72 0.659
Spiral 0.0152 - - - -45.7
Roll subsidence -2.22 - - - 0.313
III. Previous Research
Presenting motion cues to a pilot in simulated flight has often been shown to increase pilot
performance
18,25
and the pilots’ subjective impressions of simulator fidelity.
26
A large variety of
studies have been conducted on the influence of individual motion cueing components on pilot
control performance, pilot motion perception and simulator fidelity ratings. A selection of studies
which focused on lateral motion cueing and are thus of interest for aircraft decrab maneuvers, is
discussed here.
5 of 35
III.A. Helicopter Simulation Experiments
A series of experiments of particular interest to the current study were concerned with helicopter
simulation. Schroeder
27
explored motion cueing on NASA’s Vertical Motion Simulator (VMS) in
four of the six rigid-body degrees of freedom (roll, yaw, sway and heave) for multiple helicopter
control tasks, amongst which a yaw capture task. He found that lateral and vertical translational
motion platform cues had significant effects on reported simulator fidelity and pilot performance
and that they reduced pilot physical and mental workload. Yaw and roll rotational motion platform
cues were found to be less important. In particular, yaw rotational motion cues did not appear at
all useful for improving performance or reducing workload. It was also concluded that combin-
ing lateral translational motion cues with visual yaw rotational cues could induce a sensation of
physical rotation, even when the motion platform was not rotating.
Grant et al.
28
repeated the yaw capture task of the research of Schroeder in the University of
Toronto Institute for Aerospace Studies (UTIAS) Flight Research Simulator and focused on the
roles of sway translation and yaw rotation. Again, translational motion was found to improve pilot
performance. In contrast to the conclusions of Schroeder, Grant et al. found that yaw motion could
also increase performance, although additional improvements in the presence of translational mo-
tion were small. Both studies agreed on the fact that sway motion could be perceived as yaw motion
when combining it with visual yaw rotational cues. Ellerbroek et al.
14
did a second repetition of
Schroeder’s yaw capture experiment in the SRS at TU Delft and added a combined disturbance and
tracking task to further examine the individual effects of yaw and sway motion cues. In contrast to
the conclusions of the two previous experiments, it was found that yaw and sway motion showed
more equal contributions to pilot performance and fidelity ratings. Although disagreement exists
about the role of yaw, the significance of the effect of sway motion cues was demonstrated in all
three studies.
III.B. Aircraft Decrab Simulation Experiments
A second series of experiments that are of interest to the research described in this paper, investi-
gated decrab maneuvers in crosswind before landing. Groen et al.
29,30
examined the effects of both
sway, roll and yaw on pilot perception for passivedecrab maneuvers, using the Generic Fighter Op-
erations Research Cockpit Environment (GFORCE) of the National Aerospace Laboratory (NLR).
A simplified and adapted Fokker 100 aircraft model was used in this study. The research concluded
that subjective perception of motion during a decrab maneuver was dominated by sway and roll,
rather than yaw. A remarkable result in this study was that 30% of the pilots still reported to sense
motion in the condition without outside visual, even when the simulator was not moving. This
indicates that pilots sometimes tend to perceive motion out of expectation, rather than perceiving
it through their vestibular system.
6 of 35
In a follow-up study, Smaili et al.
22
examined the effects of sway, roll and yaw on subjective
pilot perception, simulator fidelity rating, pilot workload and pilot compensation for both passive
and active decrab maneuvers. In this research, a Fokker 100 aircraft model was simulated in
the Generic Research Aircraft Cockpit Environment (GRACE) of the NLR. The results showed
that reported perception of simulator motion was positively affected by platform sway for both
the pilots-flying and pilots-not-flying. Roll only had a significant influence on motion perception
for the pilots-not-flying. Platform yaw motion seemed to only have a positive effect on motion
perception in the absence of sway. No significant effects on the fidelity ratings for pilots-flying
were found.
From these previous studies it can be concluded that in relatively simple control tasks, such
as a helicopter yaw capture task, sway motion cues are likely to improve pilot performance and
simulator fidelity. For these tasks, the benefits of adding yaw and roll motion cues were shown
to be not significant. For decrab maneuvers, sway can be expected to increase the perception of
motion. No significant effects of lateral motion cueing components on reported simulator fidelity
were found.
IV. Simulator Motion Cueing
Pilot control behavior depends on the information available in the environment, perceived
through a complex multi-sensory system with contributions from visual, vestibular and proprio-
ceptive sensors. This section will discuss the generation of simulator motion cues and the motion
cueing algorithm optimization process used in this study, which aimed at a fair comparison be-
tween motion configurations and an optimal use of the available simulator workspace.
IV.A. Classical Washout Algorithm
The limited workspace available for flight simulator motion makes the generation of motion cues a
challenging problem. To compensate for the inability to provide the exact same cues in a simulator
as in a real aircraft, limitations in the human vestibular system are exploited as motion cueing
algorithms drive the motion system such that perceived differences between simulator motion and
real aircraft motion are sufficiently small.
The most widely used motion drive algorithm in commercial simulators is the classical washout
algorithm (CWA). First proposed by Schmidt and Conrad
31
and further improved and documented
by Reid and Nahon,
32,33
the CWA is characterized by the separation of high-frequency and low-
frequency motion content using linear high-pass and low-pass filters. The underlying principle of
the algorithm is that high-frequency accelerations and angular rates are able to pass through, but
low-frequency components, the components that otherwise would induce large actuator displace-
ments, are attenuated. Sustained longitudinal and lateral specific forces can be simulated by tilting
7 of 35
the simulator cabin at subthreshold rates, such that gravitational components arise in the body ref-
erence frame. The effectiveness of the CWA was demonstrated in a study by Reid and Nahon.
33
In
an analysis of the frequency response of this algorithm they showed that, when neglecting scaling
and limiting and further assuming small angles, the overall transfer function between uncoordi-
nated aircraft pitch or roll motion and the corresponding simulator motion was close to unity at all
frequencies.
¯
f
ac
¯ω
ac
K
f
K
ω
T
S→I
H
lp,xy
H
hp,φθψ
H
hp,xyz
Rate
Limit
1
s
Tilt
R
B→I
Coord.
High-Pass Specific Force Channel
Low-Pass Specific Force Channel
High-Pass Angular Rate Channel
¯g
+
+
β
LP
β
HP
¯
β
sim
¯a
sim
Figure 3. Classical washout algorithm.
A schematic representation of the CWA is given in Figure 3. All reference frames used in
the definition of the CWA are defined in the Appendix of this paper. In Figure 3, it can be seen
that the inputs of the algorithm consist of specific forces
¯
f
ac
and angular rates ¯ω
b,ac
, computed
in the aircraft model. Note the similarity with the human vestibular system, which enables human
perception of specific forces
¯
f through the otoliths and of angular rates ¯ω through the semi-circular
canals.
34
Before filtering, the input channels are scaled with filter gains K
f
= diag(K
f
x
, K
f
y
, K
f
z
)
and K
ω
= diag(K
ω
p
, K
ω
q
, K
ω
r
) and transformed to the inertial reference frame by transformation
matrices T
S→I
and R
B→I
. Both matrices depend on the actual state of the simulator and are
therefore continuously updated. After scaling and transformation to the inertial reference frame
the specific forces and angular rates are high-pass filtered in the high-pass channels using H
hp,xyz
and H
hp,φθψ
respectively. In the low-pass specific force channel, surge and sway specific forces are
filtered by a low-pass filter H
lp,xy
and then used to tilt the cabin. An implementation of the CWA
is characterized by the combination of the selected filter gains and break frequencies.
For this study, the CWA was adapted to allow the enabling and disabling of individual degrees
of freedom. For roll and yaw this was done by employing a zero gain at the input of the algorithm.
For the sway degree of freedom a specific force input of zero would still lead to large sway move-
ments in the presence of roll, as the simulator would try to negate the roll-induced specific force
8 of 35
from gravity. This problem was solved by introducing a second scaling gain at the sway output of
the algorithm, which could be set to zero when disabling the sway channel. As a result, the pilots
could still perceive sway specific forces when sway was disabled, but these would only be caused
by the simulator’s roll angle and not by any lateral simulator accelerations.
A second adaptation to the CWA was the addition of prepositioning. With prepositioning a
larger part of the motion space can be used when the future maneuvers can be anticipated, as was
the case in this study. During the initial part of the flight the simulator was smoothly repositioned in
sway and heave with sub-threshold accelerations.
35
The amount of prepositioning was determined
as described in Section IV.C.1.
IV.B. Experimental Conditions
As stated in the introduction, the individual effects of lateral motion cueing components on pilot
perception and control will be examined in this paper. With three separate motion components
under investigation, the eight possible combinations are defined in Table 2. It should be noted that
longitudinal motion cueing (i.e., surge, heave and pitch cueing) is present in all eight conditions as
well, and their filter settings are kept constant for all conditions.
Table 2. Combinations of lateral motion cueing components.
Condition Symbol Yaw Roll Sway
No Lateral Motion no · · ·
Yaw y × · ·
Roll r · × ·
Roll + Yaw ry × × ·
Sway s · · ×
Sway + Yaw sy × · ×
Sway + Roll sr · × ×
Sway + Roll + Yaw sry × × ×
IV.C. Motion Filter Tuning
Now that the structure of the motion drive algorithm has been defined, the CWA parameters for the
different motion cueing conditions can be determined. The values of these parameters have a large
influence on the motion profile that is ultimately sensed by a pilot in the simulator.
36
A systematic
tuning approach is therefore essential for a good simulator experiment.
In the research of Smaili et al.
22
the strategy was adopted to tune the motion filter settings such
that optimal workspace use is guaranteed in the full motion condition (sry). The other motion cue-
ing conditions were obtained by setting the appropriate filter gains to zero. A disadvantage of this
tuning strategy, however, is the suboptimal (and thus less representative of real-world applications)
9 of 35
use of the available motion space in conditions other than full motion. For this reason, another
tuning strategy is adopted in this paper. First, filter gains and break frequencies were determined
such that in the yaw only (y), roll only (r) and sway only (s) conditions the simulator workspace
was used optimally. Second, the settings for the other conditions were obtained by reducing all
lateral filter gains proportionally to keep the simulator within its workspace. Motion filter break
frequencies remained unchanged in this process. The advantage of this tuning strategy is that for
each condition the simulator motion cueing makes full use of the available workspace and the
overall motion magnitude is comparable for all conditions.
To ensure limited simulator displacement through sufficient motion washout for the studied
maneuver, in combination with minimal phase distortion, the high-pass and low-pass (tilt coordi-
nation) translational filters were of second order (Eqs. 2 and 3) and the rotational filters were of
first order (Eq. 4). The damping coefficients of all filters were set to unity.
H
hp,xyz
(s) =
s
2
s
2
+ 2ζω
n,hp
s + ω
2
n,hp
(2)
H
lp,xy
(s) =
ω
2
n,lp
s
2
+ 2ζω
n,lp
s + ω
2
n,lp
(3)
H
hp,φθψ
(s) =
s
s + ω
b,hp
(4)
IV.C.1. Lateral Motion Filter Tuning
The filter gains and break frequencies of the lateral motion channels were carefully tuned, using the
approach of Gouverneur et al.,
37
which makes use of the Sinacori motion fidelity criterion,
38
the
later work of Schroeder
27
and off-line estimates of the required actuator extensions. Characteristic
motion profiles of various piloted decrab runs performed in the SRS during the testing phase of the
experiment were used for this tuning.
In Figure 4 the Sinacori motion fidelity criterion is shown for sway high-pass filter settings for
the sway only condition (s). In the figure, each point represents a filter with a unique combination
of high-pass filter gain and break frequency. For a large number of combinations the phase shift
and system gain of the filter are calculated at a frequency of 1.0 rad/s and added to the Sinacori plot.
In this figure, high system gain and low phase shift correspond with high motion fidelity.
27
Gou-
verneur et al. proposed that for each combination of high-pass filter gain and break frequency, an
assessment can be made whether the simulator actuators would violate an actuator length, velocity
or acceleration restriction for a particular motion profile. The Sinacori criterion in combination
with the approach of Gouverneur et al. and a suitable reference profile can now provide guidance
for choosing a combination of filter parameters that results in minimum phase shift and maximum
10 of 35
Gain at 1 rad/s
Phase shift at 1 rad/s
high
medium
low
No restrictions violated
Length restriction violated
0 0.2 0.4 0.6
0.8
1
0
20
40
60
80
100
120
140
160
Figure 4. Workspace boundary definition for SRS adjusting sway high-pass filter.
system gain without violating actuator restrictions. For the actuator restrictions an operational
range of ± 0.555 m was used for the SRS. Furthermore, actuator velocity was restricted at 1 m/s
and actuator acceleration at 10 m/s
2
.
37
For this experiment the actuator length was found to be the
primary restriction.
From Figure 4 the encircled point was chosen as an optimal combination of low phase shift
and relatively high system gain at which the actuator length criterion was only just violated. This
point corresponded to a system gain of 0.59 and a phase shift of 24.78
◦
for a 1.0 rad/s input signal.
The corresponding sway high-pass filter had a filter gain K
f
y
of 0.6 and a break frequency ω
n,hp
y
of 0.3 rad/s. The low-pass break frequency was defined relative to the high-pass break frequency
as recommended by Reid and Nahon
33
for a good transient response to a step change in specific
force
¯
f
ac
:
ω
n,lp
= 2 ω
n,hp
(5)
New offline simulations were performed with these filter parameters to examine the actuator exten-
sions with respect to their minimum and maximum stroke. An example of the actuator extensions
for the sway only condition (s) is given in Figure 5(a). A definition of the actuator designations is
given in the Appendix of this paper.
From Figure 5(a) it can be observed that without prepositioning the first and fourth actuator
would violate their maximum extensions due to the large lateral displacement of the simulator
cabin. This was of course expected because of the length violation found in Figure 4. By applying
a prepositioning y
pp
of 50 cm in the sway direction, an actuator profile with a buffer of 20 cm at
both sides is obtained (Figure 5(b)). This buffer of 20 cm at both sides was used for all conditions
to account for different piloting techniques. Furthermore, a prepositioning z
pp
of 10 cm in the
11 of 35
max. extension
min. extension
Time (s)
Actuator extension (m)
0 5
10
15 20 25 30 35 40 45
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
(a) Actuator lengths for decrab without prepositioning
max. extension
min. extension
Time (s)
Actuator extension (m)
l
1
and l
4
l
2
and l
5
l
3
and l
6
0 5
10
15 20 25 30 35 40 45
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
(b) Actuator lengths for decrab with prepositioning
Figure 5. Actuator extensions for an experimental decrab run for K
y
= 0.6 and ω
n,hp
y
= 0.3 rad/s, sway only condition (s). Final
prepositioning is y
pp
= 0.5 m and z
pp
= 0.1 m.
downward heave direction also showed to improve the use of workspace for this condition.
This two-stage design approach using the Sinacori fidelity criterion and the evaluation of the
actuator extensions, was performed for the yaw only (y), roll only (r) and sway only (s) conditions.
For each of the other conditions the first stage of selecting the break frequency and gain was
replaced by the selection of a proportional gain K
p
which uniformly restricts the gains on all lateral
degrees of freedom, relative to the singular motion conditions. Any required prepositioning was
then chosen by examining the actuator extension plots. Table 3 lists the resulting scaling factors.
The filter parameters for the lateral channels that were ultimately obtained are summarized in
Table 4.
Table 3. Proportional scaling for motion cueing combinations.
Condition K
p
Roll + Yaw 0.4
Sway + Yaw 0.9
Sway + Roll 0.4
Sway + Roll + Yaw 0.4
IV.C.2. Longitudinal Motion Filter Tuning
The filter gains and break frequencies of the longitudinal motion channels (surge, heave and pitch)
were taken equal for all eight conditions. For the longitudinal translational cues to have little
influence on the use of simulator workspace during the experiment, high-pass break frequencies
for the second order surge and heave filters, ω
n,hp
x
and ω
n,hp
z
, respectively, were chosen relatively
high at 2 rad/s, similar to those used by Smaili et al.
22
The low-pass break frequencies were
chosen according to Equation 5. Pitch rotation filter gain K
ω
q
and break frequency ω
b,hp
q
were
12 of 35
Table 4. Lateral motion filter settings.
no y r ry s sy sr sry
K
p
(-) · 1.0 1.0 0.4 1.0 0.9 0.4 0.4
sway
(
K
f
y
(-) · · · · 0.60 0.54 0.24 0.24
ω
n,hp
y
(rad/s) · · · · 0.30 0.30 0.30 0.30
ω
n,lp
y
(rad/s) · · · · 0.60 0.60 0.60 0.60
roll
n
K
ω
p
(-) · · 0.80 0.32 · · 0.32 0.32
ω
b,hp
p
(rad/s) · · 0.30 0.30 · · 0.30 0.30
yaw
n
K
ω
r
(-) · 1.00 · 0.40 · 0.90 · 0.40
ω
b,hp
r
(rad/s) · 0.05 · 0.05 · 0.05 · 0.05
y
pp
(m) · · · · 0.5 0.45 · ·
z
pp
(m) · · 0.1 0.1 0.1 0.1 0.1 0.1
chosen equal to the values of the experiment of Smaili et al.
22
To conclude the definition of the
filter parameters for the longitudinal channels, surge filter gain K
f
x
and heave filter gain K
f
z
were
taken equal to the nominal sway filter gain K
f
y
(i.e. 0.6) from Section IV.C.1. The longitudinal
filter settings are summarized in Table 5.
Table 5. Longitudinal motion filter settings.
K ω
n,hp
ω
n,lp
ω
b,hp
(−) (rad/s) (rad/s) (rad/s)
Surge, x 0.60 2.00 4.00 -
Heave, z 0.60 2.00 - -
Pitch, θ 0.45 - - 0.25
V. Experiment
To further investigate the mixed results found in literature for the influence of yaw and roll
rotational and sway translational motion on pilot perception and control performance in decrab
maneuvers, an experiment was designed comparable to that of Smaili et al.
22
For the experiment,
pilots were asked to perform a large number of decrab maneuvers. This section describes important
aspects of the setup of this experiment, in addition to the definition of the motion cueing already
provided in Section IV.
V.A. Apparatus
The experiment was conducted in the SIMONA Research Simulator (Figure 6)
39
of TU Delft. The
SRS is equipped with a 180
◦
× 40
◦
collimated visual display system with a refresh rate of 60 Hz.
13 of 35
(a) SRS with its hexapod motion base (b) SRS flight-deck and outside visual
Figure 6. The SIMONA Research Simulator.
Dynamics of the visual system can be regarded as a pure time delay τ
v
which was shown to have
an average value of approximately 25-30 ms.
14
The dynamics of the SRS motion system can be
described by a second-order low-pass filter multiplied with a pure delay τ
m
of approximately 30
ms.
40
The interior of the cabin is outfitted as a generic, two-person cockpit of a modern airliner.
The pilots controlled the aircraft from the left-hand seat which was fitted with a loaded control
column and rudder pedals. The dynamic characteristics of the controls were matched to the aircraft
as much as possible. Furthermore, an Inertial Measurement Unit (IMU) was installed near the
simulator’s motion reference point, such that the actual specific forces and rotational accelerations
in the simulator were known.
V.B. Task
In order to concentrate on the execution of the decrab maneuver itself, each experimental run was
initiated on final approach at a radio altitude of 350 ft to land on runway 06 of Amsterdam Airport
Schiphol (EHAM). At the start of a run the aircraft was trimmed in the landing configuration on
the glide slope for a steady Instrument Landing System (ILS) approach. Indicated airspeed was
110 kts and aircraft weight 11,000 lbs. Visual Flight Rules (VFR) conditions were applicable and
a realistic view out of the window was presented. Crosswind of 20 kts was present perpendicular
to the runway from either left or right, resulting in a crab angle of 10.5
◦
.
Pilots were asked to decrab the aircraft at a height of their choice, such that touchdown would
take place with minimum lateral velocity and lateral offset with respect to the runway centerline
and minimum heading angle error. Thresholds for these measures were defined at 1 m/s lateral
velocity, 5 m lateral offset and 5 deg heading error. When exceeding any of these thresholds,
performance was defined to be inadequate and the run was discarded and repeated. The run ended
at three seconds after touchdown, such that the experimental scenario covered the entire landing
phase. Because the aircraft model lacked lateral gear dynamics, the aircraft could slide somewhat
after touchdown and the pilots were instructed to ignore the roll-out motions just before simulator
14 of 35
freeze.
To increase the feeling for the aircraft dynamics and to add some workload in the approach
before decrab, turbulence was added based on the patchy model developed by Van de Moesdijk.
41
For the turbulence to have relatively little influence on motion cueing and control during decrab,
and to increase repeatability of the turbulence realization, the symmetrical deviations from the
Gaussian distribution function were kept small.
V.C. Participants and Instructions
Six experienced Cessna Citation II pilots participated in the experiment. The participants, all
males, had an average age of 51 years (σ = ± 14.8 years) and had an average flying experience
of 7,800 hours (σ = ± 4,200 hours). Participants were elaborately briefed before the experiment
and when extra instruction proved to be necessary, it was given during the training period. The
participants were instructed to make a landing with heading alignment, lateral offset from the
runway centerline and lateral velocity as small as possible. No instructions were given regarding
the longitudinal touchdown position or the vertical speed at touchdown.
V.D. Aircraft Dynamic Model
The DASMAT nonlinear Cessna Citation 500 aircraft model, as documented by Van der Linden,
42
was used in this study. This aircraft is a relatively light-weight twin-engine type and is smaller
in length (13.26 m) than the Fokker 100 model (35.53 m) used in the work of Groen et al.
29,30
and Smaili et al.
22
Sway cues due to yawing were therefore somewhat smaller, but rotational cues
were expected to be larger than for the Fokker 100. Given the fact that most disagreement exists
in the literature about the role of yaw and roll in simulator fidelity, this could be advantageous
for the purposes of the experiment. Furthermore, a Cessna Citation model was chosen such that
experiment results could eventually be compared with real flight data using the laboratory aircraft
of TU Delft, a Cessna Citation II. Table 1 already gave an overview of the dynamic characteristics
of the used aircraft configuration.
V.E. Independent Variables
As discussed in Section IV.C, a motion filter tuning strategy was adopted that strived for full use
of the available simulator workspace. The use of simulator workspace turned out to be similar
for each motion condition with the exception of the no lateral motion condition (no) and the yaw
condition (y), the latter due to the simulator’s ability to comfortably present full yaw motion well
within the workspace.
35
It was therefore assumed that the three motion components were inde-
pendent, as their presence or absence had no influence on the use of simulator workspace and the
overall magnitude of motion cues. For the statistical analysis of the results it was assumed that the
15 of 35
experiment had three independent variables, yaw, roll and sway, which could be either absent or
present.
The experiment was defined as a within-subject or repeated-measures design, meaning that
each participant evaluated all conditions. In addition, the wind direction was varied between left
and right during the experiment, but the effects of wind on the dependent measures were not
analyzed. To diminish the influence of wind on the results, each block of eight conditions was
combined with a random permutation of the four right and four left crosswind conditions, such
that both wind directions were presented the same number of times.
V.F. Experiment Design and Procedure
The experiment was conducted in three phases, one block of 24 training runs and two blocks of
each sixteen experiment runs. A run lasted up to three seconds after touchdown, which resulted
in an average runtime of 45 seconds. Data were recorded at 100 Hz. During training, pilots got
accustomed to the simulator environment, the aircraft dynamics and the experimental procedure.
During the experiment phase each condition was repeated four times. The order in which the
conditions were presented was randomized, such that the different conditions were distributed
evenly over the participants and such that fatigue would not lead to unwanted systematic variance
in the results.
V.G. Dependent Measures
In this experiment the influence of the independent variables on pilot performance, control activity,
pilot perception and simulator fidelity was examined, using both objective and subjectivemeasures.
Various dependent measures were identified to quantify these concepts.
V.G.1. Objective Measures
Pilot performance was determined by three groups of dependent measures. A first group of mea-
sures focused on the aircraft state at touchdown. Here, the lateral velocity, lateral position with
respect to the runway centerline and the heading alignment were considered. Furthermore, al-
though not a part of the instructions, also the position with respect to the runway threshold, and
the vertical velocity, were considered at touchdown. In a second group of measures, the entire
decrab phase was analyzed. The Root Mean Squared (RMS) was calculated for the localizer error
and the roll rate. A third group of performance measures consisted of the maximum yaw rate, the
minimum crab angle, the point of decrab initiation and the average decrab rate.
For the calculation of the RMS of certain signals, a window was defined in which every decrab
maneuver could be captured. When only considering the data within this window, the data during
the crabbed approach were neglected, while still being able to compare results independently. The
16 of 35
window was defined by means of distance with respect to the runway threshold and the boundaries
were set to include flights with both high and low decrab rates. The window runs from 200 meter
in front of and 450 behind the threshold.
For the calculation of the position of decrab initiation relative to the runway threshold and the
average decrab rate, the model illustrated in Figure 7 was used. A least-squares procedure was
used to determine the model’s parameters: the initial crab angle, the decrab initiation point and the
decrab rate.
Measured
Modelled
Distance up to threshold (m)
Heading angle (deg)
runway heading
decrab rate
decrab initiation
w.r.t. threshold
initial crab angle
-500050010001500
55
60
65
70
Figure 7. Model definition for determining decrab initiation and decrab rate.
Control activity was examined using four dependent measures. First, the RMS of the pedal
rate and the wheel rate were determined during the decrab. Here, control rates were preferred over
control deflections since the nature of decrab control would result in non-zero mean deflection sig-
nals for both pedal and wheel and the RMS would not be suited for these kind of signals. Second,
the maximum pedal deflection was determined and, third, the pedal deflection was integrated over
time within the decrab window.
V.G.2. Subjective Measures
Finally, subjective measures for this experiment were considered, i.e. a rating for motion intensity
and a rating for motion fidelity. The subjective measure of motion intensity was based on a four-
point scale, given in Table 6, similar to the scale used by Smaili et al.
22
The subjective measure
of motion fidelity was based on a three-point scale, given in Table 7, also similar to that of Smaili
et al. Note, however, that the verbal description for each point on the scale was added for this
experiment specifically and was not used in the experiment of Smaili et al. Furthermore note that
the original scale descriptions were in Dutch, the native language of all participants.
V.H. Experiment Hypotheses
Previous research, summarized in Section III, generally indicated that sway cues had a larger
positive influence on pilot performance and subjective fidelity than roll and yaw cues. Although
17 of 35
Table 6. Intensity rating scale
0.
No Motion
No lateral motion cues were perceived in the simulator
1.
Too weak
Intensity of perceived lateral motion was too weak com-
pared to expected motion cues
2.
Natural
Intensity of perceived lateral motion was natural com-
pared to expected motion cues
3.
Too strong
Intensity of perceivedlateral motion was too strong com-
pared to expected motion cues
Table 7. Fidelity rating scale
0.
Low fidelity
Lateral motion cueing differences from actual flight were
noticeable and objectionable
1.
Medium fidelity
Lateral motion cueing differences from actual flight were
noticeable, but not objectionable
2.
High fidelity
Lateral motion cues were close to those of actual flight
sway cues were expectedto be relativelysmall in this study due to the chosen aircraft type, the same
hypothesis is put forward here. Second, it is hypothesized that sway and roll cues increase motion
intensity ratings. For motion intensity ratings, Smaili et al. only found a significant influence of roll
motion cues for the pilots-not-flying. However, since a smaller aircraft was used, the hypothesis
was formulated that the role of roll cues would be larger. Third, it was hypothesized that yaw cues
were below threshold and would therefore have no influence on the dependent measures.
VI. Results
The results of the experiment are presented in this section and the effects of the experimental
conditions, as summarized in Table 2, on the dependent measures will be discussed. First, however,
some general findings of the experiment will be discussed.
VI.A. Decrab Variation
To produce consistent results which can form a solid foundation for later modeling efforts, the
experiment would benefit from a constant decrab technique which is employed equally by all par-
ticipants. This section examines the variation in piloting techniques and strategies employed in the
experiment. First, to show the spread in touchdown positions during the experiment, some results
of randomly chosen runs are shown in Figure 8. An indication of the lateral velocity with respect
to the runway centerline is given by the length of the vertical lines attached to each touchdown
point. Longitudinal spacing of the touchdown points is shown to be considerable, but this is some-
what expected since no instructions were given to the pilot concerning the longitudinal position of
touchdown.
18 of 35
Runway length (m)
Runway width (m)
Performance Threshold
Aimpoint
Lateral speed (m/s)
0
2.5
0 200 400 600 800 1000 1200 1400 1600
-10
-5
0
5
10
Figure 8. Position and indication of lateral velocity at touchdown for a number of experimental runs.
Secondly, to get an idea of the different strategies adopted by the pilots, the estimation results
for decrab initiation and decrab rates from the model of Figure 7 are plotted in Figure 9. The
data set is fitted to an exponential function using a least-squares method to show the correlation
in the figure. The resulting correlation coefficient corresponding to this fit was 0.762. It can be
seen that the execution of the decrab maneuvers had considerable variation over the different runs
and a strong correlation exists between decrab initiation and decrab rate, which could be expected.
Some decrab maneuvers were executed quite gently overa longer period of time, while others were
executed more fiercely in a matter of seconds. Section VI.B.3 further examines the influence of
the different motion conditions on these parameters.
Distance up to threshold (m)
Decrab rate (deg/s)
050010001500
0
0.5
1
1.5
2
2.5
Figure 9. Decrab initiation versus decrab rate for all experimental runs.
VI.B. Pilot Performance
The effects of the motion filter settings on pilot performance will be discussed for the three groups
of measures identified in Section V.G.1: the performance at touchdown, the performance during
the decrab phase, and other characteristics of the decrab maneuver. All performance measures
presented were adjusted for between-subject variability and tested for the normality assumption
using a Kolmogorov-Smirnov test. Furthermore, the sphericity assumption for the data was tested
19 of 35
with Mauchly’s test. Ultimately, the significance of the effects of motion filter settings on the
dependent measures was determined using a repeated-measures Analysis of Variance (ANOVA)
with three independent variables, as described in Section V.E. In the analysis, a significance level
of p < 0.05 will be used, while p < 0.1 will be regarded as a marginally significant result.
VI.B.1. Performance at Touchdown
The performance measures that are examined here are lateral velocity, lateral distance from the
runway centerline, heading alignment, vertical velocity and the longitudinal distance from the
runway threshold. All measures were determined per run at the moment of touchdown.
First, the results for lateral velocity are given in Figure 10(a). It is shown that the mean lateral
velocity at touchdown was reasonably low, i.e., well below the set threshold value of 1.0 m/s, in
all conditions. As indicated by the grouped results at the right of Figure 10(a), only roll had a
marginally significant effect on this measure (F
1,5
= 5.750, p = 0.062), with the addition of roll
motion actually increasing lateral velocity at touchdown. Next, in Figure 10(b) the lateral distance
with respect to the runway centerline is given. As could already be expected from Figure 8, the
mean touchdown points were all well within the set boundary of 5 m. Here, rather surprisingly, a
significant effect was found for the influence of yaw (F
1,5
= 9 .431, p = 0.028).
Roll Off
Roll On
Conditions
Lateral speed (m/s)
no y r ry s sy
sr sry
grouped
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
(a) Lateral velocity at touchdown
Yaw Off
Yaw On
Conditions
Lateral distance (m)
no y r
ry s sy
sr sry
grouped
1
1.5
2
2.5
3
(b) Lateral distance w.r.t. runway centerline at touchdown
Conditions
Heading alignment (deg)
no y r
ry s sy sr sry
2.5
3
3.5
4
4.5
5
(c) Heading alignment at touchdown
Figure 10. Means and 95% confidence intervals of lateral velocity, lateral distance and heading alignment at touchdown.
As a final lateral touchdown performance measure, the heading alignment at touchdown was
20 of 35
considered. The results are shown in Figure 10(c). First, it can be observed that the pilots seldomly
performed a full decrab and retained some crab angle at touchdown. Furthermore, the differences
of the heading alignment at touchdown for the different motion conditions are very small. The
ANOVA showed no significant effects for this measure.
Although no instructions were given to the pilot considering the longitudinal position or the
vertical velocity at touchdown, it is interesting to examine possible effects of the motion cueing
conditions on these quantities. The longitudinal distance, for example, can be assumed to be
correlated with the difficulty of handling the aircraft. In Figure 11 the mean values are given with
their 95% confidence intervals.
Conditions
Vertical velocity (m/s)
no y r
ry s sy sr sry
-1
-0.8
-0.6
-0.4
-0.2
(a) Vertical velocity at touchdown
Conditions
Longitudinal distance (m)
no y r
ry s sy sr sry
500
540
580
620
660
(b) Touchdown distance w.r.t. the runway threshold
Figure 11. Means and 95% confidence intervals of vertical velocity and longitudinal distance at touchdown.
By inspection, the trends in the two figures show resemblance, which is an indication of a
correlation between touchdown position and vertical velocity, i.e., the earlier the touchdown took
place, the harder the landing. A correlation coefficient r = 0.682 was found for these two measures
with a significance level of p = 0.062 using a Pearson’s Correlation Analysis. In the ANOVA, a
significant interaction was found between sway and yaw for the vertical velocity at touchdown
(F
1,5
= 7.008, p = 0.046). The same interaction shows a marginally significant effect on lon-
gitudinal distance (F
1,5
= 4.441, p = 0.089). Marginally significant interactions were found for
vertical touchdown velocity between yaw and roll (F
1,5
= 4.578, p = 0.085) and between sway
and roll (F
1,5
= 4.122, p = 0.098).
VI.B.2. Decrab Performance
Decrab performance should not only be determined at touchdown, but also during the maneuvering
period. For this analysis, the mean values of the RMS of the localizer error and the RMS of the
roll rate are considered. The starting and end point of both the localizer error and the roll rate are
determined by the decrab window defined in Section V.G.1. In Figure 12 the mean values for both
measures, together with their 95% confidence intervals are given.
No significant effects were found for effects of motion cueing on the RMS of the localizer error.
21 of 35
Conditions
RMS localizer error (10
−2
deg)
no y r
ry s sy
sr sry
1.5
2
2.5
3
3.5
4
(a) RMS localizer error
Roll Off
Roll On
Conditions
RMS roll rate (deg/s)
no y r
ry s sy sr sry
grouped
2
2.5
3
3.5
4
4.5
5
(b) RMS roll rate
Figure 12. Means and 95% confidence intervals of RMS of localizer error and roll rate within decrab window.
The RMS of the roll rate during the decrab window, however, showed to be significantly dependent
on roll motion cues (F
1,5
= 11.79 7 , p = 0.019). Roll rate significantly reduced when roll motion
cues were presented. Some pilots reported small pilot-induced roll oscillations in conditions where
simulator roll motion was absent; this effect may be related to this reduced roll rate effect found in
the statistical analysis.
VI.B.3. Other Performance Measures
A third group of performance measures that could characterize a decrab maneuver was considered.
First, the maximum yaw rate and minimum crab angle during the decrab phase were obtained for
each run. Second, the model of Figure 7 was fitted on the data of each run, such that the point
of decrab initiation with respect to the runway threshold and the average decrab rate could be
estimated.
Figure 13 shows a phase-plane portrait of a typical decrab run, in which crab angle and yaw
rate can directly be compared. In this phase-plane portrait, the maximum yaw rate r
max
and min-
imum crab angle can easily be observed. Note that when a heading overshoot took place, the
minimum crab angle was defined as negative. It is remarked here that, since yaw rate peaks could
possibly also be induced by turbulence gusts, the causes of the maximum yaw rate values were
first investigated. It was found that each yaw rate peak was indeed the result of the decrab motion.
The influence of motion cueing conditions on the maximum yaw rate and the minimum crab
angle are given in Figure 14. A Pearson’s correlation coefficient r = 0.235 was found between
maximum yaw rate and average decrab rate. However, the result was not significant (p = 0 .5 76).
No significant effects of the motion filter settings on the minimum crab angle were found as well.
For the maximum yaw rate, however, a significant effect was found for the interaction of roll and
yaw (F
1,5
= 27.527, p = 0.003) and marginally significant reductions were observed for yaw
(F
1,5
= 4.268, p = 0.094) and sway (F
1,5
= 5.028, p = 0.075).
The influence of motion cueing conditions on decrab initiation and decrab rate was analyzed
22 of 35
ψ
min
r
max
Crab angle (deg)
Yaw rate (deg/s)
time
heading alignment threshold
0 1 2 3 4 5
6
7
8 9 10 11
-2
-1
0
1
2
3
4
Figure 13. Phase-plane portrait of a decrab run.
Conditions
Max. yaw rate (deg/s)
no y r ry
s sy
sr sry
3.5
4
4.5
5
5.5
(a) Maximum yaw rate
Conditions
Minimum crab angle (deg)
no y r ry s sy
sr sry
1
1.5
2
2.5
3
3.5
4
(b) Minimum crab angle
Figure 14. Means and 95% confidence intervals of the maximum yaw rate and minimum crab angle.
next, depicted in Figure 15. Due to a violation of the normality assumption for both data sets,
the nonparametric Friedman’s ANOVA was used for these measures. No significant effect of the
motion cueing settings on one of these two measures was found, however. Furthermore, a Pearson’s
Correlation Analysis was performed to a show a possible correlation between the point of decrab
initiation relative to the runway threshold and the average decrab rate. A correlation coefficient r =
−0.170 was found, corresponding with an inverse correlation, but this result was not significant
(p = 0.688).
VI.C. Control Activity
The control activity of the pilots in the different motion conditions are examined by means of
several control measures. The mean RMS values of the pedal and wheel rates, as well as the pedal
deflection integrated over time are examined within the decrab window. Finally, the maximum
pedal deflection is also considered.
Mean RMS values of the pedal and wheel rate are given in Figure 16 together with their 95%
23 of 35
Conditions
Decrab initiation (m)
no y r ry s sy
sr sry
-200
0
200
400
600
(a) Decrab initiation w.r.t. runway threshold
Conditions
Decrab rate (deg/s)
no y r
ry s sy sr sry
0.2
0.4
0.6
0.8
1
1.2
1.4
(b) Average decrab rate
Figure 15. Means and 95% confidence intervals of the point of decrab initiation w.r.t. the runway threshold and the average decrab rate.
confidence intervals. Pedal control activity was slightly lowerin the presence of yaw, but this effect
was only marginally significant (F
1,5
= 6 .564, p = 0.051). It could be expected that wheel control
activity was influenced by roll motion cueing, but this influence was not found (F
1,5
= 0.038, p =
0.852). A significant interaction between yaw and sway was present in the data (F
1,5
= 8 .639, p =
0.032). The wheel rate RMS increased with the addition of sway when yaw was also present, but
decreased without yaw.
Yaw Off
Yaw On
Conditions
RMS pedal rate (10
−2
m/s)
no y r ry
s sy sr sry
grouped
1
1.5
2
2.5
3
3.5
(a) RMS of pedal rate
Conditions
RMS wheel rate (deg/s)
no
y r ry s sy
sr sry
16
18
20
22
24
26
28
30
(b) RMS of wheel rate
Figure 16. Means and 95% confidence intervals of RMS wheel rate and pedal rate within decrab window.
The mean values for the maximum pedal deflection and the integrated pedal signal, the latter
within the decrab window, together with their 95% confidence intervals are given in Figure 17.
The ANOVA showed that the effect of roll motion on maximum pedal deflection was marginally
significant (F
1,5
= 4.274, p = 0.094). The results also show that the interaction of sway and
yaw had a significant effect on the integrated pedal signal (F
1,5
= 16.863, p = 0 .0 09). The pedal
deflection slightly decreased with the addition of sway when yaw was also present, but increased
without yaw. This is the opposite effect from what was observed for the wheel rate RMS.
24 of 35
Roll Off
Roll On
Conditions
Max. pedal deflection (10
−2
m)
no y r ry
s sy
sr sry
grouped
5
5.5
6
6.5
(a) Maximum pedal deflection
Conditions
Integrated pedal deflection (ms)
no
y r ry s sy
sr sry
0.2
0.3
0.4
(b) Integrated pedal signal
Figure 17. Means and 95% confidence intervals of maximum pedal deflection and integrated pedal signal within decrab window.
VI.D. Subjective Ratings
Finally, in addition to the analysis of objective measures for pilot performance and control activity,
subjective ratings for motion perception and simulator fidelity were examined. Note that unlike
the objective measures, the ratings were not corrected for between-subject variability. First, the
participants were asked to rate the motion intensity on the four-point scale of Table 6. Boxplots of
the mean ratings per pilot are givenfor each condition in Figure 18(a). The figure shows the median
value, boxes extending to the 25th and 75th percentile and whiskers containing the maximum
and minimum values, unless there are outliers outside 1.5 times the interquartile range from the
edge of the boxes. These outliers are displayed as a +. Note that this four-point rating scale
is clearly ordinal in contrast with the interval scale of the dependent measures discussed before.
Since interval data is one of the assumptions for parametric tests, the repeated-measures ANOVA
was not applicable here. A non-parametric Friedman’s ANOVA was used instead, to examine
possible overall significant effects, and in the case of a significant result, Wilcoxon signed-rank
post-hoc tests were used for pair-wise comparison between conditions.
From inspection, the lateral motion intensity in the roll only (r) condition was perceived as
too strong. Furthermore, the four sway conditions appear to lead to higher ratings than in the
no motion (no), yaw only (y) and roll+yaw (ry) conditions. The Friedman’s ANOVA showed a
significant overall effect χ
2
(7) = 20.352, p = 0.005. As could be expected, the Wilcoxon signed-
rank post-hoc tests showed that this effect was mainly caused by the relatively high ratings for the
roll only (r) condition. All pairs that included this roll only (r) condition showed to be significantly
different, with significance levels between p = 0.026 (for the roll only (r) condition versus the
no motion (no) condition) and p = 0.046 (for the roll only (r) condition versus the sway only (s)
condition). One more significant result was found for the comparison between the no motion (no)
condition and the sway only (s) condition (p = 0.0 27).
For the rating of simulator fidelity, the participants were asked to use the three-point scale of
Table 7. Boxplots of the mean ratings per pilot are given for each condition in Figure 18(b). The
25 of 35
Conditions
no y r ry
s sy
sr sry
No motion
Too weak
Natural
Too strong
(a) Motion intensity ratings
Conditions
no y
r ry s sy sr
sry
Low
Medium
High
(b) Fidelity ratings
Figure 18. Motion intensity and fidelity ratings (see Table 6 and Table 7 for rating scale definition).
results show that the variance in the ratings is very high. From inspection, the combination of
sway and yaw seems to have a positive influence on fidelity ratings. Although the sway conditions
showed a slightly higher mean for the fidelity ratings, no significant overall effect was found using
a Friedman’s ANOVA. Also note the very large variation in the fidelity ratings in the roll only (r)
condition.
VII. Discussion
In general, the effects of varying lateral motion cueing on pilot behavior and motion fidelity
ratings were found to be small for the decrab maneuver studied in this paper. This can be partially
explained by the relatively small number of participants, but another aspect may be the particular
motions the pilots experienced in the different conditions. To increase insight and perhaps allow
for interpretation of some of the results, a selection of time histories of perceived lateral specific
forces and roll rotational accelerations in different motion conditions is given in Figure 19. Here,
the aircraft model cues correspond with the cues that a pilot would perceive, i.e., rotational rates
and specific forces at the pilot head, when actually flying in the aircraft. In contrast, the simulator
cues correspond with the cues provided in the simulator during the experiment. The moment of
decrab initiation, approximated with the model of Figure 7, is given in these figures as well.
From Figure 19 the effects of the filter settings for different motion cueing conditions are evi-
dent. A low break frequency and a high filter gain resulted in a high correlation between simulator
and aircraft model roll cues for the roll only condition (r). However, due to the disabling of the
specific force cancellation and the rotation around the motion reference point below the pilot, the
roll motion resulted in large fluctuations in the lateral specific force (see Figure 19(a)). In the sway
only condition (s), lateral specific forces were smaller and showed a better match with the aircraft
model cues. Furthermore, in Figure 19(c) the effect of the sway scaling factor K
f
y
= 0.6 is visible
26 of 35
time (s)
f
y
(m/s
2
)
decrab initiation
20 25 30 35 40
-1.5
-1
-0.5
0
0.5
(a) Roll only condition (r)
time (s)
˙p (deg/s
2
)
20 25 30 35 40
-20
-10
0
10
20
(b) Roll only condition (r)
time (s)
f
y
(m/s
2
)
20 25 30
35
-1.5
-1
-0.5
0
0.5
(c) Sway only condition (s)
time (s)
˙p (deg/s
2
)
20 25 30
35
-20
-10
0
10
20
(d) Sway only condition (s)
time (s)
f
y
(m/s
2
)
Aircraft model
Simulator
20 25 30 35
-1.5
-1
-0.5
0
0.5
(e) Sway+roll+yaw condition (sry)
time (s)
˙p (deg/s
2
)
20 25 30 35
-20
-10
0
10
20
(f) Sway+roll+yaw condition (sry)
Figure 19. Motion cues at pilot head in aircraft and simulator for different motion conditions.
for the sway only condition (s). Note that the roll cues in this case originated from the tilt coor-
dination, but since subthreshold tilting rates were used, the roll accelerations were negligible (see
Figure 19(d)). In the final case, the sway+roll+yaw condition (sry), both sway and roll simulator
cues roughly correlate with the aircraft cues, but here the effect of the extra scaling of the lateral
filter gains is apparent (see Table 3).
In Figure 20 the same simulator cues as in Figure 19 are given in the frequency domain to-
gether with the sensory perception thresholds. It is shown that for the roll only condition (r) both
sway specific forces and roll accelerations were superthreshold over a large frequency range. In
the sway only condition (s) the sway specific forces and roll acceleration were super- and sub-
threshold, respectively, as could be expected. Note that the perception of sway specific forces was
assumed to be much less than in the roll only case, an assumption supported by Figure 19. Finally,
both sway specific forces and roll acceleration are above threshold in the sway+roll+yaw condition
(sry). This analysis would indicate that different motion cueing conditions can indeed influence
motion perception and simulator fidelity. However, the combination of high workload and addi-
tional motion cues almost certainly increased the effective thresholds and pilots may still have had
trouble distinguishing the different motions. Pilot comments indicate that these difficulties indeed
occurred.
An overview of the statistical results for the decrab experiment is given in Table 8. On the basis
of this overview, the implications of the results will be discussed. Furthermore, the subjective
27 of 35
Simulator cues
Aircraft model cues
Mean threshold line
Frequency (rad/s)
f
y
(m/s
2
)
10
0
10
1
10
−4
10
−2
10
0
(a) Roll only condition (r)
Frequency (rad/s)
˙p (deg/s
2
)
10
0
10
1
10
−2
10
−1
10
0
10
1
(b) Roll only condition (r)
Frequency (rad/s)
f
y
(m/s
2
)
10
0
10
1
10
−3
10
−2
10
−1
10
0
(c) Sway only condition (s)
Frequency (rad/s)
˙p (deg/s
2
)
10
0
10
1
10
−4
10
−2
10
0
10
2
(d) Sway only condition (s)
Frequency (rad/s)
f
y
(m/s
2
)
10
0
10
1
10
−3
10
−2
10
−1
10
0
(e) Sway+roll+yaw condition (sry)
Frequency (rad/s)
˙p (deg/s
2
)
10
0
10
1
10
−2
10
−1
10
0
10
1
(f) Sway+roll+yaw condition (sry)
Figure 20. Fourier transforms of sway specific forces and roll accelerations versus absolute perception thresholds.
23
rating results will be compared to the results from earlier experiments by Groen et al.
29,30
and
Smaili et al.
22
Table 8. Summary of the results of the decrab experiment.
Effects
y r s r x y s x y s x r s x r x y
Lat. velocity td. · MS · · · · ·
Lat. position td. S · · · · · ·
Heading alignment td. · · · · · · ·
Vert. velocity td. · · · MS S MS ·
Long. position td. · · · · MS · ·
Performance RMS localizer error · · · · · · ·
RMS roll rate · S · · · · ·
Max. yaw rate MS · MS S · · ·
Min. crab angle · · · · · · ·
Decrab initiation · · · · · · ·
Avg. decrab rate · · · · · · ·
RMS pedal rate MS · · · · · ·
Control
RMS wheel rate · · · · S · ·
Max. pedal · MS · · · · ·
Integrated pedal · · · · S · ·
S - significant (p < 0.05)
MS - marginally significant (p < 0.10)
28 of 35
VII.A. Performance
Pilot performance metrics show little consistent variation over the different motion conditions.
Only two main effects were significant: a decrease in lateral touchdown position in the presence
of yaw and a decrease in RMS roll rate when roll was present. Yaw did not help significantly
in reducing the lateral touchdown velocity or the heading alignment, the two other performance
metrics the pilots were briefed to minimize. It can therefore not be concluded from this data that
the yaw motion (or any of the other motions for that matter) caused the pilot to perform his task
better. Of course the level of attained performance is not a measure of the simulator’s fidelity or its
usefulness for a particular task, and the pilots most certainly adapted to the different configurations
to achieve a certain level of performance. The other main effect, a decrease in aircraft roll motions
in the presence of simulator roll coincides with some pilot comments that the simulator roll helped
them minimize small pilot-induced roll oscillations. This effect is however not directly apparent in
the RMS wheel rate which would indicate that the motions were quite gentle.
In addition to these main effects, two significant interactions were found: vertical touchdown
velocitybetween sway and yaw and maximum yaw rate between roll and yaw. The vertical velocity
at touchdown decreased with the addition of swaymotion when yaw was also present, but increased
without yaw. The maximum yaw rate decreased with the addition of roll motion when yaw was
also present, but increased without yaw.
VII.B. Control Activity
No significant main effects were found for the examined pilot control variables. This is some-
what disappointing since it was expected that the ability to employ pilot modeling techniques to
differentiate between the motion conditions would initially be indicated by differences in the pilot
control signals. So far no modeling was successfully performed with the data for this experiment,
and further studies are being planned.
The only significant effects were observed in the interaction between sway and yaw on the
pedal deflection integrated over time within the decrab window and on the RMS of the control
wheel deflection rate. The pedal deflection slightly decreased with the addition of sway when yaw
was also present, but increased without yaw. The wheel rate exhibited the inverse behaviour.
VII.C. Ratings
Table 9 shows the results from the pilot ratings and compares them with the earlier studies intro-
duced in Section III. As could be expected from the reported difficulties the participants had in
discriminating the different motion conditions, no significant effects were found for the motion
fidelity rating. The motion intensity perception rating was only examined in pairs showing a sig-
nificant difference between the roll only condition versus all others and the sway only condition
29 of 35
versus the no motion condition. The relativelylarge lateral accelerations in the roll condition which
were caused by the distance between pilot head and the roll rotation axis and the absence of any
mitigating lateral cues from the sway channel, are suspected to play an important role in this result.
These problems with subjective motion ratings, especially in combination with a relatively
small number of participants, once again underline the desire to develop more objective metrics
to investigate the influence of different motion components on pilot behavior. Although this study
was not yet successful in applying such metrics, it is hoped that future studies can build on it to
achieve this goal.
Table 9. Test
Effects
y r s r x y s x y s x r
GFORCE
29,30
pnf · S S · S ·
Motion perception
GRACE
22
pnf · S S · · S
GRACE
22
pf · S · · S ·
SRS pf · S
a
S
b
N/A N/A N/A
GFORCE
29,30
pnf S S · · · ·
Motion fidelity
GRACE
22
pnf · · S · · S
GRACE
22
pf · · · · · ·
SRS pf · · · · · ·
S - significant (p < 0.05) pnf - pilots-not-flying
MS - marginally significant (p < 0.10) pf - pilots-flying
a
Roll only condition significantly larger than all others in pair-wise comparison
b
Sway only condition significantly larger than no motion in pair-wise comparison
VII.D. Reflection on Hypotheses
The first hypothesis was that sway motion would have a larger impact on performance and per-
ceived motion than roll or sway. No significant effects of sway motion were found on any of the
metrics, apart from an interaction with yaw on vertical velocity at touchdown. This hypothesis is
therefore not supported by the data.
The second hypothesis was that sway and roll motion would increase the reported motion
intensity. Although this is clearly the case for the roll only condition, this is suspected to be a
result of the implementation of this individual condition which led to large lateral specific forces
at the pilot station. The addition of roll to any of the other conditions had no significant effect. The
ratings are deemed not sufficient to support this hypothesis.
The third and final hypothesis was that yaw motion had no influence on any of the dependent
measures. This was not the case, as yaw had a significant effect on lateral touchdown position
and was involved in interactions on vertical velocity at touchdown and integrated pedal deflection
30 of 35
(with sway) and on maximum yaw rate (with roll). This hypothesis must therefore be rejected.
VIII. Recommendations
In addition to the findings of the research described in this paper, some recommendations for
future research can be put forward.
The aircraft used in this research was relatively small compared to, for example, typical jet
transport aircraft. Lateral motion profiles for decrab maneuvers will therefore be considerably
different. As a subject for further research it is recommended to investigate the effects of aircraft
size on the importance of motion cues in rotational and translational degrees of freedom.
In addition, further research into the roles of motion cueing algorithm and motion cueing strat-
egy is proposed. Other motion cueing algorithms are available that might increase the fidelity
of lateral motion cueing during decrab maneuvers. Furthermore, the influence of motion cueing
strategy on performance and control results is of interest.
A validation study should be performed to compare decrab maneuvers performed in real flight
with Delft University’s Cessna Citation II aircraft with the data from this study to investigate the
differences in pilot control behavior and aircraft response.
In the experiment simulation environment, a dynamic model for the horizontal interaction be-
tween aircraft landing gear and runway was not yet available. Lateral accelerations at touchdown
due to a landing with high lateral velocity or a large heading alignment mismatch, which can ad-
versely influence the control behavior of the pilots, were not perceivable in the simulator. The
development of such a model is therefore recommended.
To improve the consistency and significance of the results, future studies should use more
participants and see if improved training is necessary. In this study the training used all motion
configurations in random order, which perhaps prevented the pilots from developing an optimal
piloting strategy for a particular motion configuration.
Ultimately, understanding the roles of simulator cues on pilot control behavior in transient
maneuvers should result in the development of control-theoretical models for the description of
pilot control in such maneuvers. With these models, the influence of simulator motion cues on pilot
control behavior can be quantified, such that simulator fidelity can be evaluated more objectively.
Future research will therefore focus on the modeling of pilot control behavior.
IX. Conclusions
A simulator experiment was performed to investigate the influence of different motion cueing
configurations on pilot performance, control and perception during a decrab maneuver with a small
business jet. The lateral motion cueing was varied over the different configurations by combining
31 of 35
none or more of the yaw, sway and roll motions, with constant longitudinal motion cues, while
changing the lateral filter gains to make full use of the available workspace.
Few significant results were found on both performance and control behavior, while virtuallyno
significant effects were observed in the motion intensity and fidelity ratings. Participants reported
having difficulty in consistently rating the motion cues during the short and high-workload decrab
maneuver. Also a large variability was observed between and within participants in how the decrab
maneuver was performed, even though a consistent piloting strategy was strived for.
Results for the lateral motion cueing degrees of freedom show that yaw motion had an unex-
pected significant effect on lateral touchdown position, but not on lateral touchdown velocity or
longitudinal touchdown position or velocity. Sway motion had no significant effect on any of the
performance or control metrics. Roll motion had a significant reducing effect on RMS aircraft roll
rate, but not on the RMS wheel rate. Roll motion without accompanying sway or yaw motion led
to relatively large lateral specific forces at the pilot station and was rated significantly more intense
than any of the other configurations.
Appendix
X
b
X
s
Y
b
, Y
s
Z
b
Z
s
k V
(a) Aircraft reference frames
l
4
l
3
l
2
l
1
l
5
l
6
X
sim
Y
sim
Z
sim
X
i
Y
i
Z
i
(b) Simulator reference frames and actuator designa-
tions
Figure 21. Definition of reference frames.
As shown in Fig.21, the origin of the aircraft’s reference frames is located in the center of
gravity. The body frame of reference (subscript b) is fixed to the aircraft and its X- and Z-axis lie
in the plane of symmetry. The stability frame of reference (subscript s) is also fixed to the aircraft
but additionally has its X-axis aligned with the velocity vector in unperturbed flight.
32 of 35
The simulator’s reference frames have their origin in the center of the circle through the upper
gimbal points (UGP). The X-axis lies in the vertical plane of symmetry and the Z-axis points
straight down. The inertial frame of reference (subscript i) remains fixed to the world with its
origin in the UGP while the actuators are extended to half their stroke. The simulator body frame
of reference (subscript sim) moves with the simulator.
Acknowledgements
The authors would like to gratefully acknowledge the pilots who participated in the experiment
desribed in this paper for their effort and useful comments.
References
1
Hess, R. A. and Malsbury, T., “Closed-Loop Assessment of Flight Simulator Fidelity,” Journal of Aircraft,
Vol. 14, No. 1, March – July 1991, pp. 191–197.
2
Hess, R. A., Malsbury, T., and Atencio, Jr., A., “Flight Simulator Fidelity Assessment in a Rotorcraft Lateral
Translation Maneuver,” Journal of Guidance, Control, and Dynamics, Vol. 16, No. 1, 1993, pp. 79–85.
3
Zeyada, Y. and Hess, R. A., “Human Pilot Cue Utilization with Applications to Simulator Fidelity Assessment,”
Journal of Aircraft, Vol. 37, No. 4, 2000, pp. 588–597.
4
Hess, R. A. and Siwakosit, W., “Assessment of Flight Simulator Fidelity in Multiaxis Tasks Including Visual
Cue Quality,” Journal of Aircraft, Vol. 38, No. 4, 2001, pp. 607–614.
5
Zeyada, Y. and Hess, R. A., “Computer-Aided Asessment of Flight Simulator Fidelity,” Journal of Aircraft,
Vol. 40, No. 1, 2003, pp. 173–180.
6
Groen, E. L., Sma¨ıli, M. H., and Hosman, R. J. A. W., “Simulated Decrab Maneuver: Evaluation with a Pilot
Perception Model,” Proceedings of the AIAA Modeling and Simulation Technologies Conference and Exhibit, San
Francisco (CA), No. AIAA-2005-6108, 2005.
7
Zaychik, K. B., Cardullo, F. M., and George, G., “A Conspectus on Operator Modleing: Past, Present and
Future,” Proceedings of the AIAA Modeling and Simulation Technologies Conference and Exhibit, Keystone (CO),
No. AIAA-2006-6625, 2006.
8
Hess, R. A. and Marchesi, F., “Pilot Modeling With Applications to the Analytical Assessment of Flight Sim-
ulator Fidelity,” Proceedings of the AIAA Modeling and Simulation Technologies Conference and Exhibit, Honolulu
(HI), No. AIAA-2008-6682, 2008.
9
Hosman, R., van der Geest, P., and van der Zee, J., “Development of a Pilot Model for the Manual Balked Land-
ing Maneuver,” Proceedings of the AIAA Modeling and Simulation Technologies Conference and Exhibit, Chicago
(IL), No. AIAA-2009-5818, 2009.
10
Van den Berg, P., Zaal, P. M. T., Mulder, M., and Van Paassen, M. M., “Preparation for Conducting Multi-
Modal Pilot Model Identification in Real Flight,” Proceedings of the AIAA Modeling and Simulation Technologies
Conference and Exhibit, Keystone (CO), No. AIAA-2006-6630, 2006.
11
Mulder, M. and Mulder, J. A., “Cybernetic Analysis of Perspective Flight-Path Display Dimensions,” Journal
of Guidance, Control, and Dynamics, Vol. 28, No. 3, 2005, pp. 398–411.
12
Steurs, M., Mulder, M., and Van Paassen, M. M., “A Cybernetic Approach to Assess Flight Simulator Fidelity,”
33 of 35
Proceedingsof the AIAA Modelling and Simulation TechnologiesConference and Exhibit, Providence (RI), No. AIAA-
2004-5442, 2004.
13
Zaal, P. M. T., Pool, D. M., Mulder, M., Van Paassen, M. M., and Mulder, J. A., “Identification of Multi-
modal Pilot Control Behavior in Real Flight,” Journal of Guidance, Control and Dynamics, Submitted for publication,
October 2009.
14
Ellerbroek, J., Stroosma, O., Mulder, M., and Van Paassen, M., “Role Identification of Yaw and Sway Motion
in Helicopter Yaw Control Tasks,” Journal of Aircraft, Vol. 45, No. 4, July – Aug. 2008, pp. 1275–1289.
15
Zaal, P. M. T., Pool, D. M., Mulder, M., and Van Paassen, M. M., “Multimodal Pilot Control Behavior in
Combined Target-Following Disturbance-Rejection Tasks,” Journal of Guidance, Control, and Dynamics, Vol. 32,
No. 5, 2009, pp. 1418–1428.
16
Pool, D. M., Zaal, P. M. T., Damveld, H. J., Van Paassen, M. M., and Mulder, M., “Pilot Equalization in
Manual Control of Aircraft Dynamics,” Proceedings of the 2009 IEEE International Conference on Systems, Man &
Cybernetics, San Antonio, Texas, 2009, pp. 2554–2559.
17
Pool, D. M., Zaal, P. M. T., Van Paassen, M. M., and Mulder,M., “Effectsof Heave Washout Settings in Aircraft
Pitch Disturbance Rejection,” Journal of Guidance, Control, and Dynamics, Accepted for publication, September
2009.
18
Zaal, P. M. T., Pool, D. M., De Bruin, J., Mulder, M., and Van Paassen, M. M., “Use of Pitch and Heave Motion
Cues in a Pitch Control Task,” Journal of Guidance, Control, and Dynamics, Vol. 32, No. 2, 2009, pp. 366–377.
19
Van Paassen, M. M. and Mulder, M., “Identificationof Human Control Behavior.” W. Karwowski: International
Encyclopedia of Ergonomics and Human Factors, Vol. 2nd edition, London: Taylor & Francis, 2006, pp. 400–407.
20
Nieuwenhuizen, F. M., Zaal, P. M. T., Mulder, M., van Paassen, M. M., and Mulder, J. A., “Modeling Hu-
man Multichannel Perception and Control Using Linear Time-Invariant Models,” Journal of Guidance, Control, and
Dynamics, Vol. 31, No. 4, 2008, pp. 999–1013.
21
Zaal, P. M. T., Pool, D. M., Chu, Q. P., van Paassen, M. M., Mulder, M., and Mulder, J. A., “Modeling Human
Multimodal Perception and Control Using Genetic Maximum Likelihood Estimation,” Journal of Guidance, Control,
and Dynamics, Vol. 32, No. 4, 2009, pp. 1089–1099.
22
Smaili, M. H., Jansen, H., Naseri, A., Groen, E. L., and Stroosma, O., “Pilot Motion Perception and Control
During a Simulated Decrab Maneuver,” Proceedings of the AIAA Modeling and Simulation Technologies Conference
and Exhibit, Hilton Head (SC), No. AIAA-2007-6797, 20–23 Aug. 2007.
23
Heerspink, H. M., Berkouwer, W. R., Stroosma, O., Van Paassen, M. M., Mulder, M., and Mulder, J. A.,
“Evaluation of Vestibular Thresholds for Motion Detection in the SIMONA Research Simulator,” AIAA Modeling and
Simulation Technologies Conference, AIAA-2005-6502, 2005.
24
Valente Pais, A. R., Mulder, M., Van Paassen, M. M., Wentink, M., and Groen, E. L., “Modeling Human
Perceptual Thresholds in Self-Motion Perception,” Proceedings of the AIAA Modeling and Simulation Technologies
Conference and Exhibit, Keystone (CO), No. AIAA-2006-6626, 2006.
25
Pool, D. M., Mulder, M., Van Paassen, M. M., and Van der Vaart, J. C., “Effects of Peripheral Visual and
Physical Motion Cues in Roll-Axis Tracking Tasks,” Journal of Guidance, Control, and Dynamics, Vol. 31, No. 6,
2008, pp. 1608–1622.
26
Reid, L. D. and Nahon, M. A., “Flight Simulation Motion-Base Drive Algorithms: Part 3. Pilot Evaluations,”
Tech. Rep. 319, University of Toronto Institute for Aerospace Studies, Dec. 1986.
27
Schroeder, J. A., “Helicopter Flight Simulation Motion Platform Requirements,” Tech. Rep. TP-1999-208766,
NASA, July 1999.
34 of 35
28
Grant, P. R., Yam, B., Hosman, R. J. A. W., and Schroeder, J. A., “Effect of Simulator Motion on Pilot Behavior
and Perception,” Journal of Aircraft, Vol. 43, No. 6, Nov. – Dec. 2006, pp. 1914–1924.
29
Groen, E. L., Smaili, M. H., and Hosman, R. J. A. W., “Simulated Decrab Maneuver: Evaluation with a Pilot
Perception Model,” Proceedings of the AIAA Modeling and Simulation Technologies Conference and Exhibit, San
Francisco (CA), No. AIAA-2005-6108, 15–18 Aug. 2005.
30
Groen, E. L., Hosman, R. J. A. W., and Smaili, M. H., “Perception Model Analysis of Flight Simulator Motion
for a Decrab Maneuver,” Journal of Aircraft, Vol. 44, No. 2, 2007, pp. 427–435.
31
Schmidt, S. F. and Conrad, B., “Motion Drive Signals for Piloted Flight Simulators,” Tech. Rep. CR-1601,
NASA, May 1970.
32
Reid, L. D. and Nahon, M. A., “Flight Simulation Motion-Base Drive Algorithms: Part 1. Developing and
Testing the Equations,” Tech. Rep. 296, University of Toronto Institute for Aerospace Studies, Dec. 1985.
33
Reid, L. D. and Nahon, M. A., “Flight Simulation Motion-Base Drive Algorithms: Part 2. Selecting the System
Parameters,” Tech. Rep. 307, University of Toronto Institute for Aerospace Studies, May 1986.
34
Hosman, R. J. A. W., Pilot’s perception and control of aircraft motions, Ph.D. thesis, Faculty of Aerospace
Engineering, Delft University of Technology, 1996.
35
Beukers, J. T., Stroosma, O., Pool, D. M., Mulder, M., and Van Paassen, M. M., “Investigation of Pilot Percep-
tion and Control During Decrab Maneuvers in Simulated Flight,” Proceedings of the AIAA Modeling and Simulation
Technologies Conference and Exhibit, Chicago (IL), No. AIAA-2009-6030, 2009.
36
Grant, P. R. and Reid, L. D., “Motion Washout Filter Tuning: Rules and Requirements,” Journal of Aircraft,
Vol. 34, No. 2, 1997, pp. 145–151.
37
Gouverneur, B., Mulder, J. A., Van Paassen, M. M., Stroosma, O., and Field, E. J., “Optimisation of the
SIMONA Research Simulator’s Motion Filter Settings for Handling Qualities Experiments,” Proceedings of the AIAA
Modeling and SimulationTechnologiesConferenceand Exhibit, Austin (TX), No.AIAA-2003-5679,11–14Aug. 2003.
38
Sinacori, J. B., “The Determination of Some Requirements for a Helicopter Flight Simulator Facility,” Tech.
Rep. CR-152066, NASA, 1977.
39
Stroosma, O., Van Paassen, M. M., and Mulder, M., “Using the SIMONA Research Simulator for Human-
Machine Interaction Research,” Proceedings of the AIAA Modelling and Simulation Technologies Conference and
Exhibit, Austin (TX), No. AIAA-2003-5525, 11–14 Aug. 2003.
40
Berkouwer, W. R., Stroosma, O., Van Paassen, M. M., Mulder, M., and Mulder, J. A., “Measuring the Per-
formance of the SIMONA Research Simulator’s Motion System,” Proceedings of the AIAA Modeling and Simulation
Technologies Conference and Exhibit, San Francisco (CA), No. AIAA-2005-6504, 2005.
41
Van de Moesdijk, G. A. J., “Simulation of Patchy Atmospheric Turbulence, Based on Measurements of Actual
Turbulence,” Memorandum M-240, Delft University of Technology, The Netherlands, Oct 1975.
42
Van der Linden, C. A. A. M., “DASMAT: the Delft University Aircraft Simulation Model and Analysis Tool,”
Tech. Rep. LR-781, Delft University of Technology, 1996.
35 of 35
