International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 1

ISSN 2229-5518

Application of Modified Shuffled Frog Leaping

Algorithm for Robot Optimal Controller Design

Mohammd Pourmahmood Aghababa1, M.E.Akbari2, A.M. Shotorani3, R.M.Shotorbani4
1. Department of Electrical Engineering, Ahar Branch, Islamic Azad University, Ahar, Iran, m-Pourmahmood@iau-Ahar.ac.ir
2. Department of Electrical Engineering, Ahar Branch, Islamic Azad, University, Ahar, Iran, m-Akbari@iau-Ahar.ac.ir
3. Department of Electrical Engineering, Azerbaijan University of Tarbiat Moallem, Iran, a.m.shotorbani@gmail.com
4. Department of Mechanical Engineering, Tabriz University, Iran, r.m.shotorbani@yahoo.com

AbstractIn this paper, a modified shuffled frog leaping (MSFL) algorithm is proposed to speed up the convergence of the standard shuffled frog leaping (SFL) method. The MSFL approach is based on an adaptive accelerated position changing of frogs. This modification causes a fast convergence rate and consequently achieving a rapid adaptive algorithm. The proposed method is used to design the optimal controller parameters for a five bar linkage manipulator robot. Simulation results have verified the effectiveness and robustness of the proposed method in practical issues. The features and the advantages of MSFL algorithm, such as escaping from local optima traps, global optimization, good robustness, simple mechanism and fast convergence, would make MSFL method as a promising optimization approach.

Index TermsShuffled frog leaping algorithm, Optimization approach, Convergence rate, PID controller, Five bar linkage manipulator robot.

1 INTRODUCTION

—————————— ——————————
volutionary algorithms (EAs) are stochastic optimization methods that mimic the metaphor of natural biological evolution and/or the social behavior of species. Shuffled
Frog Leaping (SFL) is one of the new EAs that has been pro- posed by Eusuff and Lansey for determining optimal discrete pipe sizes for new pipe networks and for network expansions [1]. It is based on evolution of memes carried by the interactive individuals and a global exchange of information among themselves. The SFL works based on memetic evolution (transformation of frogs) and information exchange in the population. Frogs which are the hosts of memes (consisting memotype like gene in chromosome in GA) search the particle with the highest amount of food in a swamp by improving their memes. This characteristic can be used in an intelligent manner in control systems.
PID (Proportional-Integral-Derivative) control is one of the earliest control strategies. It has been widely used in the in- dustrial control field. Its widespread acceptability can be rec- ognized by: the familiarity with which it is perceived amongst researchers and practitioners within the control community, simple structure and effectiveness of algorithm, relative ease and high speed of adjustment with minimal down-time and wide range of applications where its reliability and robustness produces excellent control performances. However, successful applications of PID controllers require the satisfactory tuning of three parameters (which are proportional gain (KP), integral time constant (KI) and derivative time constant (KD)) according to the dynamics of the process. Unfortunately, it has been quite difficult to tune properly the gains of PID controllers because many industrial plants are often burdened with prob- lems such as high order, time delays and nonlinearities [4].
Traditionally, these parameters are determined by a trial and error approach. Manual tuning of PID controller is very tedious, time consuming and laborious to implement, espe- cially where the performance of the controller mainly depends on the experiences of design engineers. In recent years, many tuning methods have been proposed to reduce the time con- sumption on determining the three controller parameters. The most well known tuning method is the Ziegler-Nichols tuning formula [3]; it determines suitable parameters by observing a gain and a frequency on which the plant becomes oscillatory.
Considering the limitations of the Ziegler-Nichols method and some empirical techniques in raising the performance of PID controller, recently artificial intelligence techniques such as fuzzy logic [5, 6], fuzzy neural network [7] and some sto- chastic search and optimization algorithms such as simulated annealing [8], genetic algorithm [9, 10, 11], particle swarm op- timization approach [4], immune algorithm [12] and ant colo- ny optimization [13] have been applied to improve the per- formances of PID controllers. In these studies, it has been shown that these approaches provide good solutions in tuning the parameters of PID controllers. However, there are several causes for developing improved techniques to design PID con- trollers. One of them is the important impact it may give be- cause of the general use of the controllers. The other one is the
enhancing operation of PID controllers that can be resulted from improved design techniques. Finally, a better tuned op- timal PID controller is more interested in real world applica- tions.
This paper proposes the MSFL technique as a new optimi- zation algorithm. The proposed method is applied for deter- mining the optimal values for parameters of PID controllers. Here, we formulate the problem of designing PID controller as an optimization problem and our goal is to design a controller with high performance by adjusting four performance index- es, the maximum overshoot, the settling time, the rise time and the integral absolute error of step response. After design- ing PID controllers for some simple benchmark transfer func- tions, an optimal PID controller is designed for a five bar lin- kage manipulator robot using MSFL algorithm. The advantag- es of this methodology are that it is a simple method with less computation burden, high-quality solution and stable conver- gence specifications.
The rest of this paper is organized as follows. In section 2, first an overview of SFL algorithm is given. Then, the mod- ified version of SFL is introduced. Section 3 deals with the formulation of the optimal PID controller designing problem through defining a new cost function. In section 4, simulation results of optimal PID controller designing procedure are giv- en for some benchmarks. Then, dynamic equations of the five- bar-linkage manipulator robot are illustrated following by optimal controller design using proposed MSFL. Finally, the paper ends with some conclusions in section 5.

2 MSFL ALGORITHM

In this section, the original SFL is briefly reviewed. After- wards, we propose MSFL as an enhanced SFL algorithm.

2.1 SFL algorithm

SFL algorithm, introduced by Eusuff and Lansey for water distribution system optimization, is a metaheuristic for solving optimization problems [1]. SFL is a population based coopera- tive search metaphor inspired by natural memetics. The algo- rithm uses memetic evolution in the form of influencing of ideas from one individual to another in a local search. Concep- tually, the local search is similar to particle swarm optimiza- tion. A shuffling strategy allows the exchange of information among local searchers, leading them toward a global optimum [1].

In SFL, the population consists of a set of frogs (solutions)
partitioned into subsets, referred to as memeplexes. Different
memeplexes are considered as different cultures of frogs, each
performing a local search. Within each memeplex, the indi-
vidual frogs hold ideas, that can be influenced by the ideas of
other frogs, and evolve through a process of memetic evolu- tion. After a defined number of memetic evolution steps, ideas
are passed among memeplexes in a shuffling process [2]. The local search and the shuffling processes continue until some predefined convergence criteria are satisfied [1].

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 2

ISSN 2229-5518

In general, a SFL works as follows. First, an initial popula-

G( s)Kp K1 / s KD s

(5)
tion of P frogs is created randomly. Afterwards, the frogs are
sorted in a descending order according to their fitness. Then,
the entire population is divided into m memeplexes, each con- taining n frogs. In this process, the first frog goes to the first memeplex, the second frog goes to the second memeplex, frog

m goes to the mth memeplex, and frog m+1 goes back to the first memeplex and so on. Within each memeplex, the frogs with the best and the worst fitnesses are identified as Xb and Xw, respectively. Also, the frog with the global best fitness is identified as Xg. Then, a process is applied to improve only the frog with the worst fitness and (not all the frogs) in each cycle.

Accordingly, the position of the frog with the worst fitness is adjusted as follows [1]:
Chang frog position:

Di=rand × (Xb - Xg) (1) New position:

Xi+1 =Xi +Di where -Dmax Di Dmax, (2) where rand is a random number between 0 and 1, and Dmax is the maximum allowed change in a frog’s position. If this process produces a better solution, it is replaced for the worst frog. Otherwise, the calculations in equations (1) and (2) are repeated but with respect to the global best frog (i.e. Xb is re- placed by Xg). If no improvement is possible, then a new solu- tion is randomly generated to replace the worst frog. Hence, the calculations continue for a specific number of iterations [1]. Accordingly, the main parameters of SFL are: number of frogs P; number of memeplexes; number of generation for each memeplex before shuffling; number of shuffling iterations; and maximum step size.

2.2 Modified shuffled frog leaping algorithm

The main drawback of SFL algorithm is slow convergence, closely related to the lack of adaptive acceleration terms in the position updating formula. In equation (1), rand determines the movement step sizes of frogs through the Xb and Xw posi- tions. In the standard SFL, these step sizes are random num- bers between 0 and 1 for all frogs.
In each cycle, the value of the objective function is a crite- rion that presents the relative improvement of a frog move- ment with respect to the previous one. Thus the difference between the values of the objective function in consequent iterations can represent the frog acceleration. Therefore, posi- tion changing formulae turns to the following form.

Di = rand × C × (f(Xb ) - f(Xw) ) × (Xb - Xw) (3) New position

Xi+1 = Xi + Di (4)

where C  (0,Cmax] is a constant, Cmax is a case dependant upper limit, f(Xb) and f(Xw) are the best and the worst fitness func-
where KP, KI and KD are the proportional gain, integral and derivative time constants, respectively. For designing an op- timal PID controller, a suitable objective function that represents system requirements, must be defined in the first step. A set of good control parameters KP, KI and KD can pro- duce a good step response that will resultant in minimization of performance criteria. The optimal PID controller parameters that minimize the performance indexes are designd using the proposed MSFL algorithm. This section deals with defining the objective function. Then an efficient procedure is proposed to design an optimal PID controller. Finally, the effectiveness and efficiency of the proposed MSFL algorithm in designing optimal PID controller is tested using some benchmark plants. All the computations are implemented with Mat- lab®/Simulink®.

3.1 Objective function definition

In the design of a PID controller, the performance criterion or objective function is first defined based on some desired speci- fications and constraints under input testing signal. Some typ- ical output specifications in the time domain are overshoot, rise time, settling time, and steady-state error. In general, three kinds of performance criteria, the integrated absolute error (IAE), the integral of squared-error (ISE), and the integrated of time-weighted-squared-error (ITSE) are usually considered in the control design under step testing input, because they can be evaluated analytically in the frequency domain. It is worthy to notice that using different performance indices probably makes different solutions for PID controllers. The three integral performance criteria in the frequency domain have their own advantages and disadvantages. For example, a dis- advantage of the IAE and ISE criteria is that their minimiza- tion can result in a response with relatively small overshoot but a long settling time. Although the ITSE performance crite- rion can overcome the disadvantage of the ISE criterion, the derivation processes of the analytical formula are complex and time-consuming [4]. The IAE, ISE, and ITSE performance crite- ria formulas are as follows:
In this paper, another time domain performance criterion defined by
tions that are found by the frogs in each memeplexs. Similar to

min

W ( K)= (1/( 1+ e- α ))×(T +T )+

the original SFL, if the process produces a better solution, the

K s r

α α

(9)
worst frog is replaced by the better one. Otherwise, the calcu-

(e /( 1+ e

)).( Mp + Ess )

lations in equations (3) and (4) are repeated with respect to the global best frog instead (i.e. Xg and f(Xg) replace Xb and f(Xb), respectively). If no improvement is possible, then a new solu- tion is randomly generated to replace the worst frog.
The proposed modification term, (f(Xb) - f(Xw)), called adap- tive coefficient, causes an adaptive movement. In each itera-
tion, the modification term defines the movement size, adap- tively. Therefore, the adaptive coefficient decreases/ increases the movement size relative to being closer/farther from the optimum point, respectively. By means of this method, posi- tion changing can be updated adaptively instead of being fixed or changed linearly. Therefore, using the adaptive coeffi- cient, the convergence rate of the algorithm will be increased rather than being performed by proportional large or short steps. So, the above modification accelerates the convergence of the algorithm. This new version is called modified SFL (MSFL). The main characteristics of MSFL algorithm are: adap- tive movements, fast convergence, better diversification ability and escaping from local optima. Finally, the proposed MSFL is still a general optimization algorithm that can be applied to any real world continuous optimization problems.

3 MSFL ALGORITHM FOR DESIGNING PID CONTROLLER

The PID controller is used to improve the dynamic response and to reduce the steady-state error. The transfer function of a PID controller is described as:
is used for evaluating the PID controller, where K is [KP, KI, KD], and  [-5, 5] is the weighting factor. The optimum se- lection of depends on designer’s requirements and the cha- racteristics of the plant under control. We can set to be smaller than 0 to reduce the overshoot and steady-state error. On the other hand, we can set to be larger than 0 to reduce the rise time and settling time. Note that, if is set to 0, then all performance criteria (i.e. overshoot, rise time, settling time, and steady-state error) will have the same worth.

3.2 MSFL based PID Controller

For designing an optimal PID controller, determination of vec- tor K with regards to the minimization of performance index is the main issue. Here, the minimization process is performed using the proposed MSFL algorithm. For this purpose, step response of the plant is used to compute four performance criteria overshoot (Mp), steady-state error (Ess), rise time (Tr) and setting time (Ts) in the time domain. At first, the lower and upper bounds of the controller parameters should be spe- cified. Then a population of frogs is initialized, randomly in the specified range. Each frog represents a solution (i.e. con- troller parameters K) that its performance index should be evaluated. This work is performed by computing Mp, Ess, Tr, and Ts using the step response of the plant, iteratively. Then, by using the four computed parameters, the performance in- dex is evaluated for each frog according to these performance criterions. Now the main procedure of MSFL algorithm starts

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 3

ISSN 2229-5518

as follows. The frogs are sorted in a descending order accord- ing to their performance index. Then, the entire population is divided into m memeplexes, as mentioned earlier in section 2. Within each memeplex, Xb, Xw and Xg are determined. Then, the frog with the worst fitness is improved using the modified process mentioned in section 2.2. Afterwards, the superseding frogs are created, by the procedure mentioned in the section
2.2. This process is repeated until a satisfactory of a stopping criterion. In this stage, the frog corresponding to Xg is desig- nated as the optimal vector K.
The flowchart of designing PID controller based on MSFL is

shown in Fig. 1.

start

Set initial conditions and parameters

function.
In order to obtain the optimal PID controller parameters, in
each experiment, MSFL and SFL are run 10 times with 200
iterations. The best solutions of different case studies, which have the optimal or near optimal PID controller parameters, are summarized in Table 2. Using these PID controller para-
meters, the unit step response of each case study was obtained as shown in the figurs 2-5. From the simulation results, it can be found that MSFL is better than SFL, considering the best performance index value. Moreover, MSFL produces the smooth curve for the output in conjunction with little fluctua- tion and small overshoot.

TABLE 1

SUMMARY OF SIMULATION RESULTS OF BENCHMARK PLANTS

algorithm P I D Mp Ts Tr Ess cost

Generate a random population of frogs

For each frog, calculate the step response of plant

Calculate Mp, Ess, Tr and Ts of plant’s step response

MSFL 9.6 10 8.6 4.7 1.905 1.266 0 3.9355 37

1

SFL 8.56 9.6 5.76 8.6 2.92 1.10 0 6.3100 123

MSFL 2.8 0.99 1.41 2.25 1.605 0.97 0 2.4125 45

2

SFL 3.2 1.1 1.6 5.5 2.96 0.875 0 4.6675 121

MSFL 2 0.8 1.5 1.5 2.04 1.255 0 2.3975 57

3

SFL 2.3 1.4 1.4 2.7 3.53 0.96 0 7.1900 154

MSFL 2.5 1 1.1 3.7 1.575 0.912 0 3.0935 66

4

SFL 2.2 1.3 1.2 9.4 5.05 0.96 0 7.7050 161

Calculate the objective function of frogs

1.2

Step response

MSFL

SFL

1

Run the proposed MSFL algorithm, as described in the pseudocode

0.8

No

Stop condition satisfied

Yes

0.6

0.4

0.2

stop

Fig. 1. The flowchart of MSFL based PID Controller design procedure

0

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 2. Comparison of step response of the G1(s) plant using MSFL and

SFL methods

3.3 Optimal PID Controller for Typical Transfer

Functions

In order to verify the modified SFL's favourable performance, comparison experiments have been carried out for the follow- ing four typical control plants. The transfer functions of four control systems are given as follows.

1.2

1

Step response

MSFL

SFL

G ( s )

1
); 0≤K ≤10, 0≤K ≤10, 0≤K ≤10

0.8

1 s3  6s2  7s  10 P I D

0.6

G ( s )

27
; 0≤K ≤5, 0≤K ≤5, 0≤K ≤5

0.4

2 ( s  1)( s  3 )3 P I D

0.2

G ( s )

3e0.1s

; 0≤K ≤5, 0≤K ≤5, 0≤K ≤5

0

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

3 ( s2  1.2s  1 )( s  3 )

G ( s )

P I D

e0.1s

;

Fig. 3. Comparison of step response of the G2(s) plant using MSFL and

SFL methods

4 ( s  1)( 0.5s  1)( 0.25s  1)( 0.125s  1)

0≤KP≤5, 0≤KI≤3, 0≤KD≤3
Different settings were evaluated to determine suitable val- ues for parameters of MSFL and SFL algorithms. A population of 100 frogs, 10 memeplexes, and 5 iterations per memeplex were found suitable to obtain good solutions for both SFL and MSFL algorithms. The maximum number of iterations for all experiments is considered to be 100. Also, is set to 0 to have the same merit for all performance criteria in the objective

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 4

ISSN 2229-5518

1.2

Step response

MSFL

SFL

1

0.8

2 22 h 2

21 1

q 2

where g is the gravitational constant and

0.6

M11 = I 1 + I 3 + m1 d 2
+ m3 d 2 + m4 I 2
(12)

11 11 c1

c 3 1

0.4

M22 = I 2 + I 4 + m2 d 2 + m3 l 2 + m4 d 2

(13)

0.2

11 11

c 2 2 c 4

0

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 4. Comparison of step response of the G3(s) plant using MSFL and

SFL methods

M12 = M21 = (m3dc3l2 - m4dc4 l1) cos(q1- q2) (14)

It’s noticed from (12)-(14) that for

m3 lc3 l2 = m4 lc4 l1 (15)

we have M12 and M21 equal to zero, that is, the matrix of inertia is diagonal and constant. Hence the dynamic equations of this manipulator will be

T1 = (M11+ I 1 ) q + g(m1 lc1 + m3 lc3 + m4 l1 )cosq1, (16)

h 1

4 APPLICATION OF MSFL FOR ROBOT OPTIMAL

CONTROLLER DESIGN

In this section, an optimal PID controller is designed for a five- bar-linkage manipulator robot. The results of MSFL method are compared by the results of the standard SFL algorithm.

4.1 Optimal PID Controller for five-Bar-Linkage

Manipulator Robot

To show the efficiency and desirable performance of the pro- posed algorithm in designing optimal PID controllers, a well known Mechatronics application, i.e., a robot is considered. The examined robot configuration is a five-bar-linkage. Dy- namic equations of the robot are described in the following subsection. Afterwards, MSFL algorithm for an optimum PID controller is utilized.

4.1.1 Dynamic Equations of five-bar-Linkage

Manipulator Robot


In recent years, there has been a growing interest in the design and control of lightweight robots. Several researchers have studied the modelling and control of a single link flexible beam [15]. Fig. 6 shows the 5 bar linkage manipulator built in our robotics research lab. Also, Fig. 7 depicts the five-bar lin- kage manipulator schematic where the links form a parallelo- gram. Let qi, Ti and I i be the joint variable, torque and hub inertia of the ith motor, respectively. Also, let Ii, li, dCi and mi be the inertia matrix, length, distance to the centre of gravity and mass of the ith link, correspondingly.

Fig. 6. Planar presentation of robot

The dynamic equations of the manipulator are [17]:

T2 = (M22 + I h ) q2 + g(m1 lc2 + m3 l2 + m4 lc4) cosq2, (17)

Fig. 7. Planar presentation of the robot

Notice that T1 depends only on q1 but not on q2. On the other hand T2 depends only on q2 but not on q1. This discussion helps to explain the popularity of the parallelogram configura- tion in industrial robots. If the condition (15) is satisfied, then we can adjust the two rotations independently, without wor- rying about interactions between them.

4.1.2 Simulation Results

Having 2 motors, the manipulator specification consisting of mass, length and centre of gravity of links are given in Table 2. The main purpose is designing an optimal PID controller for each of motors to control their rotations, with good perfor- mance. Using equations (16) and (17), five-bar-linkage mani- pulator robot is easily simulated using Matlab and Simulink. The block diagram of the five-bar-linkage manipulator robot with PID controller for motor 1 is shown in Fig. 8. The block diagram for motor 2 is similar to this figure. A population of
80 frogs, 8 memeplexes, and 4 iterations per memeplex were found suitable for obtaining good solutions for both SFL and MSFL algorithms. The maximum iteration of all experiments is considered equal to 200. Also, is set to 0 for all performance criteria to have the same merit in the objective function.
The following process is done to determine the optimal
values of the PID controller parameters (i.e., vector K). First,
the lower and upper bounds of the three controller parameters
are selected as 0 and 30, respectively. Then, all frogs of the
population are initialized, randomly. Each frog K (the control-
ler parameters) is sent to Matlab® Simulink® block and the values of four performance criteria in the time domain, i.e., Mp, Ess, Tr and Ts are calculated iteratively. Afterwards, the
objective function is evaluated for each frog according to these performance criteria. Then, the procedure of MSFL algorithm is performed, as illustrated in the flowchart of Fig. 3. At the end of any iteration, the program checks the stop criterion. When one termination condition is satisfied, the program stops and the latest global best solution Xg is the best solution of K.
Fig. 9 illustrates the step response without PID controllers for two motors. Figures 10 and 11 show the step response of rota- tion for motors 1 and 2, respectively. The simulation results of the best solution are summarized in Table 3. These results demonstrate that cost function is converged rapidly.

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 5

ISSN 2229-5518

TABLE 2

FIVE-BAR-LINKAGE MANIPULATOR DATA


Link Mass (Kg) Length (m) C of G (m)

1 0.288 0.33 0.166

2 0.0324 0.12 0.06

3 0.3702 0.33 0.166

4 0.2981 0.45 0.075

1.2

1.1

1

0.9

0.8

0.7

0.6

Step response

MSFL

SFL

0.5

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 8. Block diagram of the motor with PID controller.

In conclusion, MSFL algorithm has rapid convergence cha- racteristic and is highly effective in solving the optimal tuning problem of PID controller parameters.

Fig. 11. Comparison of step response of the motor #2 angel using MSFL

and SFL methods.

TABLE 3

SUMMARY OF SIMULATION RESULTS OF FIVE-BAR-ROBOT MOTORS

1.4

1.2

Step response

motor 1

motor 2

1

0.8

0.6

0.4

0.2

0

0 2 4 6 8 10 12 14 16 18 20

Time (sec)

5 CONCLOUSION

In this paper a modified SFL algorithm (MSFL) was proposed to improve the performance of the standard SFL algorithm. Different benchmarks were used to illustrate the mentioned advantages. Dealing with this problem, a new time domain performance criterion was proposed. In all case studies, MSFL performed better than SFL approach which exposed MSFL as a

Fig. 9. Step response of the robot motors without PID controller

Step response

1.2

MSFL

SFL

1.1

1

0.9

0.8

0.7

0.6

0.5

0 1 2 3 4 5 6 7 8 9 10

Time (sec)

Fig. 10. Comparison of step response of the motor #1 rotation using MSFL

and SFL methods.

promising optimization method. The optimal controller design of the five-bar-linkage manipulator robot has been considered, as a practical application. The proposed method was imple- mented for tuning the controller for the robot. High promising results demonstrate that the proposed algorithm is robust, efficient and can obtain higher quality solution with better computational efficiency.

REFERENCES

[1] M.M. Eusuf, KE. Lansey, ―Optimization of water distribution net- work design using the shuffled frog leaping algorithm.‖ J Water Re- sour Plan Manage, Vol 129, no.3, pp. 210–225, 2003.

[2] S.Y. Liong, M.d. Atiquzzaman., ―Optimal design of water distribu- tion network using shuffled complex evolution.‖ J Inst Eng, Singa- pore, vol. 44, no. 1, pp. 93–107, 2004.

[3] J.G. Ziegler and N.B. Nichols, ―Optimum settlings for automatic

controllers,‖ Trans. On ASME., vol. 64, pp. 759-768, 1942.

[4] Z.L. Gaing, ―A Particle Swarm Optimization Approach for Optimum Design of PID controller in AVR system,‖ IEEE Transactions on Energy Conversion, vol. 9, no. 2, pp. 384-391, 2003.

[5] Z.Y. Zhao, M. Tomizuka, and S. Isaka, ―Fuzzy gain scheduling of PID controllers,‖ IEEE Trans. System, Man, and Cybernetics, vol. 23, no. 5, pp. 1392-1398, 1993.

[6] A. Visioli, ―Fuzzy logic based set-point weight tuning of PID control- lers,‖ IEEE Trans. System, Man, and Cybernetics – Part A: System and Humans, vol. 29, no. 6, pp. 587-592, 1999.

[7] S.Y. Chu, C.C. Teng, ―Tuning of PID controllers based on gain and phase margin specifications using fuzzy neural network,‖ Fuzzy Sets and Systems, vol. 101, no. 1, pp. 21-30, 1999.

[8] G. Zhou and J.D. Birdwell, ―Fuzzy logic-based PID autotuner design using simulated annealing,‖ Proceedings of the IEEE/IFAC Joint Sympo- sium on Computer-Aided Control System Design, pp. 67 – 72, 1994.

[9] D.P. Kwok and F. Sheng, ―Genetic algorithm and simulated anneal-

ing for optimal robot arm PID control,‖ Proc IEEE Conf. Evolutionary

IJSER © 2011

http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 2, Issue 11, November-2011 6

ISSN 2229-5518

Computation, pp. 707–713, 1994.

[10] R.A. Krohling and J.P. Rey, ―Design of optimal disturbance rejection PID controllers using genetic algorithm,‖ IEEE Trans. Evol. Comput., vol. 5, pp. 78–82, 2001.

[11] P. Wang and D.P. Kwok, ―Optimal design of PID process controllers

based on genetic algorithms,‖ Control Engineer Practice, vol. 2, no. 4, pp.641-648, 1994.

[12] D.H. Kim, ―Tuning of a PID controller using a artificial immune net- work model and local fuzzy set,‖ Proceedings of the Joint 9th IFSA World Congress and 20th NAFIPS International Conference, vol. 5, pp.

2698 – 2703, 2001.

[13] Y.T. Hsiao, C.L. Chuang, and C.C. Chien, ―Ant colony optimization for designing of PID controllers,‖ Proceedings of the 2004 IEEE Confe- rence on Control Applications/ International Symposium on Intelligent Control/International Symposium on Computer Aided Control Systems De- sign, Taipei, Taiwan, 2004.

[14] J.P. Huissoon and D. Wang, ―On the design of a five-bar-linkage

manipulator‖ Robotica, vol. 9, pp 441-446, 1991.

[15] D. Wang and M. Vidyasagar, ―Modeling of a five-bar-Linkage Mani- pulator with One Flexible Link,‖ in Proc. IEEE Int. Symp, subject, Tur- key, pp. 21–26, 1988.

[16] Asada, Haruhiko, Youcef-Toumi, and Kamal, ―Analysis and Design

of a Direct-Drive Arm with a Five-Bar-Link Parallel Drive Mechan- ism,‖ Journal of Dynamic Systems, Measurement, and Control, 1984.

[17] D.Wang, J.P.Huissoon and K.Luscott, ―A teaching robot for demon- strating robot control strategies,‖ manufacturing research corporation of

Ontario, 1993.

IJSER © 2011

http://www.ijser.org