Professional Documents
Culture Documents
Design and Implementation of PI and Fuzzy PID Supervisory Controllers For A Flexible Link Robot
Design and Implementation of PI and Fuzzy PID Supervisory Controllers For A Flexible Link Robot
Design and Implementation of PI and Fuzzy PID Supervisory Controllers For A Flexible Link Robot
Abstract—In this paper, PID controllers are designed to controller, error and its time-derivative are inputs. In the
enhance the trajectory tracking performance of a flexible link comparison of the PI and fuzzy PID controller it can be seen
robot. Firstly, a PI controller is used where conventional that fuzzy PID response is faster. The experimental results
Ziegler –Nichols method adjusts the controller parameters. In confirm that fuzzy PID has better control on the vibrations of
addition, Fuzzy PID controller is designed in which fuzzy rules flexible link robot.
are utilized on-line to tune the PID controller parameters This paper has been organized as follows: Section II
based on the current value of error and its first time-derivative. reviews modeling of the flexible link. Section III presents PI
Also, effects of several defuzzification methods on the closed- and fuzzy logic that is used to control of the flexible link.
loop system are investigated in this paper. A flexible link robot
Simulation and experimental results are shown in section IV.
is implemented and controllers are applied. Experimental
results confirm the performance of the fuzzy PID controller.
Finally the conclusion is presented in section V.
II. FLEXIBLE LINK DYNAMIC MODELING
Keywords—fuzzy supervisory PID control; flexible link
robot; PID controller; Ziegler –Nichols Euler–Bernoulli beam theory simplifies the linear theory
of elasticity which provides a means for calculating the load-
I. INTRODUCTION carrying and deflection characteristics of beams.
In the past two decades, flexible robots have been an In the following we focus on the modeling of a single
interesting research area in engineering. Flexible-link robotic flexible link manipulator which is fixed at one end (hub) and
manipulators have many advantages compared to is driven by a motor. The other end is free to flex in
conventional heavy and rigid robots such as lower cost, horizontal plane. The length of the beam is much greater
lower energy consumption, smaller actuators, a larger work than its width, so that the beam is restricted to oscillate
space, higher operational speed, a larger payload-to- A schematic representation of the single link flexible
manipulator ratio. Apparently, flexible link manipulators manipulator (SLFM) is shown in Fig. 1.
have a complicated structure so that their dynamical analysis
and control is much more complex comparatively. Moreover, Y
the number of controlled variables for a flexible-link
manipulator is less than the number of mechanical degrees of
freedom; therefore the control objectives are double in this
case. Firstly, a flexible manipulator controller must achieve θg
the same motion path as a rigid manipulator. Secondly, m1
naturally stable vibrating modes must kept stable [1]. u
Control strategy for flexible arm does not only include
vibration control of the arm with reasonable accuracy but
achieving fast and accurate positioning. Sliding mode control
(SMC) [2], fuzzy logic control (FLC) [3, 4, 5], linear
quadratic regulator (LQR) [6], and nonlinear control methods
[7].are common control algorithms for flexible link robot.
This paper begins with a short discussion of modelling X
the behavior of a flexible beam. The analysis of a flexible Figure 1. Diagram of flexible link.
system begins with developing a model that describes the
position of each point in the system relative to a suitable For an angular displacement θ and an elastic deflection u,
inertial reference frame. The model is used to simulate the the total (net) displacement y(x,t) of a point along the
robot response for a desire input. For this robot a classic PI manipulator located at a distance x from the hub is defined as
and a fuzzy PID are designed. The parameters of PI are tuned a function of both the elastic deflection u(x, t) measured from
via Ziegler Nichols Method (ZN) [8]. In the fuzzy PID the line OX and rigid body motion θ(t)[9].
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.
ݕሺݔǡ ݐሻ ൌ ߠݔሺݐሻ ݑሺݔǡ ݐሻ (1) ¹ସ ൌ ͳͲǤͻͻͷଶ ൈ ʹͻǤͶ ൌ ͵ͷʹǤʹ ՜ ସ ൌ
ͷǤʹͻ
The deflection of any point on the beam is given by the
Euler-Bernoulli beam equation by neglecting the effects of The structure studied is made of strain steel. The flexible
shear deformation and rotary inertia. beam with mass m1 is assumed to be concentrated in its
Equation (2) gives the fourth-order partial differential middle point. Strain gauge is used for dynamic deflection
equation (PDE), which states the dynamic equation measurement. The location for the strain measurements
describing the motion of the flexible manipulator [9]. is should be equally spread along the link. Due to relatively
Young Modulus ( ൌ ʹǤͳ ൈ ͳͲଵଵ ܰΤ݉ଶ ), and ² is the small changes in resistance during measurement, the strain
beam density. gauges are usually wired as a Wheatstone-bridge, and the
output voltage is measured by an instrumentation amplifier.
߲ ସ ݕሺݔǡ ݐሻ ߲ ଶ ݕሺݔǡ ݐሻ (2) Full bridges with four strain gauges are preferred for bending
ܫܧ ߩ ൌͲ measurements because they are self-compensated for
߲ ݔସ ߲ ݐଶ
temperature effects and torsional forces. Fig. 2 shows the
Before developing a dynamic model for the arm, it is implemented robot.
necessary to find the natural modes of the arm. The natural
frequencies of vibration are obtained from the following
equation [10].
(3)
ܫܧ ܫܧ
ܹ ൌ ߚଶඨ ൌ ሺߚ݈ሻଶ ඨ
ߩܣ ߩ ݈ܣସ
refers to the moment of inertia of the beam cross section
for a rectangular cross section and
271
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.
Integrating the equation four times for each of the two PI algorithm is used in order to compute the control
intervals in which the beam is divided by the middle mass m1 signal that actuated the robot based on the following formula.
yields the followings
௧
ଷ ଶ
ݕଵ ሺݔሻ ൌ ܽଷ ݔ ܽଶ ݔ ܽଵ ݔ ܽ ǡ א ݔሾͲǡ ܮΤʹሿ (12) ݑሺݐሻ ൌ ܭ ݁ሺݐሻ ܭூ න ݁ሺݏሻ݀ݏ
(18)
ଷ ଶ
ݕଶ ሺݔሻ ൌ ܾଷ ݔ ܾଶ ݔ ܾଵ ݔ ܾ ǡ א ݔሾܮΤʹ ǡ ܮሿ (13) ݁ሺݐሻ ൌ ݎሺݐሻ െ ݕሺݐሻ
By considering the set of continuity conditions for where, ݑሺݐሻ is the control input, ݁ሺݐሻ, and ݕሺݐሻ are error and
constants terms ୨ and ୨ , dynamic conditions, and the output signals, respectively. ܭ and ܭூ stand for controller
geometrical conditions that are presented in paper [12], the parameters, which need to be tuned.
state space equations of the robot are B. Fuzzy PID Control
Ͳଷൈଷ ܫଷൈଷ (14) PI controller has a simple control structure which is easy
െͳͻʹߜ ͳͻʹߜ െͶͺߜ to understand but the response of PI controller is not fast and
ݔሶ ௦ ൌ ൦
Ͷͺߩ െͲߩ ͳͺߩ Ͳଷൈଷ
൪ ݔ௦ it robustness to un-modeled dynamics is not enough. To
overcome this problem fuzzy PID controller is designed.
െͳʹߛ ͳͺߛ െߛ
Ͳଷൈଵ Fuzzy logic controllers (FLCs) are based on experience
of a human operator and they are increasingly applied to
Ͷͺߜ many systems with nonlinearity and uncertainty.
൦ ൪ߠ
െߩ
ߛ kp
Fuzzy Logic
ݕ௦ ൌ ሾܫଷൈଷ Ͳଷൈଷ ሿݔ௦ ሾͲଷൈଵ ሿߠ (15) Controller
kd
alpha
e
୬ൈ୬ is the Identity matrix of dimension݊ ൈ ݊, the state d
vector is ୱ ൌ ൣ©ଵ ǡ ©୲ ǡ © ǡ ©ሶ ଵ ǡ ©ሶ ୲ ǡ ©ሶ ൧ , ୱ is output P
Reference +
୍ ୍ ୍ angle I Flexible link
vector, and constants ߜ ൌ and ¤ ൌ ൌ య haveǡ ² - Scope
୫భ ୫ሚ D
been defined for the sake of concision. Therefore, transfer
function between the output and the input of the robot is Figure 4. The Simulink model of supervisory fuzzy PID controller for
single flexible link robot.
ߠ௧ ሺݏሻ െߩሺ ݏସ ሺͶߛ െ ͳͻʹߜሻ ݏସ െ ͵ͺͶߜߛሻ (16)
ൌ A fuzzy controller measures the outputs of the process
ߠ ሺݏሻ ο and continuously controls the process actions. The fuzzy
οൌ ݏ ሺߛ Ͳߩ ͳͻʹߜሻ ݏସ (17) controller utilizes a form of quantification of inexplicit
ሺͻߛߩ ͺߛߜ ʹ͵ͲͶߩߜሻ ݏଶ information (input fuzzy sets) and a knowledge base to
generate control force to be applied to the system. Simulink
ʹ͵ͲͶߛߩߜ
model of the closed-loop system with fuzzy controller is
depicted in Fig. 4.
Eq. (18) represents the dynamics of the flexible beam. Fuzzy PID is a technique where PID controller
III. CONTROL SCHEME parameters (gains) are tuned during control process, using
the fuzzy rule base. It is not always known as an adaptive
The control scheme proposed in this paper is based on control, but it enlarges the operation area of a simple PID
classic PI and fuzzy PID controllers. controller and performs well also for a nonlinear plant with
A. PI Control unpredictable parameter variations.
In the proposed work, a fuzzy supervisor with two
PI (Proportional – Integral) control is one of the classical inputs/three outputs is used to schedule the controller
control methods which is mostly used in industry because of parameters according to predefined parameter values with
its performance and simple structure. PI control block respect to changing operation conditions. The PID
diagram of flexible link robot is shown in Fig. 3. parameters are adjusted by using fuzzy inference, which
yields a nonlinear mapping from the error (e) and error
Gain derivative (݁ሶ) to PID parameters.
Reference +
The benefit of this quantification is that the fuzzy sets can
angle + + Flexible link
Scope
be expressed by a unique linguistic expression, such as
- medium, small, or large. The linguistic representation of a
Gain 1/S
fuzzy set is also known as a term, and a collection of such
terms represents a term-set, or a library of fuzzy sets. Fuzzy
control turns a linguistic control strategy into an automatic
Figure 3. PI controller block diagram for flexible link robot. control strategy.
272
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.
It is considered thatܭ , ܭௗ are in predetermined ranges
[ܭǡ ,ܭǡ௫ ] and [ܭௗǡ ,ܭௗǡ௫ ]. For convenience,ܭ
and ܭௗ are normalized into the range between zero and one
by the following linear transformation:
ܶ ൌ ߙܶௗ (20)
ܭ ൌ ܭ Τሺߙܶௗ ሻ ൌ ܭଶ Τሺߙܭௗ ሻ (21) After calculating output of fuzzy supervisor, the PID
controller parameters are computed from the following
The parameters ܭሖ , ܭሖௗ and ߙ are determined by a set of equations [14]:
fuzzy rules such as
If ݁ሺ݇ሻ is ܣ and ο݁ሺ݇ሻ is ܤ then ܭሖ is ܥ , ܭሖௗ is ܦ and ܭ ൌ ൫ܭǡ௫ െ ܭǡ ൯ܭሖ ܭǡ (22)
ߙ ൌ ߙ , ݅ ൌ ͳǡʹǡ ǥ ǡ ݉. ܭௗ ൌ ൫ܭௗǡ௫ െ ܭௗǡ ൯ܭሖௗ ܭௗǡ
According to the Ziegler-Nichols PID tuning rule, the ܭ ൌ ܭଶ Τሺߙܭௗ ሻ
integral time constant ܶ is always taken four times larger
than the derivative time constant, so α is equal to 4. The set of rules which define the relation between the
Fig. 5 illustrates the fuzzy membership functions of e and input and output of fuzzy controller are defined using the
݁ሶ . Each input has seven triangular and trapezoidal
linguistic variables. All the 25 rules governing the
membership functions. The membership functions of ݇ᇱ ,݇ௗᇱ , mechanism of each output are explained in Tables Ȥ to ȤȤȤ.
and alpha are depicted in Figs. 6 and 7.
TABLE I. FUZZY TUNING RULES FOR ܭሖ
ο܍ሺ۹ሻ
N NM NS ZO PS PM PB
B
NB B B B B B B B
NM S B B B B B S
܍ሺ۹ሻ NS S S B B B S S
ZO S S S B S S S
PS S S B B B S S
PM S B B B B B S
PB B B B B B B B
Figure 5. Membership functions of e and ݁ሶ .
TABLE II. FUZZY TUNING RULES FOR ܭሖௗ
ο܍ሺ۹ሻ
N NM NS ZO PS PM PB
B
NB B B B B B B B
NM S B B B B B S
܍ሺ۹ሻ NS S S B B B S S
ZO S S S B S S S
PS S S B B B S S
PM S B B B B B S
PB B B B B B B B
273
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.
TABLE III. FUZZY TUNING RULES FOR Ȝ methods are presented. Based on Fig. 9, the LOM
ο܍ሺ۹ሻ defuzzification method in the fuzzy controller yields better
performance (such as eliminating overshoot, rise time, and
NB NM NS ZO PS PM PB
steady state error) [13].
NB 2 2 2 2 2 2 2 In order to practically validate simulation results, a
NM 3 3 3 3 2 3 3 prototype flexible link robot is implemented. By applying a
܍ሺ۹ሻ NS 4 3 3 2 3 3 4 small disturbance to the robot, the vibrations are damped in
ZO 5 4 3 3 3 4 5 smaller time intervals time. Fig. 10, shows control input
signal of fuzzy PID controller.
PS 4 3 3 2 3 3 4
PM 3 3 2 2 2 3 3
PB 2 2 2 2 2 2 2
IV. RESULTS
The proposed fuzzy PID controller is compared by
Ziegler-Nichols tuned PI controller, and fine-tuned PI
controller (which is tuned by MATLAB/SIMULINK auto
tuning option). Results are presented for comparison in Fig.
8. Based on this figure, Ziegler-Nichols based PI controller
does not yield satisfactory results. However, MATLAB
tuned PI and fuzzy PID controllers have better performance.
As figure illustrates, for comparing PI and fuzzy supervisory
Figure 10. Vibration signal of the closed-loop system by fuzzy PID
control, tuning PID parameters with fuzzy method improves controller.
the tracking error of the reference signal.
V. CONCLUSION
In this paper, a study on several PID controllers is done
to improve the tracking performance of flexible link robot.
As shown in the simulation, the fuzzy PID controller has
better performance. Fuzzy PID with LOM defuzzification
method improves the reference signal tracking error of the
closed-loop system. The experimental results are presented
which confirm the supervisory of the fuzzy PID controller.
REFERENCES
[1] M. Moallem, R.V. Patel, and K. Khorasani, “Nonlinear tip-position
tracking control of a fexible-link manipulator: theory and
experiments,” in Automatica, vol. 37, 2001, pp 1825-1834.
Figure 8. Position of closed-loop system by ZN tuned PI, MATLAB [2] J.F. Peza-Solís, G. Silva-Navarro, and N. R. Castro-Linares,
tuned PI and fuzzy PID controllers. “Trajectory tracking control in a single flexible-link robot using finite
differences and sliding modes,” Journal of Applied Research and
Technology, vol. 13, 2015, pp. 70-78.
[3] J.M. Renno, “Inverse dynamics-based tuning of a fuzzy logic
controller for a single-link flexible manipulator,” in Journal of
Vibration and Control, vol. 13, no. 12, 2007, pp. 1741-1759.
[4] P. Sooraksa, and C. Guanrong, “Mathematical Modeling and Fuzzy
Control of a Flexible-Link Robot Arm,” Mathl. Comput. Modelling,
vol. 27, no. 6, 1998, pp. 73-93.
[5] K.Y. Kuo a, and J. Linb, “Fuzzy logic control for flexible link robot
arm by Singular perturbation approach,” Applied Soft Computing,
Vol. 2, 2002, pp. 24–38.
[6] R. Dinesh Singh, and Deepika, “ Modelling, stability analysis and
control of flexible single link robotic manipulator,” in International
Journal of Advanced Research in Electrical, Electronics and
Instrumentation Engineering, vol. 3, issue 2, 2014, pp. 7390-7401.
[7] A. Shawky, D. Zydek, Y. Z. Elhalwagy, and A. Ordys, “Modeling
Figure 9. Position of closed-loop system by fuzzy PID controller by and nonlinear control of a flexible-link manipulator,” Applied
different defuzzification methods. Mathematical Modelling, vol. 37, issue. 23, 2013, pp. 9591-9602.
[8] G. Mallesham, and A. Rajani, “ Automatic Tuning of PID Controller
In addition, the effect of defuzzification method is Using Fuzzy Logic,” 8th International Conference on Development
studied. In the Fig. 9, only centroid and LOM defuzzification and Application Systems, Suceava, Romania, May 25-27, 2006.
274
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.
[9] M.O. Tokhi and A.K.M. Azad, “Flexible robot manipulators using disturbance observers,” Robotics and Autonomous Systems, vol.
modelling, simulation and control,” The Institution of Engineering 60, 2012, pp. 532–547.
and Technology, London, United Kingdom, 2008. [13] A. Ilyas, S. Jahan, and M. Ayyub, “Tuning of conventional PID and
[10] Singiresu S. Rao, Mechanical Vibrations, 5th ed., University of fuzzy logic controller using different defuzzification techniques,”
Miami, 2011, pp. 721-727. International Journal of Scientific & Technology Research, vol. 2,
[11] S. Kilicaslan, M. Kemal Özgören, and S. Kemal Ider, “Hybrid force Issuse. 1, January 2013.
and motion control of robots with flexible links,” Mechanism and [14] Zhen Yu Zhao, “Fuzzy gain scheduling of pid controller,” IEEE
Machine Theory, vol. 45, 2010, pp. 91–105. Transactions on Systems Man and Cybernetics, vol. 23, no. 5,
[12] R. Morales a, and V. Feliu, V. Jaramillo, “Position control of very September/October 1993.
lightweight single-link flexible arms with large payload variations by
275
Authorized licensed use limited to: Universitaetsbibliothek der RWTH Aachen. Downloaded on March 31,2021 at 07:19:13 UTC from IEEE Xplore. Restrictions apply.