Inverse kinematic analysis of 4 DOF pick and place arm robot manipulator using fuzzy logic controller

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 10, No. 2, April 2020, pp. 1376~1386
ISSN: 2088-8708, DOI: 10.11591/ijece.v10i2.pp1376-1386  1376
Journal homepage: http://ijece.iaescore.com/index.php/IJECE
Inverse kinematic analysis of 4 DOF pick and place arm robot
manipulator using fuzzy logic controller
Tresna Dewi1
, Siti Nurmaini2
, Pola Risma3
, Yurni Oktarina4
, Muhammad Roriz5
1,3,4,5
Electrical Engineering Department, Politeknik Negeri Sriwijaya, Indonesia
2
Intelligent System Research Group, Faculty of Computer Science, Universitas Sriwijaya, Indonesia
Article Info ABSTRACT
Article history:
Received Sep 9, 2019
Revised Oct 17, 2019
Accepted Oct 27, 2019
The arm robot manipulator is suitable for substituting humans working in
tomato plantation to ensure tomatoes are handled efficiently. The best design
for this robot is four links with robust flexibility in x, y, and z-coordinates
axis. Inverse kinematics and fuzzy logic controller (FLC) application are for
precise and smooth motion. Inverse kinematics designs the most efficient
position and motion of the arm robot by adjusting mechanical parameters.
The FLC utilizes data input from the sensors to set the right position and
motion of the end-effector. The predicted parameters are compared with
experimental results to show the effectiveness of the proposed design and
method. The position errors (in x, y, and z-axis) are 0.1%, 0.1%, and 0.04%.
The rotation errors of each robot links (θ1, θ2, and θ3) are 0%, 0.7% and
0.3%. The FLC provides the suitable angle of the servo motor (θ4)
responsible in gripper motion, and the experimental results correspond to
FLC’s rules-based as the input to the gripper motion system. This setup is
essential to avoid excessive force or miss-placed position that can damage
tomatoes. The arm robot manipulator discussed in this study is a pick and
place robot to move the harvested tomatoes to a packing system.
Keywords:
Degree of freedom
Forward kinematicsl inverse
kinematics
Fuzzy logic controller
Copyright © 2020 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Siti Nurmaini,
Intelligent System Research Group, Faculty of Computer Science,
Universitas Sriwijaya,
Jalan Raya Prabumulih-Inderalaya, Palembang, Indonesia.
Email: siti_nurmaini@unsri.ac.id
1. INTRODUCTIO
Due to its location on the equator, Indonesia is blessed with tropical climate and abundance of
biodiversity to support Indonesia economic growth in the agriculture and plantation sector. One of
the essential commodities of Indonesia plantation is tomatoes. A tomato requires special handling during
the harvesting time to ensure only the ripe ones picked, and none of them are crushed or damaged.
The efficiency and quality improvement are made possible by the application of digital farming employing
a robot to raise the quality and hygiene of the harvested fruit [1-9].
The arm robot manipulator is the most suitable type of robots to be applied in plantation and
agriculture for harvesting and packaging. The arm can be customized to imitate the human’s arm motion
from one point to others during harvesting. Robot motions can be designed using an inverse kinematics
method to generate the desired trajectory, and the robot follows the generated trajectory. The inverse
kinematics output is the ideal parameters and angles of robot links to ensure the smooth motion during
harvesting time [10-21]. The suitable end-effector of robot applied as a harvesting robot is a gripper.
Moreover, in order to achieve gripper’s smooth motion, artificial intelligence (AI) is applied to utilize
the input from the attached sensors at a robot’s system. The commonly used AI is the fuzzy logic controller
(FLC) [22-25] and the neural network (NN) [26-28]. Many kinds of research have applied inverse kinematics
to generate a robot trajectory. However, most of the previous research did not apply AI to ensure

Int J Elec & Comp Eng ISSN: 2088-8708 
Inverse kinematic analysis of 4 DOF pick and place arm robot manipulator using fuzzy … (Tresna Dewi)
1377
the smoothness of the robot [1-23]. Therefore, it is necessary to design a ready to applied arm-robot in
handling the harvested tomatoes. This design can be facilitated by carefully planning the mechanical
parameters in inverse kinematics design and simulated the resulting design in SciLab. The motion
smoothness can be achieved by implementing FLC to the end-effector design.
This paper discussed the design and control of a pick and place arm robot applied in tomato
packaging equipped with a gripper at the end-effector utilized to pick and place a target. In this study,
the considered target is a tomato to pick from a starting position and place into a belt conveyor.
The mechanical parameters of the arm robot manipulator are designed and predicted using the inverse
kinematics method. The designed parameters achieved from inverse kinematics are used to generate robot
trajectory, and the smoothness of gripper motion is ensured by applying FLC utilizing the input from
proximity sensors attached to the starting position of target and robot’s end-effector. The first proximity
sensor is for detecting the availability of tomato. The second sensor is to sense the distance between
the gripper and the tomato. The parameters from inverse kinematics also decide the angle of gripper opening
to ensure the right gripping position in holding the tomato; therefore, no damage caused during gripping and
placing the tomato. The novelty of this research is by designing a 4 DOF arm robot manipulator using inverse
kinematics analysis and combined with an FLC to ensure the smoothness of gripper motion. The input and
experimental results are compared to show the effectiveness of the proposed method. The experiment is
conducted by employing the robot to pick the harvested tomato from a box to a belt conveyor. The designed
robot is relatively cheap and straightforward, only employing two ultrasonic sensors as the proximity sensors.
2. RESEARCH METHOD
This study proposes the combination of inverse kinematics method and FLC to control a 4 DOF arm
robot manipulator applied in the digital farming system. The proposed method ensures the right design of
the mechanical system and the smoothness of the end-effector motion. The elaboration of the proposed
method is shown as a schematic diagram in Figure 1, while robot detail and the experimental setup are shown
in Figure 2. The primary input is the proximity sensors data where the first proximity sensor is to detect
target availability, and the second proximity sensor to detect the distance between the end-effector and
the target. The end-effector in this study is a gripper. The first proximity sensor is attached to the target pool,
and the second proximity sensor is attached to the gripper. These sensors positions are shown in Figure 2(a),
and Figure 2(b) shows the target in the final position or at the beginning of a belt conveyor. The proximity
sensors are the input to FLC in term of rules-based, which is dicussed in FLC design. FLC is employed to
decide the decision to move the robot based on the input from proximity sensors attached to the robot system.
The rules-based decides the rotation angles of end-effector’s motor rotation (θ4) related to mechanical design
in kinematics analysis.
The 4 DOF arm robot manipulator considered in this study is shown in Figures 2 and 3, where each
link is powered by a servo motor. Each servo motor rotation creates the angle or orientation of the robot,
and the angles are named θ1, θ2, and θ3, respectively, as presented in Figure 3(a) and 3(b). 1, 2, and 3 are
solved by kinematics analysis or widely known as inverse kinematics method. The kinematics analysis
method consists of forward and inverse kinematics of a robot. This method is conducted to design the motion
and pose of a robot based on the ideal mechanical condition of the applied robot such as the length of
the links presented in Figure 3(c). The output of inverse kinematics gives the precise parameters to design
the robot’s mechanics and motions. The physical meaning of inverse kinematics outputs is to regulate how
much the servo motors must rotate to ensure the right position of the robot and the gripping angle (θ4) applied
while holding the target.
Figure 1. The schematic diagram of the proposed method

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 10, No. 2, April 2020 : 1376 - 1386
1378
(a) (b)
Figure 2. Detail robot and experimental setup,
(a) Robot detail and target positions, (b) Target in Final Position
(a) (b) (c)
Figure 3. Four DOF arm robot applied in this study, (a) Robot in coordinate frames,
(b) Robot in initial position, (c) Robot dimension
2.1. Kinematics analysis method
Kinematics is a method of modeling robot motion without considering the force applied to move
the robot. Kinematics analysis is divided into forward and inverse kinematics. By applying kinematics
modeling, the robot design can be efficient by considering how the robot moves in x, y, and z-axis. The arm
robot considered in this study is shown in Figure 3. 1 is the angle between the first joint angle and the base,
2 and 3 are the second and third angles, and 4 is the angle of end-effector or gripper. The initial position of
the arm robot manipulator is shown in Figure 3(b).
2.1.1. Forward kinematics modeling
Kinematics analysis of arm robot manipulator starts from forward kinematics. Forward kinematics
utilizes mechanical parameters to calculate robot configuration and inverse kinematics to reverse
the derivation to acquire the desired configuration. The generic form of forward kinematics of an arm robot
manipulator is given by
𝑇3
0
= 𝐹𝑜𝑟𝑤𝑎𝑟𝑑 𝐾𝑖𝑛𝑒𝑚𝑎𝑡𝑖𝑐 [𝑃], (1)
While the inverse kinematics of an arm robot manipulator is given by
[𝑃] = 𝐹𝑜𝑟𝑤𝑎𝑟𝑑 𝐾𝑖𝑛𝑒𝑚𝑎𝑡𝑖𝑐 𝑇3
0
(2)
Where [P] is the desired position in x, y, and z-axis. 𝑇3
0
is the kinematics solution of the robot. Inverse
kinematics results are the angle of robot’s motion links by modeling robot parameters based on the target

1379
position, the length of the link, the angle between links and those parameters becomes the input to get
the transformation matrices. Those matrices are the forward kinematics required to get inverse kinematics.
The forward kinematics uses parameters to calculate the robot configuration, and the kinematic inverse
reverses this calculation to determine the parameters in achieving the desired configuration.
There are two ways to solve inverse kinematics of a robot that is using the inverse kinematic
trigonometry method and Denavit Hartenberg (DH) shown in Table 1. In this study, the mechanical
parameters are achieved by deriving DH analysis to get the transformation matrix. The robot parameters from
Figure 3(a) are sorted in Table 1, and the transformation matrix is given by
𝑇 = [
𝐶𝜃1 −𝑆𝜃1 0 0
𝑆𝜃1 𝐶𝛼1 𝐶𝜃1 𝐶𝛼1 −𝑆𝜃1 0
𝑆𝜃1 𝑆𝛼1 𝐶𝜃1 𝑆𝛼1 𝐶𝜃1 0
0 0 0 1
]1
0
, 𝑇 = [
𝐶𝜃2 −𝑆𝜃2 0 0
𝑆𝜃2 𝐶𝜃2 0 0
0 0 1 0
0 0 0 1
]1
2
(3)
𝑇 = [
𝐶𝜃3 −𝑆𝜃3 0 𝛼2
𝑆𝜃3 𝐶𝜃3 0 0
0 0 1 0
0 0 0 1
]2
3
, and 𝑇 = [
𝐶𝜃4 −𝑆𝜃4 0 𝛼3
𝑆𝜃4 𝐶𝜃4 0 0
0 0 1 0
0 0 0 1
]3
4
where Cθn is cos θn and Sθn is sin θn, and
Table 1. Denavit hartenberg parameters
NO θn dn−1 an−1 αn−1
1 θ1 0 0 α1
2 θ2 0 a2 0
3 θ3 0 a3 0
4 θ4 0 a3 0
2.1.2. Inverse kinematics
Inverse kinematics is the mathematical calculation for finding the parameters in forward kinematics,
and by knowing the best value for each parameter in Table 1, the desired position and orientation of the robot
is achieved. The derivation of inverse kinematics is a challenging mathematical task, especially when
the angles are more than 2. In this study, the inverse kinematics is only to find 1 to 3. The parameter 4 is
related to the end-effector, and FLC decides the end-effector motion based on the input from the proximity
sensors attached to the robot system.
Inverse kinematics in this paper is calculated by an algebraic method to find the reverse of (3) and
achieve 1 to 3. 1 to 3 calculation are given by (4), (5), and (6).
𝐶𝜃2 =
𝑝𝑥2+𝑝𝑦2−𝑎1
2−𝑎2
2
2𝑎122
(4)
therefore,
𝜃2 = 𝑎𝑡𝑎𝑛2(±√1 − (
2−𝑎2
2
2𝑎122
)
2
,
2−𝑎2
2
2𝑎122
) (5)
The angle created by the motion link 1 relative to the base (1) and the third link (3) are
𝜃1 = 𝑎𝑡𝑎𝑛2(𝑝𝑥
2
+ 𝑝 𝑦
2
− (𝑎2 𝑐𝑜𝑠𝜃2 + 𝑎1)2
𝑎2 𝑐𝑜𝑠𝜃2 + 𝑎1) (6)
𝜃3 = ±𝑎𝑡𝑎𝑛2(𝑝𝑥
2
+ 𝑝 𝑦
2
− (𝑎2 𝑐𝑜𝑠𝜃12 + 𝑎1)2
𝑎2 𝑐𝑜𝑠𝜃12 + 𝑎1)
2.2. Fuzzy logic controller design
FLC was introduced by Lofti Zadeh as a controller for a plant. Fuzzy sets are the logic based on
the input coming from sensors the fuzzy logic derived from the way human thinks instead of the exact
mathematical modeling. The rules-based for FLC application is the probability of possible conditions that
might occur in the system. In this study, the rules are set based on the position of a target relative to
proximity sensors detection as the input for the system, and the rules-based govern the robot whether to grab
the target, start to move, or stay in idle position. The rules-based set in study is shown in Table 2, where

 ISSN: 2088-8708
1380
inputs are the detection data from proximity sensors (sensors’ installation positions are shown in Figure 2)
and the robot base angle (θ1) shown in Figure 3(a). The more rules set, the smoother gripper motion is.
The membership function chosen in this study is given by
𝜇 = {
0 𝑖𝑓 𝑥 ≤ 𝛼
(𝑥 − 𝛼)/(𝛽 − 𝛼) 𝑖𝑓 𝛼 ≤ 𝑥 ≤ 𝛽
1 𝑖𝑓 𝑥 ≥ 𝛽
(7)
where α and β is the upper and lower boundary indicating “Near” and “Far” distance between the tomato
(target) and the gripper. α is between 0 to 5 cm, and β is in the range of 11 to 20 cm.
Based on the place of installation, the proximity sensors rules are divided into rules for proximity
sensor 1 and sensor 2, as shown in Figure 2(a). The first proximity sensor is attached to the picking place to
indicate the availability of tomato to be grabbed by the robot and placed on belt-conveyor. Proximity sensor 1
rules are in “Near” or “Far”; “Near” is when the distance is ≤ 5 cm, and “Far” is > 5cm. Proximity sensor 1
output data is related in robot motion in picking and placing a tomato. The rules for proximity sensor 2 are
set to be “Near,” “Medium,” and “Far.” “Near” is set when the proximity sensor detects the target to be 0 to 5
cm from the robot. “Medium” is set in the range 6 to 10 cm and “Far” is in 11 to 20 cm. The data from
proximity sensor 2 is related to the servo motor rotation and gripper angle (4) whether it should be “Small”,
“Medium”, and “Big”.
The rules in Table 2 are converted into the membership function as shown in Figure 4. Membership
inputs are shown in Figure 4(a) and output (gripper motion) is given in Figure 4(b). FLC outputs are
the possibilities of gripper’s final motion in grabbing the tomato, and are in term of “Small,” “Medium,” and
“Big.” These terms related to the degree (angle) of servo motors rotation and robot motions. “Small” means
the servo motor’s degree in the range of 20o to 25o, and the robot is in the state of picking and placing
the target or manipulated object. ”Medium” is when the servo rotation in a range of 25o to 25o, and the robot
is in initial motion or starting to pick and place the target. ”Big” is when the servo rotation in the range of
37o to 60o and the robot is in standby mode. The elaboration of how to relate the inputs to the output is by
fuzzification shown in Figure 5, and how the input in Figure 5(a) can cause the motion in Figure 5(b).
Table 2. Rules based fuzzy logic to controller 4DOF arm robot manipulator
NO
INPUT OUTPUT
Proximity Sensor1 Proximity Sensor2 BaseAngle(1) Gripper (4)
1 Near Near Small Small
2 Near Near Medium Small
3 Near Near Far Small
4 Near Medium Small Small
5 Near Medium Medium Medium
6 Near Medium Far Medium
7 Near Far Small Medium
8 Near Far Medium Medium
9 Near Far Far Medium
10 Medium Near Small Medium
11 Medium Near Medium Medium
12 Medium Near Far Medium
13 Medium Medium Small Medium
14 Medium Medium Medium Medium
15 Medium Medium Far Medium
16 Medium Far Small Medium
17 Medium Far Medium Medium
18 Medium Far Far Big
19 Far Near Small Big
20 Far Near Medium Big
21 Far Near Far Big
22 Far Medium Small Big
23 Far Medium Medium Big
24 Far Medium Far Big
25 Far Far Small Big
26 Far Far Medium Big
27 Far Far Far Big

1381
(a) (b)
Figure 4. Membership function of input and output,
(a) Input membership function, (b) Output membership function
(a)
Membership function for output number 1 named Motor Gripper
Big
Medium
Small
0.8
37
0.35
Motor Gripper Angle
0.15
Mu(MotorGripperAngle)
(b)
Figure 5. Input and output fuzzification, (a) Input membership function, (b) Output membership function
3. RESULTS AND ANALYSIS
Data collecting method in this study is divided into two steps, arm robot motion data; taken using
inverse kinematics to detect the precise position of tomato in Cartesian coordinates frames, and Gripper data;
taken by calculating how much the servo rotation angles produced during picking and placing tomatoes.
The arm-robot positions in Cartesian coordinates frames are converted into the angles of every robot joints
(1, 2, and 3) in Figure 3a, and the calculation of 1, 2, and 3 is given in eq. 4 to 6. The initial position
of the arm robot manipulator is shown in Figure 3(b). Table 3 presents the robot positions reference in
Cartesian coordinates frame (x, y, and z-axis) as the input to kinematics analysis to calculate 1, 2, and 3,
and the actual robot positions in x, y, and z-axis By inverse kinematics calculation, arm robot position
relatives to the target and the angles of every joint are acquired by knowing the length of links and other
mechanical parameters presented in Table 1. The inverse kinematics results are compared to the real data
from experiments to show the effectiveness of the inverse kinematics analysis in this study. Tables 4 and 5
show the error comparison between the data from inverse kinematics and the actual data taken from
the experiment. The error is calculated by
% Error =
Data from Inverse Kinematics
Actual Data from Experiment
× 100% (8)

 ISSN: 2088-8708
1382
Table 3. Input and output positions of arm robot manipulator
No Input
Cartesian (cm)
Output
Cartesian (cm) Orientation Angle (o)
x y z x y z 1 2 3
1 9 2 0 8.9 2.1 0 0 74.6 53.4
2 9 3 0 9.1 3.2 0 0 86.5 59.9
3 9 4 0 9.1 3.8 0 0 100.4 67
4 9 5 1 9 4.9 1.1 6.34 118.4 76
5 9 6 1 8.9 6.1 1.1 6.34 121.5 80
6 10 4 1 10.2 3.9 1.2 5.71 91.8 59.1
7 10 4 5 10.1 3.9 5.1 26.6 91.8 59.1
8 11 4 5 11 3.9 4.9 24.4 84.9 52,1
9 12 4 5 12.1 4 5.2 22.7 79 45.5
19 10 4 5 9.9 4 5 26.6 91.8 59.1
Table 4. Arm robot error positions in Cartesian coordinates
No Input Output Error ErrorPercentage
x y z x y z x y z x y z
1 9 2 0 8.9 2.1 0 0.1 0.1 0 1.1% 5% 0%
2 9 3 0 9.1 3.2 0 0.1 0.2 0 1.1% 6.7 % 0%
3 9 4 0 9.1 3.8 0 0.1 0.2 0 1.1% 5% 0%
4 9 5 1 9 4.9 1.1 0 0.1 0.1 0% 2% 10%
5 9 6 1 8.9 6.1 1.1 0.1 0.1 0.1 1.1% 1.6% 10%
6 10 4 1 10.2 3.9 1.2 0.2 0.1 0.2 2% 2.5% 20%
7 10 4 5 10.1 3.9 5.1 0.1 0.1 0.1 1% 2.5% 2%
8 10 4 5 11 3.9 4.9 0 0.1 0,1 0% 2.5% 2%
9 10 4 5 12.1 4 5.2 0.1 0 0.2 0,8% 0% 4%
10 10 4 5 9.9 4 5 0.1 0 0 1% 0% 0%
Error Average 0.1% 0.1% 0.04%
Table 5. Error percentage between the angles achieved from inverse kinematics and experiment
No Inverse Kinematics Prediction Output From Experiments Error Percentage
Angle (o) Angle (o) Angle (o)
1 2 3 1 2 3 1 2 3
1 0 74.6 53.4 0 74.6 53.4 0 0 0
2 0 86.5 59.9 0 86.5 59.9 0 0 0
3 0 100.4 67 0 100.4 67 0 0 0
4 6.34 120 76 6.34 118.4 76 0 1.4% 0
5 6. 34 120 80 6.34 121.5 80 0 1.25% 0
6 5.71 90 60 5.71 91.8 59.1 0 2% 1.5%
7 26,6 90 60 26.6 91.8 59.1 0 2% 1.5%
8 24.4 84.9 52.1 24.4 84,9 52.1 0 0 0
9 22.7 79 45.5 22.7 79 45.5 0 0 0
10 26,6 91.8 59.1 26.6 91.8 59.1 0 0 0
Error Average 0% 0.7% 0.3%
The comparison of robot references and actual positions are shown in Figure 6(a) where Xi, Yi, and
Zi are the positions input to inverse kinematics, and X0, Y0, and Z0 are the actual position. The errors between
the references and actual positions are shown in Figure 6(b) where Error_Pos_X, Error_Pos_Y,
and Error_Pos_Z are the error position in x, y, and z-axis respectively. Table 4 and Figure 6 indicate that
the errors are very small (0.1%, 0.1%, and 0.04% in x, y, and z-axis respectively).
The comparison of robot’s orientation between the predicted by inverse kinematics and actual
orientation from the experiment is shown in Figure 7(a) where 𝑇1𝑖, 𝑇2𝑖, and 𝑇3𝑖 are 1, 2, and 3 reference
inputs respectively, and 𝑇10, 𝑇20, and 𝑇30 are the actual 1, 2, and 3 achieved from the experiment.
The error between the predicted and actual orientation is given in Figure 7 where %T1, %T2, and %T3 are
the error in 1, 2, and 3, respectively. Table 5 and Figure 7 show that the errors are very small (0%, 0.7%,
and 0.3% for 1, 2, and 3, respectively). The errors in robot positions and orientations occur due to
the servo gears do not allow the motor to rotate precisely and the rounding calculation that unavoidable to
the limited microprocessor memory.

1383
(a) (b)
Figure 6. The comparison of robot’s references and actual,
(a) Comparison of robot position references and actual, (b) Comparison error
(a) (b)
Figure 7. The comparison of robot’s orientation between the predicted by inverse kinematics and
actual orientation achieved by experiment, (a) Comparison of robot orientation references and actual,
(b) Comparison error
Ten experiments were conducted to compare the arm robot error position in Cartesian coordinates
and angle of orientation achieved from inverse kinematics and experiments. In experiment 1, the Cartesian
reference input in x, y, and z-axis are 9, 2, and 0, respectively. This coordinates position is the target to be
achieved by arm robot manipulator. The real positions from experiments are 8.9, 2.1, and 0 for x, y, and
z-axis and 0o
, 74.6o
, and 53.4o
for 1,2, and 3, respectively. The error found in this experiment is ≤ 0.1 cm.
In experiment 5, the Cartesian reference input in x, y, and z-axis are 9, 6, and 1, respectively.
This coordinates position is the target to be achieved by arm robot manipulator. The real positions from
experiments are 8.9, 6.1, and 1.1 for x, y, and z-axis and 6.34o
, 121.5o
, and 80o
for 1, 2, and 3,
respectively. The error found in this experiment is 0.1 cm. In experiment 7, the Cartesian reference input in x,
y, and z-axis are 10, 4, and 5, respectively. This coordinates position is the target to be achieved by arm robot
manipulator. The real positions from experiments are 10.1, 3.9, and 5.1 for x, y, and z-axis and 26.6 o
, 91.8 o
,
and 59.1o
for 1,2, and 3, respectively. The error found in this experiment is 0.1 cm.
In experiment 9, the Cartesian reference input in x, y, and z-axis are 12, 4, and 5, respectively.
This coordinates position is the target to be achieved by arm robot manipulator. The real positions from
experiments are 12.1, 4, and 5.2 for x, y, and z-axis and 22.6o
, 79o
, and 45.5o
for 1, 2, and 3, respectively.
The error found in this experiment is ≤ 0.2 cm. In experiment 10, the Cartesian reference input in x, y, and
z-axis are 10, 4, and 5, respectively. This coordinates position is the target to be achieved by arm robot
manipulator. The real positions from experiments are 9.9, 4, and 5 for x, y, and z-axis and 26.5o
, 91.8o
, and
59.1o
for 1,2, and 3, respectively. The error found in this experiment is ≤ 0.1 cm.

 ISSN: 2088-8708
1384
As presented in Table 2, FLC is responsible for rotating the motor to move the gripper (end-
effector). The artificial intelligence is necessary to create robot motion in picking and placing tomato.
The defined smoothness is including controlling the force applied to the robot to ensure the robot do not
damage tomatoes as the targets in this study. Table 6 shows the comparison between FLC rules-based in
Table 2 and the experimental results data. If the angle of the motor gripper is between 0o
-25o
is the angle,
then the robot is gripping the tomato. If the angle of motor gripper is between 25o
-33o
is the angle,
then the robot is starting to move or initial motion. If the angle of the motor gripper is between 37o
-60o
is
the angle then the robot is in the standby position. 60o
is the position when the gripper is in full open, and 0o
is the angle when the gripper is completely closed.
In sample 1; If Proximity sensor1: Near AND Proximity sensor2: Near AND Motor1 angle: Small
then Gripper: Picking and Placing (Small). This condition is where the gripper is 20o
open or in rules-based is
Small. In sample 2; If Proximity sensor1: Near AND Proximity sensor2: Far AND Motor1 angle: Medium
then Gripper: Initial Motion (Medium). This condition is where the gripper is 27o
open or in rules-based is
Medium. In the sample 3; If Proximity sensor1: Medium AND Proximity sensor2: Far AND Motor1 angle:
Medium then Gripper: Initial Motion (Medium). This condition is where the gripper is 33o
open or in rules-
based is Medium. In sample 4; If Proximity sensor1: Far AND Proximity sensor2: Far AND Motor1 angle:
Small then Gripper: Initial Motion (Medium). This condition is where the gripper is 60o
open or in rules-
based is Medium. The input of FLC and the data taken from experiment in Table 6 show that the gripper
angle (4) or output data is following the rules set in FLC input shown in Table 2. The difference is in data
no 10 of Table 6, and this condition might occur due to the size of the tomato. Figure 8 shows the experiment
of picking and placing tomatoes, from the tomatoes pool to the belt-conveyor shown in Figure 2. Figure 8(a)
shows the robot in initial motion; where it is taking the green tomatoes, Figure 8(b)-(e) show the process of
picking and placing tomatoes, and Figure 8(f) shows the robot in the standby position. The experimental
results show that the proposed method to design arm robot manipulator is effective proven by the small errors
in position and orientation of robot achieved by simulation design and real application. The designed robot is
cost-effective and more accessible to apply by any farmer.
Table 6. Comparison between the input FLC and the experiment results of 4
o Input Proximity Sensor 1 Input Proximity Sensor 2 Input Rules Based End Effector Angle(4)ino
1 Detected 0-5 cm Small 20o−25o(Small)
5 Detected 6-10 cm Medium 25o−33o(Medium)
10 No detection 0-5 cm Big 37o−60o(Big)

1385
(a) (b) (c)
(d) (e) (f)
Figure 8. Robot picking and placing green and red tomato from initial motion and standby, (a) Initial motion,
(b) Picking and placing, (c) Picking and placing, (d) Picking and placing, (e) Picking and placing, (f) Standby
4. CONCLUSION
The digital farming application allows efficient and effective harvesting assisted by a robot.
The current technology makes robot production cost cheaper, and easier for the farmer to apply it.
Robot trajectory generation can be achieved by inverse kinematics analysis. Inverse kinematics analysis
produces the position and orientation of a robot as the trajectory references. The end-effector in this study is
a gripper that suitable to grab the harvested tomatoes (target). Gripper motion is designed by FLC to increase
motion smoothness and the amount of force applied to the target to avoid excessive power in grabbing
the target. Ten experiments were conducted to show the effectiveness of the proposed method.
The experimental results show that the actual robot position achieved from experiments are similar to
the reference inputs from inverse kinematics analysis. The position errors are 0.1%, 0.1%, and 0.04% in x, y,
and z-axis respectively, and orientation errors are 0%, 0.7%, and 0.3% for 1, 2, and 3 respectively.
The errors in robot positions and orientations occur due to the servo gears do not allow the motor to rotate
precisely, and the rounding calculation that unavoidable to the limited microprocessor memory.
By comparing the FLC rules and experiment data, it shows that the gripper motions also follow the rules
from FLC.
REFERENCES
[1] M. Stoelen, K. Krzysztof, V. F. Tejada, N. Heiberg, C. Balaguer, and A. Korsaeth, “Low-Cost Robotics for
Horticulture: A Case Study on Automated Sugar Pea Harvesting,” 10th European Conference on Precision
Agriculture (ECPA), 2015. DOI: 10.3920/978-90-8686-814-8 34.
[2] T. T. Nguyen, E. Kayacan, J. De Baedemaeker, W. Saeys, “Task and Motion Planning for Apple Harvesting
Robot,” IFAC Proceedings Volumes, Vol. 46, Issue. 18, pp. 247-252, 2013. https://doi.org/10.3182/20130828-2-
SF-3019.00063.
[3] G. Wang, Y. Yu, and Q. Feng, “Design of End-Effector for Tomato Robotic Harvesting,” IFAC-PapersOnLine,
Vo. 49, Issue 16, pp. 190-193, 2016. https://doi.org/10.1016/j.ifacol.2016.10.035

 ISSN: 2088-8708
1386
[4] W. Ji et al.,“Automatic Recognition Vision System Guided for Apple Harvesting Robot Computers and Electrical
Engineering,” Computers & Electrical Engineering, Vol. 38, No. 5, pp. 11861195, 2012.
https://doi.org/10.1016/j.compeleceng.2011.11.005
[5] C. Xiong, C. Xiaoman, X. Zou X, “A Method for Identification and Matching of the Picking Point for Mature
Litchi under Structural Environment,” Journal of Applied Biotechnology & Bioengineering, Vol. 3, Issue 6, pages.
4, 2017. Retrieved: http://medcraveonline.com.
[6] S.S. Mehta and T. F. Burks, “Vision-Based Control of Robotic Manipulator for Citrus Harvesting,” Computers and
Electronics in Agriculture, Vol. 102, pp.14658, 2014. http://dx.doi.org/10.1016/j.compag.2014.01.003.
[7] J. Li, M. Karkee, Q. Zhang, K. Xiao, and T. Feng, “Characterizing Apple Picking Patterns for Robotic Harvesting”,
Computers and Electronics in Agriculture, Vol. 127, pp.633640, 2016.
http://dx.doi.org/10.1016/j.compag.2016.07.024.
[8] P. M. Blok, R. Barth, and W. van den Berg, “Machine Vision for a Selective Broccoli Harvesting Robot,” IFAC-
PapersOnLine, Vol. 49, Issue. 16, pp. 6671, 2016. http://dx.doi.org/10.1016/j.ifacol.2016.10.013.
[9] G. Feng, C. Qixin, and M. Nagata, “Fruit Detachment and Classification Method for Strawberry Harvesting
Robot,” International Journal of Advanced Robotic Systems, Vol. 5, Issue 1, pp. 4148, 2008.
https://doi.org/10.5772/5662.
[10] T. Dewi, P. Risma, Y. Oktarina, and M. T. Roseno, “Perbandingan Desain dan Pemodelan Tangan Robot Lima Jari
sebagai Sistem Underactuated (In Bahasa),” Annual Research Seminar 2016, Vol 2, No. 1, pp. 139-145, 2016.
[11] M. Ceccarelli and E. Ottaviano, “Kinematic Design of Manipulators,” Robot Manipulators, Marco Ceccarelli (Ed.),
InTechopen, pp. 49-72, 2008. ISBN: 978-953-7619-06-0.
[12] H. Lin, Y. Liu, and Y. Lin, “Intuitive Kinematic Control of a Robot Arm via Human Motion,” in Proceeding of
37th National Conference on Theoretical and Applied Mechanics (37th NCTAM 2013) & The 1st International
Conference on Mechanics (1st ICM), Procedia Engineering, vol. 79, pp. 411-416, 2014.
[13] A. B. Rehiara, “Kinematics of Adept Three Robot Arm,” Robot Arms, Prof. Satoru Goto (Ed.), pp. 21-38,
InTechopen, 2011, ISBN: 978-953-307-160-2.
[14] P. Donelan, “Kinematic Singularities of Robot Manipulators,” Advances in Robot Manipulators, Ernest Hall (Ed.),
InTechopen, pp. 401-416, 2010. ISBN: 978-953-307-070-4.
[15] W. Song, and G. Hu, “A Fast Inverse Kinematics Algorithm for Joint Animation,” in Proceeding of 2011 Inter-
national Conference on Advances in Engineering, Procedia Engineering, vol. 24, pp. 350-354, 2011.
[16] T. Balkan, M. K. Ozgoren, and M. A. S. Arikan, “Structure Based Classification and Kinematic Analysis of Six-
Joint Industrial Robotic Manipulators,” Industrial Robotics: Theory, Modelling and Control, Sam Cubero (Ed.),
InTechOpen, pp. 149-184, 2006. ISBN: 3-86611-285-8
[17] S. Zodey, and S. K.Pradhan, “Matlab Toolbox for Kinematic Analysis and Simulation of Dexterous Robotic
Grippers,” Procedia Engineering, Vol. 97, pp. 1886-1895, 2014. https://doi.org/10.1016/j.proeng.2014.12.342
[18] S. Kucuk and Z. Bingul, “Robot Kinematics: Forward and Inverse Kinematics,” Industrial Robotics: Theory,
Modelling, and Control, Sam Cubero (Ed.), InTechopen, pp. 117-148, ISBN: 3-86611-285-8.
[19] I. A. Sultan “Inverse Position Procedure for Manipulators with Rotary Joints,” Industrial Robotics: Theory,
Modelling and Control, Sam Cubero (Ed.), InTechopen, pp. 185-210, 2006. ISBN: 3-86611-285-8.
[20] K. E. Clothier, and Y. Shang, “A Geometric Approach for Robotic Arm Kinematics with Hardware Design,
Electrical Design, and Implementation,” Journal of Robotics, vol. 2010, Article ID 984823, 10 pages, 2010.
doi:10.1155/2010/984823.
[21] A. Mohammed, B. Schmidt, L. Wang, and L. Gao, “Minimizing Energy Consumption for Robot Arm Movement,”
Procedia CIRP, Vol. 25, pp. 400-405, 2014. https://doi.org/10.1016/j.procir.2014.10.055.
[22] A. El-Sherbiny, M. A. Elhosseini, and A. Y.Haikal, “A comparative study of soft computing methods to solve
inverse kinematics problem,” Ain Shams Engineering Journal, Vol. 9, No. 4, 2018, pp. 2535-2548,
https://doi.org/10.1016/j.asej.2017.08.001.
[23] M. M. U. Atique, M. R. I. Sarker, and M. A. R. Ahad, “Development of an 8DOF quadruped robot and
implementation of Inverse Kinematics using Denavit-Hartenberg convention,” Heliyon, Vol. 4, No. 12, e01053,
2018, 0 https://doi.org/10.1016/j.heliyon.2018.e01053.
[24] S. Nurmaini, “Intelligent Navigation in Unstructured Environment by using Memory-Based Reasoning in
Embedded Mobile Robot,” European Journal of Scientific Research, Vol. 72, No. 2, pp. 228-244, 2012.
[25] S. Nurmaini, B. Tutuko, K. Dewi, V. Yuliza, and T. Dewi, “Improving posture accuracy of non-holonomic mobile
robot system with variable universe of discourse,” TELKOMNIKA Indonesian Journal of Electrical Engineering,
Vol 15, No. 3, pp. 1265-1279, 2017. http://dx.doi.org/10.12928/telkomnika.v15i3.6078.
[26] T. Dewi, P. Risma, and Y. Oktarina, “Fuzzy Logic Simulation as Teaching-Learning Media for Artificial
Intelligence Class,” Journal of Automation, Mobile Robotics & Intelligent Systems, Vol. 12, No. 3, pp. 59-65, 2018.
DOI: 10.14313/JAMRIS 4-2013/59.
[27] S. Nurmaini and F. Setianingsih, "Enhancement of The Fuzzy Control Response with Particle Swarm Optimization
in Mobile Robot System," 2018 International Conference on Electrical Engineering and Computer Science
(ICECOS), Pangkal Pinang, 2018, pp. 73-78.doi: 10.1109/ICECOS.2018.8605221.
[28] A. Duka, “ANFIS based Solution to the Inverse Kinematics of a 3DOF Planar Manipulator,” 8th International
Conference Interdisciplinary in Engineering, INTER-ENG 2014, Procedia Technology, Vol. 19, pp. 526-533, 2015.
https://doi.org/10.1016/j.protcy.2015.02.075.

Inverse kinematic analysis of 4 DOF pick and place arm robot manipulator using fuzzy logic controller

More Related Content

Similar to Inverse kinematic analysis of 4 DOF pick and place arm robot manipulator using fuzzy logic controller (20)

More from IJECEIAES (20)

Recently uploaded (20)

Inverse kinematic analysis of 4 DOF pick and place arm robot manipulator using fuzzy logic controller