See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261505293

Modeling and control for a 6-DOF platform manipulator

Conference Paper · March 2013

DOI: 10.1109/ICEESA.2013.6578432

CITATIONS READS

7 2,531

4 authors, including:

Jouini Marwa Mohamed Sassi

Ecole nationale supérieure d'ingénieurs de tunis Université de Tunis
7 PUBLICATIONS 50 CITATIONS 7 PUBLICATIONS 69 CITATIONS

SEE PROFILE SEE PROFILE

Anis Sellami
Université de Tunis
159 PUBLICATIONS 1,581 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jouini Marwa on 13 November 2018.

The user has requested enhancement of the downloaded file.

 Modeling and Control for a 6-DOF Platform Manipulator
Marwa Jouini, Mohamed Sassi, Neji Amara and Anis Sellami

Research unit C3S, ESSTT, University of Tunis, 5 av. Taha Hussein BP 56 – 1008 Tunis, Tunisia
[email protected] ; [email protected] ; [email protected] ; [email protected]

Abstract —This paper deals with the modeling and the control of manipulator. Chifu Yang [4] developed in 2008 a PID
a parallel robot with six degree of freedom (dof). The controller with gravity compensation with the feedback of
mathematical model of the 6-DOF parallel manipulator includes
cylinder length of platform. Lee and Kim [5] presented a
dynamics model which is on the Lagrange method. The model is
built in generalized coordinate system. The kinematics model is model based on sliding mode control for the Stewart platform.
based on the closed-form solutions. The latter has six electric In 2006, Iqbal and Bhatti [6] developed a control design for
actuators at six legs. The model-based controller is presented with tracking and regulation of a robot platform without any
feedback of platform positions. Two control laws of the actuators knowledge of the system’s mass properties in presence of
positions of the robot are proposed: PID control and Sliding nonlinearities.
Mode Control (SMC). Simulation results are given to show the The novelties in this paper is modeling of the robot, and make
comparison performance in term of robustness.
a comparative study between PID control and sliding mode
control of the actuators positions of robot platform.
Keywords: Parallel manipulator, Model kinematics, Model
dynamics, Platform, Sliding mode control, PID control.
This paper is structured as follow. We first define the studied
I. INTRODUCTION system. Afterward, kinematics and dynamics are explained in
section II. Then, section III, deals with the PID control and the

G ENERALLY, the parallel link manipulators provide

sliding mode control of the six electric actuators of the
manipulator. Simulation results are discussed in section IV and
finally some comments conclude the work in section V.
better accuracy, higher rigidity, higher load-to-weight ratio,
and more uniform load distribution than the serial
manipulators. Parallel robot adaptation in various fields has II. SYSTEM MODELING
given rise to different geometries of robots, with 3, 4, 5 or 6 The 6 - DOF electric driven parallel manipulator composed by
degrees of freedom. In the context of this paper, we will cover a fixed base (down platform), a moveable platform (upper
a particular interest in parallel robots hexapods [1] (composed platform) and six legs stretch, as shown in figure 1.
of 6 identical kinematic chains connecting a base platform), in
particular, robots with electric actuators that drive the hexapod
robot. The Stewart platform manipulator is a 6-DOF Body handheld
mechanism with two bodies connected together by six
extensible legs [2], [1]. The equations, forward kinematics and
dynamics of parallel manipulators are very complicated and
difficult to solve. In recent years, many research works have
been conducted on the dynamics and kinematics of the Gough-
Stewart platform manipulator. Geng and al. [3] developed
Lagrange equations of motion, regarding the geometry and
inertia distribution of the manipulator. The Lagrange
formulation is well structured and can be expressed in closed
form, but a large amount of symbolic computation is needed to Fixed base
find partial derivatives of the Lagrangian in this method. The
forward kinematics and inverse kinematics models are
described with closed-form solution and Newton-Raphson Fig.1: 6-DOF parallel manipulator.
method. The proper coordination of the actuators length
Α. Kinematic Model
enables the top plate to follow the desired trajectory with high
accuracy. Kinematics is the science of motion that treats the subject
Recent researches have been focused on the control of parallel without regard to the forces that cause it [7].

U.S. Government work not protected by U.S. copyright

The kinematics of 6-DOF Stewart-Gough platform mechanism
⎪⎧ K t I ( t ) = J Ω ( t )

include inverse kinematics and forward kinematics. ⎨ (4)
The length of each actuator of Stewart platform for a given ⎪⎩U ( t ) = K e Ω ( t ) + ( RI ( t ) + LI ( t ))
orientation can be determined using inverse kinematics and
can be written as: The actuators are represented by DC motors, where U the
control is input voltage, I is the current, L is the inductance
d ρi dq of winding, R is the resistance of the armature, J is the inertia
= J −1 (1)
dt dt of motor, K e , K t are the coefficients of speed and torque, and
d ρi Ω ( t ) = θ ( t ) is the rotational speed of the motor shaft.
Where ∈ R 6×1 is the vector of joint velocity of the
dt
platform, q = [ x y z Φ θ ψ ] T is the generalized coordinates III. CONTROL DESIGN
vector of the platform, x, y, z , are the platform centre of
mass Cartesian coordinates, Φ, θ , Ψ , are the platform Euler The purpose of the control structure is to calculate the length
6×1 of six legs for each position of the actuator of the robot. The
angles and q is the R vector of the operating speed , i =1..6
overall model of the system is shown in figure 2:
is the iterative number, J −1 is the R6×6 transition matrix for
speeds of operational space to the joint space (inverse q
Jacobian). Dynamic model

Inverse kinematics of parallel manipulator is different from θi

serial manipulator. The length of leg of platform can be solved θid Controller Actuators Constants ρi
by closed-form solution, and it can be described as: + -
JJJJG JJJJG JJJG JJJG
ρ i = Ai B i = Ai O + OC + P .CB i (2)

Where ρi is the length of leg of platform, P is matrix of Fig.2: Global model.

transformation from body coordinates to global coordinates,

A. PID control
Ai and Bi are respectively the connection points of the jacks
in the base and the mobile platform, O is the origin of the In order to suitably settle the PID controller, first we need to
absolute coordinate system and C is the center linked to the calculate the transfer function of the closed loop servo loop,
moving frame movable platform. the block diagram used is as follows:

B. Dynamic Model Ωi
Positions Ui DC motors 1 θi
+ regulator s
The dynamic equation of the platform considers inertia and -
Coriolis. Lebret in [8] developed the dynamic equation using θid
Lagrange method as:
Fig.3: Block diagram of the servo loop.
M (q)q + C (q, q )q + G (q ) = τ (3)
The transfer function of the PID is written in the form:
Where M ∈ R 6×6
is an inertial matrix, G ∈ R 6×1
is the vector of K K + Ps + K d s 2
H PID (s) = P + i + K d s = i (5)
s s
the gravity terms, C ∈ R 6×6 is the Coriolis /centripetal matrix,
From the transfer function of the controller and the process,
and τ ∈ R 6×1 is the applied torque vector. Some relevant the transfer function of a closed loop in canonical form is the
properties are given as below: following:
Property1: K i K PK K K
M is a symmetric and positive definite matrix for all q ∈R . + s + d s2
H ( s) = T T T
Property 2: KKi KP ( KK d + 1) 2 3 (6)
+ s+ s +s
q and q are bounded. T T T

C. Actuators model 1 RJ
Where k = k and T=
ke kt .
The robot controller is based on the parallel electric actuators. e
The state space of the robot actuator platform as follows:
The denominator of the equation (6) can be written in the
form: d + S θi
θ id +
Motors
D( s) = (s + wn )(s 2 + 2ξ wn s + wn2 ) (7)
dt
+
- +
λ Switch surface
Where ξ and wn are positive constants.
Fig.4: Block diagram of sliding mode.
PID controller parameters are attained as follows:
IV. SIMULATION RESULTS
⎧ Tw2 (1 + 2ξ )
⎪P = n
⎪ K Simulation has been performed in-order to examine the
⎪⎪ Twn3 effectiveness of the proposed controller design. The robot
⎨ Ki = (8) platform has six legs, six motors containing six regulators. The
⎪ K
simulation parameters for the 6-DOF electric Stewart platform
⎪ Twn (1 + 2ξ ) − 1
⎪ Kd = K
are given with the following nominal parameters: mass of
⎪⎩ platform m=1kg, R =1.05 Ω , L = 0.5 mH , I =17.2 A
B. Sliding mode control −1 −1
J = 1.6 kg .m 2 , ke = 1 Vrad s , U = 9V , kt = 1.77 N .m.A

The main objective of the design approach of sliding mode and q = (0.9m,0.9m,0.5m,0.3rad ,0.2rad ,0.3rad )t .
control is to force the error of the robot actuator positions and The proposed control procedure is simulated using the
its derivative to zero. The switching surface design comprises Matlab/Simulink environment. Then to test the robustness of
the construction of the sliding function. Positions of errors and the proposed controls, we study the influence of parameter
derivatives are the coordinates in the selected robot platform. variations on the performance of the position settings. The
parameters are varied simultaneously at the motor start. The
The robot actuator position errors are introduced by: value of the inertia of the motor is thus multiplied by 0.5, the
resistance of the armature is multiplied by 2 and the
e = θid − θi (9) inductance of winding is multiplied by 3.
6×1 6×1
Where θi ∈\ is the position vector and θid ∈\ is the A. Simulation with the PID controller
desired positions vector.
The simulation result of the PID control is shown in figure 5.
Linear sliding surfaces S ∈\6×1 are introduced in terms of The figure shows the evolution of the real position with
errors and its derivative positions are given below: respect to the desired position in the absence of parameters
variation.
S = λ e + e (10)
1

Where λ ∈\6×6 is a diagonal strictly positive matrix. Equation

0.5

(10) is characterized by deviations from the desired state. The 0

(rad)

sliding surfaces are achieved and maintained when S = 0.

-0.5
The simple input control to reach S = 0 through a control
equivalent, is given by: -1
desired position
position
-1.5
⎧⎪U eq = −1 si S <0 0 5 10 15 20 25 30

⎨ (11) Time (s)

⎪⎩U eq = +1 si S >0
Fig.5: Position of one actuator.

Where U eq is the equivalent control signal system, to B. Simulation with the sliding mode controller
compensate the system dynamics.
The simulation result of the sliding mode control is shown in
figure 6. The figure shows the evolution of the real position
The complete control circuit block diagram of these actuators
of manipulator is shown in figure 4. Here U is adjusted to with respect to the desired position in the absence of
positive and negative voltages for -1 and +1 values. parameters variation.
 1

0.5

0
(rad)

-0.5

-1
desired position
position
-1.5
0 5 10 15 20 25 30
Time (s)
Fig.6: Position of one actuator.

C. Robustness test

The comparison between the PID control and the sliding mode
control is introduced in the following figure in presence of Fig.8: Lengths of the six legs for the sliding mode control.
parameters variation:
From these figures we note that sliding mode control ensures
1
the stability of each length of the platform after a certain time.

0.5 V. CONCLUSION

0
The robot platform system studied in this paper is a
manipulator with a closed kinematic chain. The latter is more
complicated for control and modeling. We perform a
-0.5 performance study between PID control and sliding mode
control of robot platform actuators. Simulation results show
the stability of the system. In meantime, the sliding mode
-1 Desired position control gives high robustness performance in comparison with
PID the PID control.
Sliding mode
-1.5
0 5 10 15 20 25 30
time (s)
APPENDIX
Fig.7: Comparison between the PID and sliding mode.
Here we explore each component of the Stewart platform
Following these simulations, the PID controller and the sliding dynamic equation. The inertial matrix M can be written as:
mode controller are compared. These two methods gives us
good results. In fact, the six actuator positions of the robot
platform towards the six desired values. ⎡m 0 0 0 0 0 ⎤
The main difference between both methods is in the ⎢0 m 0 0 0 0 ⎥⎥
⎢
robustness. In fact, the variation of the parameters of the ⎢0 0 m 0 0 0 ⎥
actuator (DC motor) shows that the PID controller is not M =⎢ ⎥
robust to parameter uncertainties unlike the sliding mode ⎢ 0 0 0 P11I1 P12 I 2 P13 I3 ⎥
controller, which guarantees good results. Then we can ⎢ 0 0 0 P21 I1 P22 I 2 P23 I3 ⎥
⎢ ⎥
conclude that the sliding mode controller is better and more ⎣⎢ 0 0 0 P31 I1 P32 I 2 P33 I 3 ⎦⎥
robust than the PID controller. Where

D. Simulation of the global model P is the transition matrix, P = R ( yr ,θ ).R ( xr , Φ ).R ( zr , Ψ )

After the test of robustness we note that the sliding mode
control is robust. The simulation of the global model of the
manipulator platform is presented in the following figures,
which confirms the stability of the six lengths of the robot legs
ρi .
hh
 ⎡ P11 P12 P13 ⎤ ⎡ cos θ cosψ − sin Φ sin θ sinψ REFERENCES
P = ⎢⎢ P21 P22 P23 ⎥⎥ = ⎢⎢ − cos Φ sinψ
⎢⎣ P31 P32 P33 ⎥⎦ ⎢⎣sin θ cosψ + sin Φ cos θ sin Ψ [1] D. Stewart, “A platform with six degree-of-freedom,” in Proc, Inst.
Mech. Eng., vol. 180, pp.371–386, 1965.
[2] V. E. Gough and S. G. Whitehall,“Universal type test machine,” in
cos θ sinψ + sin Φ sin θ cos Ψ − cos Φ sin θ ⎤ Proc.9th Int. Tech. Congress FISITA, pp.117–137, 1962.
⎥ [3] Z. Geng and L.S. Haynes, “On the dynamic model and kinematic
cos Φ cosψ sin Φ ⎥ analysis of a class of Stewart platforms,” Robotics and
Autonomous Systems, pp.237-254, 1992.
sin θ sin Ψ − sin Φ cos θ cos Ψ cos Φ cos θ ⎥⎦ [4] Y.Chifu.He., Jingfeng and H. Junwei, “Modeling and Simulation of
6 Dof Parallel Manipulator Based on PID Control with Gravity
Compensation in Simulink/ADAMS,” IEEE International Workshop
⎡ I1 0 0⎤ on modeling Simulation and Optimization, 2008.
⎢ 0 ⎥⎥
[5] N. In Kim and C. Won Lee, “High Speed Tracking Control of
I a diagonal matrix of inertia, I = ⎢ 0 I 2 Stewart Platform Manipulator via Enhanced Sliding Mode
⎢⎣ 0 0 I3 ⎥⎦ Control,” proceeding of the IEEE International Conference on
Robotics and Automation, May 1998.
[6] S.Iqbal and I.Bhatti, “Direct Sliding-Mode Controller Design for a
The Coriolis and Centrifugal matrix C can be written as: 6 DOF Stewart Manipulator,” IEEE, 2006.
[7] H.R. Ben, M. Shoham, and S. Djerassi, “Kinematics, dynamics and
construction of a planarly actuated parallel robot,”Robotics and
⎡0 0 0 0 0 0 ⎤ Computer-integrated Manufacturing, 1998, pp.163-172.180, pp.
⎢0 0 0 0 0 0 ⎥⎥
371-386, 1965.
⎢ [8] G. Lebret, K. Liu and F. L. Lewis,”Dynamic Analysis and Control
⎢0 0 0 0 0 0 ⎥ of a Stewart Platform manipulator,” J. of Robotics System, vol.
C=⎢ ⎥ 10(5), pp.629-655, 1993.
⎢0 0 0 Q11 I1 Q12 I 2 Q13 I3 ⎥
⎢0 0 0 Q21 I1 Q22 I 2 Q23 I3 ⎥
⎢ ⎥
⎣⎢ 0 0 0 Q31 I1 Q32 I 2 Q33 I3 ⎦⎥

Where

⎡ Q11 Q12 Q13 ⎤

⎡⎣ P ⎤⎦ = [Q] = ⎢Q21 Q22
 Q23 ⎥⎥
⎢
⎣⎢Q31 Q32 Q33 ⎦⎥

Moreover, the Jacobean inverse matrix can be written as:

⎡UJG T JG JJJJG ⎤
(U 1 ∧ B1C )T
⎢ 1 ⎥
⎢ JG T JG JJJJG ⎥
⎢U 2 (U 2 ∧ B2 C )T ⎥
⎢ JG T JG JJJJG ⎥
U3 (U 3 ∧ B3 C )T
J −1 = ⎢⎢ JG JG JJJJG
⎥
⎥
T
⎢U 4 (U 4 ∧ B4 C )T ⎥
⎢ JG T JG JJJJG ⎥
⎢U 5 (U 5 ∧ B5 C )T ⎥
⎢ JG T JG JJJJG ⎥
⎢⎣U 6 (U 6 ∧ B6 C )T ⎥⎦

where
JJJJG
JJG Ai Bi
Ui =
ρi

View publication stats