Third International Conference on

Advances in Control and Optimization of Dynamical Systems

March 13-15, 2014. Kanpur, India

Modeling simulation & control of 6-DOF

Parallel Manipulator using PID controller
and compensator
Amit Shukla ∗ Hamad Karki ∗∗
∗
The Petroleum Institute, Abu Dhabi, UAE (e-mail:
[email protected]).
∗∗
The Petroleum Institute, Abu Dhabi, UAE (e-mail: [email protected])

Abstract: This paper presents modeling simulation and control of stuart type 6-DOF parallel
manipulator. Dynamic equations of the parallel manipulator are derived by using Kane’s
method, inheriting all the advantages while circumventing all the disadvantages of Newtown-
Euler and Euler-Lagrangian formulations. In this modeling procedure at first all the inertial
torques due to platform, motor and actuators are calculated and then this is equated with
active torque supplied by the DC-motors in the form of electromagnetic torque. By this
force-torque balance, calculation of time varying equivalent inertia of the motor is avoided.
Normal PID principle based controllers are not very effective in achieving desired accuracy in
trajectory tracking as reported in literature. Therefore, a novel control architecture has been
proposed which not only has industry standard PID controller but also includes compensator
for drastically improving tracking performance. Since, over all system of parallel manipulator
is highly non-linear in nature with coupling terms and varying parameters, only PID controller
alone could not reduce this resultant tracking error, even after best tuning. One lag compensator
for each actuator has been designed in combination with PID controller, which resulted in
significant improvement in trajectory tracking in comparison to former control methods (only
normal PID based controller) for all the parameters of cartesian trajectory.

Keywords: Parallel manipulator, Gough-Stewart platform, Control, Parallel robots, Robotics.

1. INTRODUCTION et al. (2000)). Though Euler-Lagrangian formulation is

well structured and can be expressed in close form but
In 1965 Stewart has proposed an manipulating structure requires large amount of symbolic and numeric compu-
(D.Stewart (2009)) which has 6-DOF general motion in tations to derive and evaluate partial derivatives of the
space, from where standard 6-DOF parallel manipulator lagarangian (B.Dasgupta and T.S.Mruthyunjaya (1998a)).
has got its name Stewart Platform (SP). From 1960s till Newton-Euler approach requires computations of all the
now parallel mechanisms have got many applications such constraints and moments between the links though these
as flight simulators, machine tool, parallel robots, pre- calculations are not involved in the control of manipulator
cise positioning, VR technologies and medical treatment (M.J.Liu et al. (2000)). Whereas Kane’s method based
fields etc (Y.Patel and P.George (2012)). Unlike the serial derivations for dynamic model is a kind of combination of
manipulator, which have open-loop structure, the parallel both Lagrange and Newton-Euler method, which inherits
manipulator has close loop architecture. The serial manip- all the advantages of these methods but still remains free
ulator, having serially connected links, can sweep through from above mentioned disadvantages (Wang et al. (2010)).
larger workspace with dexterous manoeuverability but suf-
fers from instability with increasing load and vibrations atTo avoid calculation of time varying inertia of the ma-
higher speed. It also suffers from poor positioning capabil-nipulator in motor’s reference frame, first torque required
ity in the workspace. Force-output-to-manipulator-weight to drive manipulator is calculated and this is equated
ratio of parallel manipulator is increased substantially by with electromagnetic torque supplied by the DC-motors as
they are only source for any torque in the system. Many
using its close-loop and multiple parallel links architecture
(Liu et al. (1993)). control strategies has been published focussing on the
hydraulic actuator driven parallel manipulator but very
All the methods proposed in literature for solution of few on the electrical motor driven. Since, overall dynamics
dynamics of the parallel manipulator can be placed in of the parallel manipulator is highly non-linear, therefore
three main categories, first Lagrangian methods (Z.Geng PID controller were unable to given good tracking perfor-
et al. (1992); S.Bhattacharya et al. (1997, 1998)) , second mances. Though in literature there were some proposed
Newtown-Euler method (B.Dasgupta and T.S.Mruthyunjaya modified PID controller for reducing tracking error but
(1998b) B.Dasgupta and P.Choudhury (1999) W.Khalil their performance was not satisfactory. This work proposes
and S.Guegan (2004)) and third Kane’s method (M.J.Liu

978-3-902823-60-1 © 2014 IFAC 421 10.3182/20140313-3-IN-3024.00015

2014 ACODS
March 13-15, 2014. Kanpur, India

a novel way of using compensator to improve tracking

performance in comparison to previously published works.
This paper is organized mainly in three sections. Section
2 gives detailed description of kinematics and dynamics
of the 6-DOF parallel manipulator along with dynamical
analysis of the actuators. Dynamic equations of DC-motor
and its control are also presented in this section. At last,
to develop complete multibody dynamic model of 6-DOF
parallel manipulator, all these dynamic equations are used
in form of force-torque balance in manipulator’s reference
frame. In the next Section 3 overall control architecture
for manipulator is described and at last Section 4 presents (a) Parallel manipu- (b) Vector diagram
all the simulation results related to tracking of a desired lator
cartesian trajectory.
Fig. 1. Six-DOF parallel manipulator used in present re-
search work (a), and Vector representation of manip-
ulator with one actuator.
2. DYNAMIC ANALYSIS OF THE PARALLEL
MANIPULATOR ! 
ωx 1 0 −sθ φ̇
!
ω = ω y = R 0 cφ cθsφ  θ̇  (4)
A parallel manipulator is the platform, and end effector ωz 0 −sφ cθcφ ψ̇
connected with multiple of linear mechanism. Figure 1(b) Actuator vector is defined by following equation:
presents schematic diagram of a 6-DOF parallel manip-
ulator having a fixed base, a movable platform and a ai = u + Rsfi − B i , (5)
extendable actuator. To fully describe the motion of mov- Where ai is the vector along actuator length, u is the
able platform, a inertial reference frame is fixed at origin end-effector vector, sfi is the position vector of joint with
shown in figure 1(b). Actuators are used to closely connect the actuator in upper platform frame and B i is the vector
movable platform and fixed base frame. Upper and lower representing joint of the actuator in the base frame.
part of the actuator is connected to movable platform and
base frame respectively with coordinate points (si ) and There is a relationship between every upper joint velocity
B i . Where parameter subscript i stands for ith actuator. (v s,i ) and generalized velocity (q̇) of the end-effector and
When actuators change their length in synchronized way it can described in terms of jacobian as follows:
there is a change in position of the movable platform. This v s,i = u̇ + ω × (Rsi ) = J si ,q q̇, (6)
change can be calculated by knowing new joint coordinates
of the actuators and movable platform. 2.1 Dynamics of the actuators
Coordinates of the new position of actuator’s upper joint
Upper platform of the manipulator is supported by six
is sfi =(sfix , sfiy , sfiz ) after parallel displacement (x, y, z)
actuator links and these actuators are driven by coreless
and rotation (φ, θ, ψ) along and around x, y, z axes DC-motor. Every actuator in its configuration has two
respectively from its initial position si =(six , siy siz ). It parts, firstly lower part is a lead ball screw, which is
can be written in terms of rotation matrix as follows: connected to the motor rotor by a synchronous belt and
 f 
six six
!
x
! secondly upper part, which is a hollow cylinder with groves
f
 s  = R six + y (1) at inner surface. With every rotation of lower screw part
iy
f s z (rotating part), upper part (actuating part) of the actuator
siz iz
moves forward or backward depending on the direction of
Where R is a rotation matrix (Y.X.Su and B.Y.Duan rotation. Rotational speed of the motor rotor and related
(2000)) and index i = 1, 2...6. actuator velocity has following relationship:
cψcθ cψsθsφ − sψcφ cψsθcφ + sψsφ
! Lb ηN
v a,i = · · ω m,i , (7)
R = sψcθ sψsθsφ + cψcφ sψsθcφ − cψsφ (2) 60 γ
−sθ cθsφ cθcφ Where v a,i is the translational velocity of the actuator
Where c = cos and s = sin along the actuator length, ηN is the rotational efficiency,
γ is the gear reduction ratio and ωm is the rotational speed
This parallel manipulator has six independent links to of the motor.
enable six freedom of motion (x, y, z, φ, θ, ψ) for the
movable platform. Figure 1(a) presents a picture of 6-DOF Translational velocity of the actuator can also be described
parallel manipulator used in present work. in terms of generalized platform velocity:
v a,i = aTi,n · v s,i = J θi ,q q̇, (8)
q̇ = [u̇T ωT ]T (3) ai
T
Where ai,n = kai k and J θi ,q is the Jacobian between
Where u̇ = [ẋ, ẏ, ż] is translational velocity of the end actuator velocity and generalized velocity of the end-
effector and ω is the generalized rotation velocity of the
effector.
end effector. Relationship between ω and Euler angles is
defined as follows: Angular acceleration αm,i can be defined further as:

422
2014 ACODS
March 13-15, 2014. Kanpur, India

When inertia of the load torque is high in comparison

to light actuators, inertial forces due to actuators can be
neglected but when both are comparative, effect of the
actuator forces can not be ignored anymore, as in present
case of the experimental manipulator setup shown in figure
1(a). The gimbals are assumed to rotate frictionlessly and
inertia of actuator around actuating axis is negligible.
There are two basic parameters to define both the parts
of actuator, first mass of the and inertia around the axis
orthogonal to the actuating direction of the actuator.
For rotating and actuating part these parameters will be
(mr , ir ) and (ma , ia ) respectively.
The inertial forces of the actuators mainly have three
parts, firstly inertial mass forces, secondly influence of
inertia and thirdly gravitational forces. These forces will
be projected in the direction of the generalized velocities
of the upper joints of the actuators with the platform.
Fig. 2. Schematic diagram of a actuator
f sas,i = J Tac,s,i ma v̇ ac,i , (16)
αm,i = ω̇ m,i
= J Tac,s,i ma J ac,s,i v̇ s,i + J Tac,s,i ma J̇ ac,s,i v s,i , (17)
= K · (aTi,n v̇ s,i + ȧTi,n v s,i )
! = M as,i v̇ s,i + C as,i v s,i , (18)
(I − ai,n aTi,n )
=K · (aTi,n v̇ s,i + v Ts,i v s,i ), (9) Now total inertial forces at upper gimbal points due to
kai,n k movement of inertias of rotating and actuating parts can
(I−ai,n aT ) be derived by using lagrangeian method and resultant
Where kai,n k
i,n
= P ai,n is the projection to the plane relationship can be described as follows (S.H.Koekebakker
with normal vector ai,n . (2001)):
The acceleration of the upper actuator joint can be de-
(ia + ir ) (aTi,n v s,i )
scribed by further differentiating equation (6): f sia,ir,i = (P ai,n v̇ s,i − P ai,n v s,i ),(19)
2 kai,n k
kai,n k
v̇ s,i = ü + ω̇ × (Rsi ) + ω × (ω × (Rsi )) = M ia,ir,i v̇ s,i + C ia,ir,i v s,i , (20)
= J ai ,q q̈ − kωk2 P ω ai , (10) At last the gravitational forces on the actuating part and
Where P ω = (I − ω n ω Tn ). rotating part are calculated and they are projected along
the generalized velocities of the platform:
Angular velocity of the actuator perpendicular to the
actuator length can be described as follows: f sa,g,i = J Tac,i ma g, (21)
v s,i
ω s,i = an,i × , (11) f sr,g,i = J Trc,i mr g, (22)
kai k
If rotating part of the actuator has center of gravity (COG) 2.2 Influence of motor systems’ dynamics
at rc and the actuating part has COG at ac, velocities of
these COG’s can be described, in terms of upper joint All the actuators are actively driven by lead-ball screw,
velocity v s,i , as follows: which in turn driven by coreless DC-motor. Rotating part
v rc,i = ωs,i × (rrc · an,i ) = J rc,s,i v s,i , (12) of the actuator is connected to DC motor by a synchronous
v ac,i = v s,i + ω s,i × (−rac · an,i ) = J ac,s,i v s,i , (13) belt. When motor rotor rotates around its axis it also
Since acceleration of the COGs of the actuator parts also forces screw to rotate around its own axis. Since motor
generate inertial forces there calculation is important and is fixed on the actuator, so along with this self rotation,
should be written in terms of platform motion: motor also rotates with rotating actuator part. For the first
kind of rotation we define combined inertia of the motor
d d
v̇ ac,i = v ac,i = (J ac,s,i v s,i ) = J ac,s,i v̇ s,i + J̇ ac,s,i v s,i , rotor and screw (snail) as ims in motor’s rotor reference
dt dt frame, for the second kind of motion we define total mass
(14)
and inertia as msm and ism in the reference frame of screw.
Differentiation of the Jacobian J ac,s,i can be calculated as
follows (S.H.Koekebakker (2001)): The inertial torque generated by the motion of motor and
actuator can be described as follows:
d rac,i
J̇ ac,s,i = (I − P ai,n )
dt kai,n k T ms,i = ims αm,i ,
rac !
= (v T an,i P ai,n ) (15) (I − ai,n aTi,n )
kai,n k2 s,i = ims K · (aTi,n v̇ s,i + v Ts,i v s,i ),
rac kai,n k
+ (P ai,n v s,i aTn,i + ai,n v Ts,i P ai,n ),
kai,n k2 = M ms,i v̇ s,i + C ms,i v s,i , (23)

423
2014 ACODS
March 13-15, 2014. Kanpur, India

This inertial torque can be projected along the generalized Where F a is the active force vector applied on the platform
velocities of the platform as follows: by the six actuators, R is the rotation matrix, An is
f sms,i = J Tθi ,q T ms,i = J Tθi ,q M ms,i v̇ s,i + J Tθi ,q C ms,i v s,i , the coordinate matrix of the normal actuator vectors
(24) An = (a1,n , a2,n ....a6,n ), S p is the coordinate matrix of
The inertial forces due to combined movement of inertias the actuator joints on the platform S p = (s1 , s2 ....s6 ),
of motor and actuator can be described at the upper joint M p (χ) is the mass matrix of the movable platform in
of the platform as follows: base frame, mp is the mass of the upper platform, ω̃ is the
skew symmetric matrix related to angular velocity ω, I p =
ims RI pp RT is the inertia matrix transformation with I pp as the
f sims,i = (M ia,ir,i v̇ s,i + C ia,ir,i v s,i ), (25)
iac + irc inertia matrix in the platform frame, C p (χ, q̇) contains
= M ims,i v̇ s,i + C ims,i v s,i (26) centripetal and coriolis force components involved in the
platform movement and Gp is the overall gravitational
Finally, calculation of inertial force due to gravity can be force on the moving platform.
defined as follows:
f sms,g,i = J Trc,s,i mms,i g, (27) 2.4 Complete dynamics of the parallel manipulator including
actuators
2.3 Multi-body dynamics of the Manipulator using Kane’s
equations Total inertial forces generated by the motion of actuators
and motors can be described as follows:
To solve dynamics of the parallel manipulator, one re- 6
X
quires partial derivative of energy function which lead to f ∗t = J Tsi ,q (f sas,i + f sia,ir,i + f sa,g,i + f sr,g,i
enormous complexity but with Kane’s method this diffi- i=1
(36)
culty can be circumvent easily. The principle of virtual 6
X
work is an efficient method to develop dynamic equa- + f sms,i + f sims,i ) + f sms,i
tions for inverse dynamics of the parallel manipulator as i=1
shown in (C.D.Zhang and S.M.Song (1993), J.Wang and It can be observed from equation (35) that left-hand-side
C.M.Gosselin (1998)). Every external force applied on the of the equation has all the active forces and right-hand-side
body can be decomposed in to a force passing through cen- of the equation has inertial forces. After combining all the
ter of gravity of the system and a moment around it. And inertial forces, generated due to motion of the actuators
then by applying principle of virtual work, Kane’s equation and motor systems, total inertial forces applied on the
for the parallel manipulator has been derived, where sum parallel manipulator can be written as follows (Yang et al.
total of active forces/torques and inertial forces/torques (2010)):
are zero (T.R.Kane and D.A.Levinson (1985), Yang et al. f ∗ = f ∗t + M p (χ)q̈ + C p (χ, q̇)q̇ + Gp , (37)
(2009)).

F ac + F ∗in = 0 (28) Final equation describing the dynamics of the parallel

manipulator can be described as follows:
M ac + =0 M ∗in (29)
J θi,q (χ)F a = M tp (χ)q̈ + C tp (χ, q̇)q̇ + Gtp , (38)
Where F ac and F ∗in are the generalized active forces and
inertial forces respectively, applied on the manipulator’s Where F a ,M tp , C tp and Gtp are defined as follows:
platform in the direction of generalized velocities. This can 2π
Fa = · γ · ηT · T em , (39)
be further elaborated as follows (Wang et al. (2010)): Lb
X 6
F ac = An · F a + mp · G (30) M tp =M p + (J Tsi ,q (M as,i + M ia,ir,i + M ms,i ))J si ,q
i=1
F in = −mp · ü (31) 6
For angular motion of the platform active moment and
X
+ J Tθi ,q M ims J si ,q
inertial moment are calculated in base frame: i=1

p
(40)
M ac = RS × An · F a (32) 6
X
M in = −I p · ω̇ − ω × I p · ω (33) C tp =C p (χ, q̇)q̇ + J Tsi ,q (C as,i + C ia,ir,i + C ms,i )J si ,q
i=1
By combining all the equations from (28) to (33) complete
6 6
dynamic model of the parallel manipulator can be derived X 2
X
as follows: + J Tθi ,q C ims J si ,q q̇ − kωk J Tθi ,q M ims P ω si
     i=1 i=1
An mp I 0 ü 6
Fa = − kωk2
X
(J Tsi ,q (M as,i + M ia,ir,i + M ms,i ))P ω si
RS p × An 0 I p ω̇
     (34) i=1
0 0 u̇ mp G (41)
+ +
0 ω̃I p ω 0 6
X
In short equation (34) can be written in following form: Gtp = Gp + J Tsi ,q (f ss,i + f ss,i ) (42)
J θi,q (χ)F a = M p (χ)q̈ + C p (χ, q̇)q̇ + Gp , (35) i=1

424
2014 ACODS
March 13-15, 2014. Kanpur, India

Fig. 3. Representation of complete dynamics for parallel manipulator driven by coreless DC-motors

Table 1. Parameter values of the coreless-DC the parallel manipulator. In the motor driven parallel ma-
motor. nipulator calculation of time varying equivalent inertia Jeq
in equation (46) is difficult in real time. Therefore, instead
PMSM Parameter Symbols Values
Rated output power Pn 5.3 W
of directly calculating Jeq in DC-motor frame, torque-force
Rated speed ωn 8200 rpm balance equation is written in the parallel manipulator’s
Torque constant Kt 0.0117mNm/A frame and electromagnetic torque, produced by the DC-
Back-emf constant Ke 0.00123V/rpm motors, is just taken as input vector in to the dynamic
Armature resistance Ra 3.6 Ω equation (38) derived for the parallel manipulator.
Armature inductance La 0.00022 mH
Maximum current Ia,max 3.4 A Over all control architecture of the parallel mechanism in
Rotor magnetic flux λf 0.1252 Wb this work can be described in three stages, first control
Moment of inertia J 1.85 kgm2 of the motor rotor position (θm ), second speed control
of the motor rotor (ωm ) and third current control of the
2.5 Control mechanism for the DC motor motor windings (ia ). Figure 3 presents schematic diagram
of the manipulator along with its three stage controllers
Every actuator of the parallel manipulator is driven by a and actuators. At first, a desired cartesian trajectory
coreless DC-motor. Benefit of using coreless (or ironless) with surge, sway, heave, roll, pitch and yaw is defined
motor rotors are numerous but two of them are outstand- which is converted in to joint space by using inverse
ing such as reduced weight of overall motor and faster kinematics of the parallel manipulator. Here joint space of
acceleration of the rotor with smaller mechanical time the manipulator is actuators’ translational speeds in the
constant. Operating principle of coreless DC-motor is same direction of their actuation. From speed of the actuator,
as normal DC motor. speed of the corresponding motor rotor is calculated which
is further integrated to calculate angular position of the
Electromagnetic torque Tem is produced by the motor by
motor rotor. Desired (θd ) and actual position (θa ) of the
circulating armature current ia in its stator windings and
motor rotor are compared and by using PID based position
they are related by torque-constant as follows:
controller desired speed of the motor is calculated:
Tem = Kt ia (43) Z
d
Similarly, induced emf ea depends only on the speed of the ωd = (Kp1 + Ki1 dt + Kd1 ){θd − θm } (47)
dt
motor rotor ωm and is related to each other by voltage- This desired speed (ωd ) is compared against actual speed
constant Ke of the motor rotor (ωa ) and passed through a PID based
ea = K e ωm (44) speed controller to generate desired current in the motor
When va is applied on the motor windings, it overcomes windings.
induced emf ea and causes current ia to flow in the motor Tref
Z
d
windings. By applying Kirchoff’s voltage law foloowing id = = (Kp2 + Ki2 dt + Kd2 ){ωd − ωa } (48)
Kt dt
equation can be written:
This desired current (id ) is compared against actual cur-
dia rent (ia ) flowing in the motor windings and a PID based
va = ea + Ra ia + La (45)
dt current controller is operated on the current difference
On the mechanical side, electromagnetic torque Tem pro- to generate reference voltage to be applied on the motor
duced by the motor overcomes the mechanical torque TL windings to generate required current.
to produce acceleration: Z
d
dωm 1 vd = (Kp3 + Ki3 dt + Kd3 ){id − ia } (49)
= (Tem − TL ) (46) dt
dt Jeq When this current ia flows throw motor windings it
Where Jeq is the equivalent inertia of the DC-motor and generates electromagnetic torque Tem , which is further
the mechanical load. Here Jeq is the time varying inertia supplied in to dynamic equation of the manipulator. Each
and which is very difficult to calculate in real time. actuator has three control loops, and there are total
six actuators so there are total eighteen control loops.
3. INTEGRATED SYSTEM AND CONTROL Complete process of the control is explained schematically
in figure 3 and equations (38) - (42).
Actuators driven by the DC-motors are sole source of Parameters of these PID controllers are fine tuned by
active forces and torques applied on the upper platform of trial and error method to achieve the best tracking per-

425
2014 ACODS
March 13-15, 2014. Kanpur, India

Fig. 4. Schematic diagram of parallel manipulator describing its complete dynamics and control loops with compensator

formance. Later it has been observed that in-spite of best Table 2. Parameter values of the parallel ma-
tuning there was always some tracking error between the nipulator.
desired trajectory and actually achieved by the controller.
Since, complete dynamics of the 6-DOF parallel manip- parallel manipulator Parameter Symbols Values
Total height of the manipulator H 0.227 m
ulator is highly non linear along with strong coupling
Upper joint radius ra 0.110 m
terms and varying parameters, linear control based PID Lower joint radius rb 0.345 m
controllers by themselves were unable to mitigate these Upper joint spacing La 37◦
errors. Therefore, to improve performance of the tracking Lower joint spacing Lb 16◦
controller, in (Y.X.Su et al. (2004)) Su has proposed usage Middle actuator length lio 0.250 m
of nonlinear PID control based technique. With the same Upper platform mass mp 1.4235 kg
purpose in (C.Yang et al. (2011)) researchers have pro- Actuator’s actuating part mass ma 0.37 kg
posed a novel feedback linearization mechanism. In (Meng Actuator’s rotating part mass mb 0.06 kg
et al. (2012)) also proposed a PID controller based control Actuator’s actuating part inertia ia 0.00055 kgm2
Actuator’s rotating part inertia ib 0.00009 kgm2
scheme but overall results are not very satisfactory in most
Reduction ratio from actuator to motor γ 16
of the cases. In (Meng et al. (2012)), model based im- Lead of ball screw Lbs 0.002 m
proved controller has been used, which essentially modifies
required torque reference to the PMSM motors. In contrast The maximum allowed displacement for manipulator
to this approach in present work, one lag compensator for platform in cartesian space is as follows (Xmax =
each actuator is proposed in addition to three staged PID 0.02m, Ymax = 0.02m, Zmax = 0.017m, φmax = 10◦ , θmax =
controllers as shown in figure 4. This approach is lot more 10◦ , ψmax = 20◦ ). The desired cartesian trajectories given
simpler than all other previously mentioned techniques to platform is described as follows:
from implementation and computational efficiency point 
of view.  Surge : X = χ1 = 0.015 sin(2π × 0.2t)m,
Sway : Y = χ2 = 0.015 sin(2π × 0.2t)m,




By using proposed compensators (S.Nise (2011)), desired 
Heave : Z = χ3 = 0.012 sin(2π × 0.2t)m,
rotational speed of the motors (ωd∗ ), as seen in figure 4, is S(t) = (51)
 Roll : φ = χ4 = 5◦ sin(2π × 0.2t),
modified as follows: ◦
pitch : θ = χ5 = 5 sin(2π × 0.2t),



yaw : ψ = χ6 = 15◦ sin(2π × 0.2t),
  
1 + T1 s
Z 
∗ d
ωd = (Kp1 +Ki1 dt+Kd1 ){θd −θm } (50)
1 + T2 s dt Figure 5 shows the comparison of desired trajectory and
Usage of lag-Compensators has significantly improved the actually achieved trajectory, and this is done in task space
tracking performance of the controller with removal of lag so that comparison can be made in realistically visible
between desired and actual trajectories. This novel ap- space.
proach gives even better results on comparison to approach As it has been already explained in Section 3 that PID con-
presented in (Meng et al. (2012)). All the tracking results trollers alone could not mitigate resultant error between
with and without compensators and degree of improve- desired and actual trajectory. This led to introduction of
ment will be analyzed in next section. compensators, which is used to modify desired rotational
speed of the DC-motors. It has been found that usage of
lag-compensators have significantly improved the tracking
4. SIMULATION RESULTS performance by minimizing the tracking error as shown
in figure 6. Where subscript WC and NC stand for with
In present research work parallel manipulator has two compensator and without compensator respectively. Error
distinct system, firstly whole mechanical structure of the in trajectory parameters are calculated as follows:
platform and secondly electric motor, and their respective |Aχ,d − Aχ,a |
Error(in%) = × 100 (52)
parameters are defined in tables 1-2. To evaluate accuracy Aχ,d
of the complete model and efficacy of its control architec- Where Aχ,d and Aχ,a stands for maximum amplitude of
ture, several trajectories have been given for the tracking.
desired and actual trajectory parameters respectively.
Trajectories have been given in cartesian space but finally
control is implemented in joint space of the manipula- To further analyze the benefit of using compensator based
tor rather than task space for simplicity of the control. controllers than simple PID control, the peak to peak
Tracking performance is the best criteria to evaluate the amplitude errors are used to calculate evaluation standard
modeling procedure and effectiveness of the control. as suggested in (Meng et al. (2012)):

426
2014 ACODS
March 13-15, 2014. Kanpur, India

(a) Surge motion (b) Roll motion

(c) Sway motion (d) Pitch motion

(e) Heave motion (f) Yaw motion

Fig. 5. Trajectory tracking long and around X-axis (a)-(b), Y-axis (c)-(d), Z-axis (e)-(f) respectively

(a) Surge motion (b) Roll motion

(c) Sway motion (d) Pitch motion

(e) Heave motion (f) Yaw motion

Fig. 6. Error in trajectory parameters along and around the X-axis (a)-(b), Y-axis (c)-(d), Z-axis (e)-(f) respectively

|Eχ,N C − Eχ,W C | Table 3. Errors in tracking trajectories with

Ipp = × 100 (53)
Eχ,N C and without compensator.
Where Eχ,W C and Eχ,N C stands for maximum amplitude Error in parameters ENC (%) EW C (%) Ipp (%)
of error curve under the control mechanism with and with- Surge X-axis 6.23 0.86 86.2
out usage of the compensator respectively. Total improve- Sway Y-axis 6.08 0.51 91.7
ment is calculated in % rather than absolute value for the Heave Z-axis 5.91 0.82 86
purpose of comparison with previously published results. Roll angle 6.12 0.18 91
Pitch angle 5.90 0.18 90.7
Before and after usage of the compensators, the maximum
Yaw angle 6.07 0.75 87.7
error in all the six parameters of the cartesian trajectory,

427
2014 ACODS
March 13-15, 2014. Kanpur, India

are described in table 3. Error data tabulated in table 3 freedom stewart platform driven by permanent magnet
shows how well compensators have mitigated the errors synchronous motors. Industrial Robot: An International
in tracking. PID controllers along with compensators have Journal, 39(1), 47–56.
reduced error in all the cartesian parameters around six M.J.Liu, C.X.Li, and C.N.Li (2000). Dynamics analysis of
times in comparison to cases, where there were no com- the gough-stewart platform manipulator. IEEE Trans.
pensators. Though it can be observed in figures 6(a)-6(f) Robot. Automat., 16(1), 94–98.
that initially there are some oscillations in trajectory but S.Bhattacharya, D.N.Nenchev, and M.Uchiyama (1998).
they get settled very soon less than in a second and they An on-line estimation scheme for generalized stewart
can be reduced further with fine tuning of PID parameters platform type parallel manipulators. Mech. Mach.
in conjunction of compensator parameters. Theory, 33(7), 957–964.
S.Bhattacharya, H.Hatwal, and A.Ghosh (1997). An on-
5. CONCLUSION line estimation scheme for generalized stewart platform
type parallel manipulators. Mech. Mach. Theory, 32(1),
This paper presents dynamic modeling procedure of the 79–89.
DC-motor driven 6-DOF parallel manipulator. For sim- S.H.Koekebakker (2001). Model Based Control of a Flight
plicity of derivation and inclusion of inertial forces ap- Simulator Motion System. Ph.D. thesis, Delft University
plied by the actuators, Kane’s method based modeling of Technology, The Netherlands.
technique is used. By establishing force-torque balance in S.Nise, N. (2011). Control Systems Engineering. John Wi-
manipulator’s reference frame, calculation of time varying ley & Sons, Inc., California State Polytechnic University,
equivalent inertia of the motor is successfully avoided. Pomona, 6th edition edition.
Performance of the modeling and control technique is T.R.Kane and D.A.Levinson (1985). Dynamics: Theory
evaluated by supplying desired cartesian trajectory for and Applications. McGraw-Hill, New York.
tracking. Due to high non-linearity of the dynamic equa- Wang, Y., Zheng, S., Jin, J., Xu, H., and Han, J. (2010).
tions of parallel manipulator simple linear PID controllers Dynamic modeling of spatial 6-dof parallel manipulator
were unable to give desired accuracy in tracking by al- using kane method. In Int. Conf on E-Product E-Service
ways leaving some lag error. By using a compensator for and E-Entertainment, 1–5.
each actuator desired speed is modified, which finally led W.Khalil and S.Guegan (2004). Inverse and direct dy-
to highly improved tracking performance, tracking error namic modeling of gough-stewart robots. IEEE Tran.
improvement goes higher than 86% for every parameter of on Robo. and Auto., 20(4), 754–762.
the cartesian trajectory. Yang, C., Huang, Q., He, J., Jiang, H., and Han, J. (2009).
Model-based control for 6-dof parallel manipulator. In
REFERENCES Int. Asia Conf on Informatics in Control, Automation
and Robotics, 81–84.
B.Dasgupta and P.Choudhury (1999). A general strategy
Yang, C., Huang, Q., Ye, Z., and Han, J. (2010). Dynamic
based on the newton-euler approach for dynamic for-
modeling of spatial 6-dof parallel robots using kane
mulation of parallel manipulator. Mech. Mach. Theory,
method for control purposes. In Int. Conf on Intelligent
34(6), 801–824.
Human-Machine Systems and Cybernetics (IHMSC),
B.Dasgupta and T.S.Mruthyunjaya (1998a). Closed-form
volume 2, 180–183.
dynamic equations of the general stewart plateform
Y.Patel and P.George (2012). Parallel manipulators
through newton-euler approach. Mech. Mach. Theory,
applications-a survey. Modern Mechanical Engineering,
33(7), 993–1012.
2, 57–64. doi:10.4236/mme.2012.23008.
B.Dasgupta and T.S.Mruthyunjaya (1998b). A newton-
Y.X.Su and B.Y.Duan (2000). The mechanical design
euler formulation for the inverse dynamics of the stewart
and kinematics accuracy analysis of a fine tuning stable
platform manipulator. Mech. Mach. Theory, 33(8),
platform for the large spherical radio telescope. Mecha-
1135–1152.
tronics, 10(7), 819–834.
C.D.Zhang and S.M.Song (1993). An efficient method
Y.X.Su, B.Y.Duan, and C.H.Zheng (2004). Nonlinear
for the inverse dynamics of manipulators based on the
pid control of a 6-dof parallel manipulator. IEE Proc.
virtual work principle. Journal Robo. Sys., 10(5).
Control Theory and Appl., 151(1), 95–102.
C.Yang, Zheng, S., Peter, O., , Huang, Q., and Han, J.
Z.Geng, L.S.Haynes, J.D.Lee, and R.L.Carroll (1992). On
(2011). Approximate feedback linearization control for
the dynamic model and kinematic analysis of a class of
spatial 6-dof hydraulic parallel manipulator. The Open
stewart platforms. Robot. Autonom. Syst., 9(4), 237–
Mech. Eng. Journal, 5, 117–123.
254.
D.Stewart (2009). Mobile robots for offshore inspection
and manipulation. In Proceedings of the Institution of
Mechanical Engineers, volume 180, 371–386.
J.Wang and C.M.Gosselin (1998). A new approach for the
dynamic analysis of parallel manipulators. Multibody
System Dynamics, 2, 317–334.
Liu, K., J.M.Fitzgerald, and F.L.Lewis (1993). Kinematic
analysis of a stewart platform manipulator. IEEE Tran.
on Indust. Elec., 40(2), 282–293.
Meng, Q., Zhang, T., He, J., Song, J., and Chen, X.
(2012). Improved model-based control of a six-degree-of-

428