Professional Documents
Culture Documents
Applying The Objective Motion Cueing Test To A Classical Washout Algorithm
Applying The Objective Motion Cueing Test To A Classical Washout Algorithm
Applying The Objective Motion Cueing Test To A Classical Washout Algorithm
R.J.A.W. Hosman 4
AMS Consult, Delfgauw, The Netherlands
and
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
S.K. Advani 5
IDT, Breda, The Netherlands
As part of the new ICAO 9625 simulator qualification guidelines a new method to
quantify simulator motion systems has been proposed: the Objective Motion Cueing Test.
This paper describes the authors’ experiences in applying this test to several motion
configurations of a hexapod research flight simulator. It includes lessons learned which are
hoped to assist other practitioners in performing and interpreting the OMCT.
I. Introduction
F LIGHT simulators are widely used for pilot training and research. In both domains it is important to faithfully
reproduce the combined aircraft-pilot system in the simulator. To correctly stimulate the pilot as part of that
combined system, high-end simulators use elaborate cueing devices such as visual, control loading and motion
systems. For training simulators the cueing devices are subject to qualification standards, as prescribed by the
relevant authorities. Many of these standards are based on those published by the International Civil Aviation
Organization (ICAO)1.
In the newest version of the ICAO standards, a testing method is proposed for motion systems: the Objective
Motion Cueing Test (OMCT). It is described as the Frequency-Domain Motion Cueing System Performance Test in
Attachment F of Amendment 3 of ICAO Manual 9625. The purpose of this test is to supplement subjective
evaluation of the simulator’s physical motion performed in the course of the qualification process with a more
objective metric2,3,4.
In a research environment, where qualification for training is usually not an issue, the OMCT can provide a
useful way of documenting the configuration and capabilities of a particular simulator and compare it to simulators
used in other studies. In this way it can support the scientific discussion about the role and usefulness of motion in
flight simulation, which has often been hampered by uncertainties in the precise characteristics of the motion system
used for a particular study.
With today’s high quality motion hardware, the pilot’s perception in the simulator is now predominantly shaped
by the structure and configuration of the motion cueing algorithm, causing earlier tests on just the robotic
1
Researcher, Control and Simulation Division, Faculty of Aerospace Engineering, O.Stroosma@tudelft.nl, P.O. Box
5058, 2600 GB Delft, The Netherlands, Senior Member AIAA.
2
Associate Professor, Control and Simulation Division, Faculty of Aerospace Engineering,
M.M.vanPaassen@tudelft.nl, P.O. Box 5058, 2600 GB Delft, The Netherlands, Member AIAA.
3
Professor, Control and Simulation Division, Faculty of Aerospace Engineering, M.Mulder@tudelft.nl, P.O. Box
5058, 2600 GB Delft, The Netherlands, Senior Member AIAA.
4
Director, AMS Consult, R.Hosman@AMSConsult.demon.nl, Dijkgraafstraat 26, 2645 KN, Delfgauw, The
Netherlands, Senior Member AIAA.
5
President, IDT, S.Advani@idt-engineering.com, Overkroetenlaan 162, 4832 KB, Breda, The Netherlands, Senior
Member AIAA.
1
American Institute of Aeronautics and Astronautics
characteristics of the motion system hardware and its controller, such as for instance described in AGARD-AR-1445,
to fall short in painting a complete picture. The OMCT expands the view by exciting the combined motion cueing
algorithm and motion base hardware in one degree of freedom and with one frequency at a time to construct a linear
transfer function from aircraft model output to motion base response. It also defines some unwanted cross effects to
be minimized, such as simulator pitch motion due to aircraft surge. However, the essentially linear and per-axis tests
can not capture the full characteristics of the motion cueing algorithm if it has significant non-linearities, such as
adaptive behaviour, tight position or rate limits, or if it leverages knowledge of expected patterns in the aircraft
motions, such as the relation between yaw and sway in aircraft where the pilot is located far in front of the aircraft’s
centre of gravity.
Many motion bases in use today use a motion cueing or washout algorithm that is based on the so-called
classical washout algorithm, described by Reid and Nahon6. It also forms the basis for a set of algorithms running on
the SIMONA Research Simulator (SRS)7 at the Delft University of Technology in the Netherlands.
This paper describes the application of the OMCT to a typical classical washout algorithm running on the SRS,
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
starting with an overview of the OMCT and the classical washout algorithm itself. An overview is given of the
technical implementation on the SRS, as well as of the analysis process. Example results are given of how the
OMCT was used to investigate some configuration choices impacting the cueing algorithm, followed by some
lessons learned and conclusions.
The frequency range to be investigated currently ranges from 0.1 rad/s to 15.8 rad/s (see Table 2). To allow a
correct analysis using Fourier transforms, the frequencies may have to be adapted slightly to properly fit into the
measurement time, depending on the number of periods in the measurement and the simulator’s sampling rate. The
amplitudes of the driving signals are nominally 1 deg/s and 1 m/s², but the rotational rates are reduced for very low
and very high frequencies to avoid non-linear effects as much as possible.
Since during operation of the simulator the pilot is the main sensor of interest, the measurements should
primarily relate to the motions he or she experiences. For this reason the reference point for both the driving signal
and the measured signal is a point at the pilot station, referred to in this paper as the OMCT reference point. This
point is located 0.35 m below the Design Eye Reference Point of the simulator. Consequently the OMCT “aircraft”
will pitch around the OMCT point for test 1 and 2, and the simulator rotations will be measured around the same
point in the simulator. Any specific forces measured at the OMCT point due to the simulator rotating around a
different point (e.g. the Upper Gimbal Centroid) will thus be considered spurious cues, to be included in the results
of cross effect test 2. Section VI will further explore this issue.
2
American Institute of Aeronautics and Astronautics
Table 2. OMCT input signals For measuring the
motion base response a
Frequency Frequency Translational Rotational amplitude number of sensors can be
number [rad/s] amplitude Attitude Rate Acceleration employed. Ideally an
[m/s²] [deg] [deg/s] [deg/s²] Inertial Measurement Unit
1 0.100 1.0 6.000 0.600 0.060 (IMU) is mounted at the
2 0.158 1.0 6.000 0.948 0.150 OMCT measurement point
3 0.251 1.0 3.984 1.000 0.251 to directly measure the
4 0.398 1.0 2.513 1.000 0.398 specific forces and
5 0.631 1.0 1.585 1.000 0.631 rotational rates at the
6 1.000 1.0 1.000 1.000 1.000 required location. Practical
7 1.585 1.0 0.631 1.000 1.585 limitations may inhibit this
8 2.512 1.0 0.398 1.000 2.512 solution, so an IMU at a
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
3
American Institute of Aeronautics and Astronautics
LIS
g
Tilt coordination channel
TS
𝑘𝑠 2 𝑠
𝐻𝐻𝑃 (𝑠) = 2 𝑠+𝜔 (1)
𝑠 2 +2𝜁𝜔𝑛 𝑠+𝜔𝑛 𝑏
The filter can be characterized by its gain k, its 2nd order break frequency ωn, its damping ratio ζ and its 1st order
break frequency ωb.
The return to neutral effect after a step input on acceleration can be proven with the Final Value Theorem in Eq.
(2).
1 1 𝑘𝑠 2 𝑠 1 1
lim𝑡→∞ 𝑥(𝑡) = lim𝑠→0 𝑠𝑋(𝑠) = lim 𝑠 �𝐻𝐻𝑃(𝑠) � = lim 𝑠 � �
2 𝑠+𝜔 𝑠 𝑠 2 =0 (2)
𝑠→0 𝑠 𝑠2 𝑠→0 𝑠 2 +2𝜁𝜔𝑛 +𝜔𝑛 𝑏
Here x is the position of the simulator, 1/s the Laplace transform of the step function and the final 1/s² a double
integrator to transform from acceleration to position. Even though a step in specific force is not uncommon (e.g. the
take-off run), a step in rotational acceleration is rarely encountered, so the filters in the rotational channels usually
have a lower order.
The remaining major components of the classical washout algorithm are related to the frames of reference the
algorithm works with. As the pilot is oriented and moving with the aircraft and simulator, the perceptional frame of
reference will be the body-fixed frame. The body frame would therefore be the obvious choice to perform the
filtering in. However, any rotations during the washout phase will lead to a drift in simulator position in inertial
space. For instance, an initial surge onset in the body frame while nose-up, followed by a washout while nose-down
will result in a vertical position offset upwards. In normal operations this would not be acceptable, so the washout
filtering is performed in the inertial frame of reference, which itself is independent of the simulator’s orientation. For
this purpose, a transformation from the specific force inputs from the aircraft’s body frame to the simulator’s inertial
reference frame (denoted by LIS) is introduced before the high-pass filter. By modifying the specific force input in
this way, an additional feedback between simulator attitude and the translational channel is introduced, which
complicates the algorithm’s workings somewhat. One effect that Reid and Nahon8 already identified is the fact that
4
American Institute of Aeronautics and Astronautics
second order surge and sway high-pass filters actually function as third order filters due to the tilt coordination
interacting with this transformation.
In a similar vein, the rotational high-pass filtering is done on the simulator’s Euler angle (β) rates or
accelerations instead of on the body rotational rates (ω). For this purpose the transformation matrix denoted by Ts in
Figure 1 is used.
An aspect of the filter implementation that will be shown to have an impact on the OMCT results is the location
of the reference point AA. On the input side the aircraft’s specific forces and rotations are calculated for this point.
This means that any aircraft rotation around a point other than AA, e.g. the aircraft’s centre of gravity or the OMCT
reference point (if not equal to AA), will introduce additional accelerations and thus specific forces in point AA. On
the output side the same holds true. Any rotations coming from the algorithm, e.g. due to the rotational or the tilt
coordination channel, will occur around point AA. In points other than AA, e.g. the OMCT reference point,
additional accelerations and specific forces will be present. Due to non-linear effects in the algorithm and generally
differences in the filter parameters in the involved channels, a pure rotation around a point other than AA on the
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
aircraft side will usually not result in a pure rotation around the same point on the simulator side.
The choice of point AA has been discussed by Reid and Nahon6. They have shown that for a hexapod motion
system the most economical choice for the motion cueing reference point is the Upper Gimbal Centroid Point
(UGP). This point lies in the centre of the circle through the upper gimbal attachment points. Around this point the
largest rotations can occur for a given actuator stroke length, making it an attractive choice for the reference point in
many motion systems. Section VI will show the impact this choice has on OMCT results, compared to choosing the
OMCT reference point as the motion cueing reference point (AA).
The implementation of the algorithm in the SRS differs from the algorithm described by Reid in that it allows for
the selection of the reference point AA. Other differences are the use of acceleration setpoints as outputs of the
algorithm, instead of position setpoints. This is needed for the motion controller driving the SRS hardware. As a
consequence, the tilt coordination angles need to be differentiated twice before being added to the Euler angle
accelerations coming out of the rotational channel. The input of that channel is also comprised of angular
accelerations, instead of angular rates, so the transformation to Euler angle accelerations also uses the time
derivative of the transformation matrix TS.
This brief overview of the classical washout motion cueing algorithm has introduced some of the characteristics
and choices that can be made when tuning a motion cueing algorithm such as this. The remainder of this paper will
show how this algorithm was investigated using the OMCT in a research environment, starting with the
implementation of the OMCT functionality in the existing motion subsystem.
IV. Implementation
The software architecture of the SRS has been described by Stroosma et al.7 and has a modular publish/subscribe
architecture. The motion subsystem consists of a central module which subscribes to a channel containing the
aircraft model motion outputs (specific forces, rotational rates and accelerations), provides this data to the motion
cueing algorithm, protects the hexapod’s excursion limits and drives the motion system’s hardware. For logging and
monitoring purposes it publishes a number of channels containing the setpoints used to drive the motion hardware,
as well as the reported simulator position and attitude, as reconstructed from the measured actuator extensions. A
separate module is connected to an onboard IMU and publishes the measured inertial data, as will be described in
section IV.B. All these modules run at a rate of 100 Hz.
To incorporate the OMCT into this architecture a signal generator module was implemented to generate the
required motion commands and publish them on the correct motion channel. A custom logging module is
responsible for recording all relevant channels for later analysis. No changes to the core motion modules are
necessary to run OMCT in this configuration.
A. Signals
The signals driving the motion system are individual sinusoids, as prescribed by the OMCT. To avoid transients
polluting the measurement, the signal generator adds a linear fade-in on the acceleration signal that smoothly
increases both acceleration, velocity and position. This ensures that the simulator oscillates around the neutral
position, inside the most linear part of its envelope. The fade-in period is followed by the measurement period,
which is of equal length to the fade-in and contains an integer number of periods of the sinusoid for proper Fourier
analysis. To ensure this the frequencies of the input signal have to be adjusted slightly, as mentioned earlier. The
difference between the used frequency and the one prescribed by OMCT is never larger than 1 per cent, with the
largest deviation at the higher frequencies. The number of periods in the measurement phase is chosen depending on
5
American Institute of Aeronautics and Astronautics
the frequency, with a minimum of 3 for low frequencies, a nominal duration of 1 minute for mid-range frequencies,
and a maximum of 30 periods for high frequencies. This is a compromise between the need to dampen out
transients, having a sufficient number of periods for analysis and minimizing the time needed to do the
measurement.
B. Measurements
The OMCT analysis only needs the inputs to the cueing algorithm and the measured output of the motion
hardware. For analysis purposes, it can be useful to also record the cueing algorithm’s output, which serves as the
command to the motion hardware. The influence of the motion’s hardware can thus be investigated separately from
the cueing algorithm itself.
In the SRS implementation two sensors are available: an IMU to measure specific forces and rotational rates and
the actuator length measurements. Some low-level processing is performed before the data is recorded. The actuator
lengths are used to reconstruct the platform’s position and attitude through the forward kinematics. The IMU data is
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
V. Analysis
The OMCT examines in the frequency response of the combined cueing algorithm and motion hardware. The
use of Fourier analysis is therefore a natural choice for the analysis. By Fourier transforming the data from the
measurement phase, as described earlier, calculating the complex quotient of the relevant input and output Fourier
coefficients at the driving frequency, and taking its amplitude and phase, a data point in the OMCT plot can be
constructed. All data points for a specific test can then be plotted in a Nichols or Bode plot. For the SRS tests
MATLAB® was used to perform the analysis.
Although some of the OMCT tests are quite straightforward, some required additional data processing or further
assumptions and simplifications. The next sections will present the tests in all six degrees of freedom and any issues
encountered during analysis. All equations will in the following use the subscript PS for measurements at the pilot
station in the simulator and PA for driving signals at the pilot station in the “aircraft”.
A. Pitch
Tests 1 and 2 examine the pitch and surge response due to aircraft pitch. Although formally two separate tests,
they can be combined into a single measurement and separated in the analysis, as is the case for the other axes.
The pitch test is somewhat different from the others because it prescribes an uncoordinated pitch manoeuvre.
With such a manoeuvre it actually combines two channels, the pitch and surge channel, which contains the gravity
component associated with the aircraft’s changing pitch angle. Although many pitch manoeuvres in flight are in fact
coordinated, with the aircraft decelerating proportionally to the gravity component and leading to no or a negligible
surge specific force, some pitch manoeuvres are indeed uncoordinated. The most important example is the first part
6
American Institute of Aeronautics and Astronautics
of the take-off rotation. It can be argued that this is a critical pitch manoeuvre, especially for training simulators,
which justifies its inclusion as a test in the OMCT. The OMCT results for uncoordinated and coordinated
manoeuvres do differ quite extensively, as will be shown in section VI.
The analysis for test 1 is a straightforward comparison between “aircraft” pitch rate and simulator pitch rate, as
measured with the IMU (Eq. (3)).
𝑞𝑃𝑆
𝐻1 (𝜔) = (𝜔) (3)
𝑞𝑃𝐴
Test 2 examines a cross effect, the surge acceleration due to pitch acceleration. In this case the output is
acceleration, according to Eq. (4), and not specific force.
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
𝑥 (𝑡) 𝑥 (𝑡)
𝑎𝑃𝑆 = 𝑓𝑃𝑆 − 𝑔 sin 𝜃𝑃𝑆 (𝑡) (4)
The frequency response to be minimized is defined by Eq. (5) and is expressed in m/deg.
𝑥
𝑎𝑃𝑆
𝐻2 (𝜔) = (𝜔) (5)
𝑞̇ 𝑃𝐴
In the SRS measurements it was quite difficult to determine this frequency response due to its small amplitude
and the need to combine two sensors, i.e. the (delayed) IMU for the surge specific force and the actuator extensions
for the pitch angle. Although the surge acceleration setpoint is usually quite small, the measured acceleration can be
an order of magnitude larger, especially at high frequencies. It can be difficult to determine if this is due to
deficiencies in the motion hardware, the sensors or both.
B. Roll
In tests 3 and 4 the simulator is driven in the roll channel in a coordinated way, meaning there is no gravity
component in the sway channel. The transfer functions are given in Eqs. (6) and (7) and can be readily deduced from
the IMU measurements. The cross effect in Eq. (7) is expressed in m/deg.
𝑝𝑃𝑆
𝐻3 (𝜔) = (𝜔) (6)
𝑝𝑃𝐴
𝑦
𝑓𝑃𝑆
𝐻4 (𝜔) = (𝜔) (7)
𝑝̇ 𝑃𝐴
C. Yaw
Test 5 examines the yaw channel, which is not coupled to any of the other axes and should provide a clean
response. It is therefore ideally suited as a first test in any OMCT implementation, together with the equally
uncoupled heave axis. The transfer function in Eq. (8) can be found using IMU measurements.
𝑟𝑃𝑆
𝐻5 (𝜔) = (𝜔) (8)
𝑟𝑃𝐴
D. Surge
For tests 6 and 7 the inputs are a specific force on the surge channel. Due to the tilt coordination providing low-
frequency cues using gravity, a pitch rotation is usually present. This rotational acceleration itself is spurious and
should be minimized, an effect tested by test 7. The main effect, surge due to surge, is defined by Eq. (9) and can be
found from IMU measurements.
7
American Institute of Aeronautics and Astronautics
𝑥
𝑓𝑃𝑆
𝐻6 (𝜔) = 𝑥 (𝜔) (9)
𝑓𝑃𝐴
The cross effect is defined by Eq. (10) and is based on IMU measurements with additional processing needed to
determine the rotational acceleration. Due to the mixing of rotations and specific forces this transfer function is not
dimensionless and is expressed in deg/m.
𝑞̇ 𝑃𝑆
𝐻7 (𝜔) = 𝑥 (𝜔) (10)
𝑓𝑃𝐴
E. Sway
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
Tests 8 and 9 are very similar to the surge tests, with a main effect and cross effect to assess false cues due to tilt
coordination. The transfer functions in Eqs. (11) and (12) can be determined with IMU measurements and
postprocessing to determine the rotational acceleration. Again, the cross transfer function has the unit of deg/m.
𝑦
𝑓𝑃𝑆
𝐻8 (𝜔) = 𝑦 (𝜔) (11)
𝑓𝑃𝐴
ṗ PS
H9 (ω) = y (ω) (12)
fPA
F. Heave
The final test introduces a heave signal and measures the simulator response, as defined in Eq. (13).
fzPS
H10 (ω) = (ω) (13)
fzPA
The specific forces can be measured with an IMU. Similar to the yaw test the heave test is relatively uncoupled
and straightforward to analyse. As such it is a good first check on an OMCT implementation.
To illustrate the use of the OMCT implementation in the SRS and the analysis method described in this section,
the following sections will discuss an application of OMCT to a set of classical washout algorithms. Although the
issues described in the current section were also encountered in the application, the focus will lie on discussing
typical results and ways the OMCT can help investigating the effects of some of the choices to be made when
configuring a classical washout algorithm.
VI. Results
The classical washout algorithm described in section III was implemented into the OMCT test environment from
section IV. A typical configuration of the algorithm was selected to evaluate the OMCT process. Additionally, the
effect of changing one of the design choices identified in section III, the algorithm’s reference point, was
investigated using the OMCT methodology. This section will present the results of these tests and illustrate how the
OMCT can be used to gain insight in a particular’s motion cueing algorithm’s characteristics.
The reference filter’s characteristics are summarized in Table 3. The parameters were selected for clarity and
conformance to the limitations of the SRS motion base. No claims are made as to their usefulness or fidelity in any
particular application.
As a first step, the OMCT can be used to examine one of the uncoupled axes, e.g. the yaw axis. For this
configuration, a second order behaviour should be visible with a low frequency phase lead of 180 degrees and a high
frequency gain of 0.6. Figure 2 shows the commanded and measured results in the modified Nichols plot, while
Figure 3 shows the same results as a Bode plot, which preserves the frequency information.
8
American Institute of Aeronautics and Astronautics
Table 3. Example filter parameters The shape in the figures
Gain k 2nd Order 2nd Order 1st Order 2nd Order
conforms to expectations and
HP break HP HP break LP break
also shows the influence of
frequency damping frequency frequency
the motion base hardware. At
ωn [rad/s] ζ [-] ωb [rad/s] ωLP [rad/s]
high frequencies the motion
Roll (P) 0.6 0.8 1.0 0.0 -
controller has a known
Pitch (Q) 0.7 0.8 1.0 0.0 -
tendency to overshoot and
Yaw (R) 0.6 1.0 1.0 0.0 -
suffers from phase lag due to
Surge (X) 0.7 1.0 1.0 0.0 2.0
the time delay. The delay of
Sway (Y) 0.6 1.0 1.0 0.0
the IMU mentioned in2.0
Heave (Z) 0.5 2.5 1.0 section IV.B was corrected
0.2 -
for in this figure by using the
phase lag between the yaw setpoint and the simulator yaw angle from actuator excursion measurements for
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
150
Phase distortion [deg]
100
50
-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Modulus
A more elaborate test is the surge due to surge (test 6). It shows the interaction of the low- and high-pass filters
with only limited attenuation and phase distortion in the overlap area. The commanded and measured results from
the IMU are shown in Figure 4 and Figure 5.
Again, the motion hardware’s tendency to overshoot at high frequencies, as well as the time delay in the system
are apparent from the difference between the two responses.
9
American Institute of Aeronautics and Astronautics
Yaw Rotation
0
10
-1
commanded
10
modulus
measured
-2
10
-3
10
-1 0 1 2
10 10 10 10
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
frequency [rad/s]
200
phase [deg]
100
-100
-1 0 1 2
10 10 10 10
frequency [rad/s]
40
Phase distortion [deg]
20
0
ω
-20
-40
-60
-80
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Modulus
Figure 4. Surge response, commanded and measured
10
American Institute of Aeronautics and Astronautics
Surge Translation
-0.1
10
commanded
modulus
measured
-0.8
10
-1 0 1 2
10 10 10 10
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
frequency [rad/s]
100
50
phase [deg]
-50
-100
-1 0 1 2
10 10 10 10
frequency [rad/s]
In addition to this application of the OMCT to examine the response of the algorithm, the response of the motion
hardware and the difference between the two, the OMCT can also be used to investigate some choices that can be
made in the configuration of the algorithm. As discussed in section III, the selection of the algorithm’s reference
point is an interesting topic. The remainder of this section will study this topic by using offline simulations in
MATLAB®. No actual measurements or motion hardware responses will be used and the results will be plotted in
the modified Nichols plot for clarity.
The SRS implementation of the classical washout algorithm has the option to select the algorithm’s reference
point. Two specific points are interesting in this regard: the OMCT reference point 0.35 m below the DERP, and the
Upper Gimbal Centroid (UGP), which in the SRS lies 1.2075 m below the DERP. The OMCT point has the
advantage of filtering in the same position the signal is inserted in, without any additional specific forces entering
the algorithm due to any displacements. On the other hand, the UGP has the advantage of allowing the most efficient
rotations and thereby permitting larger gains on the rotation channels.
The effect of the different points, although seemingly minor (a translation of just 85 cm), is nevertheless quite
visible. Figure 6 shows the effect on the surge due to surge test (test 6).
As can be seen, especially at high frequencies, the UGP version of the algorithm attenuates more than the OMCT
version and never reaches the selected gain of 0.6 again. The effect is even more pronounced on the pitch due to
pitch test (test 1), as shown in Figure 7.
The OMCT version shows a low phase distortion and moderate attenuation over all frequencies thanks to the tilt
coordination influence at low frequencies. However, due to the interaction between the additional longitudinal
specific force from the translation from the OMCT point to the UGP and the tilt coordination channel, the pitch
response of the UGP version is significantly amplified at high frequencies. This effect is usually not desired.
11
American Institute of Aeronautics and Astronautics
Surge due to surge - reference point
100
UGP
80 OMCT
60
Phase distortion [deg]
40
20
ω
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
-20
-40
-60
-80
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Modulus
Figure 6. Surge response, influence of algorithm reference point
20
Phase distortion [deg]
10
0
ω
-10
-20
-30
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Modulus
Figure 7. Uncoordinated pitch response, influence of algorithm reference point
12
American Institute of Aeronautics and Astronautics
Pitch due to pitch coordinated - reference point
180
ω UGP
160
OMCT
140
120
Phase distortion [deg]
100
80
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
60
40
20
-20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Modulus
Figure 8. Coordinated pitch response, influence of algorithm reference point
It can be argued that this particular effect may be an artefact of the way the pitch test is done, i.e. in an
uncoordinated fashion with additional surge specific forces due to gravity. To test this hypothesis, a further test was
done on a coordinated version of the test, as shown in Figure 8. It can be seen that also in this case the high
frequency amplification effect is present in the UGP version. Noteworthy is also the shape of the coordinated pitch
transfer function, which now much more resembles the standard high-pass characteristics already encountered in the
yaw test.
Based on these findings from the OMCT, it was decided that for the SRS the UGP will not be used for the
classical washout algorithm’s reference point. In this way the OMCT has helped provide a better insight in the
workings of a seemingly simple motion cueing algorithm with some surprisingly complex interactions. Additionally
it can be used to more comprehensively document the motion system’s configuration for a particular experiment,
beyond just the selected parameters (gains, break frequencies etc.). Together with the OMCT’s ability to investigate
the influence of the motion hardware in a standardized way, these aspects makes it a valuable addition to the toolbox
of our researchers and students interested in motion cueing.
13
American Institute of Aeronautics and Astronautics
• The implementation of OMCT into the SRS’ software architecture posed no significant problems due to
its modular nature. However, we can imagine that depending on a simulator’s software architecture,
there may be complexities and interdependencies that make a proper OMCT implementation more
difficult.
• The main difficulty of the SRS OMCT implementation lies in the use of the sensors. The presence of
additional latencies in the IMU and its location away from the OMCT reference point complicated the
analysis. For some motions (e.g. pitch) and frequencies it was found that the actually measured specific
forces at the IMU had phase shifts up to 180 degrees with respect to the reconstructed specific forces at
the OMCT point. Also, the different contributions to the specific force, such as the gravity component,
the inertial acceleration and the parasitic accelerations due to rotations, were sometimes of comparable
magnitude but heavily shifted in phase. The resulting specific force can thus become quite sensitive to
any deviations, such as latencies in the sensors.
• The interpretation of especially the main effects in terms of magnitude and phase is in line with earlier
Downloaded by TECHNISCHE UNIVERSITEIT DELFT on December 12, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2013-4834
methods by Sinacori9 and Schroeder10, who evaluated the high-pass filter part of the algorithm at 1
rad/s. Although no absolute OMCT criteria yet exist to judge a particular algorithm or configuration,
relative evaluations can still be made. The cross effects are more difficult to judge, also due to the lack
of criteria.
VIII. Conclusion
The Objective Motion Cueing Test aims to broaden the evaluation of simulator motion bases from purely robotic
hardware tests to more integrated tests which also take into account the motion cueing algorithm. By incorporating
the cueing, it provides a way to compare different motion systems and configurations. With its limitations of
uncoupled input axes and assumed linearity it cannot paint the complete picture, but it nevertheless provides a
standardized metric that is more elaborate than previous methods, e.g. the Sinacori/Schroeder criteria.
The use of OMCT in an academic research environment as described in this paper has two main aspects. First,
the metric can be used to investigate different cueing algorithms, configuration choices or hardware systems. An
example was given of how the choice of algorithm reference location impacts the resulting motions. Secondly, the
documentation of a particular simulator’s motion system and its comparison to other simulators can be done more
comprehensively. It was shown that details matter and a mere characterization of the motion system as using a
classical washout algorithm with given parameters may not be enough to properly assess and compare the results of
different simulator studies.
To enhance the OMCT’s usefulness, especially in the qualification for training area, the metric needs to be
complemented with criteria, i.e. boundaries in terms of magnitude and phase which delineate acceptable
configurations. Defining these boundaries requires extensive knowledge of the use of simulators for training, pilot
perception and pilot control behaviour, and will remain an important challenge for the future.
References
1
Anon., “Manual of Criteria for the Qualification of Flight Simulation Training Devices”, ICAO Document 9625, 3rd ed.,
2009.
2
Hosman, R. J. A. W., and Advani, S. K., “Status of the ICAO Objective Motion Cueing Test,” Flight Simulation Research,
New Frontiers Conference Proceedings, Royal Aeronautical Society, London, 2012.
3
Advani, S. K., and Hosman R. J. A. W. , “Revising Civil Simulator Standards – An Opportunity for Technological Pull,”
AIAA Modeling and Simulation Technologies Conference and Exhibit, AIAA, Washington, DC, 2006.
4
Advani, S. K., Hosman R. J. A. W., and Potter, M., “Objective motion fidelity qualification in flight training simulators,”
AIAA Modeling and Simulation Technologies Conference and Exhibit, AIAA, Washington, DC, 2007.
5
Anon., “Dynamic Characteristics of Flight Simulator Motion Systems,” AGARD Advisory Report 144, Advisory Group for
Aerospace Research and Development, North Atlantic Treaty Organization, Neuilly sur Seine, France, 1979.
6
Reid, L. D., and Nahon M. A., “Flight simulation motion-base drive algorithms: Part 1 – Development and testing of the
equations,” University of Toronto, Institute of Aerospace Studies, UTIAS Report No. 296, 1985.
7
Stroosma, O., Van Paassen, M.M., and Mulder, M., “Using the SIMONA Research Simulator for Human-Machine
Interaction Research,” AIAA Modeling and Simulation Technologies Conference and Exhibit, AIAA, Washington, DC, 2003.
8
Reid, L. D., and Nahon M. A., “Flight simulation motion-base drive algorithms: Part 2 – Selecting the system parameters,”
University of Toronto, Institute of Aerospace Studies, UTIAS Report No. 307, 1986.
9
Sinacori, J. B., “The Determination of Some Requirements for a Helicopter Research Simulation Facility,” NASA-CR-
152066, 1977.
10
Schroeder, J. A., “Helicopter Flight Simulation Motion Platform Requirements,” NASA-TP-1999-208766, 1999.
14
American Institute of Aeronautics and Astronautics