Robot Arm Control Using Hybrid intelligent
Abstract: The design of a fuzzy logic control suffers from select parameters of the membership functions, scaling factors, defuzzification
action, Inference engine and base rules. speaking generally, such prosecutions are executed by traditionally techniques which do not assure
an robust fuzzy control system design. There are various techniques introduced in literatures that used Genetic Algorithms to optimize a
fuzzy logic control component. . In this paper, the suggested control law consists of Fuzzy Logic Control (FLC) tuning via Geneticn Algorithm
(GA). The FLC used because it is efficient tools for control of nonlinear and uncertain parameters systems. GA is mainly presented to find a
simultaneous near optimum design of the membership functions, scaling factors, defuzzification Method, Inference engine and control base
rules. GA with different fitness functions in a form of the cumulative response error which are widely used as an efficient optimization
technique. This paper also introduce a new methodology with new multi-objective function to improve fuzzy control parameters based on
Genetic Algorithm techniques. The dynamic model of the robot manipulator its done by differential equations, these equations are hardly
nonlinear, parameters uncertainty and time varying with multiple input and multiple output (MIMO).The manipulator robot and the fuzzy
genetic control are modeled in MATLAB SIMULINK; the manipulator robot model driven nonlinear controller to draw a circle in the space
with and without parameters uncertainties. The proposed techniques showed that the proposed fuzzy controller gives superior response in the
output performance. When the parameter uncertainties including in the system, given satisfactory response.
Keywords: Genetic optimization, PUMA 560 Robot Manipulator, MATLAB SIMULINK fuzzy logic
Received: June 21, 2022. Revised: September 2, 2023. Accepted: October 4, 2023. Published: November 2, 2023.
1. Introduction
Intelligent control technologies provide solutions to
traditional approaches by leveraging insights from intelligent
biological systems. Such ideas can either come from experts,
to solve an automated control problem or by looking at how a
biological system works and translating it into techniques
suitable in solving control problems [1].
In [2] a process system is intelligent if it is able to improve
its response in case of uncertainty. The main characteristics of
intelligence are learning or adaptation and past experiences in
general, improvement and error development and self-
searching for the most effective values
Which achieves the machine's ability to check and modify
its behavior in a limited sense using the following methods:
1.1 Artificial neural networks (ANN)
ANN are computational models that attempt to make
mathematical representations of the structure of the brain,
consisting of a set of elements to perform simple
manipulations that communicate through weighted
connections [3].
The ANN are characterized via its patterns between the
neurons which working as processing units, its method of
determining the weights on the connections (called its
training) by back propagations, and its activation function .
1. 2 Fuzzy logic system (FLS)
In 1965, Zadeh introduced his research on the fuzzy logic
system, and Zadeh was the first designer to use fuzzy group
theory (FS) and fuzzy logic (FL), and he previously presented
the concept of the language variable in 1973. Zadeh also
proposed fuzzy control system by contrast, Conventional
control may take inputs/outputs between false and true as a
percentage, in traditional set theory the function definition
does not matter but the number belongs to the set or not, yes
or no and also zero or one takes on the value, this approach is
inconvenient in many applications practical and life like
lifetime group or temperature group, but element in fuzzy
group has variable values between minimum(0) and
maximum (1), that declare that elements of these sets not only
represent (1) or (0) values but also represent the percentage of
each input between these two values.
Among the many applications of fuzzy systems has been
the area of FLCs. FLC is a rule base system that acts as a
closed-loop control with its brain and is useful in executing
complex, poorly defined operations, particularly those that
can be controlled by an experienced human operator without
knowledge of underlying system dynamics. Fuzzy control has
recently been applied to many fields of industrial applications
. FLCs have been used in nonlinear control systems with
unpredictable range of operating pointss. The block diagram
of a elements of fuzzy control is showed in Figure 1.
Fig. 1. Basic comcept of fuzzy logic control.
The fuzzy logic control is consisting of the following four
elements:
a) A fuzzification :a fuzzification term that converts real-
world crisp values into fuzzy sets through membership
functions.
SAMAH ABDELSALAM
Higher Institute for Technology
and Science, Regdalleen,
LIBYA
TWFIQ H. ELMENFY
Department of Electrical
Engineering,
Faculty of Engineering
University of Benghazi, LIBYA
MONA MOHAMMED MOSA
Aljeagdif
Department of Medical Engineering
Collage of Medical Technology
Benghazi, LIBYA
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
153
Volume 2, 2023
b) Rule base (RB) :The rule base is composed of IF–
THEN rules, from which an inference emgine is created.
. A standard expression of RB with rule baes are
represented as:
Rule m: IF = and = = THEN y =
where , . . . , are inout variables and y is control
action. , . . . , and (i = 1, . . . , m) are, respectively,
the linguistic variables for , . . . , and y in the universal
of discourse of , . . . , .
c) Data base (DB): The data base is created via the
appropriate membership functions of linguistic variables ,
. . . , and that convert crisp inputs into fuzzy sets.
Bell-shaped, triangle and trapezoidal membership functions
are used.
d) Inference engine: Operators within rule base form the
inference mechanism. In general, the rules of the AND rules
(with the lowest value) or OR (with the maximum value) are
used as connecting operators between the casevariables.
e) Defuzzification for making a output of FLC action,
defuzzification is the calculating the output from inference
results for all activated logic rules into an explicit output.
Selecting of defuzzification method are some effect on control
action of FLC.
1.3 Genetic algorithms (GAs)
Referencing Darwin's evolutionary theory, computer
scientist John Holland proposed genetic algorithms in his
paper on a general method for creating programmed solutions
through evolutionary adaptive processes (Park, 1995).
A GA is a global optimization method used to solve problems
for optimal performance. The GA process in Fig. 2 consists
of developing an initial set of points, that is, an initial set of
randomly generated individuals (chromosomes), and coding
solutions to an optimization technique to going through
appropriate final solutions. This process is carried out by
repeating a large ensemble of fixed size N. The evolution of
successive ensembles from a generation to a
〖〗,the generation depends on the processes of
selection, crossover and mutation, inspired by evolutionary
biology:
Selection: A population is selected from each successive
generation to raise a new generation. There are two methods
of selection that are most commonly used and the most
common is to choose the roulette wheel and the random rest
without choosing an alternative.
a) Crossover: The replication crossover operator
produces good generations of parents selected by a
computer probabilistic method from among the
remaining surviving individuals.
b) Mutation: This process denotes changes in the
genetic sequence (mutation) with a probability of
Pm in some selected chromosomes to break out of
the local optima and confirm the accessibility of
the overall solution space..
These processes working on all members of the
generation. For each generation, GA selects the best
individuals according to adaptability or suitability which is
an index function
.
Fig. 2. Principle of a genetic algorithm.
a) Intelligent hybrid systems are a combination of two
or more computing paradigms, constructed in such
a way that the advantages of one method are used to
compensate for the disadvantages of another. some
techniques or methods
b) They are integrated as follows: Neural networks
that used to designing fuzzy logic.
c) Fuzzy systems which used to designing variuos
neural networks.
d) Genetic algorithms which used to designing of
fuzzy logic systems and selected his parameters.
Genetic algorithms in automatically training and
generating neural networks.
Fig.3. Intelligent hybrid systems.
GFLC are one of the most successful hybrid algorithms.
But GFLC recently. It is popular and has gained a lot of
interest in recent studies. Some of them have been mentioned
previously, and it is worth noting that there is no agreed-upon
method between the two areas, and we have incorporated a
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
154
Volume 2, 2023
special method for Park, Foran, Byrne and Loudini in order
for us to finally arrive at the complete GFS model. A method
will be explained in details in the next Section.
In our research, we present the application of GFLC
method to design control algorithm using Simulink non-linear
robot 3 DOF PUMA 560 which is short for (Three Degree of
Freedom Programmable Universal Manipulator for
Assembly) which was released in the past as the first modern
application of robot and became very common, it is an
industrial robot My application, this robot has 6 degrees of
freedom with 6 rotational joints, Puma 560 was used for
research in the past and it was a very popular laboratory robot,
we used a lot in research because they are well studied and
they are well known parameters that were described as "white
mice" in research that studied bots [4].
2. Mathematical Modelling of the
Puma Robot
All the parameters and the hardly nonlinear model used
Langrange Euler formulation to derive this model of the 6
DOF PUMA 560 used in the research can be found in [5]
The 3DOF PUMA robot has the same configuration space
equation general form as its 6DOF convenient. In this sort,
the last three joints are blocked so they keep their initial states
and parameters while the robot is moving.
Fig.4. Debavit and Hartenberg Coordinate frames of the 3DOF PUMA
Robot .
Using the formation equation of the robot, and by
appointed the last joints as zero always, we can define a
general equations that allows us to use PUMA robot as a 3
DOF robot.
To find the kinematics of the 3-DOF robot, a new Debavit
and Hartenberg DH coordinate system is established, and a
homogenous transformation matrix relating the coordinate
frame to the first coordinate frame is developed. However,
the 3 DOF PUMA will have the same kinematics of its 6-
DOF convenient with
4
q
,
5
q
and
6
q
to zero.
1. For the formation space equation of the robot:
M
(q).q
B(q).[
C(q).[
2
]
G(q)
(1)
Where,
: nx1 torques vector .
q : nx1 position vector ,
M (q) : nxn inertia matrix of the manipulator,
C() is coriolis torques matrix
G () is centrifugal torque matrix
[˙ ˙] is joint velocity vector that it can give by: [1˙ .
˙2, ˙1. ˙3, … . , ˙1. ˙, ˙2. ˙3, … . . ].
[˙]2 is vector, that it can given by: [1˙ 2, ˙22, ˙32, . ].
We set = =0 this yields
[ ...0...0...0
[ ]=
[
[
B(q).
+
.
And
The angular acceleration is found as to be
Now let I=
-[
B(q).[
C(q).[
2
]
G(q)
]}
(2.15)
=
+
[
=
[
=
=
[
=
[
=
These equations tell us that in instruction to ensure
and
keep their zero values, it is better to set
; so by holding the control torques of the last three
joints as
=
[
And
0 the last three joints are blocked at their initial
states.
3. Genetic Fuzzy Systems
The primary drawbacks of GA when using the Pittsburgh
approach are that it is a more computationally convenient
process due to the larger chromosome size required, thus
increasing the search base for GA, with each chromosome
indicating an entire FLC. The goal is to reduce the required
chromosome length. A number of assumptions have been
made regarding the FLC whose performance is desired
The following sections will describe the most commonly
used methods and assumptions for encoding chromosome
member parameters and rules with the fewest number of bits
inspired by [6] [7][8][9][10].
3.1 Decoding Rule Base
Trying to code the base of the FLC connected to the FLC
inputs are the error e(t), the error derivative ∆e(t), and the
output is the control action, u(t). Each input from these three
controlled variables is evaluated on a universe of natural
discourse (UOD) level from [-1 1] starting assumptions that
are set as follows:
a) The Number of Fuzzy Sets (NFS ) for the controller
is fixed at five (NB, NS, Z, PS PB.(
b) The universe of discourse is symmetrical about the
central.
c) The magnitude of the control signal
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
155
Volume 2, 2023
Coordinated with the magnitude of the input signal.
These generalizations can be used decreasing the number
of bits required for the base rule. This can be summarized to
build a rule base by knowing the following parameters [7]:
a) The grid spacing parameters [PSG_(e), PSG_e] are
the meaning of the distribution spaces of the fuzzy
groups. on the [x, y] axis of each of the inputs to
generate the base grid of the rule table and the [x,
y] axis of the intersection produces network nodes
that populate the outputs corresponding to each
rule rule
b) The following angle [0:90] specifies the slope of
the line through the origin on which the initial
points are placed.
c) Consequent line order, (CLO) is defines order of
consequent on the italic line .The CLO can be in
two sequences ( NB-NS-Z-PS-PB or PB-PS-Z-NS-
NB) and in effect doubles the range of possible
consequent line angles to 0-360) .
d) The Grid Spacing parameters of output (PSG_∆u)
is represent distribution of spaces for the fuzzy sets
on the italic line .
After calculating of the all coordinates of the points
(network nodes and output points), it is possible to proceed to
the assignment by specifying the minimum distance between
all the distances separating each node in the network from all
other output points located on the straight line. Then we select
the nearest output point for each node in the network.
Determining the locations of the initial points and the grid
spacing parameters of the inputs we will find by the way
presented by Loudini [10]. Which ensures that the PS values
are between [-1 1] as shown below.
The Ci principles are then defined in terms of the divergence
coefficient PSG as follows
:
where
FS
Descriptive examples of Ci computation are given in Table I
(five FSs) for different values of the spacing parameter.
TABLE I. Ci for PSG for five FSs [10].
To understand the decision table deriving procedure, detailed
example is given below.
The PSG_(e )= 0.5.
The PSG_∆e = 1.
The PSG_∆u = 2.
Angle = 30 . The construction parameters are given for the
previous ones. The networks and their consistent decision
tables are shown in Fig. 5. Note that the nodes are indicated
by red stars and the output points by blue circles. Purple
arrows are an example of minimum distances between output
points and network nodes related to FS mapping to the
decision table
.
(a) (b)
Fig.5. Example: (a) Grid constitution for decision table construction. (b)
Derived decision table[10].
3.2 Decoding Membership Functions
In the studying to design the FLC membership functions with
the two inputs and one output, a various of assumptions are
made in admiration of the distribution of fuzzy sets across the
UOD for each fuzzy variable. These assumptions are:
The internal and central UOD range of membership
functions could adopt either triangular ,two sided Gaussian,
sigmoidal, Gaussian curve, Generalized bell, product of two
sigmoidally, Trapezoidal .shaped membership function
shapes .
a) Outer UOD range MFs is unbounded s-shaped or z-
shaped.
b) A half overlaping is maintained between adjacent
membership functions.
c) The apex for (trimf), or plateau for (trapmf), are
coincident with zero segments of other adjacent and
non-adjacent MF within the UOD.
d) the number of fuzzy sets for the MFs fixed at five
sets (NB, NS, Z, PS PB).
e) Member of centers are distributed uniformly in the
UOD.
f) Using the presumption that have been invoked
above we are able to find the MFs for inputs/outputs
by finding the three parameters to minimizing the
chromosomes size.
for inputs/output by finding the three parameters to
minimizing the chromosomes size.
The membership function shape is specified by and
as follows:
First, we have to choose 'Trapezoidal' Membership
Function and find the four parameters using two variables,
The MF shape is specified by and as follows:
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
156
Volume 2, 2023
a) determines the type of MF
shape as follows:
• 1: 'dsigmf';
• 2 : gauss2mf.
• For 3: gaussmf.
• 4: Gbellmf.
• Psigmf.
• trimf.
b) determines the symmetric space
with respect to the centre of the MF as
shown in figure 6 .
Fig. 6. Trapezoidal shape MF .
The author can replace the find PSG which used in
creating rules base by degree of MF-center shift to effect MF
to reduced or expand it,by using MF Compounding Field
(CF) and is fired to update the MF position parameters of
each MF by raising them to the power of CF [7]:
(4) (3.3)
When
is new MF position parameters of each MF
is old MF position parameters of each MF .
And the value of CF is effect MF compression or
expansion:
a) for CF < 1 : Z-MF is compressed, NB and NS
expand.
b) for CF > 1 : Z-MF expands, NB and NS compress.
c) for CF = 1 : uniform MF distribution.
Fig. 7. MF GA-chromosome segment (triangular MFs with non-uniform
distribution)[7]).
Fig.8. MF GA-chromosome segment (trapezoidal MFs with uniform
distribution)[7].
3.3 Decoding defuzz Method
The Genetic Algorithm attempts to optimize the
scaling factors (defuzzMethod type) of the
inputs of the fuzzy controller. They are included in the
GA-chromosome each consisting of 3-bits, which are
decoded to obtain the values DFM , determines the type
of defuzzMethod as follows:
a) if DFM =1 then defuzzification is 'centroid'.
b) if DFM =2 then defuzzification is 'bisector'.
c) if DFM = 3 then defuzzification is 'mom' .
d) if DFM = 4 then defuzzification is 'som'
e) if DFM = 5 then defuzzification is 'lom' .
3.4 Decoding andMethod
There are two options for specifying and method
(ANDM) which are specified by one bit as follows
a) if ANDM= 1 then and method is min.
b) if ANDM=2 then and method is pord.
3.5 Decoding or Method
There are three choices for specifying or method (
ORM) which are specified by one bit as follows:
a) If ORM =1 then method is max.
b) If ORM =2 then method is pord.
3.6 Decoding aggMethod
There are three choices for specifying aggMethod
(AGGM) which are specified by two bit as follows:
a) If AGGM =1 then agg Method is max.
b) if AGGM =2 then agg Method is sum .
c) if AGGM =3 then agg Method is probor .
3.7 Decoding FLC Scaling Factors
The genetic algorithm also attempts to optimize the
gains of scaling factors of the and inputs and
gain of output of the fuzzy logic control. The , ,
are included in the GA chromosome, which are decoded to
yield the appropriate gain blocks of the Simulink model used
to evaluate each controller.
3.8 Measure of FIS performance GA Objective
Function
After the fuzzy controller initial solution and begins the
iterative evaluation of the generated new solutions by an
objective (Index) function (OF ). The mathematical equation
of (Of) in control applications found that using the following
OF:
OF= *sum((e).^2.*time) (5)
Where is the error.
4. Simulations and Results
The case study is to track a trajectory with the shape of
circle. The figure 9 shows the GFL controlled robot system
which includes the GFL controlling the robot block, The
dynamic parameters and equation of the 3 DOF PUMA-560
robots are shown in the second chapter , the dynamic
equations are simulations in 3 link PUMA robot blocks,and
receiving the desired path from the path generator block. The
output of the robot block is connected to a 2-D scope showing
the desired and the actual motion of the joints, and to a 3-D
scope showing the real and the desired path followed.
The 3D scope block is used to draw the trajectory followed
by the robot in space. The inputs of this graphical block are
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
157
Volume 2, 2023
the actual and the desired coordinates of the end-effector of
the robot, while its output is a 3-D graph showing the actual
path followed in blue and the desired path in green. The 3-D
scope block is based on the “Multitrack 3D Simulink Scope”.
The joint signals as extended as the desired path of the end-
effector drawing a circle in the space are shown in figure 9 ;
these signals enter the GFL controller that generates the
corresponding torques needed to control the robot.
Fig. 9. GFL controlled robot system.
4.1 GA optimization of the FLC at 3 DOF
PUMA robot
The genetic algorithm which used in this research is
used to search of the fuzzy logic controller parameters based
on the method described in third section . The tuning
method is created MATLAB M-files .
Checking the output performance and redesign the fuzzy
system for many times to reach to an satisfactory results.
The OF at instant t
is
(4.2)
Used as a multi-objective function of the system performance
to give a better performance indicator of a control system
response , the GA looks for the tuned of the FLC parameters
to reach the designer specifications. Solutions with low OF
are considered as the fittest. After many iteration, we have
adopted the parameter encoding shown in Table II .
TABLE II. Parameters adopted for encoding at 3DOFPUMA robot.
Parameter
Interval
Number of encoding bits
,
[0,50]
7
[0,100]
7
,
[0.25,4]
4
ca
[0,((15*pi)/16)]
4
cs
[0.5,2]
4
clo
[0,1]
1
[0,0.25]
3
cf
[0.5,2]
4
[1,6]
3
DFM
[1;5]
3
ANDM
[0,1]
1
ORM
[0,1]
1
AGGM
[0,3]
2
Several execution threads were executed with all GA
parameters being different. The best result was obtained from
the sum of errors =0.896. The corresponding GA
characteristics are illustrated in Table IV .
TABLE IV GA adopted parameters at 3DOFPUMA robot.
Population size
16
Number of generations
1000
Selection method
Roulette wheel selection
Crossover method
1-point slicing
Crossover
Crossover probability
0.5
Mutation probability
0.05
The GFLC design parameters and the main characteristics
(MFs and decision table) are shown in Table V , Figure 10
and Table VI .
TABLE V GFLC design parameters at 3DOFPUMA robot.
Parameter
value
Joint 1
Joint 2
Joint 3
27.55905
41.338582
38.18897
29.92125
31.496062992125985
6.692913
6.45669
84.409448818897640
17.37007
0.25
1.5
0.75
0.5
2
1
2
3.25
1.5
ca
0.7854
0.5890
0.7854
cs
1.4
1.4
0.6
clo
0
0
0
0.25
30.214
0.1071
0.0714
0.1071
0.25
output
0.1071
0.2143
0.2143
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
158
Volume 2, 2023
Cf
1.1
1.9
1.4
Cf
1.6
1.1
2
output
0.7
1.2
1.4
DFM
mom
bisector
bisector
ANDM
min
min
min
ORM
max
max
MAX
AGGM
sum
sum
SUM
TABLE VI .a The Rule table of the GFLC for first joint
1
e1
NB
NS
Z
PS
PB
NB
NB
NB
NB
NS
NS
NS
NB
NB
NB
NS
Z
Z
NS
NS
Z
PS
PS
PS
Z
PS
PS
PB
PB
PB
PS
PS
PB
PB
PB
TABLE VI .b The Rule table of the GFLC for second joint .
e 2
2
NB
NS
Z
PS
PB
NB
NB
NB
NB
NS
NS
NS
NB
NB
NS
NS
Z
Z
NS
NS
Z
PS
PS
PS
Z
PS
PS
PB
PB
PB
PS
PS
PB
PB
PB
TABLE VI.c The Rule table of the GFLC for third joint .
e3
NB
NS
Z
PS
PB
NB
NB
NB
NB
NS
NS
NS
NB
NS
NS
Z
Z
Z
NS
NS
Z
PS
PS
PS
Z
Z
PS
PS
PB
PB
PS
PS
PB
PB
PB
Fig.10. Optimal membership functions of the GFLC for
3 DOF PUMA robot
The figure 11 shown the following Trajectories in the
simulation combined together with the desired spatial
circle trajectory, and Actual versus Desired joint angles
of the system.
Fig.11. Actual versus desired joint angles and trajectory.
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
159
Volume 2, 2023
4.2 Adding uncertainty at 3DOF PUMA robot
The 3 DOF Puma 560 with parameters uncertainty It
will be interpreted as changes in inertia values . The
Inertial Constants with parameters uncertainty are thus
introduced as (,Im2 = 4.71 + 0.01,Im5 = 0.179 +0.004,
,Im6 = 0.193 + 0.01) kg.m2.The figure 4.14 shown the
Actual versus Desired joint angles of the system with
uncertainty added, it was a result OF total errors =
0.9305 , The difference is 0.0365 for not adding the
uncertainty to the system. .The tracking error in task
space is shown in Figure 12 .
Fig. 12. Real Actual versus Desired joint angles and trajectory of the
system with uncertainty added
5. Conclusion
This work has been introduced to study of advanced
control systems based on fuzzy logic controls and genetic
algorithm optimization technique which applied to the
end‐point 3 DOF PUMA 560 robot positioning of a
planar three .
Due to the nonlinear characteristics and parameter
changes in real environments, strong model uncertainty,
and very complex, MIMO tracking control of a robot arm
system is challenging. FLC was used in this research
because it is a large scale system for controlling the
nonlinear and uncertain system of parameters.
The performance of proposed GFLC has shown quite
satisfactory performances I terms of settling time, rise
time, steady state error, and settling in the endpoint
position of the manipulator arm. The composite
intelligent controller has also demonstrated robustness
capabilities against additional system uncertainty. In
other words, the retuning design approach by genetic the
algorithm offers a complete and fast way to design a
robust controller. But there are many cons of using GLFC
which can be summarized in the following points
:
a) It takes several attempts to select parameters for GA.
b) Determine the appropriate field of research for each
variable that represents FLC.
c) Writing the code for GFLC has the difficulties of
representing and dealing with FLC variables because they
are type structure. d) It takes a long time to run the GFLC
program , as it sometime needs more than 48 continuous
hours at MATLAB SIMULINK to make a computer and
it will not be able to bear this load , which makes it stop
before the search process is completed .running the
algorithm for too long is inefficient while stopping the
algorithm too early can result in an infeasible or sub-
optimal solution. In the context of a multimodal FLC
solution space, the use a restart strategy for GA in this
study provided a successful solution to the problem of
algorithm termination .The work stage is divided into
several stages, and each research phase takes the last
population and uses it in the initial population for the next
stage.
References
[1] Passino, K. M., Yurkovich, S., & Reinfrank, M. (1998). Fuzzy
control (Vol. 42, pp. 15-21). Reading, MA: Addison-wesley.
[2] Medsker, L. R. (2012). Hybrid intelligent systems. Springer Science &
Business Media.
[3] Muhammad R., Razali, Ahmad A. (2022). A hybrid controller method
with genetic algorithm optimization to measure position and angular for
mobile robot motion control Front Robot AI.
[4] Corke, P. I., & Khatib, O. (2011). Robotics, vision and control:
fundamental algorithms in MATLAB (Vol. 73, p. 2). Berlin: Springer.
[5] Armstrong, B., Khatib, O., & Burdick, J. (1986, April). The explicit
dynamic model and inertial parameters of the PUMA 560 arm.
In Proceedings. 1986 IEEE international conference on robotics and
automation (Vol. 3, pp. 510-518). IEEE.
[6] Mazhari, S. A., & Kumar, S. (2008). PUMA 560 optimal trajectory
control using genetic algorithm, simulated annealing and generalized pattern
search techniques. International Journal of Electrical and Computer
Engineering, 2(5), 830-839.
[7] Byrne, J. P. (2003). GA-optimization of a fuzzy logic controller. School
of Electronic Engineering.
[8] Mohammad V. and Ahmed F. (2018). Energy efficiency
in the robot arm using genetic algorithm. Conference of AI &
Robotics and 10th RoboCup Iranopen International
Symposium (IRANOPEN).
[9] Mazhari, S. A., & Kumar, S. (2008). PUMA 560 optimal
trajectory control using genetic algorithm, simulated
annealing and generalized pattern search
techniques. International Journal of Electrical and Computer
Engineering, 2(5), 830-839.
>@/RXGLQL00RGHOOLQJDQGLQWHOOLJHQWFRQWURO
RI DQ HODVWLF OLQN URERW PDQLSXODWRU,QWHUQDWLRQDO
-RXUQDORI$GYDQFHG5RERWLF6\VWHPV
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
160
Volume 2, 2023
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
The authors equally contributed in the present
research, at all stages from the formulation of the
problem to the final findings and solution.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
Conflict of Interest
The authors have no conflicts of interest to declare
that are relevant to the content of this article.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
International Journal of Applied Sciences & Development
DOI: 10.37394/232029.2023.2.16
Twfiq H. Elmenfy,
Mona Mohammed Mosa, Samah Abdelsalam
E-ISSN: 2945-0454
161
Volume 2, 2023
