Sei sulla pagina 1di 4

Fuzzy Logic Tuned PID Controller for Stewart Platform Manipulator

Dereje Shiferaw Professor R. Mitra


Adama University Indian Institute of Technology Roorkee
Department of Electrical Engineering Department of Electronics and Computer Engineering
Adama, Ethiopia Roorkee, India
derejdec@iitr.ernet.in,dnderejesh@gmail.com rmtrafec@iitr.ernet.in

Abstract— The Stewart platform manipulator is a 6DOF [3] the performance of a conventional PID controller tuned by
manipulator having higher stiffness and precision compared to Ziegler Nichols method, a fuzzy logic PID like controller and a
serial manipulator. The control of this nonlinear MIMO system has PID controller whose gains are tuned by fuzzy logic have been
been a hot research issue in the past few decades and various compared and it was verified that tuning the gains of PID
adaptive and robust controllers have been proposed. In this paper, controller using fuzzy logic results in a better controller when the
we present a model less adaptive controller using fuzzy logic. The controlled system has nonlinearity. In [13] the use of fuzzy logic
gains of the joint space PID controllers for each leg are tuned using for tuning PID controller for non linear systems with H infinity
fuzzy logic and hence the system can effectively handle the tracking performance is well reported and the paper has shown that
problem of variation of platform mass due to loading conditions. using fuzzy logic to tune PID controller gains will give a better
The input to the fuzzy logic is a weighted sum of errors in the leg
tracking accuracy for nonlinear systems.
lengths and hence the coupling between the legs, which is
In most of the above fuzzy logic based tuning algorithms, the
drawback of independent leg PID control, is solved. Simulation
fuzzy rule base is obtained from step response of the controlled
results show that the controller has a good adaptive performance
and achieves a better tracking accuracy than simple PID controller. system with the objective of minimizing settling time. In our
implementation, the fuzzy logic is used to tune the gain of the PID
online and moreover a weighted sum of errors is taken as input of
Keywords- Adaptive Controller, Fuzzy Logic, PID controller, the fuzzy logic to consider the coupling between the legs.
Stewart platform manipulator Simulation results have shown that the controller drives the
) Stewart platform in a desired trajectory with a very good precision.
The paper is organized as follows: section two discuss the
kinematic modeling of the robot, section three gives a brief
I. INTRODUCTION overview of fuzzy logic and the design of the controller. Section
four is about simulation results and then conclusion follows.
The Stewart platform manipulator is a 6DOF parallel manipulator
having a fixed base and a moveable platform connected together .
by six extensible legs, see Fig.1. It has advantages of high
precision positioning capacity, high structural rigidity, and strong II. GEOMETRIC AND KINEMATIC MODELING
carrying capacity. Potential areas of application include flight and For geometric and kinematic modeling, the method proposed by
other motion simulators, light weight and high precision [12] is used. The centers of the universal and spherical joints are
machining, data driven manufacturing, dexterous surgical robots denoted by Bi (i =1, 2 … 6 ) and Pi (i = 1, 2 , … 6) respectively.
and active vibration control systems for large space structures Reference frames Fb and Fp are attached to the base and the
[4][14][15]. The control of this highly coupled nonlinear dynamic platform as shown in Fig.1. The position vector of the center of
system has been a hot research issue in the last few decades. universal joints Bi in frame Fb is bi and the position vector of the
center of spherical joints Pi in frame Fp is pi. Let r= [rx, ry, rz] be the
The various types of controllers designed for this platform can be position of the origin Op with respect to Ob and also let R denote
the orientation of frame Fp with respect to Fb. Thus the Cartesian
broadly classified as model less and model based controllers [6][7].
space position and orientation of the moveable platform or end
The model based controllers such as computed torque controllers
effector is specified by X=[rx, ry, rz,, α, β,  ] where the three
[1], sliding mode controllers [10] and passivity based controllers
angles α, β,  are three rotation angles that constitute the
[15] have the potential of giving higher precision in trajectory
transformation matrix R.
tracking and global asymptotic stability can be achieved if the
Then length of leg i is the magnitude of the vector BiP i which is
dynamic model parameters are identified exactly. But the given by
complexity of the control algorithms has delayed their use in
practical application and PID controllers are still in use. Bi Pi  Rpi  r - bi  q i
(1)
PID controllers, which are mostly implemented in joint space in This is the inverse geometric formula that gives the length of
SISO, do not guarantee high performance because of the highly each leg for a given desired position and orientation of the end
nonlinear dynamics of the system and variation of the payload [1]. effecter. The direct geometric model which gives the position r=[rx
Therefore we propose here to use fuzzy logic to tune the gains of ry rz] and orientation angles α, β,  for a given measured value of
PID controller to improve its performance. Fuzzy logic controllers qi, i=1, 2, …, 6 is nonlinear and is solved using numerical
have been used for various applications and [3][8] [13][16] discus methods[5].
the use of fuzzy logic for tuning the gains of a PID controller. In
The inverse kinematic model gives the velocity of the active Hence the parameters to be tuned are the three gains Kp, KD and
joint q for a given end effecter linear and angular velocity. KI. There are different tuning algorithms designed for linear
systems that aim at achieving control specifications such as set
 b vp  point following, disturbance rejection, robustness to model
q  J p  b 
b
uncertainty and rejection of measurement noise. The traditional
 ω p  tuning algorithms try to achieve one of these specifications. In the
(2) current application the system is required to follow a desired
Where bJp is the Jacobean matrix of the platform and is given by
trajectory, which means control objective is not regulation but
tracking, and moreover the system is required to operate at
a 31 (R.P1  r)X a 31  different payload conditions. Therefore the control system should
a (R.P2  r)X a 32 
be robust to model uncertainty and for that we use fuzzy logic to
 32 tune the three gains.
b
J p  .. .. .. .. 
  Fuzzy logic provides a formal methodology for representing,
a 36 (R.P6  r)X a 36  (3) manipulating and implementing a human heuristic’s knowledge
about how to control a system. The heuristic’s knowledge is
 
embedded into the controller in the form of IF…THEN rules.
Hence the fuzzy PID controller will have IF…THEN rules and an
and inference engine. The inference engine uses the rules to produce an
B i Pi output for a given input. The rules are of the following type.
a 3i  If E is PS and ER is NB then  is PS
B i Pi Where E and ER are inputs to the fuzzy logic system,  is the
(4)
output, PS and NB are linguistic values of the inputs and output.

Error, E and rate of error, ER are inputs to the controller and the
output  is the change in control gains. The terms PS and NB stand
for positive small and negative big respectively. Then a decision is
made about the current output value using an inference mechanism
and the rule base which contains IF…THEN rules. A final output
interface part converts the decision to a numerical value. The fuzzy
logic tuned PID control system discussed in this paper is shown in
Fig.2.

For the fuzzy logic gain tuning, error E and rate of error ER are
each divided into five linguistic values of negative big (NB),
negative small(NS), Zero(Z), positive big(PB) and positive small
(PS). The membership functions used are symmetric triangular
functions and they are shown in Fig.3. The output of the fuzzy
controller α used to change the gains of the PID controller as given
by the following algorithm.
K p k  1  K p (k)  α(k)e(k)K p 0
(7)
K D (k  1)  K D (k)  α(k)e (k)K D (0) (8)
Figure 1 Generalized Stewart Platform
K I (k  1)  K I (k)  α(k)e(k)e (k)K I (0) (9)
III. DESIGN OF CONTROLLER The starting value for the three terms is taken from a separate
A PID controller in its standard form is given as PID controller tuned manually and which gives acceptable output..
 det  1 
t
u t   K p et   Td   et dt 
IV. SIMULATION RESULTS

 dt Ti 0  For the simulation study of the performance of the


(5) controller, a typical 6-6 geometry Stewart platform with the
Multiplying the gain term outside the bracket by the geometric parameters given in table I is implemented using
derivative and integral time constants, the equation can also simmechanics tool box of MATLAB. The fuzzy logic tuner
be written as is also implemented using fuzzy logic toolbox of MATLAB
and three mfile sfunctions were written to implement the
det 
t
u t   K p et   K D  K I et dt
discrete difference equation given in (4)-(6). The trajectory
dt 0
 taken to test the performance of the controller is heave or
(6)
Figure 2 Fuzzy tuned joint space PID control of a single
leg

uplift motion in z direction for 2.5 sec followed by circular Figure 4 Graph of membership functions for error rate,
motion in the x-y plane. ER
.
To make the trajectory twice differentiable, the trajectories.
are joined with parabolic blend. The desired trajectories and
the actual trajectory for the simple PID and fuzzy tuned PID
is given in Figs. 6-8.

From the simulation results it can be seen that the


controller can track the desired trajectory to a better degree
of accuracy than the simple PID controller. The gains of the
controller used for the simulation are KP= 8.5x107,
KD=9x105, and KI=100 and the tracking error for these
values is given in Fig. 9 and 10.

Figure 5 Membership functions for output variable

TABLE I
Geometric Specifications of Stewart platform
Joint positions
Base

Platform

Base radius 0.8m


Platform radius 0.5m

Figure 3 Graph of membership functions for error E


CONCLUSION REFERENCES
A conclusion section is not required. Although a
conclusion may review the main points of the paper, do not [1] A. Ghobakhloo, M. Eghtesad and M. Azadi, “Position Control of
replicate the abstract as the conclusion. A conclusion might Stewart-Gough Platform Using Inverse Dynamics with Full
Dynamics,” IEEE Conference 9th Int. Workshop on Advanced Motion
elaborate on the importance of the work or suggest Control (AMC’06), PP.50-55, Istanbul-Turkey 2006.
applications and extensions
[2] A.Visioli,, ”Tuning of PID Controllers Using Fuzzy Logic”, IEE
Proc.-Control Theory Appl., Vol. 148, No. I , PP.1-8, January 2001
[4] B. Dasgupta and T. S. Mruthyunthaya, “The Stewart Platform
Manipulator: a review,” Mechanism and Machine Theory, vol. 35,
No. 1 PP. 15-40, 2000

[5] Charles C. Nguyen, Zhen-Lie Zhou and Sami S. Antraz, “Efficient


Computation of Forward Kinematics and Jacobean Matrix of A
Stewart Platform Based Manipulator,” Proceedings of the 1997 IEEE
Int. Conf. on Robotics and Automation vol. 1, PP.869-874.

[6] Dongya Zhao, Shaoyuan Li, and Feng Gao, “Fully Adaptive
Feedforward Feedback Synchronized Tracking Control for Stewart
Platform Systems”, International Journal of Control, Automation, and
Systems, vol. 6, no. 5, PP. 689-701, October 2008

[7] Fathi H. Ghorbel, Olivier Chételat, Ruvinda Gunawardana,


andRoland Longchamp, Modeling And Set Point Control Of Closed
Chain Mechanisms, Theory And Experiment, IEEE Transactions On
Control Systems Technology, Vol. 8, No. 5, PP. 801-815, September
2000
Figure 6 Three dimensional view of the desired
trajectory [8] Hyeong-Pyo Hong, Suk-Joon Park, Sang-Joon Han, Kyeong-Young
Cho, Young-Chul Lim, Jong-Kun Park, Tae-Gon Kim, “A Design of
Auto-Tuning PID Controller Using Fuzzy Logic”,

[9] Lee Hyung San and Myug-Chul Han, “ The Estimation for Forward
Kinematics of Stewart Platform Using the Neural Network,” Proc. Of
the 1999 IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pp.
501-506.

[10] Nag-In Kim and Chong-Won Lee,” High Speed Tracking Control of
Stewart Platform Manipulator Via Enhanced Sliding Model
Controller”, Proceeding of the 1998 International Conference on
Robotics and Automation, PP. 2716-2721, Leuven Belgium, May
1998

[11] S. Iqbal, A.I. Bhatti, “Dynamic Analysis and Robust Control Design
Figure 7 Task Space tracking errors for x, y and z for Stewart Platform With Moving Payloads”, Proceedings of the
17th World Congress The International Federation of Automatic
directions Control,PP. 5324-5329, Seoul, Korea, July 6-11, 2008

[12] Wisama Khalil & Ouarda Ibrahim, “General Solution for the
Dynamic Modeling of Parallel Robots”, Journal of Intelligent Robot
Systems, Vol. 49, PP.19-37,2007

[13] Wen-Shyong Yu, “Adaptive Fuzzy PID Control for Nonlinear


Systems with H∞ Tracking Performance”, 2006 IEEE International
Conference on Fuzzy Systems, Sheraton Vancouver, Canada, PP.
1010-1015,
July 16-21, 2006

[14] Y. X. Su and B. Y. Duan, “The Application of the Stewart Platform in


Large Spherical Radio Telescopes,” Journal of Robotic Systems
vol.17 No.7, 2000, PP. 375-383.

[15] Yung Ting, Yu-Shin Chen and Shih-Ming Wang, ”Task Space
Control Algorithm for Stewart Platform,” Proc. Of the 38 th Conf. on
Decision and Control, PP. 3857-3862, Dec. 1999, Phoenix, Arizona-
USA
Figure 8 Joint space tracking performance
[3] A. Omran A. Omran, G. El-Bayiumi, M. Bayoumi, and A. [16] Zhen-Yu Zhao, Masayoshi Tomizuka, and Satoru Isaka, “Fuzzy Gain
Kassem,”Genetic Algorithm Based Optimal Control for a 6DOF Non Scheduling of PID Controllers”, IEEE Transactions On Systems,
Redundant Stewart Platform”, International Journal of Mechanical, Man, And Cybernetics. Vol. 23, No. 5. September/october 1993
Industrial and Aerospace Engineering, Vol.2, No.2, PP. 73-79, 2008

Potrebbero piacerti anche