Parallel Robotic Manipulation via Pneumatic Artificial Muscles
Dimitris Gryparis, George Andrikopoulos and Stamatis Manesis
Department of Electrical and Computer Engineering, University of patras, Eratosthenous str., Rio, Greece
Keywords: Pneumatic Artificial Muscles, Parallel Manipulator, PID Control.
Abstract: In this article, a 6 Degree-of-Freedom (DOF) parallel manipulator, actuated by Pneumatic Artificial Muscles
(PAMs), is being presented. Incorporated in a two Stewart Platform-based design, the novel manipulator’s
motion capabilities are being examined through kinematic analysis, while the open-loop operation
characteristics and performance of the manipulator’s control via a multiple PID-based scheme are being
experimentally evaluated.
1 INTRODUCTION
Over the last decade, there has been an increasing
scientific interest in robotic manipulators that are
lightweight, safe and compliant (Calabria et al.,
2012; Radojicic & Surdilovic, 2009; Denkena et al.,
2008). In such cases, the selection of the type of
actuation that will power the manipulator is of
utmost importance. The Pneumatic Artificial Muscle
(PAM) has drawn the attention of the scientific
community regarding its merits for utilization in
biorobotic, medical and industrial applications
(Andrikopoulos, Nikolakopoulos & Manesis, 2011).
By mimicking the operation and properties of the
organic muscle, the PAM provides a suitable
solution for safer user interaction, as well as more
strong and natural motion through inherent
compliance, absence of mechanical parts, as well as
an impressive power-to-mass ratio. The most
utilized PAM-type is the McKibben Artificial
Muscle which was invented by the physician Joseph
L. Mckibben in the 1950s and was incorporated in
artificial limps (Stewart, 1965).
So far, similar PAM-actuated manipulator
approaches have included one Degree-of-Freedom
(DOF) per platform concepts (Calabria et al., 2012),
Stewart-based platforms via the utilization of 3
PAMs conically incorporated (Radojicic &
Surdilovic, 2009) and via more complex pneumatic
actuation mechanics (Denkena et al., 2008).
Parallel manipulator aproaches with the
utilization of other actuation means have included 3-
DOF parallel manipulators actuated by servomotors
(Cazalilla et al., 2014), (Khosravi et al., 2014),
(Ning et al., 2006) and 6-DOF Stewart-based
platforms actuated via hydraulic actuators placed at
the base of the robots (Guo, et al., 2007), (Pi et al.,
2010).
This article presents the development and control
of a parallel manipulator with 6 DOFs, which was
implemented in a two Stewart Platform-based
design. In this novel approach, each platform
consists of four PAM actuators in parallel
configuration, which are being operated as
antagonistic pairs, while a pneumatic cylinder is
being incorporated in the center of each platform,
thus, achieving 3 DOFs per moving platform.
The presented manipulator possesses the
advantage of incorporating more DOFs than the
manipulators found in related literature, thus
possessing enhanced motion capabilities, improved
motion range and increased relative workspace. In
addition, the cascaded configuration of the two
platforms is enhanced by the placement of all the
PAMs on the lower segment of the manipulator in
order to provide a novel mechanical approach on the
actuated motion of this parallel robotic structure.
The PAM-actuated manipulator is being depicted in
Figure 1.
In the following sections, the experimental setup
components and the kinematic analysis of the
manipulator’s structure are being presented, followed
by a presentation of the open-loop operation
characteristics of the manipulator. Finally, the
performance of the setup’s control via a multiple
PID-based scheme is being experimentally evaluated.
29
Gryparis D., Andrikopoulos G. and Manesis S..
Parallel Robotic Manipulation via Pneumatic Artificial Muscles.
DOI: 10.5220/0005008700290036
In Proceedings of the 11th International Conference on Informatics in Control, Automation and Robotics (ICINCO-2014), pages 29-36
ISBN: 978-989-758-040-6
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
2 EXPERIMENTAL SETUP
COMPONENTS AND
KINEMATIC ANALYSIS
2.1 Setup Components
Figure 1: The parallel robotic manipulator.
The structure utilizes eight PAMs that have been
incorporated on the lower segment of the
manipulator. Four PAMs are utilized for the
actuation of the lower platform, which are placed
closer to the center of the ground plate and are
characterized by 170 mm nominal length and 20mm
inner diameter. Additional four PAMs are being
utilized for the actuation of the upper platform,
which are placed in the outer formation of the
ground plate and are characterized by 220 mm
nominal length and 20mm inner diameter. These
PAMs are being connected to the upper platform via
Teflon flexible shafts. All utilized PAMs are
manufactured by Festo AG & Co. KG and feature a
maximum pulling force of approximately 2000 N
and a weight of less than 300 g.
In order to augment the rigidity of the
construction and increase the DOF motion
capabilities, two non-revolute pneumatic cylinders
were placed in the center of each platform. The
operational range of each cylinder is 0-6 bar and
their nominal length is 300 mm. The non-rotating
pistons are connected to the platforms via universal
(cardan) joints. These joints allow rotations only
around x or y axis.
Twelve Festo proportional pressure regulators
are being utilized in order to provide compressed air
to each muscle and to each pneumatic cylinder
independently. In addition, two dual axis
inclinometers manufactured by Level Developments
Ltd were utilized featuring an effective angle
measurement range of in both x and y axis, in order
to provide the necessary angle feedback for the
control loop that is being described in the sequel.
Furthermore, two input/output cards manufactured
by Measurement & Computing are being utilized
for the interconnection of the pressure regulator and
sensor equipment with the personal computer. The
interface between the setup and the computer system
components is being composed via LabVIEW
software.
2.2 Forward Kinematic Analysis
A simplified axial representation of the manipulator
is being presented in Figure 2. It must be noted that
due to the design concept of placing all PAMs at the
ground plate, the upper platform is being rotated by
45
o
with respect to the z axis of the lower platform in
order to achieve ameliorated performance and more
compact design implementation.
Figure 2: Simplified axial representation of the
manipulator.
2
n
d
metallic
plate
Flexible
shafts
PAMs
1
s
t
metallic
plate
Immobile
ground plate
Proportional
Regulators
Pneumatic
Cylinder
ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics
30
The symmetrical design concept of the manipulator
and the utilization of universal joints, which allow
independent movements in relation to the x-y axes,
provide the advantage of considering the PAM-
induced motion in each plane for the both platforms
as decoupled, thus simplifying the kinematic
analysis of each platform. In this way, it is possible
to study the forward kinematics of each plane
independently without introducing complexities due
to motion coupling. The following kinematics
analysis is based on a geometric analysis, as is
commonly practiced in parallel robotic theory. The
symbolic representation A(i,j)
indicates the point
where the i-th PAM (i {1,…4}) is attached to the j-
th link (j {1,2}) in the first. The immobile ground
plate denoted as 0, the middle plate as 1 and the
upper plate as 2, yields the kinematic analysis of a
single antagonistic segment presented in Figures 3
and 4. The symbolic representations O
q
, q {0,1,2}
and
'
w
O
, w {1,2} indicate the geometric center of
the q-th metallic platform and the w-th universal
joint’s geometric center respectively.
Figure 3: Geometrical analysis of a single antagonistic pair
of PAMs in the y-z plane of the lower platform.
Figure 4: Geometrical analysis of a single antagonistic pair
of PAMs in the x-z plane of the lower platform.
To simplify the analysis of the manipulator, the
lengths of the pneumatic cylinders are being
considered as constants and are noted as l
1
and l
2
respectively.
Furthermore, the angle in x and y axis
are being denoted as
i
and ,
i
respectively, where i
indicates the lower (i=1) or the upper (i=2) platform.
To further simplify the mathematical analysis,
the following notations are being utilized:
'
01 1
OO l
&
'
12 2
||OO l
(1)
111 1
OA c
&
'
11 2
|O |Oc
(2)
212 3
||OA c
(3)
In addition, constraints posed by the structural
properties of the design ensure that:
'
11
||OO
=
'
22
||OO
=c
2
(4)
2.2.1 Kinematic Analysis of the Lower
Platform
The methodology used to obtain the coordinates of
the endpoints of the manipulator’s platforms’ is
being presented below in the following equations.
The coordinates of the middle point of the straight-
line segment, which is being formulated from the
aforementioned endpoints, are the coordinates of the
platforms’ centers. The coordinates of the points
'
1
O
and O
1
are geometrically derived from Figure 3.
'
11
(0,0, )lO
(5a)
121121
(0, sin(), cos())clcO

(5b)
There is only one unique straight line that passes
from both
'
1
O
and
1
O in the y-z plain in the form of
11
zkyb
where k
1
is the slope of the line and b
1
is a static constant. The straight line that intersects
'
1
O
is being described by:
11
tan( )zy l

(6)
The straight line that intersects the
points
'
11
A
and
1
O , and by design is vertical to (6) is
being presented below:
12 11
ztan()cos()yc l

(7)
The following y-plane coordinate equations are
being geometrically deduced via Figure 3 as:
(1,1) 2 1 1 1
sin( ) c cos( )
A
yc

(8a)
(2,1) 2 1 1 1
sin( ) c cos( )
A
yc

(8b)
ParallelRoboticManipulationviaPneumaticArtificialMuscles
31
and substitution of (8a) and (8b) in (7) give the z-
plane geometrical equations:
(1,1) 2 1 1 1 1
cos( ) sin( )
A
zc c l

(9a)
(2,1) 2 1 1 1 1
cos( ) sin( )
A
zc c l

(9b)
From Figure 4, the coordinates of O
1
can be derived
as:
12 1 2 11
(c sin( ),0,c cos( ) )
xx
Ol

(10)
There is only a unique defined straight line that
passes from both
'
1
O
and
'
1
O
in the x-z plane, the
form of which is
23
zkxb where k
2
stands for
the line’s slope and b
3
for the constant coefficient:
11
tan( )zx l

(11)
The straight line that intersects both
1
O and
31
A
is
perpendicular to (11) based on the manipulator’s
design properties, thus giving:
112 1
z tan( ) cos( )xlc

(12)
The following x-plane coordinate equations are
being geometrically deduced via Figure 4 as:
(3,1) 2 1 1 1
sin( ) cos( )
A
xc c

(13a)
(4,1) 2 1 1 1
cos( ) c cos( )
A
xc

(13b)
Substitution of (13a) and (13b) in (12) gives:
211(3,1) 1 1
cos( ) sin( ) l
A
zc c

(14a)
211(4,1) 1 1
cos( ) sin( ) l
A
zc c

(14b)
Thus, Equations (8), (9), (13) and (14) can be
utilized in order to compute the first platform’s
center Ο
1
coordinates (x,y,z) via the following
geometrical representation:

1(3,1)(4,1)
4
xA A
Ox x
(15a)

1(1,1)(2,1)
4
yA A
Oy y
(15b)

1 (1,1) (2,1) (3,1) (4,1)
4
zA A A A
Oz z z z
(15c)
2.2.2 Kinematic Analysis of the Upper
Platform
Utilization of the exact same methodology and by
introducing the coordinates of the point
1
O as the
new starter point, the coordinates of
'
2
O
are being
deduced in order to compute later the coordinates of
O
2
:
'
2122
sin( )( ) 2
x
Olc

(16a)
'
2122
sin( )( ) 2
y
Olc

(16b)
'
2 (1,2) (2,2) (3,2) (4,2)
4
zB B B B
Oz z z z
(16c)
where
(, )
B
ij
z
are auxiliary points in the three
dimensional space utilized in order to simplify the
geometrical analysis and their coordinates are being
derived as:
(1,2) 1 2 2 1 1 1
cos( )( ) c sin( )
B
zlc l

(17a)
(2,2) 1 2 2 1 1 1
cos( )(c ) sin( )
B
zlcl

(17b)
(3,2) 1 2 2 1 1 1
cos( )( ) sin( )
B
zclcl

(17c)
(4,2) 1 2 2 1 1 1
cos( )( ) c sin( )
B
zlc l

(17d)
Following the previous methodology and by rotating
the computed coordinates by 45
o
with respect to the
z axis, posed by the manipulator’s upper platform
design, the coordinate equations of the upper
platform’s respective geometrical points are being
computed as:


'
(1,2) 2 2 2 3 2
'
(1,2) 2 2 2 3 2
'
(1,2) 2 2 2 3 2
2
sin( ) c cos( )
2
2
sin( ) c cos( )
2
sin( ) c cos( )
Ay
Ay
Az
xOc
yOc
zOc






(18a)


 
'
(2,2) 2 2 2 3 2
'
(2,2) 2 2 2 3 2
'
(2,2) 2 2 3 2 2
2
sin( ) c cos( )
2
2
sin( ) c cos( )
2
cos sin
Ay
Ay
Az
xOc
yOc
zc c O






(18b)
 

 

 
'
(3,2) 2 2 2 3 2
'
(3,2) 2 2 2 3 2
'
(3,2) 2 2 2 3 2
2
sin cos
2
2
sin cos
2
cos sin
Ax
Ax
Az
xOcc
yOcc
zOc c






(18c)
 

 

 
'
(4,2) 2 2 2 3 2
'
(4,2) 2 2 2 3 2
'
(4,2) 2 2 2 3 2
2
sin cos
2
2
sin cos
2
cos sin
Ax
Ax
Az
xOcc
yOcc
zOc c






(18d)
Thus, the coordinates of the endpoint O
2
of the
manipulator are being computed as presented below:
ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics
32



2 (1,2) (2,2) (3,2) (4,2)
2 (1,2) (2,2) (3,2) (4,2)
2 (1,2) (2,2) (3,2) (4,2)
4
4
4
xA A A A
yA A A A
zA A A A
Ox x x x
Oy y y y
Oz z z z



(19)
3 ANTAGONISTIC OPERATION
PROPERTIES AND CONTROL
ALGORITHM FORMULATION
3.1 Antagonistic Open-Loop Operation
of PAM Pairs
The law that governs the antagonistic motion of the
PAM pairs i=1,2 and i=3,4 of the j-th platform link
in the presented manipulator applies an initial state
of pressure P
0,j
in the PAMs of the j-th link, which
is increased by a quantity ΔP in those specified by
i=1,3 and is decreased by the same quantity in the
their antagonistic ones specified by i=2,4. The
aforementioned antagonistic law is being presented
below:
,0,
(1)
w
ij j k
P
PP
(20)
where k indicates the pair of PAMs and is being
defined as:
1, for 1, 2 and 1
2, for 3, 4 and 1
3, for 1, 2 and 2
4, for 3, 4 and 2
ij
ij
k
ij
ij
and w=i+1.
In this way, all spectrum of the operating
pressure range is being exploited and this law is
expected to give the maximum motion range, but,
due to the nonlinearities and intense hysteresis
phenomena during the inflation and deflation states
of the PAMs, it fails provide a smooth antagonistic
operation. The symmetric around P
0,j
nature of the
pressure spectrum, which is posed by (20), does not
lead to symmetrical length alterations due to the
highly nonlinear relationship describing PAM’s
length with relation to the operating pressure.
In order to achieve a smoother antagonistic
behavior and to avoid the various phenomena that
are caused by the nonlinearities and hysteretic
behavior of the PAMs a modified antagonistic law is
being utilized with the appropriate insertion of
corrective coefficients as shown below:
,0,
,
(1)
w
k
ij j
ij
P
PP
C

(21)
With the experimentally derived corrective terms C
i,j
regarding the i-th
PAM attached to the j-th platform
link, it is made possible to compensate for the
PAM’s nonlinearities and improve the overall
antagonistic cooperation between PAMs. The
utilized PAMs operate in a range of 0-6 bar of
pressure and the experimental values of the
coefficients are being derived for initial pressure
state P
0,j
=3 bar in all the PAMs. The coefficient
values are being displayed in Table 1.
Table 1: Corrective coefficients.
i j
ΔP[bar]
1 1.5 2 2.5 3
C
i,j
1
1
1 1.2 1.92 2 2.1
2 1.04 1.12 1.39 1.69 1.8
3 1.28 1.48 1.92 2.13 2.31
4 1 1.23 1.49 1.68 1.91
1
2
1 1.05 1.15 1.64 2.05
2 1.09 1.22 1.7 2.09 2.25
3 1.2 1.43 1.21 2.12 2.41
4 1.12 1.21 1.48 1.69 2
3.2 Control Algorithm Formulation
The symmetric nature of the manipulator’s design
and the setup’s components (non-revolute cylinders,
universal joints) leading to the analysis
simplification of decoupled movements in the x, y
axis of both platforms, led to the utilization of four
independent PID controllers for controlling each
manipulated value ΔP
k
of the k-th pair of PAMs. The
formulated multiple PID control structure is
presented in Figure 5.
Figure 5. Multiple PID-based Control Scheme.
The following PID law is being utilized for every
independent controller:
ParallelRoboticManipulationviaPneumaticArtificialMuscles
33
,, D,
0
() () ( ) ()
t
kPkkIkk kk
d
Pt K e t K e d K e t
dx


(22)
where e
k
(t) stands for the error signal imposed by the
k-th pair of PAMs and is being formulated by
subtracting the angle values θ
j
, φ
j
that are being
measured by the inclinometers from the reference
values θ
j,ref
, φ
j,ref
that are being provided by the user,
respectively.
As presented in Figure 5 the closed loop control
is being performed via four independent PID
controllers, each for every PAM antagonistic pair.
Every PID receives as input the respective error
signal and the control effort produced is utilized as
the pressure quantity ΔP
k
regarding the k-th PAM
pair. Finally, the antagonistic law (20) is being
computed and the antagonistic pressure signals are
being supplied into the PAM-actuated manipulator
system. The control structure’s goal is to bring the
manipulator to the reference pose governed by the
set-point platform angles.
4 SIMULATED PROPERTIES
AND EXPERIMENTAL
RESULTS
4.1 Open-Loop Performance
Characteristics
In the open-loop operation of the manipulator, the
user can choose moving each platform
independently or both platforms combined as it is
being depicted in Figure 6. The initial height of the
manipulator can be adjusted by the proper selection
of the initial pressure feeds of the cylinders P
j,cyl
and
the respective initial pressure states P
0,j
of the
utilized PAMs.
Forward kinematic analysis of the platform’s
movement has been simulated in order to produce
the theoretically derived workspace of the
manipulator for initial PAM lengths l
PAM
,
(j=1)
=15 cm
and l
PAM
,
(j=2)
=18.5 cm initial cylinder lengths
l
CYL,j
=27 cm, which correspond to P
0,j
=3 bar initial
PAM pressures and P
CYL,j
=1.5 bar initial cylinder
pressures, respectively, It has to be noted that
throughout the simulation trials the length of the
cylinders remained unaltered, blocking any
movements in the z
j
axes, an approach that has been
also followed in the following experimental trials.
The simulated workspace of the central point of
the lower platform, as computed from (15a)-(15c),
whereas the simulated workspace of the upper
platform’s central point is being derived by utilizing
the same methodology presented in (19). The total
workspace, featuring the upper and lower platforms’
workspace in green and blue color respectively, is
being displayed in Figure 7.
(a) (b) (c)
Figure 6: The robotic manipulator during (a) movement of
the lower platform, (b) movement of the upper platform
and (c) movement of both platforms.
The simulated workspace of the central point of
the lower platform, as computed from (15a)-(15c),
whereas the simulated workspace of the upper
platform’s central point is being derived by utilizing
the same methodology presented in (19). The total
workspace, featuring the upper and lower platforms’
workspace in green and blue color respectively, is
being displayed in Figure 7.
4.2 Closed-Loop Performance Results
In this Subsection the performance of the multiple
PID-based scheme in controlling the manipulator
setup is being evaluated. All experimental results
presented have been performed with initial PAM
lengths l
PAM
,
(j=1)
=15 cm and l
PAM,(j=2)
=18.5 cm and
initial cylinder lengths l
CYL,j
=27 cm, which
correspond to P
0,j
=3 bar initial PAM pressures and
P
CYL,j
=1.5 bar initial cylinder pressures, respectively.
Throughout the experimental trials the length of the
cylinders remained unaltered, blocking any
movements in the z
j
axes. With the aforementioned
initial values, the platform achieves a range of
motion from -15.5
o
to +15.5
o
in both x and y axes of
the lower platform and from -9 to +9 degrees for the
upper platform.
The responses of (θ
1
, φ
1
) of the lower platform
and (θ
2
, φ
2
) of the upper platform during a set-point
experiment, as well as the disturbance rejection
capabilities of the control scheme are being
displayed in Figures 8 and 9, respectively. The PID
ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics
34
gains K
P,k
, K
I,k
and K
D,k
have been experimentally
tuned via an extensive trial-and-error search and the
values utilized for the presented experimental trials
are being displayed in Table 2.
Figure 7: Simulated workspace of the lower (blue) and
upper (green) platform.
Table 2: PID Parameter Values.
K
P,k
K
I,k
K
D,k
Lower Platform 0.019 0.0102 0.0007
Upper Platform 0.019 0.0055 0.001
As shown in Figure 8, the platform reaches its
steady state in both x and y axis in less than one
second from the time that the step inputs are applied.
The control structure’s robustness is being tested and
the system manages to cancel two large-amplitude
and short-duration disturbances, which are being
properly added on the system’s output signals in the
form of manually-induced shocks, and return to
previous tracking performance. The absolute mean
steady-state errors from the reference angles θ
1,ref
and φ
1,ref
are being kept in low percentages,
specifically 0.51% for the y axis and 0.68% for the x
axis in its steady state, respectively, which further
proves the efficacy of the PID-based control scheme.
As presented in Figure 9, the upper platform
reaches its steady state in about 1.5 seconds and the
absolute mean steady-state error from the reference
angles θ
2,ref
and φ
2,ref
are being kept in low
percentages, specifically 0.64% for the y axis and
0.50% for the x axis, respectively. The small
increase in the response time and the steady-state
error are mainly caused by the elastic behavior of the
material selected for the flexible shafts, which are
being made of Teflon – a choice made as a trade-off
between rigidity during motion and overall
performance. In addition, Figure 9 shows the control
structure’s capabilities cancelling high amplitude
disturbances that are being properly added on the
system’s output signals in the form of manually-
induced shocks.
Figure 8: Set-point tracking performance and disturbance
cancellation capabilities of the control scheme during
lower platform movements.
Figure 9: Set-point tracking performance and disturbance
cancellation capabilities of the control scheme during
upper platform movements.
In addition, the responses of the θ
1
, θ
2
, φ
1
and φ
2
angles during simultaneous operation of both
Outpu
t
Disturbances
Outpu
t
Disturbances
ParallelRoboticManipulationviaPneumaticArtificialMuscles
35
platforms during a set-point experiment, as well as
the disturbance rejection capabilities of the control
scheme are being displayed in Figure 10. It can be
easily seen that all the controllers successfully lead
the manipulator to the desired angles, which justifies
the choice of utilizing independently structured PID
controllers instead of a more complex MIMO based
scheme to counteract for coupled system variables.
The absolute mean steady-state errors of the first
platform, which are also being kept in low
percentages, 0.9% and 0.62% for the y and x axes of
the lower platform and 0.59% and 0.56% for the y
and x axes of the upper platform, respectively. The
control structure again manages to cancel the high-
amplitude and short-duration disturbances added in
the four measured angle signals and return the
manipulator to its former state.
Figure 10: Set-point tracking performance and disturbance
cancellation capabilities of the control scheme during both
platform movements.
5 CONCLUSIONS
In this article, a PAM-actuated robotic manipulator
has been presented. The research attention has been
focused on the structural novelties of the
experimental set-up, the forward kinematic analysis
of the set-up platform components, as well as its
simulated workspace characteristics. The open-loop
operation of the structure’s movement has been
tested in various motion paterns. Finally, the closed
loop control of the manipulator via a multiple PID-
based control scheme based on a decoupled
movement theorysis has been evaluated, giving
smooth and fast responses along with high
disturbance cancellation capabilites. In the future,
different control schemes will be evaluated along
with flexible shafts of different rigidities.
REFERENCES
Andrikopoulos, G., Nikolakopoulos, G. & Manesis, S.,
(2011) A Survey on Applications of Pneumatic
Artificial Muscles. 19th Mediterranean Conference on
Control and Automation (MED), 20-23 June 2011,
Corfu, Greece, pp. 1439-1446.
Calabria, M., et al. (2012) “Redundancy resolution and
control of manipulators driven by antagonistic
pneumatic muscles. 17th International Conference on
Methods & Models in Automation & Robotics
(MMAR), 27-30 August 2012, Międzyzdroje, Poland,
pp. 119–124.
Cazalilla, J. et al. (2014) Adaptive control of a 3-DOF
parallel manipulator considering payload handling and
relevant parameter models. Robotics and Computer-
Integrated Manufacturing, 30, p. 468-477.
Denkena, B. et al. (2008) New Compliant Mc-Kibben
Actuator Driven by Pneumatic Actuators as a Hexapod
Platform in Robotic Applications. 10th Workshop on
Advanced Space Technologies for Robotics and
Automation (ASTRA), 11-13 November 2008,
Noordwijk, The Netherlands.
Guo, H. et al. (2007) Cascade control of a hydraulically
driven 6-DOF parallel robot manipulator based on a
sliding mode. Control Engineering Practice, 16, p.
1055-1068.
Khosravi, M. & Taghirad H. (2014) Robust PID control of
fully-constrained cable driven parallel robots.
Mechatronics, 24, p.87-97.
Ning, K. et al. (2006) A New Wire-Driven Three Degree-
of-Freedom Parallel Manipulator. Journal of
Manufacturing Science and Engineering, 128(3), p.
816-819.
Pi, Y. & Wang X. (2010) Observer-based cascade control
of a 6-DOF parallel hydraulic manipulator in joint
space coordinate. Mechatronics, 20, p. 648-655.
Radojicic J. & Surdilovic, D. (2009) Modular Hybrid
Robots for Safe Human-Robot Interaction.
International Journal of World Academy of Science,
Engineering and Technology (WASET), 60, p. 832–
838.
Stewart, D. (1965) A Platform with Six Degrees of
Freedom. Institution of Mechanical Engineers.
180(15).
Outpu
t
Disturbances
ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics
36