Upload
infodotz
View
214
Download
0
Embed Size (px)
Citation preview
7/30/2019 21E8F20Cd01
1/6
Hybrid PD/PID Controller Design for Two-Link
Flexible Manipulators
Rasheedat M. MahamoodSchool of Mechanical, Industrial and Aeronautical
Engineering, University of the Witwatersrand,
Johannesburg, South Africa
Email: [email protected]
Jimoh O. PedroSchool of Mechanical, Industrial and Aeronautical
Engineering, University of the Witwatersrand,
Johannesburg, South Africa
Email: [email protected]
AbstractThis paper investigates the development of a hybridcollocated PD and non-collocated PID controller designed forinput tracking and vibration control of two-link flexible ma-nipulator. The two-link robot manipulator was modelled using
Lagrange and assumed mode method. The PD controller is usedfor motion tracking and the PID for vibration control. Effect ofchanging controller gains on performance is studied using twocase studies. Also studied is the effect of payload variation onthe performance of the proposed controller. The performance ofthe designed controllers is evaluated in terms of input trackingcapability, energy utilization, deflection suppression and vibrationcontrol. Results show that a simple PD-PID controller canbe effectively designed for point-to-point motion control andvibration suppression for two link flexible manipulators. Also thestudy reveals that the controller is robust to payload variation.
I. INTRODUCTION
Advantages of flexible robot manipulators over their rigid
counterparts cannot be overemphasised: they require lessmaterial, lower power consumption, have higher manipulation
speed, can use smaller actuators, are more manoeuvrable and
transportable, are safer to operate due to reduced inertia,
higher payload to robot weight ratio and most importantly
they have less overall cost. Amidst the aforementioned
advantages, the control of flexible manipulators to maintain
accurate positioning is very challenging. The flexible nature
and distributed characteristics of the system makes the
dynamics a highly non-linear one [1]. Application of flexible
link robot in industry is expected to increase only if their
performance is improved.
The control strategies for flexible manipulator systemsare classified as: feedforward (open-loop) and feedback
(closed-loop) control [1]. Open-loop control (feedforward)
[2], [3] [4], which is the simplest method does not require
any measurement from the plant for the control action to
be implemented. The problem with the open loop control
is that exact knowledge of the plant is required. Feedback
control strategies for Flexible Manipulator Systems (FMSs)
are classified as collocated and non-collocated control.
Collocated means the actuators and the sensors are at the
same location. It is used to guarantee stable control of
rigid-body motion. Non-collocated control on the other hand
means that the actuators and the sensors are at different
locations. Closed loop (feedback) control technique utilizes
an accurate real time monitoring of the plant to be controlled
for successful implementation of control action. Different
methods have been used in closed-loop form to controlflexible link manipulator. Examples include Proportional-
Integral-Derivative (PID) [5], end-point acceleration feedback
[6], [7], state feedback [8], optimal control technique [9],
robust control techniques [10], and singular perturbation
method [11], [12].
The most widely used form of industrial controllers is the
PID Controller. They constitute more than 90% of feedback
controller used today [13]. This is because it is cheap, simple
in structure, and robust in performance over a wide range
of operating conditions [14]. PID control is also good at
dealing with actuator saturation and integrator windup [15].
This is why many authors have designed controller for FMSsbased on PID control technique [13], [15], [16], [17], [18],
[19]. Tokhi and Azad [5] carried out a comprehensive study
on open loop control and a hybrid collocated proportional
derivative (PD) and non-collocated PID control strategy for
single-link flexible manipulator. Simulation and experimental
results showed a better performance in the proposed hybrid
PD-PID controllers. Cheong el al. [20] also developed a PID
composite controller for single link flexible manipulator. PD
and a disturbance observer were proposed to control the slow
dynamics while PID for fast dynamics. Experimental results
show the effectiveness of the proposed controller.
The literature shows that relatively few PID controllershave been used to control FMSs compared to their rigid
counterpart. The reason can be associated to problem of the
common tuning methods that shows sluggish responses when
applied to a non-minimum phase system like FMSs [17].
In this study, a hybrid PD-PID controller is developed for
two-link FMSs. The manipulator is modelled using Lagrange
and assumed mode methods. The PD controller is for point to
point motion control, while the PID controller is for vibration
suppression. Simulation is performed within Matlab/Simulink
environment for evaluation of the control strategies. A
unit-step response analysis is conducted, and performance
evaluation of the control strategies is performed in terms of
Proceedings of 2011 8th Asian Control Conference (ASCC) WeB1.4
Kaohsiung, Taiwan, May 15-18, 2011
- 1358 -
7/30/2019 21E8F20Cd01
2/6
Fig. 1. Two-link flexible manipulator system
reference tracking, deflection, end-point acceleration, and
input torque. Effects of varying payload on the proposed
controller is also studied. The results are presented and
discussed. The paper ends with concluding remarks.
I I . SYSTEM MATHEMATICAL MODEL
A. Modelling of robotic manipulators
Lagrangian approach is commonly used to derive the dy-
namic equations of motion of flexible multi-body systems,
although there are three main methods used in the literature:
Newton-Euler, Lagrangian approach and Hamilton approach
[21]. Assumed mode method is the most used approxima-
tion method for reducing partial differential equation (PDE)
(equation of motion) into ordinary differential equation (ODE)
[22]. The first two modes are adequate to describe the system
dynamics [23]. The model of the two-link flexible, planar,
manipulators derived by [22] is used in this study. The linksare modelled as Euler-Bernoulli beams, with proper clamped-
mass boundary conditions. It is assumed the beams elastic
deflections are small and no deflections in the axial direction.
B. Formulation of the recursive kinematics equations
Figure 1 shows a two-link flexible robot manipulator sys-
tem, both links are actuated by individual motors at the hubs.
(X0Y0), (XiYi), and (XiYi) are the inertial frame, the rigidbody moving frame, and the flexible body moving frame
associated with the ith link. i is the angular position of theith link, and yi(xi) is the transversal deflection of the ith
link (0 xi li) where li is the length of the ith link.The rigid transformation matrix and the elastic homogenous
transformation matrix due to the deflection of the link are
defined respectively as:
Ai =
cos i sin isin i cos i
and Ei =
1 yie
yie 1
(1)
where yie =yixi
xi=li
, and tan1(yie) y
ie(small
deflections assumption). The global transformation matrix
Wi transforming coordinates from X0Y0 to XiYi follows arecursion as:
Wi = Wi1Ei1Ai = Wi1Ai, W0 = I (2)
The previous absolute position vectors pi of a point along thedeflected ith link, is defined by recursive kinematics equations:
pi = ri + Wiipi, ri+1 = ri + W
ii ri+1 (3)
where
i
pi =i
pi(xi) = (xiyi(xi))T
is the position of a pointalong the deflected ith link, with respect to frame (Xi, Yi),and iri+1 =
i pi(li) = (liyi(li))T being the position of the
origin of frame (Xi, Yi). The absolute velocity of this pointpi on the links is:
pi = ri +Wiipi + W
ii pi, ri+1 = ri + W
ii ri+1+ W
ii ri+1 (4)
and ri+1 =i pi(li), with
i pi(xi) = (0 yi(xi))T. The links are
assumed inextensible in the longitudinal direction. The rates
of the recursions take the form of:
Wi =Wi1Ai + Wi1Ai
Wi = WiEi + WiEi (5)
C. Lagrangian formulation
The system total kinetic energy T is given by:
T =n
i=1
Thi +n
i=1
Tli + Tp (6)
where Thi , Tli , and Tp are the kinetic energies of the ith hub,
ith link, and the payload, respectively. The ith hub kineticenergy is given by:
Thi =1
2mhi r
Ti ri +
1
2Jhi
2i (7)
where mhi is the mass of the ith hub, Jhi is the moment of
inertia of the ith hub, and i is the absolute angular velocity
of frame (Xi, Yi):
i =i
j=1
j +i
k=1
yke (8)
The kinetic energy of the ith link is given by:
Tli =1
2
li0
i(xi) pTi (xi) pi(xi)dxi (9)
where i is the linear density of the ith link. The kinetic energy
associated with the payload is given by:
Tp =1
2mpr
Tn+1rn+1 +
1
2Jp
2i + y
ne(10)
where mp and Jp are the mass and moment of inertia of thepayload located at the end of linkn. The total potential energyU is given by:
U =n
i=1
Ui =n
i=1
1
2
li0
(EI)i(xi)
d2yi(xi)
dx2i
2dxi (11)
Ui is the elastic energy stored in the ith link, with (EI)i
being its flexural rigidity.
Computing the total kinetic energy T and potential energy U,then the Lagrangian L is given by:
L = T U (12)
- 1359 -
7/30/2019 21E8F20Cd01
3/6
D. Assumed mode shapes
The links are modelled as Euler-Bernoulli beams and they
satisfy the following equation:
(EI)i 4
yi(xi, t)x4i
+ i 2
yi(xi, t)t2
= 0, i = 1, . . . , n (13)
where yi is the deflection of the ith link. Equation (13) is a
partial differential equation satisfying the following boundary
conditions:
yi(0, t) = 0, y
i(0, t) = 0, i = 1, . . . , n (14)
Assuming a nei number of modes, deflection of each link canbe obtained by the method of separation of variables as:
yi(xi, t) =
nei
j=1
ij(xi)ij(t) (15)
where ij(t) are the time varying variable associated with thespecial mode shape function ij(x) of the i
th link. Solution
of the two variables are as follows:
ij(xi) = C1,ij sin(ijxi) + C2,ij cos(ijxi)
+ C3,ij sinh(ijxi) + C4,ij cosh(ijxi) (16)
ij(t) = exp(jijt)
= C5,ij sin(ijt) + C6,ij cos(ijt) (17)
where:4ij =
2iji/(EI)i (18)
ij is the natural angular frequency of the ith link,
C1,ij . . . C 6,ij are constants obtained from the followingboundary conditions, Eq. (14). This yields:
C2,ij + C4,ij = 0, C1,ij + C3,ij = 0 (19)
E. Dynamic equations of motion
The dynamic model is formulated using Lagrange-Euler
equation:
ddt
Lqi
L
qi= i, i = 1, . . . , n (20)
Solution of Eq. (20) yields the closed form equation:
B(q(t))q(t) + h(q(t), q(t)) +Kq(t) = (t) (21)
where q(t) =
1, . . . , n, 11, . . . , 1ne1 , . . . , n1, . . . , nnei
is a N-vector generalised coordinates (N = n +
i=1 nei), is an n-vector of generalized torques applied at the joints. B
is a positive-definite symmetric inertia matrix, h is a vector of
Coriolis and centripetal forces, and K is the diagonal stiffness
matrix. Detailed derivation of the mathematical model can be
found in [22].
Fig. 2. PD-PID controller structure for the two-link flexible manipulator
III. CONTROLLER DESIGN
The control objective for the two-link flexible manipulator
shown in Fig. 1 is to design PD collocated controllers for
each link so that the hub angles follow the reference trajecto-
ries.Also to design non-collocated PID controllers so that thevibrations of the end effectors are eliminated simultaneously.
There are two stages involved in the controller design. the
first stage involves the design of PD controllers for hub angle
motion; while the second stage is concerned with the PID
controller for the vibration control of the two links.
A. Collocated PD Controller
Similar PD control structures are used for both links, they
only differ in gains. The PD control input is given by:
uPDi(t) = Aci
KPi (id(t) i(t))Kvi i(t)
i = 1, 2
(22)
where: uPDi(t) is the PD control input, id(t), i(t) andi(t)are the desired joint angle, actual joint angle and actual joint
velocity ofith-link respectively, KPi , Kvi , and Aci are the PDcontrol proportional and derivative gains, and motor amplifier
gain for the ith-link respectively.
B. Non-Collocated PID Controller
A separate PID controller is designed for the control of end-
point elastic acceleration of each of the links; this is necessary
because of the coupling effects. The control input for the ith-link is as follows:
uPIDi(t) = KPiei(t) + KIi
ei(t)dt + KDi ei(t) i = 1, 2
(23)
where ei(t) = id i(t), id , and i(t) are the desiredand actual tip acceleration of ith-link respectively. One of thecontrol objectives is to achieve zero elastic acceleration which
corresponds to zero vibration, hence id = 0.
C. Hybrid Controller
The two control schemes are combined as shown in Fig. 2.
The motion tracking and the vibration suppression are required
to be achieved simultaneously. The total control input i(t)from each motor is given by adding (22) and (23):
i(t) =2
i=1[uPDi(t) + uPIDi(t)] (24)
- 1360 -
7/30/2019 21E8F20Cd01
4/6
D. Performance Evaluation
The performance of the proposed controller is investigated
using the performance index:
J =
1
tf(J1 + J2 + J3) (25)
J1 =
tf0
2i=1
id i(t)
imax
2+
id i(t)
imax
2 dt
(26)
J2 =
tf0
2
i=1
yid yi(t)
yimax
2+
id i(t)
imax
2dt
(27)
J3 =
tf0
2
i=1
id(t) i(t)
imax
2dt (28)
J is the overall performance index; J1 is the performance
index related to the rigid body motion tracking; J2 is the per-formance index related to the deflection and vibration control
of the manipulator; J3 is the performance index related to theoverall control inputs at the joints; tf, imax, imax, yimax,imax, and imax are the final simulation time, maximum hubangle, maximum hub angular velocity, maximum deflection,
maximum tip acceleration, and maximum actuator torque of
the link i respectively. yi(t) and i(t) are the deflection andtorque of the ith link respectively. yid is the desired hubdeflection of the ith link.
IV. SIMULATION RESULTS
Numerical simulation was carried out on the two-link
flexible manipulator in Matlab/Simulink environment to testthe performance of the proposed control scheme. The system
parameters used in the simulation are presented Table I.
TABLE IPARAMETERS OF THE TWO-LINK FLEXIBLE MANIPULATOR [22]
Symbol Parameter Value
1 = 2 Mass density 0.2 kgm3
EI1 = EI2 Flexural rigidity 1.0 Nm2
l1 = l2 Links length 0.5 mJh1 = Jh2 Hub mass moment of inertia 0.2 kgm
2
G Gear ratio 1m1 =m2 Mass of the link 0.1 kg
mp Mass of the payload 0.1 kgJl1 = Jl2 Link mass moment of inertia 0.0083 kgm
2
Jp Payload mass moment of inertia 0.0005 kgm2
The manipulator is to track a desired step response while
suppressing end-effector vibration. Ziegler-Nichols gain tuning
procedure was tried, but very poor performance was achieved.
This is because two-link flexible manipulator is a highly open
loop unstable system. Gains are tuned manually in two stages.
A. Stage 1
Initially, both PID controllers and the PD controller for
link 2 are turned off. The amplifier gains are fixed as 1
for both links. The proportional and derivative gains of link
1 are then tuned simultaneously. Unlike the Ziegler-Nichols
procedure where the proportional gain is tuned until there is
overshoot; two-link flexible manipulator is a highly openloop
unstable and highly coupled system. This method does not
work well for the gain tuning. According to [24], for optimum
performance of PID controllers, the proportional, integral and
derivative gains must be simultaneously tuned. Manual tuning
is still the most favoured method despite different types of
tuning methods available Eriksson and Wikander [25]. As the
proportional gain was tune, the derivative gain has to be tune
simultaneously in order to achieve satisfactory result. while
tuning the PD gains of the fist link, the derivative gain of
the second gain has to be tune otherwise improved result
cannot be achieved. tuning the three controller gains enable
good tracking to be achieved in link 1. This behaviour can be
attributed to the coupling in the system. Then the Proportional
gain of the second link controller is tuned systematically . As
the proportional gain of the second link is tuned the good
tracking already obtained in in link 1 is affected thereforethe previous three already tuned gains needed to be re-
tuned simultaneously until good tracking in the two links
are achieved. To study the effect of controller gains on the
performance two sets of controller gains are obtained for two
cases as shown in Table II.
TABLE IIPD CONTROLLER GAINS FOR THE TWO-LINK FLEXIBLE MANIPULATOR
Cases PD Gains
Link 1 Link 2Kp Kv Kp Kv
1 0.55 1.5 0.06 0.22 1.1 1.15 0.25 0.42
B. Stage 2
After achieving good motion tracking with the PD con-
troller, then the PID controllers are also tuned systematically
as done in stage 1. Tuning the PID tends to degrade the perfect
tracking initially obtained. After some few trials of tuning and
re-tuning, the gains in Table III are achieved for the two cases.
The proposed algorithm is realizable experimentally because
PD/PID control algorithm has been successfully implemented
for single-link flexible manipulator in the literature [20], [5].
The future work is to carry out experimental validation and
also to extend this algorithm to a 4-degree of freedom gantry
type two-link flexible manipulator. Results of the effects oftwo different sets of PD and PID controllers gains on the
response of the two-link manipulator are shown in Figs.35.
Their effects on the performance index are also studied using
(25).
TABLE IIIPID CONTROLLER GAINS FOR THE TWO-LINK FLEXIBLE MANIPULATOR
Cases PID Gains
Link 1 Link 2Kp KI KD Kp KI KD
1 2 104 2 0.1 0.1 1.5
2 0.2 103 1.5 0.1 0.1 0.5
- 1361 -
7/30/2019 21E8F20Cd01
5/6
Fig. 3. Time histories of joints angles with hybrid PD-PID controller forcases 1 and 2
Fig. 4. Time histories of end-points accelerations with hybrid PD-PIDcontroller for cases 1 and 2.
Figure 3 shows the motion tracking, joint angles of link 1
and 2 in case 1 reaching the steady state at about 5 and 7
seconds respectively. While in case 2, the joint angles reach
the steady state in 3 and 5 seconds respectively. Comparing
these two cases it is observed that the higher the PD gains
the faster the response to the input tracking. Also the higher
the PID gains the faster the vibration suppression as shown in
Fig. 4. Case 1 with higher PID settles in less than 5 seconds
as compared to case 2 which settles at about 7 seconds. The
Fig. 5. Time histories of torque inputs with hybrid PD-PID controller forcases 1 and 2.
Fig. 6. Time histories of joints angles with hybrid PD-PID controller forcase study 1 with varying payloads.
Fig. 7. Time histories of end-points accelerations with hybrid PD-PIDcontroller for case study 1 with varying payloads.
faster response achieved in case 2 is traded off for higher
torque required (maximum of 0.84 Nm for link 1 and 0.36Nm
for link 2) shown in Fig. 5 compared to maximum of 0.52 and
0.145 Nm for links 1 and 2 respectively in case 1. However
the overall performance index of case 2 (0.3389) is better
than that of case 1 (0.3976) according to the factors penalized
in the performance index in Eq. (25). Effect of variation in
payload was also studied. Figures 6 - 8 show the response of
the two-link flexible manipulator with hybrid PD collocated
and PID non-collocated control to a payload of 0.04 kg,0.08kg and the nominal 0.1 kg for the case 1. Figure 6 shows
the joint angles input tracking with the various payloads. It
can be observed that there is overlapping in the response.
The endpoint accelerations show increase in amplitude of
vibration with increase in payload but the ststem settles at
the same time (see Fig. 7). Hence variation in payload does
not affect the vibration suppression of the proposed controller.
The applied torques shown in Fig. 8 also show no significant
change . These mean there is no significant effect on controller
performance with payload variation. It can be concluded that
the proposed controller is robust to payload variation.
- 1362 -
7/30/2019 21E8F20Cd01
6/6
Fig. 8. Time histories of torque inputs with hybrid PD-PID controller forcase study 1 with varying payloads.
V. CONCLUSION
This paper developed hybrid collocated PD and non-collocated PID controller for a two-link flexible manipulator.
The PD controller is for joint motion control and the PID
controller is for endpoint vibration suppression. The proposed
hybrid controller was tested within Matlab/Simulink environ-
ment. The performance of the proposed controller has been
evaluated in terms of input tracking, vibration suppression
and control torque. Effects of payload variation on the con-
troller was also studied. Simulation result using the proposed
controller has shown that the controller is very effective for
input tracking and vibration control for a highly nonlinear and
coupled system like two-link flexible manipulator. Payload
variation does not have a significant effect on the proposed
controller; this shows that the controller is robust. It can
therefore be concluded that the proposed hybrid PD/PID
controller is capable of tracking the desired joint angle while
suppressing vibration simultaneously in the presence of pay-
load uncertainty of the two-link flexible manipulator.
ACKNOWLEDGMENT
This work is supported by the Advanced Manufacturing
Technology Strategy (AMTS), an operating unit of the Council
for Scientific and Industrial Research (CSIR) in South Africa.
REFERENCES
[1] M. O. Tokhi and A. K. M. Azad, Flexible Robot Manipulators -Modelling, Simulation and Control, Institution of Engineering andTechnology, London, 2008.
[2] M. S. Alam and M. O. Tokhi, Designing feedforward command shaperswith multi-objective genetic optimisation for vibration control of asingle-link flexible manipulator, Engineering Applications of Artificial
Intelligence, Vol. 21, No. 2, 2008, pp. 229 246.
[3] L. Consolini and A. Piazzi, Generalized Bang-Bang Control for Feed-forward Constrained Regulation, Automatica, Vol. 45, No. 10, 2009, pp.2234 2243.
[4] Z. Mohamed and M. O. Tokhi, Command Shaping Techniques forVibration Control of a Flexible Robot Manipulator, Mechatronics , Vol.14, No. 1, 2004, pp. 69 90.
[5] M. O. Tokhi and A. K. M. Azad, Control of Flexible ManipulatorSystems, Proceedings of the Institution of Mechanical Engineers, Part
I: Journal of Systems and Control Engineering, Vol. 210, No. 12, 1996,pp. 113 130.
[6] M. O. Tokhi and M. Z. Md. Zain, Hybrid Learning Control Schemes withAcceleration Feedback of a Flexible Manipulator System, Proceedingsof the Institution of Mechanical Engineers, Part I: Journal of Systemsand Control Engineering, Vol. 220, No. 4, 2006, pp. 257267.
[7] F. Khorrami and S. Jain, Non-Linear Control with End-Point Accel-eration Feedback for a Two-Link Flexible Manipulator: Experimental
Results, Journal of Robotic Systems, Vol. 10, No. 4, 1993, pp. 505 530.
[8] M. Baroundi, M. Saad and W. Ghie, State-Feedback and LinearQuadratic Regulator Applied to a Single-Link Flexible Manipulator,Proceedings of the IEEE International Conference on Robotics and
Biomimetics (ROBIO), 2009, pp. 1381 1386.[9] D. T. Pham and M. Kalyoncu, Optimisation of a Fuzzy Logic Controller
for a Flexible Single-Link Robot Arm Using the Bees Algorithm,Proceedings of the 7th IEEE International Conference on Industrial
Informatics, 2009, pp. 475 480.[10] J. G. Yim, J. S. Yeon, J. Lee, J. H. Park, S. H. Lee and J. S. Hur, Robust
Control of Flexible Robot Manipulators, Proceedings of the SICE-ICASEInternational Joint Conference, 2006, pp. 3963 3968.
[11] J. Dong and G. H. Yang, H Control Design for Fuzzy Discrete-TimeSingularly Perturbed Systems via Slow State Variables Feedback: AnLMI-Based Approach, Information Sciences, Vol. 179, No. 17, 2009,pp. 3041 3058.
[12] F. Khorrami and U. Ozguner, Perturbation Methods in Control ofFlexible Link Manipulators, Proceedings of the IEEE InternationalConference on Robotics and Automation, Vol. 1, 1988, pp. 310 315.
[13] M. T. Ho and Y. W. Tu, PID Controller Design for a Flexible LinkManipulator, Proceedings of the 44th IEEE Conference on Decisionand Control and European Control Conference 2005, pp. 6841 6846.
[14] K. H. Ang, G. Chong and Y. Li, PID Control System Analysis, Design,and Technology, IEEE Transactions on C ontrol Systems Technology, Vol.13, No. 4, 2005, pp. 559 576.
[15] B. Zain and M. O. Tokhi and S. F. Toha, PID-Based Control of a Single-Link Flexible Manipulator in Vertical Motion with Genetic Optimisation,Proceedings of the Third UKSim European Symposium on Computer
Modeling and Simulation (EMS09), 2009, pp. 355 360.[16] M. A. Ahmad, Vibration and Input Tracking Control of Flexible Ma-
nipulator Using LQR with Non-Collocated PID Controller, Proceedingsof the Second UKSim European Symposium on Computer Modeling andSimulation (EMS08), 2008, pp. 40 45.
[17] S. Z. He, S. H. Tan, F. L. Xu and P. Z. Wang, PID Self-Tuning ControlUsing a Fuzzy Adaptive Mechanism, Proceedings of the Second IEEE
International Conference on Fuzzy Systems, Vol. 2, 1993, pp. 708713.[18] M. Sasaki, A. Asai, T. Shimizu and S. Ito, Self-Tuning Control of a
Two-Link Flexible Manipulator Using Neural Networks, Proceedings ofthe ICROS-SICE International Joint Conference, 2009, pp. 2468 2473.
[19] M. A. Ahmad, M. H. Suid, M. S. Ramli, M. A. Zawawi and R. M. T. R.Ismail, PD Fuzzy Logic with Non-collocated PID Approach for Vi-bration Control of Flexible Joint Manipulator, Proceedings of the 6th
International Colloquium on Signal Processing and Its Applications(CSPA), 2010, pp. 1 5.
[20] J. Cheong, W. K. Chung and Y. Youm, PID Composite Controller andIts Tuning for Flexible Link Robots, Proceedings of the IEEE/RSJ
International Conference on Intelligent Robots and Systems, Vol. 3,2002, pp. 2212 2217.
[21] M. Farid, Dynamic Modelling and Control of Flexible Manipulators, PhdThesis, Department of Mechanical Engineering, University of Calgary,
Alberta, 1997.[22] A. De Luca and B. Siciliano, Closed-Form Dynamic Model of Planar
Multilink Lightweight Robots, IEEE Transactions on Systems, Man andCybernetics, Vol. 21, No. 4, 1991, pp. 826 839.
[23] J. M. Martins, Z. Mohamed, M. O. Tokhi, J. Sa da Costa and M. A.Botto, Approaches for Dynamic Modelling of Flexible ManipulatorSystems, IEE Proceedings on Control Theory and Applications, Vol.150, No. 4, 2003, pp. 401 411.
[24] Y. Li, K. H. Ang and G. C. Y. Chong, PID Control System Analysisand Design: Problems, Remedies, and Future Directions, IEEE ControlSystems Magazine, Vol. 26, No. 1, 2006, pp. 32 41.
[25] E. Eriksson and J. Wikander, Robust PID Design of Flexible Manipu-lators Through Pole Assignment, Proceedings of the 7th InternationalWorkshop on Advanced Motion Control, 2002, pp. 420 425.
- 1363 -