Content uploaded by Mohammad Reza Chalak Qazani
Author content
All content in this area was uploaded by Mohammad Reza Chalak Qazani on Mar 17, 2021
Content may be subject to copyright.
1
Abstract— Motion cueing algorithm (MCA) is used to
reproduce the realistic driving motion feeling for the user of the
virtual vehicle using linear acceleration and angular velocity
motion signals while respecting the physical constrains of the
simulation-based motion platform. Classical washout filter is the
most common and commercial type of the MCA because of easy
tuning, short processing time and simplicity. The classical washout
filter is a conservative method in using workspace because of the
worst-case scenario tuning technique. Also, the existing adaptive
washout filter based on time-varying cut-off frequency does not
consider the dexterity of the motion platform as considering this
factor leads to use the high acceleration areas of the platform
workspace. In this paper, an adaptive washout filter is designed
and developed based on time-varying cut-off frequency using
dexterity and direction monitoring factors to use the available high
acceleration areas of the platform for better generation of the
high-frequency part of the motion signal. The proposed adaptive
washout filter is composed of three fuzzy logic units for each mode
including longitudinal, lateral, yaw and heave and each unit has
two inputs including motion sensation error between the real
vehicle driver and user of the motion platform and the weight of
the end-effector which is a parameter based on dexterity and
current position of the end-effector. Each fuzzy logic unit has also
one output which is the selected cut-off frequency of the related
washout filter. The proposed adaptive washout filter uses the high
dexterous areas of the workspace because it considers the dexterity
of the motion platform in calculation of the end-effector weight as
the second input of the fuzzy logic units. The proposed adaptive
washout filter is validated in MATLAB/Simulink programming
language and the experimental results indicate that the proposed
adaptive washout filter performed better than existing adaptive
and classical washout filters in terms of motion sensation error
between the driver of real vehicle and the user of motion platform
as well as usage of the platform workspace.
Index Terms—adaptive motion cueing algorithm, fuzzy logic
controller, simulation-based motion platform, performance index.
I. INTRODUCTION
HE simulation-based motion platform (SBMP) are
useful research tool for studying the safety of roads, testing
new developed vehicles, and training the users of vehicles
[1-3]. Due to the constraints of the platform’s workspace, the
(Corresponding author: Moh ammad Reza Chalak Qazan i)
1,2,4,5,6 The authors are with th e Institute for Intelli gent Systems Research and
Innovation, Deakin Universit y, Geelong, VIC 3125, Australia (e-mail:
m.r.chalakqazani@gmial.c om; houshyar.asadi@deakin.edu.au;
motion signals from a real vehicle cannot be utilized straight in
the SBMP [4-7]. The motion cueing algorithm (MCA) is
utilized to reproduce motion signals of a real vehicle for the
SBMP while respecting the SBMP boundaries [8, 9]. The
significant drawback of the MCA is motion sickness caused by
the limitations of the SBMP [10].
Schmidt and Conrad [11] introduced the first MCA known
as classical washout filter. The human vestibular system [12-
14] senses the motion by detecting the special force and the
angular velocity of the SBMP via the otolith organs and
semicircle canals, respectively. Classical washout filters have
drawbacks such as constant parameters, adjustment at worst-
case scenario tunning, and disregard for the model of human
vestibular. Recently, researchers work on genetic algorithm and
swarm intelligence to optimize the tuning problem of the
classical washout filter [15-17]. This leads to generating wrong
motion cues, and higher motion sensed error between the real
vehicle and SBMP users.
Due to the shortcomings of the existing classical washout
filter, adaptive washout filter was introduced by Parrish et al.
[18] for a flight SBMP. In order to eliminate the wrong motion
cues, Ariel and Sivan [19] used an adaptive washout filter in
real-time utilizing of the human vestibular model. Later,
researchers focused more on the adaptive washout filter
because of its flexibility and better performance [20, 21]. Later,
the adaptive washout filters using fuzzy logic have been
developed to address the issues of the existing classical and
adaptive washout filters [22-26]. The fuzzy based-adaptive
washout filters can be divided in two groups including time-
varying cut-off frequency and motion compensator unit. The
first adaptive washout filter based on fuzzy logic was proposed
by Song et al. [22] to change the classical washout filter fixed
parameters in accordance with the margins of the workspace
and various situations of driving scenarios. Asadi et al. [23, 24]
modified the previous fuzzy based-adaptive washout filter [22]
for changing the cut-off frequencies by considering the model
of human vestibular and sensation error to enhance the motion
fidelity for the SBMP user. Qazani et al. [27] improved, Asadi
proposed work [23, 24], with consideration of SBMP’s
limitations along passive and active joints coordinate system
rather than end-effector coordinate system. Later, Asadi et al.
[25, 26] developed another generation of adaptive washout
shadym@deakin.edu.au; chee.lim@deakin.edu.au;
saeid.nahavandi@deak in.edu.au).
3Department of Computer Engineering, University of Kurdistan, Sanandaj, Iran (e-
mail: m.rostami@eng .uok.ac.ir)
Mohammad Reza Chalak Qazani
1
, Houshyar Asadi
2
, Member, IEEE, Mehrdad Rostami
3
, Shady
Mohamed4, Chee Peng Lim5, Saeid Nahavandi6, Fellow, IEEE
Adaptive Motion Cueing Algorithm Based on
Fuzzy Logic Using Online Dexterity and
Direction Monitoring
T
2
filter based on fuzzy logic-based motion compensator units
while considering the workspace limitations as well as motion
sensed motion error between the real driver of vehicle and the
SBMP user to generate motion correction signals when
required. Qazani et al. [28] improved, Asadi proposed work [25,
26], with consideration of active joints limitations instead of
end-effector’s limitations. Moreover, Qazani et al. [29] used the
concept of motion compensation generator unit to vary the
SBMP’s neutral position for increasing the linear workspace
limitation of the SBMP. However, the recent proposed adaptive
washout filters [18-29] are able to reduce the sensed motion
error between the driver of the real vehicle and the SBMP user,
they are not able to monitor the online dexterity. The area of the
platform workspace with ability of higher linear acceleration
generation can be achieved via monitoring the dexterity in order
to regenerate the highly efficient high frequency motion signal
for the user of the SBMP.
It is useful to use the areas of the workspace with higher
performance availability. Performance analysis is one of the
critical topics for researchers since the generation of
manipulators to evaluate the motion platform capability.
Dexterity is a value between 0-1 which shows the areas of the
workspace with higher accessible linear acceleration for the
end-effector using the less actuator effort [30]. Paul and
Stevenson [31] introduced the dexterity measurement.
Yoshikawa [32] defined the square root of the Jacobin matrix
to find the manipulability of the robots. Pedrammehr et al. [33-
35] and Qazani et al. [36, 37] used the dexterity value in order
to evaluate the performance of the parallel SBMPs such as
Stewart, Hexarot and Gantry-tau manipulators [38-40]. In Fig.
1.a-b, the Steward platform workspace with the most massive
inscribed cubic box is shown in isometric- and top-view,
respectively. The boundary of this cubic cox is utilized as a
maximum value of excursion in x-, y- and z-axis.
In this study, the fuzzy logic units are designed and
developed while considering the dexterity and motion sensation
error as inputs of the fuzzy units in order to generate the
appropriate cut-off frequencies for the washout filters. The
dexterity is employed to assess the motion platform
performance and use the higher acceleration available areas of
the platform workspace. Also, the direction detector is proposed
in this study for better usage of actuator acceleration
limitations. In addition, the SBMP is a Steward platform with
±30-centimetre linear displacement in x-axis, ±15-centimetre in
the y-axis and ±2-centimetre in the z-axis. The end-effector is
capable of moving 15 degrees around x- and y-axis and 20
degrees around z-axis while located in the neutral position.
The methodology of the developed adaptive washout filter
using fuzzy logic is described in Section II. Modeling the
SBMP SimMechanic, that is implemented in
MATLAB/Simulink programming language by considering
classical, previous adaptive and developed adaptive washout
filters are presented in Section III. Moreover, this section
provides and discusses the results for those of other washout
filters and developed adaptive washout filter. The concluding
remarks are presented in Section IV.
II. METHODOLOGY
Zadeh [41] introduced the fuzzy logic theory, and it has
been developed in control systems. Fuzzy theory based models
are one of the powerful methods for the stochastic behaviour-
based system and changeable inputs [42, 43]. Fuzzy is highly
applicable for systems when the input-output relation is not
identified [42]. The classical washout filter is composed of
washout filters using fixed parameters, including cut-off
frequency and damping ratio, which are adjusted regarding the
worst-case scenarios. Table I shows different fixed parameters
of the classical washout filter along four modes. The cut-off
frequencies of the proposed adaptive washout filters are
adjusted based on the dexterity of the end-effector in addition
to sensed motion error between the driver of real vehicle and
SBMP user to overcome the disadvantage of insufficient use of
manipulator workspace and reducing the human sensation
error, respectively. Therefore, in this paper, fuzzy theory is used
to maximize the motion fidelity of the driving SBMP.
The developed adaptive washout filter structure based on
dexterity and motion sensation error is demonstrated in Fig. 2.
In this project, the control block for varying cut-off frequency
in high- and low-pass filters has two signals that are considered
as input. These signals will be discussed in Section III.C.
A. Human Vestibular Model
Human vestibular model is the part of the human perception
system, and it is the fast sensor to sense the motion signal.
Otolith limbs and semicircle canals, that form the human
vestibular model, cause a sense of linear acceleration and
angular velocity, respectively. The organs of the otolith feel the
special force by decreasing the gravity from the linear
acceleration as follows:
=− (1)
where , and indicate special force, linear acceleration and
gravity, respectively. By considering as a seat’s rotation
matrix, the sensed linear acceleration in the coordinate system
at the centre of the seat is calculated as [12]:
=−(,,)=−
−
− (2)
where , and displays the roll-, pitch- and yaw-angle of the
end-effector around x-, y- and z-axis. In studies by Telban and
Cardullo [12] and Asadi et al. [13], the transition functions of
otolith limbs are reviewed and evaluated. In this paper, a
transition function with the best sensed special force stimulant,
, over the applied force, , is chosen that defined as follows:
Fig. 1. The most m assive inscribed cubic box inside
the Steward platform
workspace: (a) ISO-Metric V iew; (b) Top View.
3
=
()() (3)
where , and are long time fixed, the short time fixed of
the otolith limbs and neural processing term lead operator and,
= 5.3 seconds in the x-axis, = 5.33 seconds in y- and z-
axis, = 0.016 seconds and = 13.2 seconds in three
directions. The semicircle system transfer function to calculate
the sensed angular velocity,
, based on applied angular
velocity, , is as [12, 14]:
=
()() (4)
where and are a long time fixed and adaption time fixed,
and = 5.3 seconds in the x-axis, = 6.1 seconds in the y-
axis, = 10.2 seconds in the z-axis, and = 30 seconds in
three directions
B. Dexterity Analysis
The usage of the higher acceleration available area of the
platform workspace is the main concept of this study in order
to regenerate high-fidelity linear acceleration signal.
Nevertheless, the Steward platform is limited because of
passive joints rotation and leg length limitations [4]. It can stop
motion during the simulation if one of the joints moves to the
boundaries. Normally, the simulators’ engineers imagine the
SBMP workspace as a cube space with fixed values and
imitations in x-, y-, and z-axis as linear motion, and roll-, pitch-
, and yaw-angle as angular motion. It leads to losing the high-
performance area of workspace for every parallel and serial
manipulator. Furthermore, the maximum angular motion of the
manipulator depends on the central position of the end-effector
[4]. For instance, the angular motion will be maximum when
the end-effector is in its neutral position. If the manipulator is
closer to the workspace boundaries, the angular motion will be
reduced, consequently. Then, this simplified assumption is not
worthwhile anymore. In this study, dexterity is used as a new
technique in the MCA domain to monitor the end-effector
dexterity. It is used as one of the fuzzy block inputs to change
the value of cut-off frequency to generate suitable SBMP’s
motion according to the current SBMP’s situation. The
dexterity of the Stewart manipulator calculated as follows [7]:
=
‖‖‖‖ (5)
where is a Steward platform Jacobin matrix that defines the
relation between legs and the end-effector velocity. It can be
found based on the inverse kinematic problem of the
manipulator [44]. The norm is defined as:
‖‖=trace
(6)
There are multiple reasons to decrease the dexterity.
Singularity is the important crucial cause of the dramatic
decrease in the dexterity. The developed adaptive washout filter
never moves to the singular point. The fuzzy units do not let the
dexterity to reach zero which means the singular position. It
changes the inputs to enhance the dexterity and avoid the
motion platform to reach the singularity. Also, it should be
noted that, a slight reduction of dexterity has two hints for the
SBMP. Firstly, it might be close to the platform workspace
boundaries because of prismatic joints limitations. Secondly, if
this happens while the platform of the Steward platform is well
within the workspace boundaries, it proves the platform
orientation near the end of the rotational limitations.
Unfortunately, the dexterity is a scalar parameter that makes
this value useless for implementing the 6-DoF of the
manipulator. The weight parameters of the end-effector have
been defined to show the share of every DoF for every
configuration of the manipulator as follows:
,,
=,,
(7.a)
,,
=,,
(7.b)
where , and are the end-effector linear displacements
along x-, y- and z-axis. Also, , and are the angular
displacements of the end-effector along x-, y- and z-axis. Eq.
(7a-b) is the initial displacement weights of the end-effector
without implementing the dexterity. The final displacement
weights of the end-effector are defined by multiplication of the
Eq. (7a-b) to the dexterity as follows:
,,=.,,
(8.a)
,,=.,,
(8.b)
C. Fuzzy Logic Control Unit
In this study, Mamdani fuzzy blocks and membership
functions are selected to decrease sensed special force. As the
MCA inputs are nonlinear and unpredictable motion signals, the
usage of the Mamdani-type fuzzy gives the better results
compared with Sugeno-type fuzzy [45]. The proposed fuzzy-
based controller receives two inputs and generates one output.
The first fuzzy logic controller input is the weight parameter
relates to the dexterity parameter, and the second input is the
sensed motion error between the driver of real vehicle and the
SBMP user. The output of the fuzzy logic controller is the cut-
Fig. 2. The longitudinal structure of developed adaptive washout filter.
TABLE I
THE TUNING PARAMETERS OF THE CLASSICAL WASHOUT FILTER.
Index
Linear Acceleration Angular
Acceleration
High-pass
Translation
Low-pass Tilt-
coordination
High-pass
Rotational
Longitudinal
2.75
2.00
4.50
1.00
2.00
Lateral
2.75
2.00
4.50
1.00
2.00
Heave 4.25 2.00 - - -
Yaw
-
-
-
-
2.5
4
off frequency of the washout filters. The input weight parameter
of the dexterity is partitioned into five membership functions
such as very near (VN), near (N), middle (M), far (F) and very
far (VF) (Fig. 3.a). Also, the membership function of the sensed
motion error between the driver of real vehicle and SBMP user
is categorized as five sensation categories such as very negative
(VN), negative (N), zero (Z), positive (P) and very positive
(VP) (Fig. 3.b). The system of human vestibular cannot
precisely determine motion quantities; then this hypothesis is
worthwhile. The membership function of the frequency outputs
changing units is shown in Fig. 3.c-d which varies between 0.5
to 2.5 rad/sec for high-pass filter and between 4.25 to 6.25 for
the low-pass filter. They consist of five membership functions
such as very small (VS), small (S), medium (M), big (B) and
very big (VB). The triangular membership functions are chosen
as the inputs and outputs of the fuzzy module to achieve light
computational load. It should be noted that the centroid
defuzzification method is used in this study which is the most
popular method in MATLAB fuzzy Toolbox.
Fuzzy rules of the algorithm are defined to defeat the
weakness of fixed parameters and help the Steward platform to
make more efficient use of workspace for different driving
scenarios. Fig. 4.a-b demonstrate fuzzy logic rules for high- and
low-pass filters of translational and tilt coordination channels,
respectively. It is obvious that the rules of a low-pass filter are
inverse of the high-pass filter. If the manipulator is near to its
neutral position with decreasing the high-pass frequency and
raising the low-pass frequency improves the feeling of better
motion in the workspace.
D. Direction Detector Unit
The direction detector has been proposed for adaptive
washout filter for the first time in this study. Fig. 5a shows the
structure of the direction detector unit. The direction detector
unit is utilized to extract the directional motion of the end-
effector (positive or negative side) by saving the previous signal
and comparing the sign of the last and current signal with
the end-effector position. If the input signal sign and the end-
effector position are the same, it proves that the manipulator is
approaching the boundary. Otherwise, it is escaping from it. For
instance, the manipulator is on the positive direction of the x-
axis, and the input signal is positive, the manipulator moves
towards the positive side of the x-axis to get close to the
boundary. Then, the final weight factor of the end-effector,
which is found in Eq. (8.a-b) should be updated for the same
sign on the signal and end-effector position as follows:
,,=,,
+ 0.5 (9.a)
,,=,,
+ 0.5 (9.b)
The 0.5 value is added to the weight coefficient because the
manipulator will escape from going to the boundary and have
more space to exploit. Also, if the sign of the input signal and
the position of the upper platform position is reversed, the
weight factor should be updated as below:
,,=,,
(10.a)
,,=,,
(10.b)
Then, if the Steward platform is near to the workspace
limitations and the signal commands to get away the boundary,
there is no need to reduce the cut-off frequency with using this
technique. It produces a better motion signal with low sensed
motion error. Also, this assumption can lead to better usage of
the actuator speed limit to have the best motion sensation during
the simulations.
III. RESULTS AND DISCUSSION S
This section is provided in two subsection which is used to
verify the model using MATLAB software, validate the
developed adaptive washout filter with classical and pervious
adaptive washout filters developed by Schmidt and Conrad [11]
and Asadi et al. [23], respectively.
Fig. 4. The surface plot of the fuzzy logic rules for: (a) high-pass filter; (b) low
-
pass filter.
Fig. 5. The schematic structure of: (a) Direction detector unit; (b) Simulink
model of the developed adaptive washout filter.
Fig. 3.
The memberships function of: (a) Input weight factor; (b) input sensed
special force error; (c) ou tput translational; (d) output rotational.
5
A. Verification
In experiments, the Simulink model of the Steward platform
using washout filters has been designed and developed (Fig.
5.b). Initially, the Rigs of Rods (RoR) soft body physics engine
is utilized to simulate the vehicle environment (version 0.39.5)
(http://www.rigsofrods.com). Then, the outputs of RoR are
considered as input for the Simulink file. Then, motion signals
are produced in the MCA blocks. The inputs of the developed
adaptive fuzzy-based washout filter are dexterity and special
sensation force error and sensed angular velocity error by the
model block of human vestibular for both of the real vehicle
driver and the user of the SBMP.
The regenerated motion signals are imported to the Steward
platform SimMechanic model block. This block consists of leg
trajectory, PID controller, Steward platform and performance
analysis sub-blocks. Besides, there is a plant sub-block of the
Steward platform that is modelled in SimMechanic of
MATLAB software. The outputs of detectors import to the
performance analysis subplot to calculate the dexterity. This
sub-block also monitors the limitations of SBMP’s legs and
passive joints. The special sensation force and angular sensation
velocity are modelled using the transition function of the otolith
limbs and semicircle canal.
B. Validation
Figure 6.a-b shows the motion signals that are input of the
Simulink file for 25 seconds. The vehicle starts moving at 1.51
seconds of the scenario and stops moving at 24 seconds of
driving. The linear acceleration and angular velocity in three
directions are shown in Fig. 6.a-b. The scale factor is 0.23 for
linear acceleration signals and 1 for angular velocity signals to
keep the platform inside the linear and angular limitation and
allow MCAs to improve signal production. The scale factor
should be the same for classical, previous adaptive and
developed adaptive washout filters to reach the fair comparison
of MCAs [46, 47].
Recorded driving scenario results in the RoR for the
classical, previous and developed adaptive washout filters are
listed in Table II. Root mean square error (RMSE) presents the
sensed motion error between the real vehicle driver and the
Steward platform user. The reported results show that the error
of sensed motion using the proposed method is less than other
methods due to the consideration of the human vestibular model
as well as the wise usage of platform’s workspace. Another
section of Table II displays the correlation coefficient (CC) to
demonstrate the shape similarity of the SBMP user sensed
motion and the real vehicle driver motion sensation. CC is
introduced in MCA by Asadi et al. [48, 49] and it can vary
between 0 to 1 while 1 shows the highest fidelity regeneration
of the motion signals compared with the driver of the real
vehicle.
Fig. 7.a-c displays the sensed angular velocity for the driver
of real vehicle and the user of the SBMP using the classical,
previous adaptive and developed adaptive washout filters in the
x-, y- and z direction. The sensed angular velocity between the
driver of real vehicle and the user of the SBMP is much more
dependable with lower wrong motion cues using the developed
adaptive washout filter compared with pervious and classical
washout filters based on Fig. 7.a-c. Also, the error of sensed
Fig. 7. The sen sed angular velocity and sensed special force
between the real
vehicle driver and the SBMP user using classical, prev ious adaptive and
developed adaptive washout filters along: (a) roll-angle; (b) pitch-
angle; (c)
yaw
-angle; (d) x-axis; (e) y-axis; (e) z-axis.
Fig. 6. The reference signal (real vehicle motions): (a) linear acceleration; (b)
angular velocity.
6
angular velocity is under the human sensed angular threshold
unit because, the tilt coordination channel does not generate the
signals beyond the tilt rate limit. There is no significant
difference between the values of sensed angular velocity error.
However, the developed adaptive washout filter reduces the
error of sensed angular velocity between the driver of real
vehicle and the user of the SBMP 13.45%, 19.06% and 18.13%
compared with classical washout filter along x-, y- and z-axis.
Moreover, the developed adaptive washout filter decreases the
error of sensed angular velocity between the driver of real
vehicle and the user of the SBMP 15.26%, 17.95% and 13.98%
compared with previous adaptive washout filter along x-, y- and
z-axis. Fig. 7.d-f demonstrates the sensed special force of the
driver of real vehicle and the user of the SBMP using the
classical, previous adaptive and developed adaptive washout
filters in the x-, y- and z-direction. The developed adaptive
washout filter is able to accurately pursue the reference signal
shape. The CC of the sensed special force using the developed
adaptive washout filter increases 42.30%, 50.17% and 45.87%
compared with classical washout filter along x-, y- and z-
direction, correspondingly. In addition, the developed adaptive
washout filter increases the sensed special force 25.00%,
22.73% and 18.80% compared with previous adaptive washout
filter along x-, y- and z-direction, correspondingly. The
maximum value of the sensed special force error for
longitudinal movement is lower than 1.2 m/s2 through the
developed method, while it is 1.8 and 2 m/s2 for the previous
adaptive and classical washout filters, respectively.
Fig. 8.a-f shows the SBMP linear and angular displacement
of the end-effector in x-, y- and z-axis, respectively, using
classical, previous adaptive and developed adaptive washout
filters. The Steward platform with developed adaptive washout
filter moves more than classical washout filter due to the
structure of proposed adaptive washout filter. The cut-off
frequency of the high-pass filter is flexible based on the end-
effector configuration in the proposed adaptive washout filter.
Then, it can decrease the high-pass filter and use the larger
workspace compared with classical washout filter. The
proposed method changes the cut-off frequency to compensate
motion sensation error by taking the online performance
analysis of the platform into account. It should be noted that the
chance of reaching the physical limitation is a bit higher for the
proposed washout filter compared to classical washout filter.
Also, compared to the previous adaptive washout filter, the risk
of achieving physical limitation is either less or equal. It is due
to the cubic box supposition of the workspace in this method.
Fig. 9 shows the variation of the dexterity as a performance
analysis of the Steward platform. Our proposed adaptive
washout filter's dexterity is close to the classical washout filter
that has been adjusted based on the worst-case scenario. The
previous adaptive washout filter has a slight dexterity reduction
that shows the higher risk of reaching to the manipulator
limitations or singularity. The dexterity of the previous adaptive
washout filter at t=13.01 second is 0.6372, while the dexterity
of the proposed adaptive washout filter is 0.7238. It leads to
better performance and higher motion fidelity as it is shown in
Fig. 7.
TABLE II
THE RESULT OF THE CLASSICAL, PREVIOUS ADAPTIVE AND DEVELOPED
ADAPTIVE WASHOUT FILTERS.
Index
RMSE
CC
CL WF PRE
WF
PRO
WF
CL WF PRE
WF
PRO
WF
SSFx 1.2496
(m/s
2
)
0.9613
(m/s
2
)
0.7209
(m/s
2
)
0.5663 0.7403 0.8743
SSFy 1.3892
(m/s2)
0.8957
(m/s2)
0.6921
(m/s2) 0.5426 0.7325 0.8634
SSFz 0.1284
(m/s
2
)
0.0856
(m/s
2
)
0.0695
(m/s
2
)
0.0095 0.0152 0.0187
SAVx 7.4876
(deg/s)
7.6475
(deg/s)
6.4799
(deg/s)
0.4184 0.5272 0.7925
SAVy 7.8657
(deg/s)
7.7585
(deg/s)
6.3658
(deg/s) 0.3986 0.5103 0.7648
SAVz 5.2387
(deg/s)
4.9825
(deg/s)
4.2857
(deg/s)
0.3421 0.4448 0.5128
RMSE: Root Mean Square Error; CC: Correlation Coefficient;
CL WF:
Classical washout filter; PRE WF: Previous adaptive washout filter; PRO:
Developed adaptive washou t filter; SSF: Sensed special force
; SAV: Sensed
angular velocity;
Fig. 8. Linear and ang ular displacement of SBMP using classical, p revious
adaptive and developed adap tive washout filters along: (a) x direction; (b) y
direction; (c) z direction ; (d) roll-angle; (e) pitch-angle; (f) yaw-angle.
7
Finally, the parking motion scenario (sudden accelerations or
deaccelerations) is implemented to the SBMP in order to show
the relatability and repeatability of the proposed adaptive
washout filter over the classical and pervious adaptive washout
filters. Fig. 10.a presents the sensed special force for the driver
of real vehicle and the user of the SBMP using the classical,
previous adaptive and developed adaptive washout filters along
longitudinal mode. The shape similarity of the sensed special
force between the driver of real vehicle and the user of the
SBMP using developed adaptive washout filter is 0.9837, while
it is 0.8651 and 0.6502 using previous adaptive and classical
washout filter. It proves the better generation of the motion cues
using the developed adaptive washout filter with higher fidelity
over the previous adaptive and classical washout filters.
Moreover, the RMSE of the sensed special force between the
driver of real vehicle and the user of the SBMP using developed
adaptive, previous adaptive and classical washout filters are
0.2728, 0.1762 and 0.0660 m/s2. It proves the regenerations of
the less false motion cues using the developed adaptive washout
filter contrasted with the classical and previous adaptive
washout filters. Also, Fig. 10.b shows the linear motion of the
end-effector along x-axis using three investigated methods. As
it is obvious, the developed adaptive washout filter is able to
utilize the SBMP’s workspace more wisely compared with
classical and previous adaptive washout filters.
IV. CONCLUSIONS
In this paper, an adaptive washout filter based on time-
varying cut-off frequency using fuzzy logic units is developed
for a typical Stewart platform in order to use higher dexterous
parts of the workspace. The classical washout filter is a
commercial type of MCA because of simplicity, but it has some
disadvantages including fixed parameters, conservative
motions, and lack of considering human perception model in
the design which reduces motion fidelity of the platform. Also,
the previous adaptive washout filter [23] with consideration of
the sensed motion error between the driver of real vehicle and
SBMP user as well as the end-effector limitations is not able to
use the higher dexterous part of the workspace. The developed
adaptive washout filter uses the sensed motion error between
the driver of real vehicle and SBMP user and the dexterity value
to vary the cut-off frequency of the washout filters. The
dexterity of the end-effector is monitored in this study to use
the area of the platform workspace with ability of generating
higher linear acceleration signals. The performance analysis
factor is used as a physical limitation parameter of the SBMP
for the first time in the MCA domain to use the workspace of
platform efficiently via better regeneration of the high
frequency motion signal. Moreover, the direction detector has
been used for the first time in designing adaptive washout filter
to update the value of the weight parameter (which is a
parameter based on dexterity and current position of the end-
effector) to use the maximum speed of the actuator. It helps to
feel a better sensation of the movement in the SBMP. The
Stewart platform is implemented using Simulink in MATLAB
programming language. The experimental results showed that
the proposed adaptive washout filter reduced the sensed motion
error between the driver of real vehicle and the SBMP user
compared to the classic washout filter with fixed parameters
and even the previous adaptive washout filter. Moreover, the
reported results also showed an improvement in term of sensed
shape following factor. As a future study, the robust adaptive
fuzzy [50] with implementation of the external disturbances and
strict-feedback can be employed. Also, the idea of using neural
network [51-56] and supervised learning methods can be used
to design the high efficient MCA using the regenerated results
of the proposed adaptive washout filter with low computational
load and more efficiency.
REFERENCES
[1] M. R. C. Qazani, H. Asadi, S. Khoo, and S. Nahavandi, "A Linear
Time-Varying Model Predictive Control-Based Motion Cueing
Algorithm for Hexapod Simulation-Based Motion Platform," IEEE
Transactions on Systems, Man, and Cybernetics: Systems, 2019.
[2] H. Asadi, S. Mohamed, C. P. Lim, and S. Nahavandi, "Robust
optimal motion cueing algorithm b ased on the line ar quadratic
regulator method and a genetic algorithm," IEEE Transactions on
Systems, Man, and Cybernetics: S ystems, vol. 47, no. 2, pp. 238-
254, 2016.
[3] M. R. C. Qazani, H. Asadi, and S. Nahava ndi, "A decoupled linear
model predictive control-based motion cueing algorithm for
simulation-based motion platform with limitted workspace," in
2019 IEEE International Conference on Industrial Technology
(ICIT), 2019: IEEE, pp. 35 -41.
[4] J.-P. Merlet, Para llel robo ts. Springer Science & Business Media,
2006.
[5] M. R. C. Qazani, H. Asadi, and S. Nahavandi, "High-fidelity hexarot
simulation-based motion platform using fuzzy incremental
controller and model predictive control-based mo tion cueing
algorithm," IEEE Systems Journal, 2019.
[6] S. Pedrammehr, M. R. C. Qazani, H. Asadi, and S. Nahavandi,
"Control System Development of a Hexarot-Based High-G
Centrifugal Simulator," in IC IT, 2019, pp. 78-83.
Fig. 10. (a): The sensed special force between t
he real vehicle driver and the
SBMP user using classical, previous adaptive and developed adaptive washout
filters along x-
axis; (b): Linear displacement of SBMP using classical, previous
adaptive and developed adaptive washout filters along x-direction.
Fig. 9. Performance analysis of the SBMP using classical, previous
and
developed adaptive fuzzy logic-based washout filters.
8
[7] J.-P. Merlet, "Jacobian, m anipulability, condition number, and
accuracy of p arallel robots," J ournal of Mechanica l Design, vol.
128, no. 1, pp. 199-206, 2006.
[8] S. Casas, C. Portalés, J. Gimeno, and M. Fernández, "Simulation of
parallel mechanisms for motion cueing generation in vehicle
simulators using AM-FM bi-modulated signals," Mechatronics, vol.
53, pp. 251-261, 2018.
[9] M. R. C. Qazani, H. A sadi, and S. Nahavandi, "A Motion Cueing
Algorithm Based on Model Predictive Control Using Terminal
Conditions in Urban Driving Scenario," IEEE Systems Journal,
2020.
[10] J. Iskander et al., "From car sickness to autonomous car sickness: A
review," Transportation research part F: traffic psychology and
behaviour, vol. 62, pp. 716-726, 2019.
[11] S. Schmidt and B. Conrad, "The calculation of motion drive signals
for piloted flight simulators," Palo Alto, CA, USA: NASA, 1969.
[12] R. J. Telban and F. M. Cardullo, "Motion cueing algorithm
development: Human-centered linear and nonlinear approaches,"
NASA TechReport CR-2005-213747, 2005 .
[13] H. Asadi, S. Mohamed, C. P. Lim, and S. Nahavan di, "A review on
otolith mod els in human perception," Behavioural brain research,
vol. 309, pp. 67-76, 2016.
[14] H. Asadi, S. Moham ed, C. P. Lim, S. Nahavandi, and E. Nalivaiko,
"Semicircular canal modeling in human perception," Reviews in the
Neurosciences, vol. 28, no. 5, pp. 537-549, 2017.
[15] S. Casas, C. Portalés, P. Morillo, and M. Fernández, "A particle
swarm approach for tuning washout algorithms in vehicle
simulators," Applied Soft Computing, vol. 68, pp . 125-135, 2018.
[16] H. Asadi, S. Mohamed, K. Nelson, S. Nahavandi, and D. R. Zadeh,
"Human perception-based washout filtering using genetic
algorithm," in International Conference on Neural Information
Processing, 2015: Springer, pp. 401-411.
[17] H. Asadi, S. Mohamed, K. Nelson, S. Nahavandi, and M.
Oladazimi, "An optimal wa shout filter ba sed on genetic algo rithm
compensators for improving simulator driver perception," in DSC
2015: Proceedings of the Driving Simulation Conference &
Exhibition, 2015: Max Planck Institute for the Advancement of
Science, pp. 1-10.
[18] R. V. Parrish, J. E. Dieudonne , and D. J. Martin Jr, "Coordinated
adaptive washout for motion simulators," Journal of aircraft, vol.
12, no. 1, pp. 44-50, 1975.
[19] D. Ariel and R. Sivan, "False cue reduction in moving flight
simulators," IEEE Transactions on Systems, Ma n, and Cybernetics,
no. 4, pp. 665-671, 1984.
[20] L. REID and M. NAHON, "Response of airline pilots to variations
in flight simulato r motion algorithms," Journal of Aircraft, vo l. 25,
no. 7, pp. 639-646, 1988.
[21] L. Nehaoua, H. Arioui, S. Espie, and H. Mohellebi, "Motion cueing
algorithms for small driving simulator," in Robotics and
Automation, 2006. ICRA 2006. Proceedings 2006 IEEE
International Confe rence on, 2006: IEEE, pp. 3189 -3194.
[22] J.-B. Song, U.-J. Jung, and H.-D. Ko, "Washout algorithm with
fuzzy-based tuning for a mo tion simulator," Journal of Mechanical
Science and Technology, vol. 17, no. 2, pp. 221 -229, 2003.
[23] H. Asadi, A. Mohammadi, S. Mohamed, D. R. Zadeh, and S.
Nahavandi, "Adaptive washout algorithm based fuzzy tuning for
improving human perception," in International C onference on
Neural Information Processing, 2014: Springer, pp. 483-492.
[24] H. Asadi, A. Mohammadi, S. Mohamed, and S. Nahavandi,
"Adaptive translational cueing motion algorithm using fuzzy based
tilt coordination," in International Conference on Neural
Information Processing, 2014: Springer, pp. 474-482.
[25] H. Asadi, S. Mohamed, and S. Nahavandi, "Incorporating human
perception with the motion washout filter using fuzzy logic control,"
IEEE/ASME Transactions on Mechatronics, vol. 20, no. 6, pp.
3276-3284, 2015 .
[26] H. Asadi, C. P. Lim, S. Mohamed, D. Nahavandi, and S. Nahavandi,
"Increasing motion fidelity in driving simulators using a fuzzy-
based washout filter," IEEE Transactions on Intelligent Vehicles,
vol. 4, no. 2, pp. 298-308, 2019.
[27] M. R. C. Qazani, H. Asadi, T. Bellmann, S. Mohamed, C. P. Lim,
and S. Nahavandi, "Adaptive Washout Filter Based on Fuzzy Logic
for a Motion Simulation Platform with Consideration of Joints
Limitations," IEEE Transac tions on Vehicula r Technology, 2020.
[28] M. R. C. Qazani, H. Asadi, T. Bellmann, S. Perdrammehr, S.
Mohamed, and S. Nahavandi, "A New Fuzzy Logic Based Adaptive
Motion Cueing Algorithm Using Parallel Simulation-Based Motion
Platform," in 2020 IEEE International Conference on Fuzzy Systems
(FUZZ-IEEE), 2020: IEEE, pp. 1-8.
[29] M. R. C. Qazani, H. Asadi, S. Mohamed, and S. Nahavandi,
"Prepositioning of a Land Vehicle Simulation-Based Motion
Platform Using Fu zzy Logic and Neural Network," IEEE
Transactions on Vehicular Technology, vol. 69, no. 10, p p. 10446-
10456, 2020.
[30] K. Gupta and B. Roth, "Design considerations for manipulator
workspace," Journal of Mechanical Design, vol. 104, no. 4, pp. 704-
711, 1982.
[31] R. P. Paul and C. N. Stevenson, "Kin ematics of robot wrists," The
International journal of robotics research, vol. 2, no. 1, pp. 31-38,
1983.
[32] T. Yoshikawa, " Manipulability of robotic m echanisms," The
international journal o f Robotics Research, vol. 4, no. 2, pp. 3-9,
1985.
[33] S. Pedrammehr, M. R. C. Qazani, H. Asadi, and S. Nahavandi,
"Control system development of a hexarotbased high-G cen trifugal
simulator," in The 20th IEEE International Conference on Industrial
Technology IEEE-ICIT 2019, 2019.
[34] S. Pedrammehr, M. R. C. Qazani, H. Asadi, and S. Nahavandi,
"Manipulability Analysis of Gantry-Tau parallel manipulator," in
ACRA 2019: Proceedings of the Australasian Conference on
Robotics and Autom ation, 2019 : [Australian Robotics &
Automation Association], pp. 1-7.
[35] S. Pedrammehr, M. R. C. Qazani, H. Asadi, and S. Nahavandi,
"Kinematic Manipulability Analysis of Hexaro t Simulators," in
ICIT, 2019, pp. 13 3-138.
[36] M. R. C. Qazani, S. Pedrammehr, H. Abdi, and S. Nahav andi,
"Performance evaluation and calibration of gantry-tau parallel
mechanism," Iranian Journal of Science and Technology,
Transactions of Mechanical Engineering, pp. 1-15, 2019.
[37] M. R. C. Qazani, H. Asadi, S. Pedrammehr, and S. Nahavand i,
"Performance analysis and dex terity monitoring of hexapod-based
simulator," in 2018 4th International Conference on Control,
Automation and Robotics (ICCAR), 2018: IEEE, pp. 226-231.
[38] M. R. C. Qazan i, H. Asadi, and S. Nahavandi, "A new g antry-tau-
based mechanism using spherical wrist and model predictive
control-based motion cueing algorithm," Robotica, vol. 38, no. 8,
pp. 1359-1380, 2020.
[39] M. J. Tajari l, S. Pedrammehr, M. R. C. Qazani, and M. J. Nategh,
"The effects of joint clearance on the kinematic error of the hexapod
tables," in 2 017 5th RSI In ternational Conference on Robotics and
Mechatronics (ICRoM), 2017: IEEE, pp. 39-44.
[40] M. R. C. Qazani, V. Mohammadi, H. Asadi, S. Mohamed, and S.
Nahavandi, "Developm ent of G antry-Tau-3R Mechanism Using a
Neuro PID Controller," in ACRA 2019 : Proce edings of the
Australasian Conference o n Robotics and Auto mation, 201 9:
[Australian Robotics & Automation Association], pp. 1-8.
[41] L. A. Zadeh, Fuzzy logic: advanced concepts and structures. IEEE
Educational Activities Department, 1992.
[42] J.-B. Song, U.-J. Jung, and H.-D. Ko, "Washout algorithm with
fuzzy-based tuning for a motion simulator," KSME internation al
journal, vol. 17, no. 2, pp. 221-229, 2003.
[43] H. Asadi, S. H. A . Kaboli, A. Mohammadi, and M. Oladazimi,
"Fuzzy-control-ba sed five-step Li-ion battery charger by using AC
impedance technique," in Fourth international conference on
machine vision (ICMV 2011): machine vision, image processing,
and pattern analysis, 2012, vol. 8349: International Society for
Optics and Photon ics, p. 834939.
[44] M. R. C. Qazani, H . Asadi, S. Mohamed, and S. Nahavandi, "An
Inverse Kine matic-based Model Predictive Motion Cueing
Algorithm for a 6-DoF Gantry-Tau Mechanism," in ACRA 2019:
Proceedings o f the Australasian Conference on Robotics and
Automation, 2019: [Australian Robotics & Automation
Association], pp. 1-9.
[45] H. Ying, Y. Ding, S. Li, and S. Shao, "Comparison of necessary
conditions for typical Tak agi-Sugeno and Mamdani fuzzy systems
as universal approximators," IEEE Transactions on S ystems, Man,
and Cybernetics-Part A: Systems and H umans, vol. 29, no. 5, pp.
508-514, 1999.
9
[46] H. Asadi, C. P. Lim, A. Mohammadi, S. Mohamed, S. Naha vandi,
and L. Shanmugam, "A genetic algorithm–based nonlinear scaling
method for optimal motio n cueing algorithm in driving simulator,"
Proceedings of the Institution of Mechanical Engineers, Part I:
Journal of Systems and Control Engineering, vol. 232, no. 8, pp.
1025-1038, 2018 .
[47] H. Asadi et al., "A Model Predictive Control-based Motion Cueing
Algorithm using an optimized Nonlinear Scaling for Driving
Simulators," in 2019 IEEE Interna tional Conference on Systems,
Man and Cybernetics (SMC), 2019: IEEE, pp. 1245-1250.
[48] H. Asadi, "Human perception-based washout filtering," Deakin
University, 2015.
[49] H. Asadi, S. Mohamed, K. Nelson, and S. Nahavandi, "A linear
quadratic optimal motion cueing algorithm based on human
perception," in ACRA 201 4: Proceedings of Australasian
Conference on Robotics and Automation, 2014: Australian Robotics
and Automation Association, pp. 1-9.
[50] K. Sun, Q. Jianbin, H. R. Karimi, and Y. Fu, "Event-triggered robust
fuzzy adaptive finite-time control of nonlinear systems with
prescribed performance," IEEE Transactions on Fuzzy Systems,
2020.
[51] K. Sun, J. Qiu, H. R. Karimi, and H. Gao, "A novel finite-time
control for nonstrict feedback saturated nonlinear systems with
tracking error constraint," IEEE Transactions on Systems, Man, and
Cybernetics: Systems, 2019.
[52] H. D. Kabir, A. Khosravi, S. Nahavandi, and A. Kavousi-Fard,
"Partial adversarial training for neural network-based uncertainty
quantification," IEEE Transactions on Emerging Topics in
Computational Intelligence, 2019.
[53] H. D. Kabir, A. Khosravi, M. A. Hosen, and S. Nahavandi, "Neural
network-based uncertainty quantification: A survey of
methodologies and applications," IEEE access, vol. 6, pp. 36218-
36234, 2018.
[54] S. M. J. Jalali, M. Karimi, A. Khosravi, and S. Nahavandi, "An
efficient neuroevolution approach for heart disease detection," in
2019 IEEE International Conference on S ystems, Man and
Cybernetics (SMC), 20 19: IEEE, pp. 3771-3776.
[55] S. M. J. Jalali, A. Khosravi, P. M. Kebria, R. Hedjam, and S.
Nahavandi, "Autonomous robot navigation system u sing the
evolutionary mu lti-verse optimizer algorithm," in 2019 IEEE
International Conferen ce on Systems, Man and Cybernetics (SMC),
2019: IEEE, pp. 1221-12 26.
[56] H. D. Kabir, A. Khosravi, S. Nahavandi, and D. Srinivasan, "Neural
Network Training for Uncertainty Quantification Over Time-
Range," IEEE Transactions on Emerging Topics in Computational
Intelligence, 2020.
Mohammad Reza Chalak Qazani received the
Bachelor of Engineering in manufacturing a nd
production from University of Tabriz, Tabriz, Iran in
2010 and the master’s degree in robotic and mechanical
engineering from the Tarbiat Modares University,
Tehran, Iran, in 2013 .
He is currently a Ph.D. student in the Institute for
Intelligent Systems Research and Innovation (IISRI),
Deakin University. His current research interests
include model predictive control, motion cueing
algorithm and soft computing controllers.
Houshyar Asadi received the Bachelor of
Engineering degree (First Class Hons.) in electrical-
control systems in 2008, and master’s degree in
industrial Electronic and Control Engineering from
University of Malaya in 2012. He received the Ph.D.
degree in human perception-based washout filtering
using Artificial intelligence (AI) from the Institute for
Intelligent Systems Research and Innovation (IISRI),
Deakin University, Australia, in 2015.
He is currently a Senior Research Fellow with the
IISRI, Deakin University. His current research interests include Artificial
intelligence, motion-based simulator technologies, robotics, control, and human
factors.
Mehrdad Ro stami received Bachelor’s degree in
Computer Engineering from the Razi University,
Kermanshah, Iran, in 2012 and the Master’s degree in
Artificial Intelligence, in 2014, from the University of
Kurdistan, Sanandaj, Iran.
After his M.Sc., he has got 5 years’ industrial work
experience in machine learning research and 6 years’
teaching experience in different computer science
courses. His research interests include Machine
Learning, Data Mining, Pattern Recognition, Signal
Processing, and Bioinformatics.
Shady Mohamed received th e B.Sc. and M.Sc.
degrees in information technology from Cairo
University, Giza, Egypt, in 2000 and 2003,
respectively, and the Ph.D. degree in control theory
from Deakin University, Geelong, VIC, Australia, in
2009. He is currently an associated professor with the
Institute for Intelligent Systems Research and
Innovation (IISRI), Deakin University, Australia.
He is interested in interdisciplinary research
involving signal processing, control theory, huma n
biodynamics, haptics and medical imaging.
Chee Peng Lim received his PhD degree from the
University of Sheffield, UK, in 1997. His research
interests include computational intelligence-based
systems for pattern recognition, data mining, fault
detection, condition monitoring, decision support,
and process optimization. He has published over 450
technical papers in these areas.
He is currently a professor at Institute for
Intelligent Systems Research and Innovation, Deakin
University, Australia.
Saeid Nahava ndi received a Ph.D. from Durha m
University, U.K. in 1991. He is an Alfred Deakin
Professor, Pro Vice-Chancellor, Chair of Engineering,
and the Founding Director of the Institute for
Intelligent Systems Research and Innovation at
Deakin University. His research interests includ e
modeling of complex systems, robotics and haptics.
He has published over 900 scientific papers in various
international journ als and conferences.
Professor Nahavandi is Editor-In-Chief: IEEE SMC
Magazine, the Senior Associate Editor: IEEE Systems Journal, Associate Editor
of IEEE Transactions on Systems, Man and Cybernetics: Systems, and IEEE
Press Editorial Board memb er. Professor Nahavand i is a Fellow of IEEE
(FIEEE), Engineers Australia (FIEAust), the Institution of Engineering and
Technology (FIET). Saeid is a Fellow of the Australian Aca demy of
Technology and Engineering (ATSE).