Pitch Motion Perception Thresholds During Passive and Active Tasks

Abstract

Knowledge about motion perception thresholds is essential for simulator motion cueing. Thresholds are generally measured in a passive experimental setup in which subjects do not actively influence their motion. For flight simulation applications, it is useful to also investigate thresholds during control tasks, where pilots actively influence the motion they sense. In this paper, thresholds were estimated during an active control task using a pilot model parameter identification method. A comparison with conventional passive threshold measurements was made. The threshold identification method was based on a multichannel pilot model extended with a nonlinear absolute threshold element. Two experiments were performed in a flight simulator: a passive experiment to measure the sensory pitch threshold, and an active experiment with a compensatory control task to identify the active pitch threshold. In the active experiment, the gain of the inertial motion amplitude was varied and two types of compensatory control tasks were considered. For both tasks, the pitch threshold was identifiable only for high motion gain levels. The measured passive threshold was lower than other values found in literature. The threshold identified from the active control task was higher than the measured passive threshold, but it was comparable with passive threshold values reported in other studies.

Full text

Pitch Motion Perception Thresholds During Passive
and Active Tasks
A.R. Valente Pais,∗D.M. Pool,†A.M. de Vroome,‡M.M. van Paassen,§and M. Mulder¶
Delft University of Technology, Delft, The Netherlands
Knowledge about motion perception thresholds is essential for simulator motion cueing.
Thresholds are generally measured in a passive experimental setup in which sub jects do
not actively influence their motion. For flight simulation applications it is useful to also
investigate thresholds during control tasks, where pilots actively influence the motion they
sense. In this paper, thresholds were estimated during an active control task using a
pilot model parameter identification method. A comparison with conventional passive
threshold measurements was made. The threshold identification method was based on
a multi-channel pilot model extended with a nonlinear absolute threshold element. Two
experiments were performed in a flight simulator: a passive experiment to measure the
sensory pitch threshold, and an active experiment with a compensatory control task to
identify the active pitch threshold. In the active experiment, the gain of the inertial motion
amplitude was varied and two types of compensatory control tasks were considered. For
both tasks, the pitch threshold was identifiable only for high motion gain levels. The
measured passive threshold was lower than other values found in literature. The threshold
identified from the active control task was higher than the measured passive threshold, but
it was comparable to passive threshold values reported in other studies.
∗Ph.D. student, Control and Simulation Division, Faculty of Aerospace Engineering, P.O. Box 5058, 2600GB Delft, The
Netherlands; a.r.valentepais@tudelft.nl. AIAA student member.
†Ph.D. student, Control and Simulation Division, Faculty of Aerospace Engineering, P.O. Box 5058, 2600GB Delft, The
Netherlands; d.m.pool@tudelft.nl. AIAA student member.
‡MSc student, Control and Simulation Division, Faculty of Aerospace Engineering, P.O. Box 5058, 2600GB Delft, The
Netherlands; aniekdevroome@gmail.com.
§Associate Professor, Control and Simulation Division, Faculty of Aerospace Engineering, P.O. Box 5058, 2600GB Delft,
The Netherlands; m.m.vanpaassen@tudelft.nl. AIAA member.
¶Professor, Control and Simulation Division, Faculty of Aerospace Engineering, P.O. Box 5058, 2600GB Delft, The Nether-
lands; m.mulder@tudelft.nl. AIAA senior member.
1 of 32
American Institute of Aeronautics and Astronautics
A. R. Valente Pais, D. M. Pool, A. De Vroome, M. M. Van Paassen, and M. Mulder,
“Pitch Motion Perception Thresholds During Passive and Active Tasks,” Journal
of Guidance, Control & Dynamics, vol. 35, no. 3, pp. 904–918, 2012.

Nomenclature
AFR Afferent Firing Rate IPS
AiForcing function sinusoid amplitude rad
eError signal rad
fdDisturbance forcing function rad
ftTarget forcing function rad
H(jω) Frequency response function
HcControlled element dynamics
Hnm Neuromuscular dynamics
Hpe Pilot visual response
Hpx Pilot vestibular response
HSC C Semi-circular canal dynamics
i∆Input signal to threshold element IPUT
jImaginary unit -
KcControlled element gain -
KmVestibular motion gain rad/IPUT
KvCentral visual gain -
KSC C Semi-circular canal dynamics gain -
KθCueing motion gain -
nRemnant signal rad
niInteger multiple of base frequency -
o∆Output signal of threshold element IPUT
TmMeasurement time s
Tvl Central visual lead time constant s
uControl signal rad
uvVisual contribution to the control signal rad
umMotion contribution to the control signal rad
δeInput to the controlled element rad
∆abs Absolute threshold value IPUT
∆meas Measured threshold value rad/s2
ωFrequency rad/s
ωmBase frequency rad/s
ωnm Neuromuscular dynamics frequency rad/s
φiForcing function sinusoid phase shift rad
τmVestibular motion time delay s
τvCentral visual time delay s
θPitch angle rad
Θ Estimated parameter vector
ζnm Neuromuscular dynamics damping ratio -
I. Introduction
Knowledge of human motion perception and perception thresholds is fundamental for the design, opti-
mization and operation of flight simulators. Throughout the past half-century a large number of studies
on perception thresholds have been performed.1–5 However, most of these studies consisted of single-axis
motion experiments, where the thresholds were measured in the absence of other cues, that is, without vi-
sual motion or inertial motion in other degrees-of-freedom. In an aircraft simulation environment, typically
multiple motion cues are provided to the pilot. Research on thresholds in such a multi-cue environment has
shown that motion perception thresholds are affected by the presence of other motion6–8 and visual cues9
as well as workload levels.1
All of these studies, however, used passive tasks, where subjects did not actively influence the motion
for which the threshold value was determined. This is usually not the case during flight simulation where,
generally, the pilot is performing a control task that affects the vehicle and simulator motion. A few studies
2 of 32
American Institute of Aeronautics and Astronautics

have measured perception thresholds while sub jects were performing some type of manual control task.
Hosman and Van der Vaart1used a visual control task, so no motion feedback, and an auditory binary
choice task to increase subjects workload. Samji and Reid10 provided subjects with motion feedback but
their control task was in a different degree-of-freedom (DOF) than the threshold measurement.
The logical next step, to improve our knowledge of motion perception thresholds in a flight simulation
environment, is to measure them during an active control task in the same DOF. In this case, the simulator
motion will oscillate between sub- and supra-threshold amplitudes and it is not possible to rely on subjects
subjective reports of perceived motion. However, the control signal could be used to derive information
regarding the threshold value. If the supplied motion oscillates between sub- and supra-threshold amplitudes,
the threshold might affect the usage of the supplied motion and, thereby, affect the control behavior. By
measuring the pilot’s control behavior and using pilot model parameter identification techniques,11–15 the
threshold can then be identified.
Pilot model parameter identification methods have been used previously,16–20 but they did not include
a threshold estimation. De Vroome et al.21 describe a novel identification method that allows for non-
linearities in the model structure, hence allowing the use of a motion threshold in the model. In this
paper, this new identification method is used to estimate the pitch sensory threshold during an active task.
The identification method used has been described in detail in Pool et al.22 and throughout this paper
the reader will be referred to this reference for the details of the identification algorithm. The thresholds
identified during the active task are compared to measurements made during a passive, single-axis, in the
dark stimulation. The results from both the passive and the active experiment are compared to results found
in the literature.
The paper starts with an overview of the relevant literature for the development of the proposed threshold
model. Second, the experimental methods for the passive and the active experiments are described. The
threshold identification method, including the pilot model, is also briefly described here. Next, the results of
both experiments are presented and compared. Finally, the obtained results are discussed and conclusions
and recommendations are given.
II. Angular Motion Thresholds
II.A. Hosman and Van der Vaart
Hosman and Van der Vaart were among the first to measure sensory thresholds for pitch and roll rotational
motion perception over a range of stimulus frequencies.1,23 For each frequency, an acceleration motion profile
with varying amplitude was created, as depicted in Figure 1. Two threshold measurements were taken, one
for the part with increasing amplitude at the moment subjects started perceiving the motion, the upper
3 of 32
American Institute of Aeronautics and Astronautics

threshold, tup, and one for the part with decreasing amplitude when sub jects lost perception of motion, the
lower threshold, tlow. They found that the upper threshold was significantly higher than the lower threshold.
Time, s
Acceleration, rad/s2
020 40 60 80 100
tlow
tup
Figure 1. Example motion profile, used to determine upper and lower sensory threshold levels.1,5
They also found that both threshold values varied with the stimulus frequency. Based on work by other
authors they reported that the stimulation of the semi-circular canals (SCC) results in a certain afferent
nerve firing rate (AFR) and when this firing rate rises above neural noise, the motion is detected. This signal-
to-noise-ratio value corresponds to an absolute threshold that is independent of the stimulus frequency. As a
consequence the frequency characteristics of the measured thresholds had to be related to the SCC dynamics.
This relation was described using Equation (1), where HSC C represents the SCC dynamics and |HS CC (j ω)|
the gain of these dynamics, ∆abs is the absolute threshold that remains constant with frequency and ∆meas
is the frequency-dependent measured threshold.
|HSC C (jω)|=∆abs
∆meas (jω)(1)
Since the absolute threshold corresponds to an AFR value, it could be expressed in Impulses Per Second
(IPS) units. However, the actual value of the calculated absolute threshold will depend on whether the
measured threshold value is considered in rad/s2or in deg/s2. This absolute threshold value does not
represent an actual firing rate, but rather a proportional measure of the excitation of the SCC that results
from the further processing of the AFR in the central nervous system (CNS). For this reason, in this paper
the absolute threshold values are expressed in Impulse Per Unit of Time (IPUT) and for clarity purposes,
they are always calculated from inputs in rad/s2.
Based on previous work by Fernandez and Goldberg,24–26 and others,27–30 Hosman and Van der Vaart
determined a structure for the SCC transfer function and using the frequency dependency of the measured
thresholds (Equation (1)), they fitted its parameters. The resulting model is defined by Equation (2), where
4 of 32
American Institute of Aeronautics and Astronautics

¨
θis the angular acceleration presented to the subjects. The gain KS CC could not be determined from the
threshold data.
HSC C (jω) = AF R (jω)
¨
θ(jω)=KSC C
1 + 0.1097jω
1 + 5.924jω (2)
Heerspink et al.5replicated the Hosman and Van der Vaart experiment in the SIMONA Research
Simulator (SRS) for six degrees-of-freedom. They found similar results to the Hosman and Van der Vaart
measurements of pitch and roll thresholds.
II.B. Benson et al.
Benson, Hutt, and Brown4measured sensory thresholds for angular motion. The stimulus consisted of a
single sinusoid acceleration signal. The duration of the stimulus was 3.3 s, corresponding to a frequency of
1.9 rad/s. Thresholds for roll and pitch motion were determined with the subjects laying on a litter in a
supine position (that is, lying with the face upward) or with right side down, respectively. The thresholds
for the angular motion were presented in velocity units and were 2.04 deg/s for roll and 2.07 deg/s for pitch.
Taking into account that a constant was added to the velocity signal to make sure it started at zero, the
velocity thresholds given correspond to 1.94 deg/s2and 1.97 deg/s2in acceleration units, respectively.
Benson et al. also investigated the influence of stimulus frequency on yaw thresholds. Again using a
single sinusoid acceleration signal, they varied the stimulus duration between 0.9 and 20 s, corresponding
to a frequency range between 0.3 and 7 rad/s. They found that the measured angular velocity thresholds
decreased with increasing frequency. A linear regression of the inverse of the threshold values showed a 5.9
dB/decade slope, from which Benson et al. concluded that the detection process was dependent on angular
velocity and angular acceleration.
The scatter of the data did not allow them to properly fit a regression line other than first order. However,
they mention that a third order regression does show a similar behavior to the known SCC dynamics. When
comparing their results for yaw with those of Hosman and Van der Vaart for pitch and roll, they observed
that the latter were one order of magnitude lower. Benson et al. provided as possible explanation for this
difference the fact that during Hosman and Van der Vaart’s experiment there might have been oculogyral
cues due to the illuminated cockpit. However, comparing the results of Hosman and Van der Vaart to those
of Heerspink et al., who used a dark cockpit and blindfolded their subjects, no large differences can be
found, leaving some doubt about whether the presence of oculogyral cues explains the lower roll and pitch
thresholds. One other factor that might explain the difference between Benson’s yaw thresholds and Hosman
and Van der Vaart’s pitch and roll thresholds is that during pitch and roll motion not only the SCC but
perhaps also the otoliths are stimulated, making it easier to detect the tilting motion.
5 of 32
American Institute of Aeronautics and Astronautics

II.C. Groen and Bles
Groen and Bles9investigated the influence of body tilt to enhance the perception of visually-induced linear
acceleration in a simulator, a method commonly known as tilt coordination. They tilted subjects in pitch
(with the center of rotation on the subject’s head) while presenting synchronous visual linear acceleration.
The pitch motion was sinusoidal (back and forth) with frequencies between 0.25 and 2 rad/s. The visual
stimulus was also sinusoidal and had amplitude values of 0.44, 0.88 and 1.76 m/s2. Pitch thresholds were
obtained, expressed in tilt angle, velocity and acceleration for all the combinations of frequencies and visual
amplitudes. They found that the pitch thresholds expressed in terms of velocity were almost constant with
frequency and increased slightly with increasing visual acceleration. The value of these thresholds was in
the neighborhood of 3 deg/s, which is slightly higher than the sensory thresholds measured by Benson et
al. and almost seven times higher than pitch sensory thresholds calculated from Heerspink et al. data (0.44
deg/s for a frequency range between 0.6 and 6 rad/s).
In a related study, Zaichik et al.6and Rodchenko et al.7also measured pitch thresholds, however, not
in the presence of visual cues, but while subjecting volunteers to simultaneous heave acceleration. Pitch
thresholds increased from around 0.7 deg/s without heave motion to 1.9 deg/s with a heave motion of 0.05
g (in a simulator6) and from around 1.5 deg/s without heave motion to 2.2 deg/s with a heave motion of 0.5
g (in flight7).
III. Frequency-Domain Description and Threshold Model
III.A. Frequency Description
Despite the methodological differences in the three studies presented in the previous section, one main
conclusion can be drawn: the frequency dependency of the angular motion thresholds follows the inverse
dynamics of the SCC. In the work of Hosman and Van der Vaart this relationship was described using
Equation (1). For a better visualization, the pitch threshold values measured by Hosman and Van der Vaart
and by Heerspink et al. are plotted together with the dynamics of the SCC in Figure 2. For this, it is first
necessary to calculate the value of the absolute threshold, which can be computed by re-writing Equation (1)
in the form of Equation (3):
∆abs = ∆meas (jω)|HSC C (jω)|(3)
The measured thresholds (∆meas ) used are those of Heerspink et al., as these have been collected in the
SRS, which is also used in the present paper, hence being suitable as reference data.
6 of 32
American Institute of Aeronautics and Astronautics

The SCC dynamics used in the calculations are presented in Equation (4) and correspond to the transfer
function from Equation (2), with the gain KSCC chosen such that the total transfer function has a gain of
1 at the frequency of 1 rad/s.
HSC C (jω) = 5.97 1 + 0.1097j ω
1 + 5.924jω (4)
Taking the measured thresholds (∆meas ) and multiplying those with the gain of the SCC transfer function
(HSC C ) for the corresponding frequency, in theory, a constant value should be found for ∆abs . In practice,
the variance in the threshold measurements accounts for some variation in the calculated value of ∆abs. For
practical purposes, ∆abs is calculated here as the average of all obtained values. Both sides of Equation (1)
can now be plotted as shown in Figure 2. In the figure the threshold data were obtained from References 1
and 5.
|HSCC |, dB
Frequency, rad/s
SCC dynamics
Hosman and Van der Vaart
Heerspink et al.
10−210−1100101102
-20
-15
-10
-5
0
5
10
15
20
Figure 2. Inverse of the pitch threshold values and the SCC transfer function gain.
In the work of Groen and Bles,9the threshold values, presented in terms of angular velocity, were fairly
constant at different frequencies. The constant velocity threshold is an expected result if we consider that
the SCC acts as an integrator between the frequencies of 0.2 and 9 rad/s. This means that although the SCC
are stimulated by angular acceleration, their output is proportional to velocity in this range of frequencies.
Hence, velocity thresholds that remain constant across different stimulus frequencies are consistent with
acceleration thresholds that follow the inverse of the SCC dynamics. To illustrate this, the threshold data
from this study, in acceleration units, are plotted in Figure 3 alongside the SCC transfer function gain. The
threshold values for the three visual linear acceleration conditions were scaled using as reference the absolute
threshold calculated from the data of Heerspink et al., as explained above.
The threshold data from Groen and Bles show a frequency variation that approximately follows the
dynamics of the SCC, but with a different gain. This gain difference can be attributed to the fact that in
7 of 32
American Institute of Aeronautics and Astronautics

|HSCC |, dB
Frequency, rad/s
SCC dynamics
Heerspink et al.
Groen and Bles:
0.44 m/s2
0.88 m/s2
1.76 m/s2
scaled SCC dynamics
10−210−1100101102
-60
-50
-40
-30
-20
-10
0
10
20
Figure 3. Inverse of the pitch threshold values and the SCC transfer function gain.
that experiment participants were subject not only to motion but also to visual information. The visual cue
raised the pitch thresholds without modifying the frequency description.
III.B. Threshold Model
Pilot model parameter identification has been successfully used previously to assess control behavior during
different experimental conditions.11–15, 19, 31–35 In these studies, participants are asked to perform a compen-
satory tracking task while their visual and motion cues settings are manipulated. The visual and inertial
motion cues and the corresponding pilot control activity are recorded during these sessions. The visual and
inertial signals become inputs to a mathematical model of the pilot and the model-generated pilot control
signal will be compared to the measured signals. By means of numerical optimization, the parameters of
the pilot model are varied until the model-generated pilot control signals approximate the measured data.
The final identified model parameters are objective measures of the adopted behavior during the different
experimental runs.
The pilot models used in these methods generally consist of a visual and an inertial path or channel and
neuromuscular dynamics. So, if an identification method is to be used, first the threshold mechanism has to
be modeled and included in the pilot model motion channel.
Based on the results from previous studies a threshold model was developed. This model has been
described elsewhere8, 12, 36–39 but it has never been used for identification purposes. The model consists of
a dead zone applied to the stimulus signal after the SCC transfer function, see Figure 4. In this approach
there is one threshold value for all frequencies that determines the width of the dead zone. Within the dead
zone the output of the threshold model, o∆, is zero. Outside, the output is the same as the input, i∆.
8 of 32
American Institute of Aeronautics and Astronautics

¨
θi∆o∆
i∆
o∆
HSC C (jω)∆abs
−∆abs
Figure 4. Angular threshold model, including the dynamics of the SCC.
Some assumptions were made regarding this model. First, it was assumed that the threshold process
occurs after the SCC, that is, it is a threshold applied to the signal-to-noise ratio of the AFR. One other
possibility would be that the threshold is caused by “stiction” of the cupula and hence, actually happens be-
fore the SCC dynamics. However, independently of the physical process causing the threshold, the frequency
dependency present in the measured thresholds can always be modeled by a frequency-invariant value after
the SCC. Second, it was assumed that the threshold is a single value, implying no distinction was made
between upper and lower threshold levels. Third, for stimuli above the threshold value the output of the
threshold model is assumed to equal the input.
IV. Experimental Method
Two experiments were performed, a passive and an active experiment. In the passive experiment, sensory
thresholds for pitch motion were measured during pitch motion stimulation in the dark, using a similar
method to Heerspink et al.5and Valente Pais et al.8These data provided a baseline for comparison with the
data from the active experiment. In the latter, pitch motion thresholds were estimated from control input
data collected while subjects performed a pitch attitude manual control task. The active pitch threshold
for each participant was estimated by fitting a pilot model, including the threshold model, to the measured
control task data.
IV.A. Apparatus
The experiments were conducted in the SIMONA Research Simulator (SRS) of the Delft University of
Technology, shown in Figure 5.
Figure 5. The SRS.
The SRS is a high-fidelity, full-motion simulator with a hydraulic six
DOF motion base which allows for a maximum displacement of ±24 deg
in pitch and a maximum pitch rate of ±35 deg/s. The motion system time
delay is 35 ms and the head-down visual display time delay is 25 ms.40
For a detailed description of the SRS motion system capabilities and the
computer architecture and software used, please refer to References 41–43.
The SRS is equipped with an inertial measurement unit (IMU) to measure
9 of 32
American Institute of Aeronautics and Astronautics

all specific forces and rotational accelerations. The IMU is a six DOF Inertial Science Inc. ISIS IMU (Rev.
C), consisting of three solid-state RRS75 rate gyros and three solid-state accelerometers.
IV.B. Subjects
Eight subjects, seven males and one female with an average age of 29.4 years and standard deviation of 7.4
years, participated in both experiments. None of the sub jects reported SCC defects. Only subject 1 had
participated in a passive threshold experiment before. Four subjects were highly experienced with the type
of active control task in the SRS (subjects 1, 2, 5 and 6 performed three or more similar experiments). The
remainder of the subjects had performed a similar task once, except for subject 4.
IV.C. Overall Procedure
The experimental trials were divided in two blocks of four hours. Each subject performed each block on a
different day. Only subject 1 performed both experimental blocks in one day, with a one hour lunch break
in between. The first experimental block consisted of the passive experiment and the first part of the active
experiment. The second experimental block started with the final part of the active control task, followed
once again by the passive experiment. The passive experiment was performed twice, before and after the
active experiment, so we could observe if there was any effect of habituation to the simulator, fatigue or even
time and day, on the measured sensory thresholds.
Between the passive and the active experiment there was a short break. During the active experiment
there was a 15 minute break every 30 minutes.
IV.D. Passive Experiment
IV.D.1. Experimental Design
The passive experiment was a one-way repeated measures design. Pitch thresholds for increasing amplitude
motion (upper thresholds as defined in Section II.A) were measured at two different frequencies: 1 and 4
rad/s. The small number of frequency levels chosen was a compromise between preventing the experiment
from becoming too long and tiring for the participants and being able to determine with some confidence the
value of the absolute threshold. Each subject repeated each experimental condition four times. To prevent
subjects from randomly indicating perceived motion, three runs without any motion were added. The total
of 11 experimental runs were randomized per subject.
10 of 32
American Institute of Aeronautics and Astronautics

IV.D.2. Motion Signals
The pitch motion signal consisted of an increasing amplitude sinusoid. The measurement runs began with a
random time interval between 5 and 25 seconds before the motion started. Thereafter, the motion gradually
increased for 94.25 seconds, equivalent to 15 periods for the 1 rad/s condition and 60 periods for the 4 rad/s
condition. An example of motion profiles for both frequencies is given in Figure 6. The runs without any
inertial motion had a duration of 69 seconds.
q, deg/s
t, s
Commanded
Measured
Pitch rate threshold, Ref. 5
0 10 20 30 40 50 60 70
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
(a) 1 rad/s signal.
q, deg/s
t, s
53 54 55 56 57 58 59
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
(b) Zoom in of a 1 rad/s signal.
q, deg/s
t, s
Commanded
Measured
Pitch rate threshold, Ref. 5
0 10 20 30 40 50 60 70
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
(c) 4 rad/s signal.
q, deg/s
t, s
59 60 61
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
(d) Zoom in of a 4 rad/s signal.
Figure 6. Example of commanded and measured pitch rate signals for the two tested frequencies.
In both experiments, compensation was used for the specific forces arising from the distance between
the simulator’s axis of rotation and the sub ject’s head, that is, the simulator rotated around the subject’s
head. However, in an attempt to keep the motion stimuli as simple as possible, no compensation was done
for the specific forces due to the orientation with respect to gravity. Comparing two studies performed in
the SRS simulator, the work of Heerspink et al.,5where no compensation was used, and the work of Valente
Pais et al.,8where compensation was used for frequencies higher than 0.6 rad/s, it can be seen that without
11 of 32
American Institute of Aeronautics and Astronautics

compensation thresholds were lower but that the frequency description between 0.6 and 10 rad/s still matched
the dynamics of the SCC. This being the case, the threshold model is still valid and since no compensation
was used for both the passive and the active task, it was not expected that compensation related issues
would affect the comparison between the thresholds obtained in the two parts of the experiment.
Figure 7 illustrates the longitudinal specific forces at the subject’s head for the 1 rad/s and 4 rad/s signals.
The longitudinal specific force due to tilt with respect to gravity was calculated using the commanded pitch
angle. This signal is indicated in the figure as the “Calculated at the pilot’s head” signal. The “Measured
at the pilot’s head” signal was obtained by transforming the specific forces and pitch rate measured by the
IMU, to the pilot’s head position. The error signals displayed in Figure 7b and Figure 7d correspond to the
difference between the calculated and the measured signals. Since the simulator motion was designed such
that the pilot’s head was the center of rotation, these signals should be close to zero.
fx, m/s2
t, s
Calculated at pilot’s head
Measured at pilot’s head
Surge threshold, Ref. 5
0 10 20 30 40 50
-0.10
-0.05
0.00
0.05
0.10
(a) 1 rad/s signal.
error fx, m/s2
t, s
Measured at pilot’s head
Surge threshold, Ref. 5
0 10 20 30 40 50
-0.10
-0.05
0.00
0.05
0.10
(b) Error for a 1 rad/s signal.
fx, m/s2
t, s
Calculated at pilot’s head
Measured at pilot’s head
Surge threshold, Ref. 5
0 10 20 30 40 50 60 70 80 90
-0.10
-0.05
0.00
0.05
0.10
(c) 4 rad/s signal.
error fx, m/s2
t, s
Measured at pilot’s head
Surge threshold, Ref. 5
0 10 20 30 40 50 60 70 80 90
-0.10
-0.05
0.00
0.05
0.10
(d) Error for a 4 rad/s signal.
Figure 7. Longitudinal specific forces at the pilot’s head due to tilt with respect to gravity.
IV.D.3. Procedure and Subjects’ Instructions
Subjects were seated in the right-hand seat in the simulator cabin and were instructed to close their eyes.
They were wearing a headset with active noise cancellation on which aircraft engine noise was played.
Participants were asked to sit upright and avoid making any head movements. A handheld button was used
to record the subjects’ answers. The cabin lights were off and there were no instruments or outside visual
12 of 32
American Institute of Aeronautics and Astronautics

displays. Communication with the control room was possible at all times and participants were informed of
the start and end of each run.
Participants were told they were to experience several experimental runs which consisted of an increasing
amplitude pitch rotational motion. They were also told there were runs without any motion and that they
should not let themselves be tricked by this “no motion condition”. This was thought to keep subjects
motivated and help the experiment supervisor to detect subjects that did not understand or were unable to
perform the perception task. Subjects were instructed to press the button as soon as they perceived any
motion, even if they could not yet identify the direction of the motion. After pressing the button, they were
asked to indicate the frequency of the stimulus by saying ‘backward’ and ‘forward’ synchronized with the
perceived motion. This information was used by the experiment supervisor to monitor subjects’ performance.
Too many wrong answers would indicate a subject that would press the button too early, before he or she
would really be detecting the motion. At the end of each run, subjects received feedback about whether or
not they perceived the frequency of the motion correctly.
IV.D.4. Data Analysis
The sensory threshold value for each run was determined from the button press. The maximum pitch
acceleration before the button press was taken as the threshold value. From the measured sensory thresholds
the absolute thresholds were calculated using Equation (3).
IV.E. Active Experiment
IV.E.1. Active Threshold Determination
The control task used for the determination of the active threshold was a compensatory tracking task. In
this type of task the pilot is required to continuously minimize the tracking error, e, which is the deviation
from a desired aircraft state. The tracking error is presented on a visual display, as depicted in Figure 8.
e
Figure 8. Compen-
satory tracking task
display.
This error can be induced by disturbing the controlled system’s output, by pro-
viding a target signal that the pilot has to follow, or both. Frequently, both options
are done simultaneously, since this allows for discrimination between the visual and
the vestibular channels during the identification procedure.12 Also in this experi-
ment the compensatory tracking task consisted of a disturbance-rejection task and
a target-following task. Previous experiments12, 13, 31, 32 have shown that pilot con-
trol behavior is influenced by the presence of inertial motion during compensatory
tracking tasks. Since the determination of motion perception thresholds requires a
13 of 32
American Institute of Aeronautics and Astronautics

control task where inertial motion affects pilot control, the compensatory tracking task was thought to be
suitable.
The dynamics of the system to be controlled were defined by a double integrator, given in Equation (5),
with a gain Kc= 4. This type of controlled element was selected because the effect of inertial motion on
control behavior is more evident for this type of system than for systems with more stable dynamics.31
Hc(jω) = Kc
(jω)2(5)
IV.E.2. Pilot Model
Pilot control behavior during compensatory tracking tasks can be modeled with a multi-channel pilot model
as depicted in Figure 9. In the model, pilot responses to visual and vestibular motion stimuli are modeled
separately. Two forcing functions, a target-following signal, ft, and a disturbance-rejection signal, fd, are
used to make identification of both channels possible.12 Furthermore, a remnant signal, n, is added to the
control signal, u, to model the unknown nonlinear portion of pilot behavior. For this study, the model was
extended with the model of the absolute threshold described in Section III.B. Figure 9 shows the pilot model
used including the threshold model, which is depicted with the symbol ∆abs . Please note that throughout
this paper all frequency describing functions use the variable jω, instead of s= (σ+jω), to indicate that
the models are to be used within a control loop with time-stationary properties and that the pilot behavior
describing functions are valid for non transient inputs only.
Hnm(jω )
e−jωτv
Kv(1 + jωTvl)
(jω)2e−j ωτm
HSCC (j ω)Km
∆abs
Hc(jω)
Hpx
Hpe
e
θ
ftuv
um
nfd
u θ
i∆o∆
+
−
+
−
+++δe
Kθ
Figure 9. Detail of visual and vestibular motion channels.
Humans adapt their control behavior such that the open-loop dynamics around the crossover frequency
can be described by a single integrator and a time delay.44 When the controlled element is a double integrator,
humans have to generate lead,11, 44 which is modeled in the visual motion channel, Hpe.
The vestibular motion channel, Hpx, consists of the vestibular motion gain, Km, the SCC dynamics,
HSC C , and a time delay constant, τm. The SCC dynamics used are those described in Equation (4).
Finally, the neuromuscular dynamics present in both channels are given by Equation (6), where ωnm is
the natural frequency and ζnm the damping ratio.
14 of 32
American Institute of Aeronautics and Astronautics

Hnm (jω) = ω2
nm
ω2
nm + 2ζnmωnm jω + (jω)2(6)
IV.E.3. Forcing Functions
Two different control tasks were used, one in which the target-following task (ft) was dominant and one in
which the disturbance-rejection task (fd) was dominant. The reason for using these two control tasks was
the different effect of motion reported for these different types of control tasks.17,18, 20, 33 This could possibly
result in a different effect of the threshold on pilot control behavior, and hence, on the identifiability of the
threshold for both tasks.
For a condition with a dominant target-following task, the disturbance signal amplitude was scaled with a
factor of 0.5 and vice-versa. Furthermore, the phases of both forcing functions were chosen such as to prevent
the control task from starting with a too high acceleration. The resulting forcing function properties are
listed in Table 1, where niis the integer multiple of the base frequency, the base frequency is ωm= 2π/Tm
and Tmis the measurement time (81.92 s), ωiis the actual frequency, Aiis the amplitude and φiis the
phase, of each sine. The amplitudes for both forcing functions in the table do not include the scale factor
used to make either of the tasks dominant.
Table 1. Forcing function properties.
fdft
niωiAiφiniωiAiφi
−rad/s rad rad −rad/s rad rad
5 0.3835 0.0031 -1.2726 6 0.4602 0.0876 6.0641
8 0.6136 0.0071 -3.0424 9 0.6903 0.0764 3.3242
11 0.8437 0.0115 0.1229 13 0.9971 0.0613 6.2580
17 1.3039 0.0193 1.5607 19 1.4573 0.0430 5.7018
28 2.1476 0.0284 -1.0470 29 2.2243 0.0250 1.6352
46 3.5282 0.0363 2.3832 47 3.6049 0.0120 4.7074
59 4.5252 0.0407 -0.4468 61 4.6786 0.0081 2.3684
82 6.2893 0.0488 1.0625 83 6.3660 0.0052 1.3033
106 8.1301 0.0590 -1.5272 107 8.2068 0.0038 0.7234
137 10.5078 0.0755 -1.0093 139 10.6612 0.0029 0.1830
178 13.6524 0.1033 -2.8089 179 13.7291 0.0024 0.5258
211 16.1835 0.1308 -2.3834 213 16.3369 0.0021 2.9077
IV.E.4. Experimental Design
The active experiment had a two-way repeated measures design. The two independent variables were the
type of control task and the motion gain.
The control task type was determined by the forcing functions used, which resulted in a predominantly
target-following task or in a predominantly disturbance-rejection task.
The motion gain determined the scaling of the inertial motion provided to the subjects. Previous studies
have shown that the amount of inertial motion available influences the subjects’ control strategies.12, 13, 31, 32
15 of 32
American Institute of Aeronautics and Astronautics

Even for tasks where the amplitude of motion is well above the sensory threshold, the use that subjects
make of the inertial cues seems to depend on the motion gain. This pilot adaptation might have an effect
on the identifiability of the threshold value. Moreover, for too low a motion gain the majority of the time
participants will be subject to sub-threshold stimulation, whereas for too high a motion gain there will
be little sub-threshold stimulation. For identification purposes it is desirable to find a balance between
sub- and supra-threshold stimulation. The amplitudes and frequencies of the motion provided to subjects
is dependent, to a certain extent, from subjects’ control behavior. From an experiment design point of
view, the motion stimulation may be influenced by the design of the forcing functions and the choice of the
motion gains applied to the inertial motion (but not to the visual). From experience with other experiments
using tracking tasks, for the designed forcing functions, a motion gain of one results in reliable identification
results. Decreasing the motion gain from one to zero in a certain number of steps should result in a few
conditions where the motion has the right balance of sub- and supra-threshold stimulation to allow for a
reliable identification of the threshold parameter. For this reason, four motion gains were chosen: 0.25, 0.5,
0.75 and 1. Only the motion gains were changed and there was no other form of washout or filtering applied
to the inertial cues.
Together with the two control tasks, the four motion gain conditions resulted in a total of 8 experimental
conditions. Typically 3 to 5 runs per condition, presented in random order, were sufficient for training. In
the measurement phase, each condition was performed five times. Depending on the number of training
runs, a total of 64 to 80 runs per subject was performed, of which 40 were measurement runs.
IV.E.5. Procedure and Subject’s Instructions
Similar to the passive experiment, subjects were seated in the right-hand pilot seat of the SRS, but now
used the electronic side stick, which was located on their right, for the pitch tracking task. Participants
were asked to sit upright as much as possible and avoid making head movements. They were wearing a
headset with active noise cancellation and the same engine sound was used to cancel out simulator noise.
Subjects were supplied with a compensatory flight display, as shown in Figure 8, presented on a 15 inch LCD
display mounted in front of them. The remainder of the LCD displays, the cabin light and the outside visual
projectors were all switched off. Subjects were encouraged to pursue the best control performance possible
and to maintain this level throughout the runs. Participants were informed of the start and end of each
experimental run. Each experimental run lasted 100 seconds of which the first 18.08 seconds were considered
as run-in time and the last 81.92 seconds were used for identification. After each run, subjects received
feedback about their performance in terms of the error signal Root Mean Square (RMS). Consistency of
subjects’ control activity was monitored with the RMS of the control signal.
16 of 32
American Institute of Aeronautics and Astronautics

Whereas the four motion gain conditions were presented to the subjects in random order throughout
both training and measurement runs, the two control task conditions were presented in two separate blocks.
Each block started with a training phase, in which subjects could get acquainted with the experiment and
reach consistent performance and control activity. Half of the participants (subjects 1, 4, 5 and 7) started
with the conditions of the disturbance-rejection task and the other half (subjects 2, 3, 6 and 8) with the
target-following task.
IV.E.6. Data Analysis
For each experimental run, the controlled element pitch attitude, θ, its first and second derivatives, the
tracking error signal, e, and the control signal, u, were recorded. The recorded time histories were used to
estimate the parameters of the multi-channel pilot model defined in Section IV.E.2.
From the multi-channel pilot model, eight parameters were identified: the visual gain Kv, the visual
lead time-constant Tvl, the visual time delay τv, the vestibular gain Km, the vestibular time delay τm, the
absolute threshold ∆abs, the neuromuscular damping factor ζnm and the neuromuscular natural frequency
ωnm.
The nonlinearity introduced in the model by the threshold element excluded the use of linear identification
methods such as Fourier coefficients12 or linear time-invariant models.14 Instead, Maximum Likelihood
Estimation (MLE) was used for identification of the model parameters. The MLE method combined a
genetic algorithm optimization with an unconstrained steepest-descent optimization,15 to reduce the chance
of converging to a local minimum. The algorithm and its performance on the identification of the parameters
of the proposed pilot model with threshold has been described in detail in Pool et al.22
IV.F. Hypotheses
It is expected that the threshold values measured with the passive experiment correspond, per subject, with
the ones estimated during the active experiments; that is, the subject with the relatively lower threshold
during the passive experiment also shows the lower values during the active experiment.
It is hypothesized that the average threshold value during the active task will be different from the value
during the passive task. However, no hypotheses are formulated to whether the active threshold will be
higher or lower than the passive threshold. On the one hand, during the active experiment subjects perform
a control task that increases workload which could lead to higher threshold values.1,10 On the other hand,
the subjects’ control input actively influences the provided motion. This gives subjects an extra cue on
what type of pitch motion to expect which may lower the threshold value. It is also important to note that
whereas in the passive experiment subjects have their eyes closed, in the active experiment they are looking
17 of 32
American Institute of Aeronautics and Astronautics

at the visual display. Although the visual display is the only source of light in the cabin, it has been shown
that even in dimly lit environments, due to the occurrence of oculogyral illusion, the inertial thresholds are
lower than in complete darkness.37
V. Results
V.A. Passive Experiment
Among the conditions tested, there were also runs without motion. Moreover, subjects were asked to indicate
the direction of motion after they had pressed the button. Such tactics could help exclude subjects that
did not understand or were unable to perform the perception task. All subjects were capable of reporting
the motion frequency, although occasionally they showed a 180 deg phase reversal, that is, they would say
“backward” when in fact the simulator was tilting forward. In these cases, subjects were still considered
to have performed satisfactorily. Button presses during “no motion” conditions were seldom and all eight
subjects were considered to have understood the task.
The measured sensory thresholds and the corresponding absolute thresholds are shown in Figure 10,
together with the results from Heerspink et al.5
Measured thresholds, deg/s2
Before
After
Total
Heerspink et al.
Absolute thresholds, -
1 rad/s 4 rad/s Absolute 0
0.002
0.004
0.006
0.008
0.01
0
0.5
1
1.5
2
2.5
(a) Means and standard deviations of measured and abso-
lute threshold values.
|HSCC |, dB
Frequency, rad/s
SCC dynamics
Passive experiment
Heerpink et al.
10−210−1100101102
-20
-15
-10
-5
0
5
10
15
20
(b) SCC model with inverse measured sensory threshold
means and standard deviations.
Figure 10. Mean pitch sensory thresholds before and after the active experiment.
In Figure 10a it can be observed that for both frequencies tested, the thresholds measured before and
after the active experiment were very similar. In comparison to the results from Heerspink et al. the obtained
thresholds were 35% and 40% lower, for the 1 and 4 rad/s conditions respectively, but the standard deviations
are comparable. Two independent t-test analyses showed that for both the 1 rad/s (t(10) = 16.05, p < 0.01)
and the 4 rad/s (t(10) = 6.05, p < 0.01) conditions the mean thresholds found were significantly different
18 of 32
American Institute of Aeronautics and Astronautics

from the thresholds measured by Heerspink et al.. The absolute thresholds, calculated from the measured
thresholds at both frequencies, were 30% lower and showed a much smaller standard deviation than the
absolute threshold calculated from the Heerspink et al. data.
Figure 10b presents the results of both experiments together with the SCC dynamics. Here, the scaled
inverse of the measured thresholds are shown, as described by Equation (1). The SCC dynamics used are
described in Equation (4) and the absolute threshold value was calculated from the data from Heerspink et
al.. The measured thresholds follow the dynamics of the SCC and again it is visible that the values from
Heerspink et al. were higher.
V.B. Active Experiment
V.B.1. Pitch Acceleration Stimuli
The pitch acceleration cues available to the subjects during the active task were quite different from the
passive task. The motion stimuli frequency and amplitude depended on the forcing function and on the
tracking performance of each subject. The amplitude also depended on the motion gain condition. It is
a well known fact that during tracking tasks subjects will vary their control behavior depending on the
available motion cues.12, 13, 31, 32 For this reason, it is interesting to observe what type of motion cues were
available to subjects during the active task. As an example of typical motion cues during the active part of
the experiment, Figure 11 shows the pitch acceleration time signals, at the subject head, during 16 seconds
of one target-following run and one disturbance-rejection run, for each of the motion gains.
V.B.2. Linear Pilot Model Parameters
Using the identification method described in detail in Pool et al.,22 the parameters of the pilot model were
estimated. The estimated parameters of the linear part of the pilot model vestibular channel over the eight
subjects are shown in Figure 12. Both the vestibular gain, Km, and the vestibular time delay, τm, become
smaller with increasing motion gain. Compared to previous studies,18–20,35 the estimated vestibular time
delay was disproportionately high for the conditions with motion gains of 0.25 and 0.5. This suggests that
for these conditions the vestibular time delay parameter could not be estimated accurately. The relatively
higher vestibular gains found for motion gains of 0.25 and the larger 95% confidence interval (CI) for the
condition with motion gain of 0.5 and the disturbance-rejection task was most likely also a result of the
unreliable vestibular time delay identification.
The estimated parameters of the pilot model visual channel are depicted in Figure 13. The mean and
the 95% CI values were corrected for between-subject variability. The visual gain, Kv, and the visual time
delay, τv, increased with increasing motion gain. The higher 95% CI of the visual time delay for the lowest
19 of 32
American Institute of Aeronautics and Astronautics

Pitch acceleration, deg/s2
Time, s
disturbance-rejection task
target-following task
16 18 20 22 24 26 28 30 32
-30
-20
-10
0
10
20
30
(a) Gain = 0.25.
Pitch acceleration, deg/s2
Time, s
disturbance-rejection task
target-following task
16 18 20 22 24 26 28 30 32
-30
-20
-10
0
10
20
30
(b) Gain = 0.5.
Pitch acceleration, deg/s2
Time, s
disturbance-rejection task
target-following task
16 18 20 22 24 26 28 30 32
-30
-20
-10
0
10
20
30
(c) Gain = 0.75.
Pitch acceleration, deg/s2
Time, s
disturbance-rejection task
target-following task
16 18 20 22 24 26 28 30 32
-30
-20
-10
0
10
20
30
(d) Gain = 1.
Figure 11. Example of pitch acceleration time signals during the two tasks and the four motion gains.
motion gain conditions might be explained by the low estimation accuracy also found for the vestibular
motion parameters in those conditions.
The visual lead time-constant, Tvl, decreased with increasing motion gain. This was an expected result,
since adding motion introduces lead in the control loop, subjects have to generate less visual lead. For all
motion gain conditions the visual lead parameters are lower for the disturbance-rejection task than for the
target-following task, which reflects the different use of motion in the two different task types.
An ANOVA was performed to test the effect of the task type and the motion gain on the identified pilot
model parameters. These results are presented in Table 2.
As stated before, the motion gain and hence the available motion cues, have an effect on subjects behavior
and that can be seen in the estimated pilot model parameters. At higher motion gains, subjects seem to
make a better use of the information conveyed by the motion.
V.B.3. Threshold Parameter
The threshold parameter could not be reliably identified for all motion gain conditions. As seen in the
previous section, subjects make more use of the motion when the motion gain is high. Conversely, the lower
20 of 32
American Institute of Aeronautics and Astronautics

Km, rad/IPUT
Motion gain, -
task
0.25 0.5 0.75 1
-0.5
0
0.5
1
1.5
2
2.5
3
(a) Vestibular motion gain.
τm, s
Motion gain, -
disturbance-rejection task
target-following task
0.25 0.5 0.75 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
(b) Vestibular time delay.
Figure 12. Vestibular motion parameter means and 95% CI, corrected for between-subject variability.
Kv, -
Motion gain, -
0.25 0.5 0.75 1
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
(a) Visual gain.
Tvl, s
Motion gain, -
0.25 0.5 0.75 1
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
(b) Visual lead time-constant.
τv, s
Motion gain, -
disturbance-rejection task
target-following task
0.25 0.5 0.75 1
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3
0.32
0.34
(c) Visual time delay.
Figure 13. Visual motion parameters means and 95% CI, corrected for between-subject variability.
the motion gain, the less subjects make use of that information. Possibly, in the conditions with lower motion
gains, the amplitude of the motion cues was too low to provide the subjects with useful information, forcing
them to adopt a control strategy based solely on visual cues.
When considering the pilot model, such an effect can be seen as an elimination of the vestibular channel.
In mathematical terms, this can be accomplished by decreasing the vestibular gain to zero or by increasing
the value of the absolute threshold parameter up to the point where no motion signal passes through.
Figure 14 and Figure 15 illustrate the second option. The simulated pitch acceleration signals after the SCC
dynamics is plotted for both the the disturbance-rejection and the target-following tasks, for each repetition
of one experimental condition for one of the subjects. The passive absolute threshold and the identified
threshold parameter for that subject and that condition are also plotted.
As can be seen, for the disturbance-rejection task, at motion gains of 0.25 and 0.5 the identified threshold
value is such that the motion cues do not pass through. For the target-following task this effect is seen only
21 of 32
American Institute of Aeronautics and Astronautics

Table 2. Two-way repeated measures ANOVA results for the identified pilot model parameters, where ** is
highly significant (p<0.01), * is marginally significant (p<0.05), and −is not significant (p>0.05).
Dependent
measures
Independent variables
task Kθtask ×Kθ
df F Sig. df F Sig. df F Sig.
Km1,7 2.48 −1.32,9.25a6.62 ** 1.23,8.61a0.86 −
τm1,7 2.64 −1.31,9.22a7.35 ** 1.42,9.92a2.02 −
Kv1,7 12.70 ** 3,21 21.73 ** 3,21 1.81 −
Tvl 1,7 3.92 * 1.39,9.73a16.50 ** 1.31,9.19a1.90 −
τv1,7 23.99 ** 1.20,8.42a12.93 ** 3,21 4.01 **
∆abs 1,7 23.57 ** 3,21 28.43 ** 3,21 17.24 **
aGreenhouse-Geisser sphericity correction applied.
for the 0.25 motion gain. In both tasks, as the motion gain decreases, the identified threshold increases,
decreasing the amount of motion that passes through the vestibular channel. If there is not enough informa-
tion passing through the channel the identification of its parameters is not possible. A detailed description
of the problems encountered during the identification of the threshold parameter can be found in Reference
22. There, it also explained what the impact of a poor identification of the threshold parameter has on the
identification of the linear parameters of the pilot model.
For each condition, subjects performed five repetitions of the control task. The genetic algorithm, used as
part of the identification procedure, was applied 20 times to the time histories of each of the five repetitions.
As a measure of the consistency of the identification of the threshold parameter, one can look at the threshold
value found in each of these 20 runs. If these values are very different, spread over the search interval, it means
the genetic algorithm has converged to a different solution every time, suggesting an unreliable estimation of
the threshold parameter. To have an indication of the reliability of the identification, the standard deviation
of the estimated threshold parameter across the 20 runs was calculated. The standard deviation values were
then averaged across the five repetitions of one condition, for each subject. Figure 16 shows the average
standard deviation for each condition for all subjects.
As can be observed from Figure 16, the average standard deviations are the lowest for the conditions
with motion gain of 1 and slightly lower for the disturbance-rejection task than for the target-following task.
This result is in agreement with simulation results from Pool et al.22 They applied the identification method
to simulated data generated with different levels of remnant, see Table 3. For remnant levels equivalent to
the ones found in the present experimental data, the accuracy of the identified threshold model was found
to be at an acceptable level only for the motion gain 1 conditions.
A summary of the estimated threshold values obtained over all subjects is given in Table 4. Based on
the previously mentioned averaged standard deviations, and supported by the simulation results found in
Reference 22, only the results from the conditions with a motion gain of 1 were used to calculate the overall
22 of 32
American Institute of Aeronautics and Astronautics

Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(a) Gain = 0.25.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(b) Gain = 0.5.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(c) Gain = 0.75.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(d) Gain = 1.
Figure 14. Simulated SCC output for five repetitions of each condition during the disturbance-rejection task.
Table 3. Success rate in estimating threshold parameter from simulated control task data within 50% of the
true simulated value of ∆abs (based on simulations for 100 different remnant noise realizations).22
Task Motion gain, Kθ
0.25 0.5 0.75 1
disturbance-rejection task 0% 0% 24% 55%
target-following task 0% 0% 32% 65%
threshold value for the active experiment. For reference purposes, the identified threshold values over all
motion gains, for the disturbance-rejection and target-following tasks are also presented.
V.C. Comparison of Passive and Active Experiments Results
Figure 17a shows the mean estimated threshold values for all conditions of the active experiment. The
identified thresholds increase with a decrease in the motion gain. The higher values found at the low motion
conditions are related to the phenomenon seen in Figure 14 and Figure 15, where the threshold value appears
at the same level as the maximum amplitude of motion for that condition, thereby disabling the vestibular
channel of the pilot model.
23 of 32
American Institute of Aeronautics and Astronautics

Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(a) Gain = 0.25.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(b) Gain = 0.5.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(c) Gain = 0.75.
Simulated SCC output, IPUT
Time, s
passive threshold
identified threshold
16 20 24 28 32
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
(d) Gain = 1.
Figure 15. Simulated SCC output for five repetitions of each condition during the target-following task.
Figure 17b shows the average absolute threshold value calculated from the active and passive experiments
and from data from Heerspink et al.,5Hosman and Van der Vaart1and Valente Pais et al.8All these studies
measured passive pitch acceleration thresholds in a flight simulator. Heerspink et al. and Hosman and Van
der Vaart did not provide compensation for the specific forces arising from the tilt angle and in the study of
Hosman and van der Vaart the subject’s head was not exactly in the center of rotation. Valente Pais et al.
provided compensation for the tilt angle for frequencies above 0.6 rad/s and the subject’s head was in the
center of rotation. The threshold measured using compensation was 79% higher than the passive threshold
measured in the present study.
The absolute threshold estimated from the active experiment was 60% higher than the one measured
during the passive experiment (F(1,7) = 6.53, p < 0.05). However, when comparing both threshold values
with other studies it can be seen that the active threshold value is closer to absolute threshold values from
other passive studies.
Figure 18 shows the relationship between the measured passive thresholds and the identified active
thresholds (only at motion gain 1) for each subject, for both the disturbance-rejection and the target-
following tasks. The linear regression fit of the data, with corresponding R2value, is also shown. For the
24 of 32
American Institute of Aeronautics and Astronautics

Motion gain, -
Mean σ, IPUT
7
0.25 0.5 0.75 1
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
(a) Disturbance-rejection task.
Motion gain, -
Mean σ, IPUT
sub 1
sub 2
sub 3
sub 4
sub 5
sub 6
sub 7
sub 8
0.25 0.5 0.75 1
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
(b) Target-following task.
Figure 16. Mean standard deviation of the identified threshold parameter across the five repetitions.
Table 4. Estimated absolute pitch motion perception thresholds in IPUT units.
Disturbance-rejection Target-following Total
Motion gain Mean Std. dev. Mean Std. dev. Mean Std. dev.
0.25 0.0326*0.0064*0.0215*0.0088*0.0270*0.0095*
0.50 0.0481*0.0181*0.0201*0.0162*0.0341*0.0221*
0.75 0.0125*0.0103*0.0112*0.0093*0.0119*0.0098*
1.00 0.0089 0.0050 0.0077 0.0035 0.0083 0.0044
Active experiment 0.0255 0.0194 0.0151 0.0120 0.0083** 0.0044**
Passive experiment – – – – 0.0052 0.0021
Heerspink et al. – – – – 0.0072 0.0010
*The identified threshold values are less reliable.
** Calculated only from the conditions with a motion gain of 1.
disturbance-rejection task there was a positive relationship between the passive and the active thresholds per
subject, r= 0.681, p(one-tailed) <0.05. This means that subjects with relatively high passive thresholds
also presented above average thresholds during the active task. However, for the target-following task, the
passive and active thresholds were not significantly correlated, r= 0.1102, p(one-tailed) >0.05.
This observed difference in the correlations is difficult to explain. Tentatively, one can say that during the
target-following task most of the inertial motion presented to subjects is a consequence of their own control
actions. If some subjects are better than others in interpreting this motion and relating it with their own
inputs, this can cause an artificial lowering of their threshold, leading to a weak correlation between their
passive and their active threshold. In subjects where the largest difference is found between the thresholds
from the two tasks, subjects 2 and 5, one can see that indeed the threshold identified from the target-following
task is much lower than the threshold from the disturbance-rejection task.
25 of 32
American Institute of Aeronautics and Astronautics

Motion gain, -
Absolute threshold, IPUT
wing
disturbance-rejection task
target-following task
0.25 0.5 0.75 1
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
(a) Active experiment.
Experiment
Absolute threshold, IPUT
Heerspink et al.
Hosman and Van der Vaart
Valente Pais et al.
Passive Active
0
0.0025
0.005
0.0075
0.01
0.0125
(b) Passive and active experiments.
Figure 17. Absolute threshold values for the passive and active experiment. The bars indicate the 95% CI.
VI. Discussion
The threshold values obtained from the passive experiment described in this paper were lower than the
ones found in literature. At first one might think that the presence of specific forces due to the tilt with
respect to gravity were used by the subjects for motion detection, which could artificially lower the pitch
threshold. However, observing the values presented by Heerspink et al.5and Hosman and van der Vaart,1
which also did not compensate for the specific forces arising from the tilt angle, the thresholds obtained in
the present study are still lower.
If the relatively low threshold value cannot be attributed only to the presence of specific forces, perhaps
the specific experimental setup used may have had some influence in the obtained threshold value. In the
present experiment only pitch thresholds were measured and subjects were asked to press the response button
as soon as they felt motion. After they pressed the button they were required to identify the direction of
motion. In the experiment of Heerspink et al. more degrees of freedom were tested and the participants had
not only to indicate when they perceived motion but they were also required to identify in which DOF they
were moving and the frequency of motion. Although it is not clearly indicated in the report of Heerspink
et al. what happened if subjects wrongly identified the DOF, they do report that for pitch, the percentage
of correctly identified motion conditions was 64%. They also report that subjects always identified the
frequency of the motion stimulus correctly. In the present study, subjects were expecting pitch motion and
were not required to detect in which DOF they were moving. Perhaps the fact that subjects expected pitch
and could focus solely on this DOF lead to the lower thresholds found.
In the active experiment, the identified pilot model parameters helped understand how the different
motion gains and task type affected subjects’ control strategies. As mentioned before, it was known that
26 of 32
American Institute of Aeronautics and Astronautics

1
R2=0.4630
2
R2=0.4630
3
R2=0.4630
4
R2=0.4630
5
R2=0.4630
6
R2=0.4630
7
R2=0.4630
8
R2=0.4630
Absolute active threshold, IPUT
Absolute passive threshold, IPUT
00.002 0.004 0.006 0.008 0.01 0.012
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
0.022
(a) Disturbance-rejection task.
1
R2=0.0121
2
R2=0.0121
3
R2=0.0121
4
R2=0.0121
5R2=0.0121
6
R2=0.0121
7
R2=0.0121
8
R2=0.0121
Absolute active threshold, IPUT
Absolute passive threshold, IPUT
0 0.002 0.004 0.006 0.008 0.01 0.012
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
0.022
(b) Target-following task.
Figure 18. Relationship between passive and active thresholds for each subject.
varying the motion gain would cause the subjects to adapt their control strategy. The effect of decreasing
the motion gain could especially be seen in the visual lead time-constant. The lead time-constant increased
as the motion gain decreased to compensate for the decrease in lead information conveyed by motion. From
this, one may conclude that for the low motion gain conditions subjects rely more on information gathered
visually than through motion, whereas for the high motion gain conditions, the opposite happens. This
gradual transfer of weight between the visual and motion information contribution to the control task has
some consequences in the reliability of the identification procedure for the low motion gain conditions.
At the low motion gain conditions the values of the vestibular time delay were uncommonly large indi-
cating possible problems with the identification algorithm. By observing the threshold parameter identified
at these same conditions one can see that, in fact, the threshold value is above or close to the maximum
amplitude of the motion. This means that there is no information passing through the motion channel
of the pilot model and hence, no reliable identification of its parameters can be done. This is not just a
mathematical problem. As supported by the increasing visual lead-time constant with decreasing motion
gain, at the lowest motion conditions subjects may in fact be making insufficient use of the motion. This
affects the identification of the motion channel parameters, including the threshold parameter.
An intuitive way of analyzing the consistency of the identified threshold parameters is to look at the
spread of threshold values obtained from the identification algorithm. Also here one can see that as the
motion gain decreases, the standard deviation of the threshold values found increases, indicating a less
reliable estimation. The identification problems encountered at low motion conditions suggest that subjects
make less use of motion at low motion gain conditions. As the motion gain increases it becomes easier to
27 of 32
American Institute of Aeronautics and Astronautics

use the motion information to perform the tracking task and this is reflected in more reliable identification
of the pilot model parameters.
Although the identification was successful for motion gains of 1, 0.75 and occasionally also 0.5, the
identified thresholds at the motion gain of 1 were considered to be the most reliable ones and hence, were
chosen for comparison with the passive threshold.
The estimated active threshold was significantly higher than the measured passive threshold. The increase
from passive to active threshold was 60%. Samji and Reid10 report an increase of 51%, from a passive pitch
threshold to a threshold measured during a control task, with and without motion feedback related to the
control task. However, this was for detection of a yaw onset cue. For the detection of a yaw motion-base
return cue, the increase was 266% for a no-motion control task, and 160% for a control task with motion
feedback. An important difference between this study and the present one is that the threshold measurements
were done in yaw while the control task was performed mainly in roll. Hosman and van der Vaart1also
report increases of 26% and 70% between passive pitch thresholds and thresholds measured during a roll
control task and a combined roll control task and auditory binary choice task, respectively.
The increase from passive to active threshold varied between 26% and 266% for different types of motion
stimuli and control tasks. It seems that small changes in the experimental setup might have a big influence on
the measured threshold values. For example, the difference observed between passive and active thresholds
was smaller than the difference observed between the passive threshold and the threshold from Valente Pais
et al.,8which used a very similar experimental setup but applied compensation for the specific forces arising
from the tilt with respect to gravity. Nevertheless, the increased in the threshold value for the active task is
in accordance with what is found in literature for thresholds measured during out-of-the-loop control tasks.
VII. Recommendations
It would be interesting to further investigate the use of motion for the control task at the low motion
conditions. Intuitively one can say that if the motion usage decreases with the motion gain, at a certain low
enough motion gain, while the motion amplitude is still above the sensory threshold, the pilot might rely
solely on visual information to perform the control task. If this is the case, a few questions arise, such as: At
which level of motion does the pilot stops using the motion? Is that level related to the motion gain, to the
maximum peak amplitude, to the average motion amplitude or to the percentage of time that the motion
signal is above and below the sensory threshold?
It is also relevant to investigate the possibility of applying the identification method to other nonlinear
mechanisms, such as coherence zones,45, 46 to investigate whether objective determination of these is also
28 of 32
American Institute of Aeronautics and Astronautics

possible. However, for this purpose, it is necessary to first develop models of these nonlinear perception
processes.
VIII. Conclusions
Motion perception thresholds were estimated during an active, pilot-in-the-loop control task and were
compared to sensory thresholds measured during a passive experimental setup. The measured passive thresh-
olds were lower than the values reported in literature. The difference found might be due to the presence of
specific forces resulting from the tilt with respect to gravity and the fact that subjects were informed of the
motion DOF.
The active threshold values were higher than the measured passive thresholds. However, this difference
was not as large as observed in other studies that investigated the effect of workload on rotational thresholds.
Moreover, the estimated active threshold values were similar to other passive threshold values found in
literature. Subjects’ expectations and small methodological differences in the experimental setup might
influence the threshold value as much as the addition of the active task.
Based on the comparison of the estimated active threshold and passive threshold values found in literature,
one might say that when subjects influence their own motion stimuli, the active threshold is in the same
order of magnitude as the passive threshold.
The proposed identification method was successful in estimating the pitch motion threshold parameter
for the experimental conditions with a motion gain above 0.5, and it was the most reliable for conditions
with motion gain of 1. For lower motion gains the amplitude of the motion cues was insufficient to excite
the motion channel of the pilot model resulting in a bad estimation accuracy of the threshold parameter and
of the other pilot model parameters.
Acknowledgments
The first author is supported by a Toptalent grant from The Netherlands Organisation for Scientific
Research (NWO). The second author is supported by a grant from the Dutch Technology Foundation (STW),
the applied science division of NWO, and the technology program of the Ministry of Economic Affairs.
References
1Hosman, R. J. A. W. and Van der Vaart, J. C., “Vestibular Models and Thresholds of Motion Perception. Results of
Tests in a Flight Simulator,” Tech. Rep. LR-265, Delft University of Technology, Delft, The Netherlands, 1978.
2Benson, A. J., Spencer, M. B., and Stott, J. R. R., “Thresholds for the Detection of the Direction of Whole-Body, Linear
Movement in the Horizontal Plane,” Aviation, Space and Environmental Medicine, Vol. 57, No. 11, November 1986, pp. 1088
29 of 32
American Institute of Aeronautics and Astronautics

– 1096.
3Greig, G. L., “Masking of Motion Cues by Random Motion: Comparison of Human Performance with a Signal Detection
Model,” Tech. Rep. 313, UTIAS, 1988.
4Benson, A. J., Hutt, E. C. B., and Brown, S. F., “Thresholds for the Perception of Whole Body Angular Movement
About a Vertical Axis,” Aviation, Space and Environmental Medicine, Vol. 60, No. 3, March 1989, pp. 205 – 213.
5Heerspink, H. M., Berkouwer, W. R., Stroosma, O., Paassen, M. M. V., Mulder, M., and Mulder, J. A., “Evaluation of
Vestibular Thresholds for Motion Detection in the SIMONA Research Simulator,” AIAA Modeling and Simulation Technologies
Conference and Exhibit, San Francisco, CA, USA, August 15-18 , No. AIAA-05-6502, 2005, pp. 1–20.
6Zaichik, L. E., Rodchenko, V., Rufov, I. V., Yashin, Y. P., and White, A. D., “Acceleration Perception,” AIAA Modeling
and Simulation Technologies Conference and Exhibit, Portland, OR, USA, August 9-11, No. AIAA-99-4334, 1999, pp. 512–520.
7Rodchenko, V., Boris, S. Y., and White, A. D., “In-Flight Estimation of Pilot’s Acceleration Sensitivity Thresholds,”
AIAA Modeling and Simulation Technologies Conference, Denver, CO, USA, August 14-17 , No. AIAA-2000-4292, 2000, pp.
1–8.
8Valente Pais, A. R., Mulder, M., Van Paassen, M. M., Wentink, M., and Groen, E. L., “Modeling Motion Perception
Thresholds in Self-Motion Perception,” AIAA Modeling and Simulation Technologies Conference and Exhibit, Keystone, CO,
USA, August 21-24 , No. AIAA-06-6626, 2006, pp. 1–15.
9Groen, E. L. and Bles, W., “How to Use Body Tilt for the Simulation of Linear Self Motion,” Journal of Vestibular
Research, Vol. 14, No. 5, 2004, pp. 375–385.
10Samji, A. and Reid, L. D., “The Detection of Low-Amplitude Yawing Motion Transients in a Flight Simulator,” IEEE
Transactions on Systems, Man, and Cybernetics, Vol. 22, No. 2, March/April 1992, pp. 300–306.
11McRuer, D. T., Graham, D., Krendel, E. S., and Reisener, W., “Human Pilot Dynamics in Compensatory Systems.
Theory, Models and Experiments With Controlled Element and Forcing Function Variations,” Tech. Rep. AFFDL-TR-65-15,
Systems Technology Inc. & the Franklin Institute, 1965.
12Stapleford, R. L., Peters, R. A., and Alex, F. R., “Experiments and a Model for Pilot Dynamics with Visual and Motion
Inputs,” Tech. Rep. NASA CR-1325, National Aeronautics and Space Administration, 1969.
13Jex, H. R., Magdaleno, R. E., and Junker, A. M., “Roll Tracking Effects of G-Vector Tilt and Various Types of Motion
Washout,” Fourteenth Annual Conference on Manual Control, 1978, pp. 463–502.
14Nieuwenhuizen, F. M., Zaal, P. M. T., Mulder, M., Van Paassen, M. M., and Mulder, J. A., “Modeling Human Multi-
channel Perception and Control Using Linear Time-Invariant Models,” Journal of Guidance, Control, and Dynamics, Vol. 31,
No. 4, 2008, pp. 999–1013.
15Zaal, 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.
16Hosman, R. J. A. W. and Van der Vaart, J. C., “Effects of Vestibular and Visual Motion Perception on Task Performance,”
Acta Psychologica, Vol. 48, No. 1-3, 1981, pp. 271 – 287.
17Van der Vaart, J. C., Modelling of Perception and Action in Compensatory Manual Control Tasks, Ph.D. thesis, Faculty
of Aerospace Engineering, Delft University of Technology, 1992.
18Hosman, R. J. A. W., Pilot’s Perception and Control of Aircraft Motions, Ph.D. thesis, Delft University of Technology,
1996.
19Zaal, 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.
30 of 32
American Institute of Aeronautics and Astronautics

20Zaal, 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.
21De Vroome, A. M., Valente Pais, A. R., Pool, D. M., Van Paassen, M. M., and Mulder, M., “Identification of Motion
Perception Thresholds in Active Control Tasks,” Proceedings of the AIAA Modeling and Simulation Technologies Conference
and Exhibit, Chicago, IL, USA, No. AIAA-2009-6243, 2009, pp. 1–20.
22Pool, D. M., Valente Pais, A. R., De Vroome, A. M., Van Paassen, M. M., and Mulder, M., “Identification of Nonlinear
Motion Perception Dynamics Using Time-Domain Pilot Modeling,” 2011, Submitted to Journal of Guidance, Control, and
Dynamics.
23Hosman, R. J. A. W. and Van der Vaart, J. C., “Thresholds of Motion Perception and Parameters of Vestibular Models
Obtained from Tests in a Motion Simulator,” Tech. Rep. M-372, Delft University of Technology, Delft, The Netherlands, 1980.
24Goldberg, J. M. and Fernandez, C., “Physiology of peripheral neurons innervating semicircular canals of the squirrel
monkey. I. Resting discharge and response to constant angular accelerations,” Journal of Neurophysiology, Vol. 34, No. 4, July
1971, pp. 635–660.
25Fernandez, C. and Goldberg, J. M., “Physiology of peripheral neurons innervating semicircular canals of the squirrel
monkey. II. Response to sinusoidal stimulation and dynamics of peripheral vestibular system,” Journal of Neurophysiology,
Vol. 34, No. 4, July 1971, pp. 661–675.
26Goldberg, J. M. and Fernandez, C., “Physiology of peripheral neurons innervating semicircular canals of the squirrel
monkey. III. Variations among units in their dischargeproperties,” Journal of Neurophysiology, Vol. 34, No. 4, July 1971,
pp. 676–684.
27Young, L. R. and Oman, C. M., “Model for Vestibular Adaptation to Horizontal Rotation,” Aerospace Medicine, Vol. 40,
No. 10, 1969, pp. 1076 – 1080.
28Malcolm, R. and Melvill Jones, G., “A Quantitative Study of Vestibular Adaptation in Humans,” Acta Otolaryngologica,
Vol. 70, No. 2, 1970, pp. 126 – 135.
29Steer, R. W., “Progress in Vestibular Modeling. Part 1 - Response of the Semicircular Canals to Constant Rotation in a
Linear Acceleration Field,” Tech. Rep. 19700021417, NASA, January 1970.
30Melvill Jones, G., Handbook of Physiology, Vol. VI, Vestibular System Part I: Basic Mechanisms, chap. The Functional
Significance of Semicircular Canal Size, Springer Verlag, Berlin, Heidelberg, New York, 1974, pp. 171–184.
31Shirley, R. S. and Young, L. R., “Motion Cues in Man-Vehicle Control – Effects of Roll-Motion Cues on Human Operator’s
Behavior in Compensatory Systems with Disturbance Inputs,” IEEE Transactions on Man-Machine Systems, Vol. 9, No. 4,
1968, pp. 121–128.
32Bergeron, H. P., “Investigation of Motion Requirements in Compensatory Control Tasks,” IEEE Transactions on Man-
Machine Systems, Vol. MMS-11, No. 2, June 1970, pp. 123–125.
33Pool, 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.
34Pool, D. M., Zaal, P. M. T., Van Paassen, M. M., and Mulder, M., “Effects of Heave Washout Settings in Aircraft Pitch
Disturbance Rejection,” Journal of Guidance, Control, and Dynamics, Vol. 33, No. 1, 2010, pp. 29–41.
35Pool, D. M., Zaal, P. M. T., Damveld, H. J., Van Paassen, M. M., Van der Vaart, J. C., and Mulder, M., “Modeling Wide-
Frequency-Range Pilot Equalization for Control of Aircraft Pitch Dynamics,” Journal of Guidance, Control, and Dynamics,
Vol. 34, No. 5, September-October 2011, pp. 1529–1542.
31 of 32
American Institute of Aeronautics and Astronautics

36McRuer, D. T. and Krendel, E. S., “Mathematical Models of Human Pilot Behavior,” AGARDograph AGARD-AG-188,
Advisory Group for Aerospace Research and Development, Jan. 1974.
37Huang, J. and Young, L. R., “Sensation of Rotation About a Vertical Axis with a Fixed Visual Field in Different
Illuminations and in the Dark,” Experimental Brain Research, Vol. 41, No. 2, 1981, pp. 172–183.
38Borah, J., Young, L. R., and Curry, R. E., “Optimal Estimator Model for Human Spatial Orientation,” Annals of The
New York Academy of Sciences, Vol. 545, Representation of Three-Dimensional Space in the Vestibular, Oculomotor, and
Visual System, 1988, pp. 51–73.
39Groen, E. L., Wentink, M., Valente Pais, A. R., and Mulder, M., “Motion Perception Thresholds in Flight Simulation,”
AIAA Modeling and Simulation Technologies Conference and Exhibit, Keystone, CO, USA, August 21-24 , No. AIAA-06-6254,
2006, pp. 1–11.
40Stroosma, O., van Paassen, M. M., Mulder, M., and Postema, F. N., “Measuring Time Delays in Simulator Displays,”
AIAA Modeling and Simulation Technologies Conference and Exhibit, Hilton Head, SC, USA, August 20-23 , No. AIAA-2007-
6562, 2007, pp. 1–9.
41van Paassen, M. M. and Stroosma, O., “DUECA - Data-Driven Activation in Distributed Real-Time Computation,”
AIAA Modeling and Simulation Technologies Conference and Exhibit, Denver, CO, USA, August 14-17 , No. AIAA-2000-4503,
2000, pp. 1–7.
42Stroosma, O., van Paassen, M. M., and Mulder, M., “Using the Simona Research Simulator for Human-Machine Interac-
tion Research,” AIAA Modeling and Simulation Technologies Conference and Exhibit, Austin, TX, USA, August 11-14, No.
AIAA-2003-5525, 2003, pp. 1–8.
43Berkouwer, W. R., Stroosma, O., van Paassen, M. M., Mulder, M., and Mulder, J. A., “Measuring the Performance of
the SIMONA Research Simulator’s Motion System,” AIAA Modeling and Simulation Technologies Conference and Exhibit,
San Francisco, CA, USA, August 15-18 , No. AIAA-2005-6504, 2005, pp. 1–12.
44McRuer, D. T. and Jex, H. R., “A Review of Quasi-Linear Pilot Models,” IEEE Transactions on Human Factors in
Electronics, Vol. HFE-8, No. 3, 1967, pp. 231–249.
45Van der Steen, H., Self-Motion Perception, Ph.D. thesis, Delft University of Technology, 1998.
46Valente Pais, A. R., Van Paassen, M. M., Mulder, M., and Wentink, M., “Perception Coherence Zones in Flight Simula-
tion,” Journal of Aircraft, Vol. 47, No. 6, 2010, pp. 2039–2048.
32 of 32
American Institute of Aeronautics and Astronautics