Content uploaded by Ruud Hosman
Author content
All content in this area was uploaded by Ruud Hosman on Sep 29, 2014
Content may be subject to copyright.
Design & evaluation of spherical washout algorithm for
Desdemona simulator
Mark Wentink1, Wim Bles2 and Ruud Hosman3
TNO Defense, Safety and Security, Soesterberg, The Netherlands
and
Michael Mayrhofer4
AMST Systemtechnik GmbH, Ranshofen, Austria
In co-operation with TNO Human Factors, AMST Systemtechnik developed an advanced
simulator motion system based on a 6 Degrees-of-Freedom (DoFs) centrifuge design. The
simulator cabin is suspended in a freely rotating gimbal system (3 DoFs, >360°) which, as a
whole, can move vertically along a heave axis (1 DoF, ±1m) and horizontally along a linear
arm (1 DoF, ±4m). To provide sustained centripetal acceleration (1 DoF, 0-3g), the linear
arm can spin around a central yaw axis. Unique about Desdemona’s 6 DoFs motion
capabilities is that it can combine onset cueing along the x, y and z-axis (like a hexapod
simulator) with sustained acceleration cueing up to 3g (like a Dynamic Flight Simulator). In
addition, unusual attitudes and large attitude changes can be simulated one-to-one.
Desdemona was developed as a spatial disorientation training and research simulator with
applications in flight simulation, driving simulation and astronaut training. TNO is
responsible for the motion cueing algorithms. Since the Desdemona simulator can move and
rotate along three orthogonal axes, a classical washout scheme can be applied to transform
vehicle (e.g.: aircraft) motion into simulator motion cues. However, for Desdemona this
approach has a number of disadvantages:
•The motion space actually used for simulation would be limited to a relatively small
circular segment of about 90° (R=4m) in which a rectangular area can be fitted (the
remaining segment of 270° is not used).
•The possibility to use centripetal acceleration to simulate sustained specific force would
not be used, since sustained cues are simulated by tilting the pilot with respect to gravity
in the classical washout algorithm.
•The classical washout algorithm is based on simulator kinematics in a Cartesian Frame-
of-Reference while the kinematic design of Desdemona is spherical.
To make better use of the circular motion space of Desdemona a dedicated washout
algorithm was developed, called: Spherical Washout Filter. Instead of directly high-pass
filtering the x- and y-component of the specific forces, these components are first
transformed to radial and tangential acceleration after which the radius, the cabin yaw
angle and the central yaw rate are high-pass filtered. The spherical washout algorithm
significantly enlarges the motion space, since the simulator moves back towards a certain
base radius and not towards a fixed neutral point in space as is the case in conventional
hexapod simulation. In addition, sustained specific forces can be simulated using a
combination of tilt and centripetal acceleration.
1 Researcher, Training & Education Department, wentink@tno.nl, AIAA member.
2 Technical Manager Desdemona Program, Training & Education Department, bles@tno.nl.
3 Consultant, Training & Education Department, AIAA member
4 AMST Project leader Desdemona, Research & Development, AIAA Member
American Institute of Aeronautics and Astronautics
AIAA Modeling and Simulation Technologies Conference and Exhibit
15 - 18 August 2005, San Francisco, California AIAA 2005-6501
Copyright © 2005 by TNO Defense, Safety and Security. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
Nomenclature
0
ψ
= Earth-fixed Frame-of-Reference (FoR)
A
ψ
= FoR fixed to the pilot’s head in the aircraft, aligned with the aircraft axes
S
ψ
= FoR fixed to the pilot’s head in the simulator, aligned with the simulator cabin
1 6
ψ ψ
K
= FoR fixed to each of the six DoF of Desdemona, where
6
ψ
is fixed to the cabin
f
= Specific force vector
ˆ
f
= Perceived specific force vector
G
= G-load vector
centr
k
= Percentage of the specific force that will be simulated using centripetal acceleration
m
n
R
= Rotation matrix that rotates a point in
n
ψ
to
m
ψ
m
n
J
= Jacobian mapping that maps an incremental displacement in
n
ψ
to
m
ψ
centr
ω
= Centrifugal rate
Kf
= Specific force scaling factor
( )
K s
= Controller in Laplace notation
I. Introduction
HE Desdemona simulator is depicted in Figure 1. In this picture, the arrows indicate the separate Degrees-of-
Freedom (DoFs). The six, serial DoF are:T1. Central yaw rotation,
centr
ψ
360> °
2. Radial translation,
R
4m±
3. Heave translation,
H
1m±
4. Cabin roll rotation,
cab
φ
360> °
5. Cabin yaw rotation,
cab
ψ
360> °
6. Cabin pitch rotation,
cab
θ
360> °
Together, these six DoF can move the simulator cabin in
any arbitrary direction, and can position it in any orientation in
space.
A. Research purpose of Desdemona simulator
Originally, TNO Human Factors designed Desdemona as a
disorientation demonstrator to be used in the disorientation
training program for the Royal Netherlands Air Force. In
addition, Desdemona provides a unique motion platform to
facilitate research into motion cueing, simulator training and
effectiveness, and the design and evaluation of new motion
platforms. The simulator is expected to be operational at the
start of 2006.
This paper describes the Spherical Washout Filter that was
developed to specifically suit the novel kinematic layout of
Desdemona. To keep the discussion clear and compact, the paper describes the simulation of specific force only.
B. Need for innovative washout filter
To take full advantage of the large circular motion space of Desdemona (
4R m=
) a dedicated motion cueing
algorithm was developed. A classical washout filter, such as developed by Reid & Nahon, 1985 [1], is not really
suited because it is based on Cartesian kinematics while the inherent kinematics of Desdemona are Polar. Applying
a classical washout strategy would result in a typical ellipsoid shaped motion space that fits inside a circular segment
of at most
180°
(see Figure 2, Left). This leaves at least
180°
of unused motion space! ). Instead, what is desired is
a motion cueing algorithm that explicitly uses the Polar kinematics of the system and that actively forces the cabin
into curved trajectories such that the full circular motion space is used in simulations (Figure 2, Right).
American Institute of Aeronautics and Astronautics
Figure 1: Desdemona simulator. The separate
DoFs (two translations and four rotations) are
indicated by arrows.
Classical washout
Spherica l washout
Hexapod motion space
x
y
cabin
R
Figure 2, Left: Typical x-y motion space of a hexapod simulator driven by a classical washout
filter. The potential motion space of Desdemona is a circular band from
1R m=
to
4R m=
(
0R m=
is a singular point). Right: The motion space of the Spherical Washout Filter
occupies the complete circular band since the ellipsoid motion space moves with the cabin in a
circular direction. The classical and spherical washout principles are compared by visualizing
the direction of washout in the x-y plane by springs and dampers (two springs for classical
washout and two springs and a damper for spherical washout).
A.
B.
C.
D.
A
f
( )
lim A
K⋅
ff
, , ( )
x y z
HP s
3
6
R
, ,
cab cab cab
φ ψ θ
P
C
T
, , ( )
R H
HP s
ψ
1
s
1
s
centr
ω
∑
, ,
centr R H
ψ
, , ( )
x y z
LP s
-
( )
K s
( )
1
6
3
J−
1
s
S
f
,
cab cab
φ θ
DFS
, ,
centr cab
ω ω ψ
-
S
G
( )
cab
HP s
ψ
cab
ψ
-1
centr
ψ
R
centr
k
Figure 3: Mathematical layout of the Spherical Washout Filter, only the specific force
simulation part is shown. The input is the specific force in the aircraft, the output are the
position commands of each DoF of Desdemona. A. Washout on cabin yaw angle to position the
cabin tangential. B. Translational channel for onset cueing. C. Calculation of required
centrifugal rate and tilt angles. D. Jacobian feedback structure to correctly tilt the cabin with
respect to the G-vector.
Additionally, the motion cueing algorithm that drives Desdemona should make use of the possibility to simulate
sustained specific forces up to 2g by means of centripetal acceleration. In comparison, a classical washout filter
relies on tilt coordination to simulate sustained specific forces up to 1g. In Dynamic Flight Simulation (DFS),
American Institute of Aeronautics and Astronautics
special motion cueing algorithms have been developed that simulate the G-load in a fighter aircraft by means of
building up centripetal acceleration in large centrifuge based simulators [2]. However, these motion cueing
algorithms usually only drive 3 DoF so that translational and rotational accelerations are not independent. Ideally, a
dedicated motion cueing algorithm for Desdemona combines the onset cueing characteristics of a classical washout
filter with the sustained cueing capabilities of DFS, and incorporates a smooth transfer between the two, without the
need to switch between two separate filters.
The next section describes the Spherical Washout Filter that takes advantage of the full motion space of
Desdemona, and combines onset cueing and sustained centripetal cueing in one and the same algorithm.
II. Spherical Washout Filter design
A. The basic idea
In Figure 2, Right, the two orthogonal springs on the
lower cabin indicate the working-principle of classical
washout, since the dynamic behavior of an ideal spring
(zero length at rest) is described by a 2nd-order high-pass
filter. The dark gray ellipsoid shows the resulting motion
space of the simulator cabin (note that the centre of the
complete circular area contains a singular point). To take
full advantage of the total, circular, motion space of
Desdemona a different layout of the springs (i.e.: washout
axes) is proposed.
A damper is fixed to the central yaw axis so that the
rotation of the cabin around this axis eventually comes to a
stand still when no acceleration is applied. It is not
necessary to “pull” back the cabin to the starting point of
motion, since the motion space is circular. A damper has
1st-order, high-pass dynamics. A spring is attached
between the cabin and a certain base radial. This spring
stays aligned with the radial between the central yaw axis
and the cabin, and thus moves with the cabin around the
central yaw axis. The spring constantly pulls back the
cabin towards the base radial. A last spring is attached to
the cabin yaw axis, so that the cabin always returns to a
tangent orientation.
In advance of a more detailed description of the
Spherical Washout Filter, Figure 4 shows an example of
the trajectory of the simulator cabin under the washout
action of the damper and springs.
B. Spherical Washout Filter mathematics
The specific force that acts on the pilot in the aircraft serves as an input to the filter. The goal of a motion cueing
filter is to reproduce the perceived specific force in the simulator:
ˆ ˆ
S A
=f f
(1)
At the design stage of a motion cueing filter it is more convenient to leave out the influence of human perception
models (e.g.: dynamics of vestibular system), and to assume that the goal is:
S A
=f f
(2)
In the translational channel, B. in Figure 3, the aircraft specific force is first scaled and limited and then the
required acceleration is calculated by subtracting the G-load vector in the simulator from the specific force.
Consequently, the x,y,z-components of the acceleration are high-pass filtered (1st-order) and
3
6
R
transforms the
commanded acceleration on the simulator pilot to a local FoR (
3
ψ
) that is fixed to the radial between the central
yaw axis and the cabin, and that moves circularly with the cabin around the central yaw axis. In this FoR the local
American Institute of Aeronautics and Astronautics
Figure 4: Typical curved trajectory when
simulating a longitudinal onset cue (arrows). The
rectangles show the position and orientation of the
cabin at 1 second intervals.
Cartesian accelerations can be transformed to Polar coordinates:
3
3
3
10 0
0 1 0
0 0 1
x
centr
P P
C y C
z
aR
R T a with T
Ha
ψ
= ⋅ = −
&&
&&
&&
(3)
where
, ,
centr R H
ψ
are controlled DoFs of Desdemona actuated by separate drives.
To limit the required motion space of the simulator cabin,
, ,
centr R H
ψ
&& &&
&&
are filtered by a second high-pass filter
(1st, 2nd and 2nd order respectively). Integration of the filtered accelerations results in commanded drive velocities.
The coordination channel (C. and D. in Figure 3) calculates a certain centrifugal rate in order to simulate sustained
specific forces by means of centripetal acceleration. This rate is added to the central yaw rate coming from the
translational channel, so that the total central yaw rate equals:
( ) ( )
centr centr centr
trans coor
ψ ψ ω
= +
& &
(4)
Integrating once more results in the commanded drive positions.
To assure that the cabin orientation lines up with the tangential direction, the cabin yaw angle also rotates
depending on the amount of central yaw rotation:
( ) ( )
cab centr
HP s
ψ ψ
= ⋅ −
(5)
where
( )
HP s
is a second order high-pass filter.
In the coordination channel, low-pass specific forces are simulated using both tilt coordination and centripetal
acceleration. The input parameter
centr
k
specifies what percentage of the specific force is simulated by using tilt and
what percentage by using centripetal acceleration. The block DFS in Figure 3 calculates three specific force
components and the required centrifugal rate based on the sustained specific force vector in the aircraft:
( )
,
S A
centr centr
k
ω
′
=
f DFS f
(6)
where
A′
f
is the scaled, limited and low-pass filtered specific force in the aircraft. From
A′
f
and
centr
k
, the required
specific force components in the simulator are calculated as follows:
( )
( )
( )
( ) ( ) ( )
22
222
f
f
f
f f f
A A
S A
centr
S
centr
A
x
S A
y
A
S A A z
x y A
z
G
G k G g g
G g
R
G
ω
′ ′
′
′
′
′
′ ′ ′
=
= ⋅ − +
−
⇒ ≈
⇒ =
− − ⋅
f
f
(7)
If a certain amount of the sustained specific force is reconstructed by tilting the cabin with respect to the G-load
American Institute of Aeronautics and Astronautics
vector in the simulator, then
S A
z z
≠f f
. The output of the block DFS are the three specific force components in the
simulator and the G-load vector resulting from
g
and
centr
ω
. Next, the simulator has to be tilted in order to align the
G-load vector such that the calculated specific force components in Eq. 7 are matched in the simulator. Since the G-
load vector has full rank (
3 1×
ℜ
) it is quite cumbersome to calculate the required cabin pitch, roll and yaw angle
analytically, instead a Jacobian feedback structure [3] is used to numerically calculate the required tilt angles:
( )
( )
( )
( )
1
6
3
6 6
3 3
'
, , ,
S
cab x
S
cab y
S
cab z
S
cab cab cab
S A S
J
with J J G
and K s feedback
φ
ψφ
φ ψ θ
−
∆ ∆
∆ = ⋅ ∆
∆ ∆
=
∆ = ⋅ −
f
f
f
f f f
(8)
The output of the translational channel and coordination channel consists of all the commanded drive positions
for each DoF (as well as velocities and accelerations). Local, high-bandwidth, control loops then control each of the
drives towards the commanded values.
C. Comparison of false cues in classical vs. spherical washout
As shown in Figure 4, a longitudinal onset cue results in a curved trajectory under the action of the Spherical
Washout Filter. Additionally, false cues, not present in the aircraft, occur. In classical washout, the application of
2nd and 3rd order filters results in false cues with opposite sign compared to the original cue since the cabin has to
move back to its neutral position (see Groen et al. [4] for an example and a discussion on the effect of such false
cues). The application of a 1st order filter on the central yaw axis in spherical washout largely prevents the
occurrence of false cues with an opposite sign. In addition, due to the curvature of the cabin trajectory, the motion
space is quite large. The length of the trajectory in Figure 4 is 5.9 meters, which is quite a large motion space in
comparison to a commercial hexapod simulator that typically has a maximum stroke of about 2-3 meters.
However, actively curving the trajectory of the cabin involves lateral accelerations (i.e.: centripetal acceleration)
that are not present in the aircraft, and thus causes false cues. Lateral accelerations are applied to curve the
trajectory inwards, back towards a certain base radius. Since they are proportional to the yaw rate (
centr
ψ
&
) , the build
up of the false cue is relatively slow and cabin tilt can be applied to compensate for the lateral acceleration. The
Jacobian feedback structure in the coordination channel automatically tilts the cabin to compensate the false lateral
acceleration (ideally sub-threshold).
In classical washout, false cues occur in the same direction as the simulated cue. This is a big difference
compared to spherical washout, where the false cues occur in both the x- and y-direction as a result of a longitudinal
or lateral cue (i.e. cross-coupling between the x- and y-channel). The disturbing effect of a false cue in a direction
perpendicular to the original cue is not well known and has to be investigated once Desdemona is operational.
III. Spherical Washout Filter evaluation
A. The input profiles
Since the Spherical Washout Filter is a novel filter, and incorporates non-linear transformations, simple step
acceleration profiles serve as input profiles to demonstrate the main characteristics of the filter. More realistic
profiles would lead to more complicated responses that are difficult to relate to specific characteristics of the
Spherical Washout Filter.
In the coordination channel the tilt rate is limited to 5˚/s. In case of G-load cueing using centrifugal motion (in
the last example), the tilt rate is limited to 25˚/s. The base radius to which cabin motion is washed out is
3R m=
.
In all simulations the z-component of the input specific force is equal to gravity. The simulation results are
visualized by means of three plots: a plot of the cabin trajectory including the cabin yaw rotation, a plot of the
specific forces on the pilot and a plot of the angular rates.
B. Simulation of longitudinal step input
In this example a longitudinal specific force step input is simulated. The following equation defines the profile.
American Institute of Aeronautics and Astronautics
0 1
0.3 1
0
x
A
A
y
A
z
t s
g t s
g
∀ <
=∀ ≥
=
=
f
f
f
(9)
For comparison,
0.3g
is about equal to the acceleration of a B-747 during the take-off run. Figure 5 shows the
simulation results of the Spherical Washout Filter in two cases: 1. the filter frequency in the initial high-pass filter
on
A
x
f
is set to
2 /rad s
, 2. the frequency is set to
1 /rad s
. Decreasing the filter frequency results in a longer onset
cue and a longer trajectory of the simulator cabin. Notice that the Spherical Washout Filter actively curves the
trajectory and rotates the cabin back to the tangential direction.
-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
Desdemona trajectory, top view
Length = 2.9 m
y-position [m]
x-position [m]
0 1 2 3 4 5 6 7 8 9 10
-0.1
0
0.1
0.2
0.3
0.4
Specific force
fx,y
[g]
0 1 2 3 4 5 6 7 8 9 10
-10
0
10
20
Angular rate
ωx,y,z
[°/s]
Time [s]
ωx
ωy
ωz
x
f
y
f
-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
Desdemona trajectory, top view
Length = 5.5 m
y-position [m]
x-position [m]
0 1 2 3 4 5 6 7 8 9 10
-0.1
0
0.1
0.2
0.3
0.4
Specific force
fx,y
[g]
0 1 2 3 4 5 6 7 8 9 10
-10
0
10
20
Angular rate
ωx,y,z
[°/s]
Time [s]
ωx
ωy
ωz
x
f
y
f
Figure 5: Response to longitudinal step acceleration (
0.3g
). The rectangles show the position and orientation
of the cabin at 1 sec. intervals. Top: The initial, high-pass filter frequency is
2 /rad s
. Bottom: Larger onset
cue due to a smaller high-pass filter frequency of
1 /rad s
.
American Institute of Aeronautics and Astronautics
C. Simulation of lateral step input
A lateral step input at
3t s=
is added to the longitudinal step input in the previous example (
1 /rad s
filter
frequency). The input is:
0 1
0.3 1
0 3
0.3 3
A
x
A
y
A
z
t s
g t s
t s
g t s
g
∀ <
=∀ ≥
∀ <
=− ∀ ≥
=
f
f
f
(10)
In the second simulation in Figure 6, the sign of the lateral specific force is changed. Note that the Spherical
Washout Filter is asymmetrical, the response changes significantly when the input changes sign.
-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
Desdemona trajectory, top view
Length = 5.3 m
y-position [m]
x-position [m]
-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
Desdemona trajectory, top view
Length = 6 m
y-position [m]
x-position [m]
0 1 2 3 4 5 6 7 8 9 10
-0.1
0
0.1
0.2
0.3
0.4
Specific force
fx,y [g]
0 1 2 3 4 5 6 7 8 9 10
-20
-10
0
10
20
Angular rate
ωx,y,z
[°/s]
Time [s]
ωx
ωy
ωz
y
f
x
f
Figure 6: Response to longitudinal step (
0.3g
) followed by a lateral step at t=3s. The rectangles show the
position and orientation of the cabin at 1 sec. intervals. Top: Negative lateral step acceleration. Bottom:
Positive lateral acceleration.
American Institute of Aeronautics and Astronautics
0 1 2 3 4 5 6 7 8 9 10
-0.4
-0.2
0
0.2
0.4
Specific force
fx,y [g]
0 1 2 3 4 5 6 7 8 9 10
-20
-10
0
10
20
Angular rate
ωx,y,z
[°/s]
Time [s]
ωx
ωy
ωz
x
f
y
f
D. Simulation of G-load in combination with onset cueing
This example shows the response of the Spherical Washout Filter to a block shaped lateral G-load of
0.5g
.
Centrifugal motion is used to simulate sustained lateral specific force. Note that the same Spherical Washout Filter
generated both this example and the previous examples, with the difference that
1
centr
k=
(pure centrifugal motion)
in this example and
0
centr
k=
(pure tilt) in the previous examples (see Eqs. 6,7 ). The inward movement of the cabin
at the beginning of the trajectory, and the outward movement at the end clearly show the lateral onset cues.
-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
Desdemona trajectory, top view
Length = 21.2 m
y-position [m]
x-position [m]
0 1 2 3 4 5 6 7 8 9 10
-0.2
0
0.2
0.4
0.6
Specific force
fx,y [g]
0 1 2 3 4 5 6 7 8 9 10
-20
0
20
40
60
80
Angular rate
ωx,y,z
[°/s]
Time [s]
ωx
ωy
ωz
y
f
x
f
Figure 7: Response to a block shaped, lateral G-load of
0.5g
.
IV. Discussion
A. The evaluation criteria
Since the purpose of the motion cues generated by the Spherical Washout Filter is to invoke realistic control
behavior of the pilot in the simulator, and to enhance simulation fidelity in general, it would be optimal to analyze
the simulation results from a pilot perception and control point of view [5]. Although at TNO Human Factors
extensive knowledge of human perception models [6] and pilot control models [7] is available, in this study these
models were not (yet) applied due to the additional complexity of these models.
Instead, this paper discusses the simulation results in terms of the accuracy with which specific forces are
reproduced in the simulator and by analyzing the amount and amplitude of the false cues that potentially disturb the
simulation.
B. Evaluation of longitudinal step input simulation
The simulations in Figure 5 clearly demonstrate the main characteristics of the Spherical Washout Filter.
Initially, the longitudinal onset cue drives the cabin forward in a straight line, then, when the distance to the base
radius increases the cabin is pulled towards the base radius at
3R m=
and simultaneously rotates to a tangential
direction. The yaw rate that is build up during the onset cue is eventually washed out and the cabin comes to a
standstill at the end of the simulation. Meanwhile, in the coordination channel the sustained specific force cue
causes the cabin to slowly tilt backwards until the specific force in the simulator equals that in the aircraft.
Compared to classical washout and typical hexapod simulators, the duration of the onset cue is rather long. Even
more important, the cabin is not accelerated in opposite direction after the onset cue as is the case in classical
washout that uses 2nd- or 3rd-order high-pass filters to move the cabin back to the neutral point (in spherical washout
a 1st-order filter on the central yaw angle determines the shape of the onset cue). The Desdemona cabin is only
moved back towards the base radial and not all the way back to its starting position.
False accelerations in the lateral direction occur since the Spherical Washout Filter actively curves the cabin
American Institute of Aeronautics and Astronautics
trajectory towards the base radial. To compensate these false cues, the Jacobian feedback structure in the
coordination channel automatically tilts the cabin sideways in roll. The specific force plots show that the
compensation is quite effective; the resulting lateral specific force is small compared to the simulated longitudinal
cue. In addition to tilting the cabin, rotating it back towards the tangential direction results in false yaw motion that
is not present in the real aircraft. Although tilt motion has been limited to 5˚/s, the interaction with the yaw motion
results in somewhat higher angular rates of the pilot. This effect is best seen in the lower plots in Figure 5 where the
long onset cue (5.6m stroke!) is accompanied by a yaw rotation of the cabin of more than 90˚ and a yaw rate of up to
17˚/s.
The maximum angular accelerations on the pilot are in the order of about 5˚/s2 (not plotted). Since these are false
cues they should be minimized. For example, Van der Steen, 1998 [8], found that the perceptual threshold for yaw
and roll acceleration, while the subject is simultaneously experiencing longitudinal vection, is about 3.5˚/s2 and
5.0˚/s2, respectively. In Desdemona, the false angular accelerations occur in combination with onset cues and
sustained specific forces which might mask the false cues even more than the vection in van der Steen’s
experiments, and thus increase the perceptual thresholds.
C. Evaluation of lateral step input simulation
Although a negative lateral specific force step is added in the simulations in Figure 6, the resulting longitudinal
specific forces and angular rates are quite similar to those in the previous simulation in Figure 5. This is quite a
good result considering that longitudinal and lateral accelerations are cross-coupled when the trajectory of the cabin
is curved.
The cross-coupling effect is more obvious in the lower three plots in Figure 6 where a positive lateral
acceleration to the left is simulated. The sudden decrease in the radius at the start of the lateral onset cue (at
3t s=
)
causes a small, simultaneous decrease in the longitudinal specific force in the simulator. This effect is accountable
to the dependency of the longitudinal onset cue on the radius (i.e. tangential acceleration). The ‘bump’ in the
longitudinal acceleration at the start of the lateral onset cue is smooth and does not cause a significant jerk.
The asymmetry between the simulated trajectories in Figure 6 clearly shows the non-linearity of the filter; the
trajectories are completely different (although comparable in length).
D. Evaluation of G-load simulation in combination with onset cueing
In the simulation in Figure 7, the Spherical Washout Filter uses centrifugal motion instead of tilt in order to
simulate the sustained lateral specific force. A combination of both tilt and centrifugal motion is also possible by
choosing a value between 0 and 1 for
centr
k
. Choosing
1
centr
k=
(as in this simulation) lets the Spherical Washout
Filter behave as a Dynamic Flight Simulation (DFS) algorithm combined with onset cues like in a hexapod
simulator. In Figure 7, the onset cues at the beginning and the end of the block shaped G-load are clearly visible in
the specific force plot as well as in the trajectory plot. The cabin moves inward at the first onset cue (to the left) and
outward at the second (to the right).
Even though the maximum tilt rate was increased to 25˚/s in this simulation, there occurs a false cue in
longitudinal direction due to the initial yaw acceleration (to increase centripetal acceleration). The tilt rate is not
quite large enough for the Jacobian feedback structure in the coordination channel to compensate the occurring
tangential acceleration completely. Furthermore, it is interesting to see that at first the Jacobian feedback structure
uses both tilt and centripetal acceleration in order to build up the sustained specific force vector until the yaw rate is
high enough to use centripetal acceleration only. In the angular rate plot this can be derived from the existence of
roll motion (
x
ω
) at the beginning of the specific force cue.
Due to the centrifugal motion, a significant yaw rate of up to
70 / s°
is build up during the simulation. The
existence of rather large rotational false cues is typical for Dynamic Flight Simulation in general [2]. However, in a
number of vehicle control tasks, rather high yaw rates and lateral acceleration are coupled and occur simultaneously
as well. A good example is curve driving in a car. Whether Sperical Washout Filters are more suitable than
conventional washout filters in these conditions will be examined in due time.
V. Conclusion
Desdemona is an innovative motion platform that can combine the advantages of onset cueing in 6 DoFs with
sustained G-load cueing as in Dynamic Flight Simulation. Although Desdemona was originally designed for pilot-
in-the-loop disorientation training, the simulation goals include motion cueing research, flight simulation, driving
simulation and the specification of future moving base simulators.
American Institute of Aeronautics and Astronautics
To take full advantage of the kinematic design and the large circular motion space of Desdemona a dedicated
motion cueing algorithm was developed, called Spherical Washout Filter. The main characteristics of this novel
washout filter are that it actively curves simulator trajectories back towards a certain base radius, it incorporates a
large circular motion space and it can simulate sustained specific forces by applying both tilt and centrifugal
acceleration. Furthermore, the filter makes use of a Jacobian feedback structure to tilt the cabin correctly with
respect to the G-load vector in the simulator. Due to the feedback of the (calculated) specific force in the simulator
this structure also compensates effectively for parasitic accelerations (by means of tilt).
The simulated step responses showed that the onset cues in longitudinal direction are quite large, the ‘stroke’ can
be over 5m long. As a result a false cue due to acceleration reversal does not occur. Acceleration reversal (forward
onset cue followed by a backward acceleration) is common in conventional hexapod simulators since they have to
return to a neutral point. Pilots are sensitive to a too strong decrease or even reversal of acceleration in a specific
force step response (e.g. take-off run) [4].
Curving the trajectory of the simulator back to a certain base radius results in the occurrence of parasitic
accelerations that are usually not present in the real aircraft. Yaw acceleration and lateral (centrifugal) acceleration
are predominant as false cues in the Spherical Washout Filter. Although the false lateral accelerations can be
compensated for quite effectively by tilt it remains to be seen how sensitive subjects are to the occurrence of false
yaw accelerations. To a large extend, the effectiveness of the Spherical Washout Filter depends on the perceptual
threshold for yaw acceleration (while being immersed in a simulation).
However, in car driving simulation yaw motion and lateral acceleration are coupled when driving on a curved
road. Therefore, the Spherical Washout Filter is expected to produce especially good results in driving simulation
[9].
References
1Reid, L.D., and Nahon, M.A., Flight simulation motion-base drive algorithms: part 1 – developing and testing the equations,
Institute for Aerospace Studies, University of Toronto, 1985.
2Spenny, C.H., Liebst, B.S., Chelette, T.L., Folescu, C., Sigda, J., Development of a sustainable-G dynamic flight simulator,
AIAA Guidance, Navigation, and Control Conference, CP4075, AIAA, Washington, DC, 2000.
3Craigh, J.J., Introduction to robotics: mechanics and control, 3rd ed., Pearson Prentice Hall, Upper Saddle River, 1989, Chap.
10, 309.
4Groen, E.L., Clari, M.V.C., Hosman, R.J.A.W., Evaluation of perceived motion during a simulated takeoff run, Journal of
Aircraft, Vol. 38, No. 4, 2001, 600-606.
5Groen, E.L., Smaili, M.H., Hosman, R.J.A.W., Motion fidelity of a simulated decrab maneuver, AIAA Motion Simulation
and Control Conference, AIAA, San Francisco, CA, 2005.
6Bos, J.E. & W. Bles, Theoretical considerations on canal-otolith interaction and an observer model, Biol. Cybern., Vol. 86,
2002, 191-207.
7Hosman, R.J.A.W., Stassen, H.G., Pilot’s perception in the control of aircraft motions, Control engineering practices, Vol. 7,
No. 11, 1999, 1421-1428.
8H. van der Steen, Self-motion perception, Ph.D. Dissertation, Aeronautics and Astronautics Dept., Delft University of
Technology, Delft, The Netherlands, 1998.
9Wentink, M. & Bles, W., Driving curves in Desdemona, Motion Simulation Conference, DLR, Braunschweig, Germany,
2005.
American Institute of Aeronautics and Astronautics