2022 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET)

A Review of SCARA Robot Control System

See Han Tay 1, Wai Heng Choong 2, Hou Pin Yoong 3,*
1, 3
Biomechatronics Research Laboratory
Faculty of Engineering, Universiti Malaysia Sabah
2022 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET) | 978-1-6654-6837-4/22/$31.00 ©2022 IEEE | DOI: 10.1109/IICAIET55139.2022.9936755

Kota Kinabalu, Sabah, Malaysia

Email: [email protected] 1, * Corresponding author: [email protected] 3
2 Underwater Research Laboratory

Faculty of Engineering, Universiti Malaysia Sabah

Kota Kinabalu, Sabah, Malaysia
Email: [email protected]

Abstract—In the era of industrial revolution 4.0 (IR4.0), that requires high precision and low power consumption. To
robotics is a common automation tool to comply the repetitive achieve high accuracy, the SCARA robot requires an
and dangerous jobs for assisting the human workforce. Among advanced controller to handle and control the nonlinear
the robotic arm, the SCARA robot is one of the well-established dynamic behaviour of the robot system [12], [13]. The detail
robots in the industry due to its high speed and high accuracy. of the SCARA robot will be discussed in the next section
To achieve good control at high speed, the controller is key to including the implementation, kinematic modelling, dynamic
ensuring the SCARA robot at such a performance. This paper modelling and trajectory planning.
presents a brief review of the controllers and their pros and cons
when applying to the SCARA robot. The reviewed controllers
include PID Control, fuzzy logic control, neural networks II. SCARA ROBOT
control, sliding mode control, impedance control, adaptive SCARA robot was invented in 1979 by a Japanese
control and robust control. Besides that, the review also covers professor, Professor Hiroshi Makino [1]. The SCARA robot
the implementation of the Internet of Things (IoT) and artificial development intends to enhance the robot performance with
intelligence (AI) on the SCARA robot as the trend of IR 4.0. high accuracy and fast operation, both of which are SCARA's
benefits [14]. Therefore, the establishment of the SCARA
Keywords—robotic, SCARA robot, controller, Internet of
Things
robot had been implemented in many industries such as the
medical industry and manufacturing industry [15].
I. INTRODUCTION A. Kinematic Modelling
In the 20th century, robotics has become more important Kinematic modelling for the SCARA robot provides the
as a flexible automation tool to comply the repetitive and physical description in the form of mathematics of the robot.
dangerous jobs for reducing worker hazards and tiredness. At The kinematic modelling can be divided into two parts which
the same time, the use of robot automation also increases are forward kinematic modelling and inverse kinematic
production efficiency and consistency. The first modelling. The forward kinematic modelling for the SCARA
implementation of robots started in 1937, which is a cartesian is formulated to yield the end-effector position with the joint
robot [1]. Since that, the cartesian robot is widely been used angles as input. Therefore, the forward kinematic modelling
in the automation industry such as in agriculture, automated can apply to produce a position planning for the SCARA
storage and retrieval system (AS/RS) and marble mosaic tiling robot. By inverting the kinematic model, the desired end-
automation [2]–[4]. However, the operation range and the effector position can be used in inverse kinematic modelling
precision of the cartesian robot are poorer than other types of to yield the desired joint angles [16]. Therefore, kinematic
robots [5]. As a result of the continued improvement of modelling is essential to provide the working envelope and
researchers, the first cylindrical robot was introduced and match the function of the robot arm to its application [17]. The
installed in the United States in 1962, [6]. position of the robot arm is described via the widely used
Cylindrical robots are normally applied in the assembling Denavit-Hartenherg approach, which uses the homogeneous
industries such as welding and dealing with die-casting transformation matrix to formulate the robot's kinematics.
equipment [7]. The cylindrical robot has better accuracy [17]. The joint link homogeneous transformation matrix is
compared to the cartesian robot, but the range of motion is formulated as shown in (1).
limited to a cylindrical shape. Therefore, the spherical robot cos 𝜃 − sin 𝜃 cos 𝛼 sin 𝜃 sin 𝛼 𝑟 cos 𝜃
and articulated robot are implemented to improve the range of 𝐻 =
sin 𝜃 cos 𝜃 cos 𝛼 − cos 𝜃 sin 𝜃 𝑟 sin 𝜃
(1)
0 sin 𝛼 cos 𝛼 𝑑
motion [8]. Due to the improvement made, the spherical robot 0 0 0 1
and articulated robot are widely been used for applications that
required higher reachability, such as applications in the where n is the destination joint, 𝛼 , 𝜃 , 𝑑 and 𝑟 are the
military field, medical, material handling and manufacturing Denavit-Hartenberg parameters. The 𝛼 is the angle formed
[9]–[11]. by orthogonal projections of joint axes 𝑍 and 𝑍 onto a
plane perpendicular to the same normal, 𝜃 is the angle formed
Last but not least is the Selective Compliance Articulated by the orthogonal projections of the common normal, 𝑋 and
Robot Arm (SCARA) robot which will be mainly discussed in 𝑋 , onto a plane perpendicular to the joint axes 𝑍, 𝑑 is the
this review paper. SCARA robot has high precision and distance between 𝑋 and 𝑋 along with 𝑍 and 𝑟 is the
accuracy at the same time power efficiency as there is only 1 distance between the common normals to axes 𝑍 and 𝑍
joint which is joint 3 need to operate against gravity force. along with 𝑋 . The Denavit-Hartenberg parameter can be
Therefore, the SCARA robot is a good choice for an industry obtained from the joint link parameter of the robotic arm. As

978-1-6654-6837-4/22/$31.00 ©2022 IEEE

shown in (2), multiplying the homogeneous transformation 𝐾 𝑒(𝑡)
matrix for each joint will get the final position and the rotation +
𝑢(𝑡) 𝑦
𝑟 𝑒(𝑡)
angle for the robotic arm. 𝐾 𝑒(𝑡)𝑑(𝑡) +
Plant
+
− +
𝐻 = 𝐻 𝐻 𝐻 …𝐻 (2)
𝑑𝑒
𝐾
𝑑𝑡
B. Dynamic Modelling
Sensor
The purpose of dynamic modelling is to predict the force
and torque behaviour of the robot system with various pay- Fig. 1. General block diagram for PID controller.
loads conditions [18]. From the dynamic modelling, the
required torque, joint velocity, joint acceleration and joint
position of the SCARA robot can be calculated and used as a III. SCARA ROBOT CONTROL SYSTEM
reference to select a suitable actuator and transmission system The SCARA robot's controller is used to govern the robot
[14],[19]. To design control systems, it is possible to apply the to perform desired tasks. Different applications require
mathematical dynamic model of the SCARA robot arm's different ranges of allowable end-effector error compared to
dynamic behaviour. The best controller matching the required the desired position. The error of the robot is a performance
performance can then be selected for realization. The Newton- indicator in the robot arm design. Besides the accuracy of the
Euler equation and the Lagrangian equation are typically robot, the speed of the robot is also a performance concern, as
employed as the two methodologies for dynamic modelling of it defines the amount of time required to complete the desired
the SCARA robot. The general equation of the Lagrange task. As known that the SCARA robot has high precision and
equation and the Newton-Euler equation are shown in (3) and operation speed, thus, the controller for the SCARA is
(4) respectively. required to handle the high operation speed and high accuracy
computation. Moreover, the SCARA robot consists of a
𝑑 𝜕𝑇 𝜕𝑇 𝜕𝑈
𝜏= − + (3) nonlinear dynamic system due to friction and uncertain
𝑑𝑡 𝜕𝑞̈ 𝜕𝑞 𝜕𝑞
payload. Thus, the design and development of the controller
𝜏 = 𝑀(𝑞)𝑞̈ + 𝐶(𝑞, 𝑞̈ )𝑞̇ + 𝐺(𝑞) (4) for the SCARA robot is not an easy task. As a preliminary
effort to design and develop the controller, the review is
where τ is the torque for each joint, 𝑞 is the joint conducted on various types of controller systems used for
displacement, 𝑞̇ is the joint velocity, 𝑞̈ is the joint controlling the SCARA robot. The controllers are such as PID
acceleration, Tis the kinematic energy and 𝑈 is the potential control, robust control, neural networks control, fuzzy logic
energy. Both of the methods can estimate the dynamic control, adaptive control and sliding mode control. The
behaviour of the SCARA robot well. advantages and disadvantages of each of these control systems
for the SCARA robot are discussed and compared.
C. Trajectory planning
Trajectory planning is one of the procedures to design a A. PID Control
robot movement by defining the desired path, acceleration and PID control is a closed-loop control system as shown in
velocity of the robot arm [20]. The information computed Fig 1. The PID control signal can be formulated as the
from the trajectory planning is fed into the dynamic model to following (6).
obtain the desired torque in a time sequence manner which can 𝑑𝑒
then be used in performing the path movement by the 𝑢(𝑡) = 𝐾 𝑒(𝑡) + 𝐾 𝑒(𝑡)𝑑(𝑡) + 𝐾 (6)
𝑑𝑡
controller [21]. The controller ability and stability can be
studied by the achievement of the robot arm position compare where 𝑢(𝑡) is the control signal, 𝑒(𝑡) is the error signal, 𝑟 is
to the desired position at each time sequence point in the the desired output, 𝐾 is the proportional gain, 𝐾 is the
trajectory [22]. To achieve trajectory planning the joint space integral gain and 𝐾 is the derivative gain. Each of the gains
method and the cartesian space method can be used [23]. The is crucial as the parameters will affect the transient response
joint space method is applying Third-order Polynomial to of the output and the output will be feedback to the system to
compute the desired path, acceleration and velocity of the compare with the reference signal to yield the control signal.
robotic arm as the general formula of the third-order As the PID controller is simple, PID controller is widely been
polynomial is shown in (5). used in many fields including robots application [24], [25].
𝜃(𝑡) = 𝐶 + 𝐶 𝑡 + 𝐶 𝑡 + 𝐶 𝑡 (5) However, the PID control only has good stability in a system
that behaves linearly or near to linear [26]. Therefore, in most
where 𝜃 is the joint position for the robot arm, 𝐶 (𝑖 = 0 𝑡𝑜 3) cases, the performance of PID Control is not sufficient when
are the unknown constants, and t is the time. When the initial the system behaves non-linearly [27], [28]. Besides that, the
and final conditions of the robotic arm had been decided for PID controller also has insufficient sensitivity to counter the
the robotic arm to move from the initial position to the final high speed and high precision operation requirements [29]. To
position through a defined path, the parameters will be used to overcome the linear response of the PID controller many
compute the acceleration and velocity of the robotic arm at Membership Function Plots
each point and period on the desired path. Using the defined Near Medium Far
acceleration and velocity at each point and period on the 1
desired path, the controller will control the robot arm to 0.5
achieve the requirement of each point on the desired path. The
accuracy of the robot arm to perform the desired path lies in 0 1 2 3 4 5 6
the controller. Thus, the controller design is a very important Input variable
element in defining the performance of the SCARA robot. Fig. 2. The example of the membership function in graph form.
 𝑦
Reference Model
approaches had been adopted to actively change the PID gain
𝑦 −
based on the operating condition. The commentary adopted 𝑟
Controller
𝑢
Plant
𝑒

method is by using fuzzy logic and AI algorithms to improve +

the performance of the PID control. 𝑎
Adaptation Law

B. Fuzzy Logic Control Fig. 4. Adaptive Control System.

Based on the digital operation of 1 and 0, the Fuzzy Logic reference model, control law and adaptation mechanism as
Control is derived from mathematical logic to compute the shown in Fig 4. The components in the adaptive control
signal between 1 and 0. The signal between 1 and 0 can be system are designed according to the plant. Considering the
obtained from the input variable according to the membership plant for the SCARA robot is a continuous-time linear time-
function which was assigned as shown in Fig 2 and this invariant plant as a reference model, the reference model can
process is called the inference engine [25]. As shown in Fig 2, be modelled using the general state-space model as per (8) and
the fuzzy logic will compute the signal according to the slope (9).
of the membership function. The computed signal will be used
for controlling the robotic arm according to the output 𝑥̇ (𝑡) = 𝐴𝑥(𝑡) + 𝐵 𝑤(𝑡) + 𝐵 𝑢(𝑡) (8)
membership function as shown in Fig 3.
𝑦(𝑡) = 𝐶 𝑥(𝑡) + 𝐷 𝑤(𝑡) (9)
The PID controller gain is actively tuned according to the
weightage score computed from the inference engine. These where 𝑥̇ (𝑡) is the state of the plant, 𝑤(𝑡) is the disturbance
overcome the fixed gain that requires pre-tuning and lack of input, 𝑦(𝑡) is the final output and 𝐴 , 𝐵 , 𝐵 , 𝐶 , 𝐷12 is the
adaptability to the nonlinear behaviour of the SCARA robot constant matrixes. The adaptation mechanism which is the
[30]. Overcoming the limitation of the PID controller using adaption control law will continuously be adjusting the
the Fuzzy Logic improved the smoothness and stability of the adjustable parameter in the controller according to the
robot as reported by Oktarina and colleagues 2019. operating environment based on the reference model [33]. The
Furthermore, the implementation of fuzzy logic is simple and adaptation control law can be formulated as shown in (10)
thus it is widely used [25]. where 𝑥 (𝑡) is the reference output, 𝑟(𝑡) is the step-type
reference signal, 𝑥 (𝑡) is the integral compensator and 𝐾(𝑡),
C. Sliding Mode Control 𝐹(𝑡), 𝐻(𝑡) will be updated using (11).
The Sliding Mode Control is a nonlinear dynamic control 𝑦(𝑡)
system [13]. The operating concept of the Sliding Mode 𝑢(𝑡) = [𝐾(𝑡) 𝐹(𝑡) 𝐻(𝑡)] 𝑥 (𝑡) + 𝐺𝑥 (𝑡) (10)
Control is to control the system state variables toward a 𝑟(𝑡)
designated surface known as the sliding surface and ensure the [𝐾̇ (𝑡) 𝐹̇(𝑡) 𝐻̇(𝑡)] = 𝑒(𝑡)[𝑦(𝑡) 𝑥 (𝑡) 𝑟(𝑡) ]Φ (11)
state variable trajectory is on this designated surface. Equation
(7) is a sliding mode surface that is commonly been used to where 𝑒(𝑡) is the tracking error, Φ and 𝐺 can be assigned
design a Sliding Mode Controller. based on the behaviour of the SCARA robot. Therefore,
𝑑
Adaptive Control can perform well with unknown parameters
𝑠 = ( + 𝜆) 𝑋 (7) as the controller can adapt to the unknown parameters [34].
𝑑𝑡 As the uncertainty of the environment can be predicted by the
where 𝑋 is the system state error, 𝑠 is the sliding surface, 𝜆 is controller, the performance of the controller in terms of
the positive constant and 𝑚 is the system order. As the system accuracy is much better than the fixed gain controller such as
state gets closer to the sliding surface, 𝑠 equals zero, and 𝑋 the PID controller, especially with unknown parameters. The
asymptotically reaches zero. Besides the high accuracy, the ability to self-adapt to the unknown environment of the
Sliding Mode Control can also handle parametric adaptive controller can guarantee the stability and
uncertainties and also external disturbances [31]. With that, convergence of the controller [35].
Sliding Mode Control is capable of managing the SCARA
robot's nonlinear dynamic system. However, the disadvantage E. Robust Control
of the Sliding Mode Control is it can cause high-frequency Robust Control is a type of controller designed to enhance
oscillation in the control signal in a discontinuous control the performance of the controller in a certain range. The
system [32]. The high-frequency oscillation can lead to Robust controller idea is proposed by Evan Anderson and Paul
system instability. Dupuis [36]. Consider using the same plant in (7) and (8)
which is the linear time-invariant plant. The equation for the
D. Adaptive Control system can be derived as (12) and (13).
Adaptive Control is also one of the advanced controllers 𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (12)
which can handle nonlinear dynamics systems. The adaptive
controller generally contains four blocks, which are the plant, 𝑦 = 𝐶𝑠 + 𝐷𝑢 (13)
Membership Function Plots Uncertainty Disturbance
Stop Slow Fast +
1
𝑟 + 𝑒 𝑢 𝑦
Controller Plant
0.5 +
− +
Sensor
0 0.2 0.4 0.6 0.8 1 +
Output variable
Fig. 3. The example of the output membership function in the graph form. Fig. 5. Robust Control System. Noise
TABLE I. SUMMARY OF PROS AND CONS OF THE CONTROL Input Layer Hidden Layer Output Layer
METHOD FOR THE SCARA ROBOT.

Control Methods Characteristics Ref

Stable control. [24]
Easy to apply. [25]
Insufficient for the non-linear control [27]
PID Control system. [28]
Low precision. [29]
Unable to adapt to different operating Fig. 6. Artificial Neural Network.
conditions.
Stable control [25] achieve better precise control compared to PID Controller
Easy to apply. [30]
Fuzzy Logic
High accuracy. [19]. The mature algorithm used estimation of position and
Control motion error to actively tune the PID gain in dealing with
Able to adapt to different operating
conditions. the non-linear behaviour of the SCARA robot dynamic
Able to handle uncertain parameters. [31] [29].
Able to handle external disturbance. [32]
Sliding Mode
Control
High precision. [39]  Particle Swarm Optimization
Cause high-frequency oscillation in the
control signal PSO algorithm is a type of AI that can compute the
[33] optimal solution for a control system. PSO algorithm uses
Able to handle uncertain situations. particles to compute the best solution in a dimension
Adaptive Control [34]
High accuracy and precision.
[35] space. The particles will move randomly in the dimension
Able to handle uncertain situations in a [37] space to search for the best solution within the range.
certain range. [38] Every particle will be kept updated in its position and
Robust Control
High accuracy and precision in a certain
range. velocity by using (14) and (15).
𝑣 (𝑡 + 1) = 𝜔𝑣 (𝑡) + 𝑈 ∅ [𝑃 (𝑡) − 𝐿 (𝑡)] + 𝑈 ∅ [𝑃 (𝑡) − 𝐿 (𝑡)] (14)
where 𝐴 , 𝐵 , 𝐶 and 𝐷 are the constant matrix that will be
defined during the control system development. The idea of a 𝐿 (𝑡 + 1) = 𝐿 (𝑡) + 𝑣 (𝑡 + 1) (15)
Robust control system is to design a control system which is where 𝑣 is the particle velocity, 𝐿 is the particle's position,
included uncertainty criteria which are noise and disturbance 𝜔 is the inertial weight factor which will be defined
as shown in Fig 5. When the uncertainty criteria are included
according to the design, 𝑈 is the constant multiplier called
in the control system as shown in Fig 5, the control system is
swarm confidence, ∅ is random that is uniformly
operating as in the real situation which contains uncertainty
distributed, 𝑃 is the previous particles point that consists
input. Therefore, the Robust Control can improve the position
of the best fitness value and 𝛿 is the index of the best
error of the robotic arm when compared with PID Controller
particles and 𝑑 is the dimension space. The position and
[37]. Thus, Robust Control can be implemented in a situation
velocity of the particles will update every iteration and
that consists of uncertainty parameters [38].
most of the particles will converge to a small region. The
A summary of the control system reviewed is tabulated in
fittest vector will be selected as it will be the optimal
Table 1.
solution for the system. Therefore, the PSO algorithm can
apply to a traditional control system such as PID to define
IV. SCARA ROBOT WITH AI CONTROL AND IOT the optimal gain value for the SCARA robot control
Nowadays, the technology of the processor consists of system [40]–[42]. Moreover, the PSO algorithm consists
powerful computing speeds that give rise to AI technology. of some advantages which are good computational
By applying the AI technology to a SCARA robot controller, efficiency, stable convergence characteristics and ease of
the AI can optimise the control parameter and adapt to implementation [41]. There are also different AI
uncertain parameters for the SCARA robot such as particle algorithms that can compute the optimal solution for the
swarm optimization (PSO), genetic algorithm (GA) and neural control system.
network. Therefore, the implementation of AI technology can
 Genetic Algorithm
improve the performance of the SCARA robot. Besides AI
technology, the IoT is also a technology that can apply to the GA is another algorithm that can compute the
SCARA robot. optimal solution for the system. However, the difference
between PSO and GA is the implementation of GA is
A. AI Control normally applied with fuzzy logic control because the GA
is efficient and easy to apply for multivariable
 Neural Network Control
optimisation problems [50]. The flowchart which consists
Neural networks is widely been used in the industry of GA elements is shown in Fig 7.
to overcome the various type of operating environments.
Neural Network is comprised of node layers with the input Each of the elements plays an important role to
layer, output layer and hidden layers as shown in Fig 6. ensure the GA is mature to optimise the fuzzy logic
Every node has associated the weight and threshold for control. At the beginning of the GA, the string of
training the algorithm. In the process of training, the output parameters are the genes in a chromosome which will be
will be fine-tuned in every iteration until it is achieving defined according to the optimization problem. Next is the
desired error value. Therefore, the Neural Network initial generation which will form the initial generation
consumes time and data set to train the algorithm to with the random number. The initial generation will be
achieve desired performance. The well-trained Neural evaluated by using the fitness function as a performance
Network which is known as the mature algorithm can index to verify the optimal solution. After the performance
 Coding of Initial Fitness
Start TABLE II. SUMMARY OF PROS AND CONS OF THE AI
parameters Generation evaluation
CONTROL FOR THE SCARA ROBOT.

Yes Control Methods Characteristics Ref

End of No
End
iteration? Selection High accuracy and precision. [20]
Gentle control behaviour. [32]
Neural Network Able to adapt to different operating
Control conditions.
Required data and time to train the
Mutation Crossover Reproduction algorithm.
Optimise the PID controller. [47]
Particle Swarm Good computational efficiency. [48]
Optimization Stable converge characteristics. [49]
Fig. 7. Flowchart for Genetic algorithm. Easy to implement.
Optimise the fuzzy logic control. [43]
index had been computed, the chromosome which has a Genetic Algorithm Easy to implement. [50]
Efficient for multivariable optimisation.
better performance index will be selected for the next step
which is reproduction. However, there are some challenges to the implementation
The next generation will be formed from the mating of the IoT. The main challenge is security and privacy [49].
pool. The new generation will have a high similarity with As the IoT may involve access to personal data that could
the old generation which has a high-performance index. implicate the personal privacy of the user. Moreover, the
The new generation will randomly select a few pairs of challenge of the IoT is the platform, programming language
strings to exchange the chromosome and this process is and protocol are different between different companies [59].
called crossover. However, to avoid trapping in the local This will cause difficulty in connecting different types of
minimum, the mutation is required to randomly change the devices from other companies [49].
chromosome. The loop will be ongoing until it reaches the Last but not least is the cellular data which is the core of
number of iterations or the GA is mature. A matured GA IoT as the cellular data commonly used is the 4G which is not
can improve fuzzy logic control by computing the optimal optimised for IoT implementation [50]. Therefore, the wide
membership functions and fuzzy rules [43]. implementation of IoT lies in the solution to the above
challenges.
A summary of the control system reviewed is
tabulated in TABLE II. V. CONCLUSION AND FUTURE TREND
B. IoT In conclusion, the controller is crucial for a SCARA robot
as a driver to move the robotic arm. There are many different
As mentioned above, the purpose of advanced control and controller types, though, that can be used with the SCARA
AI control is to enhance the control system to assure the robot. The controller selection needs to match the application
SCARA robot's accuracy and stability. On another hand, the of the robot as the controller has its pros and cons. Selection
IoT play another role which is connecting the SCARA robot of controller according to the SCARA robot application can
with other devices such as computer and cloud server as the also prevent overdesign for the robot as the overdesign is a
IoT is a global network for the machine to interact between waste of resources. If the application involves linear
other machines and devices [44]. behaviour, the PID control and fuzzy control is the right
The IoT consists of three elements which are hardware, choice. As both control systems have fixed gain control
middleware and presentation [45]. Hardware is the sensor, systems. Thus, the PID control and fuzzy logic control is
actuator and communication hardware such as the SCARA suitable for fixed load robot which did not have much
robot, middleware is the cloud storage and computing tools uncertainty. On the other hand, advanced control like the
such as the cloud server and computer and the presentation is sliding mode control, adaptive control and robust control are
the tools that can be widely accessed on other platforms such suitable to handle high uncertainty situations. Therefore, the
as the smartphone or laptop [44], [45]. Therefore, the advanced control system is more suitable to be applied to the
technology of IoT can be implemented in many applications SCARA robot due to the SCARA robot uncertainty
such as health care, transport systems, industrial plants, parameters.
robotics and the military [46]. By applying the IoT, the robot In the future, the technology of AI can assist the advanced
can connect to a computer to improve its computation ability controller to achieve better performance for the SCARA
[47]. As the computation can run on an independent computer, robot. However, such powerful AI needed a very powerful
the SCARA robot does not consume energy for the computation computer. Therefore, IoT technology can link the
computation process which can decrease the energy SCARA robot with a powerful computation computer such as
consumption when the SCARA robot is mounted on a mobile a supercomputer. Furthermore, the IoT can create a more user-
robot [47]. The mobile robot only focuses on the friendly environment for the user as IoT can allow the user to
communication system to receive the control signal from the control the SCARA robot by just using a smartphone.
cloud [47]. Furthermore, the technology of IoT can improve
the performance and calibrate the SCARA robot easily by just
replacing an advanced control or an AI control with a ACKNOWLEDGMENT
computer that is linked with the SCARA robot [48]. This paper is sponsored by UMS grant (SDK0216-2020).
Moreover, the IoT also allows the user to remote control the
SCARA robot by just using a computer or even a smartphone
[48].
 REFERENCES [27] W. M. Elsrogy, M. A. Fkirin, and M. A. M. Hassan, “Speed control of DC
motor using PID controller based on artificial intelligence techniques,”
[1] W. S. Barbosa et al., “Industry 4.0: examples of the use of the robotic arm 2013 Int. Conf. Control. Decis. Inf. Technol. CoDIT 2013, pp. 196–201,
for digital manufacturing processes,” Int. J. Interact. Des. Manuf., vol. 14, 2013.
no. 4, pp. 1569–1575, 2020.
[28] W. Pambudi and J. Pelawi, “Simulasi Folding Machine Dengan Pid , P , Pi
[2] J. Barnett, M. Duke, C. K. Au, and S. H. Lim, “Work distribution of multiple , Pd Dan Fuzzy – Pd ( Proportional Differential ),” J. Sains dan Teknol., vol.
Cartesian robot arms for kiwifruit harvesting,” Comput. Electron. Agric., 1, no. 1, pp. 25–33, 2015.
vol. 169, no. December 2019, p. 105202, 2020.
[29] M. A. Al-Khedher and M. S. Alshamasin, “SCARA robot control using
[3] A. Oral and E. P. Inal, “Marble mosaic tiling automation with a four degrees neural networks,” ICIAS 2012 - 2012 4th Int. Conf. Intell. Adv. Syst. A
of freedom Cartesian robot,” Robot. Comput. Integr. Manuf., vol. 25, no. 3, Conf. World Eng. Sci. Technol. Congr. - Conf. Proc., vol. 1, pp. 126–130,
pp. 589–596, 2009. 2012.
[4] 2011 Bruce, “MACHINE CONTROL LEVEL SIMULATION OF AN [30] S. J. Huang and D. N. Yang, “Fuzzy logic controller for a SCARA robot
AS/RS IN THE AUTOMOTIVE INDUSTRY,” J. Chem. Inf. Model., vol. with synchronous network,” Int. J. Comput. Appl. Technol., vol. 10, no. 1–
53, no. 9, pp. 1689–1699, 2013. 2, pp. 15–26, 1997.
[5] T. Pham, “Position Control of Cartesian Robot,” 2019. [31] M. C. Lee, J. M. Lee, K. Son, D. S. Ahn, S. H. Han, and M. H. Lee,
[6] B. Singh and N. Sellappan, “Evolution of Industrial Robots and their “Implementation of Sliding Mode Control for SCARA Robot,” IFAC Proc.
Applications,” Int. J. Emerg. Technol. Adv. Eng. Website www.ijetae.com Vol., vol. 28, no. 5, pp. 85–89, 1995.
ISO Certif. J., vol. 9001, no. 5, pp. 1–6, 2013. [32] F. Adelhedi, A. Jribi, Y. Bouteraa, and N. Derbel, “Adaptive sliding mode
[7] R. Chaturvedi, A. Islam, and K. Sharma, “Anticipated Investigation of a control design of a SCARA robot manipulator system under parametric
Cylindrical Robot ARM by Means of Compound Materials,” vol. 7, no. 4, variations,” J. Eng. Sci. Technol. Rev., vol. 8, no. 5, pp. 117–123, 2015.
pp. 736–745, 2020. [33] D. P. Garg, “Adaptive control of nonlinear dynamic SCARA type of
[8] F. Collins and M. Yim, “Design of a spherical robot arm with the Spiral manipulators,” Robotica, vol. 9, no. 3, pp. 319–326, 1991.
Zipper prismatic joint,” Proc. - IEEE Int. Conf. Robot. Autom., vol. 2016- [34] J. A. Sarapura, F. Roberti, and R. Carelli, “Adaptive 3D Visual Servoing of
June, pp. 2137–2143, 2016. a Scara Robot Manipulator with Unknown Dynamic and Vision System
[9] Q. Zhang, Q. Jia, H. Sun, and Z. Gong, “Application of a genetic algorithm- Parameters,” Automation, vol. 2, no. 3, pp. 127–140, 2021.
based PI controller in a spherical robot,” 2009 IEEE Int. Conf. Control [35] L.-A. Dessaint, M. Saad, B. Hebert, and C. Gargour, “An adaptive controller
Autom. ICCA 2009, pp. 180–184, 2009. for a direct-drive SCARA robot: Analysis and simulation,” in [Proceedings]
[10] D. L. Fay, “A linear base articulated robot arm for surgical endoscopy,” IECON ’90: 16th Annual Conference of IEEE Industrial Electronics
Angew. Chemie Int. Ed. 6(11), 951–952., 2006. Society, 1990, no. 0, pp. 414–420.
[11] R. Jain, M. Nayab Zafar, and J. C. Mohanta, “Modeling and Analysis of [36] L. P. Hansen and T. J. Sargent, “Robust Control and Model Uncertainty,”
Articulated Robotic Arm for Material Handling Applications,” IOP Conf. Am. Econ. Rev. Pap. Proc., no. 1989, 2000.
Ser. Mater. Sci. Eng., vol. 691, no. 1, 2019. [37] [37] S. C. Zhen, Z. Zhao, X. Liu, F. Chen, H. Zhao, and Y. H. Chen, “A
[12] E. Akdoǧan and M. A. Adli, “The design and control of a therapeutic Novel Practical Robust Control Inheriting PID for SCARA Robot,” IEEE
exercise robot for lower limb rehabilitation: Physiotherabot,” Mechatronics, Access, pp. 227409–227419, 2020.
vol. 21, no. 3, pp. 509–522, 2011. [38] V. Mosquera and A. Vivas, “Robust Control for a Scara Robot With
[13] C. Hua, X. Guan, and G. Duan, “Variable structure adaptive fuzzy control Parametric Uncertainty,” no. January, 2006.
for a class of nonlinear time-delay systems,” Fuzzy Sets Syst., vol. 148, no. [39] V. I. Utkin, Sliding Modes in Control and Optimization. 1992.
3, pp. 453–468, 2004.
[40] M. T. Daş and L. C. Dülger, “Control of a SCARA robot: PSO-PID
[14] M. Shariatee, A. Akbarzadeh, A. Mousavi, and S. Alimardani, “Design of approach,” Control Intell. Syst., vol. 38, no. 1, pp. 24–31, 2010.
an economical SCARA robot for industrial applications,” 2014 2nd
RSI/ISM Int. Conf. Robot. Mechatronics, ICRoM 2014, pp. 534–539, 2014.
[41] [41] M. T. Daş, L. C. Dülger, and G. S. Daş, “Robotic applications with
Particle Swarm Optimization (PSO),” 2013 Int. Conf. Control. Decis. Inf.
[15] V. Verma, A. Gupta, M. K. Gupta, and P. Chauhan, “Performance Technol. CoDIT 2013, pp. 160–165, 2013.
estimation of computed torque control for surgical robot application,” J.
Mech. Eng. Sci., vol. 14, no. 3, pp. 7017–7028, 2020.
[42] A. Coronel-Escamilla, F. Torres, J. F. Gómez-Aguilar, R. F. Escobar-
Jiménez, and G. V. Guerrero-Ramírez, “On the trajectory tracking control
[16] R. R. Serrezuela, A. F. C. Chavarro, M. A. T. Cardoso, A. L. Toquica, and for an SCARA robot manipulator in a fractional model driven by induction
L. F. O. Martinez, “Kinematic modelling of a robotic arm manipulator using motors with PSO tuning,” Multibody Syst. Dyn., vol. 43, no. 3, pp. 257–
MATLAB,” ARPN J. Eng. Appl. Sci., vol. 12, no. 7, pp. 2037–2045, 2017. 277, 2018.
[17] J. Fang and W. Li, “Four degrees of freedom SCARA robot kinematics [43] G. Mester, “Design of the Fuzzy Control Systems Based on Genetic
modeling and simulation analysis SCARA Robot Kinematics,” Int. J. Algorithm for Intelligent Robots,” Interdiscip. Descr. Complex Syst., vol.
Comput. Consum. Control, vol. 2, no. 4, pp. 20–27, 2013. 12, no. 3, pp. 245–254, 2014.
[18] M. T. Das and L. C. Dülger, “Mathematical modelling, simulation and [44] I. Lee and K. Lee, “The Internet of Things (IoT): Applications, investments,
experimental verification of a scara robot,” Simul. Model. Pract. Theory, and challenges for enterprises,” Bus. Horiz., vol. 58, no. 4, pp. 431–440,
vol. 13, no. 3, pp. 257–271, 2005. 2015.
[19] Y. He et al., “Dynamic Modeling, Simulation, and Experimental [45] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of Things
Verification of a Wafer Handling SCARA Robot with Decoupling Servo (IoT): A vision, architectural elements, and future directions,” Futur. Gener.
Control,” IEEE Access, vol. 7, no. c, pp. 47143–47153, 2019. Comput. Syst., vol. 29, no. 7, pp. 1645–1660, 2013.
[20] B. Horne, M. Jamshidi, and N. Vadiee, “Neural networks in robotics: A [46] L. A. Grieco et al., “IoT-aided robotics applications: Technological
survey,” J. Intell. Robot. Syst., vol. 3, no. 1, pp. 51–66, 1990. implications, target domains and open issues,” Comput. Commun., vol. 54,
[21] G. Chen, Y. Yang, L. Zhai, K. Zou, and Y. Yang, “SCARA robot control pp. 32–47, 2014.
system design and trajectory planning: A case study,” Adv. Intell. Soft [47] C. M. J. M. Dourado et al., “A new approach for mobile robot localization
Comput., vol. 139 AISC, pp. 171–176, 2012. based on an online IoT system,” Futur. Gener. Comput. Syst., vol. 100, pp.
[22] J. Ma and S. Choi, “Kinematic skeleton extraction from 3D articulated 859–881, 2019.
models,” CAD Comput. Aided Des., vol. 46, no. 1, pp. 221–226, 2014. [48] M. Apostol, G. Kaczmarczyk, and K. Tkaczyk, “SCARA robot control
[23] G. Q. Ma, Z. L. Yu, G. H. Cao, Y. Bin Zheng, and L. Li, “The kinematic based on Raspberry Pi,” Proc. 2019 20th Int. Carpathian Control Conf.
analysis and trajectory planning study of high-speed SCARA robot handling ICCC 2019, 2019.
operation,” Appl. Mech. Mater., vol. 687–691, pp. 294–299, 2014. [49] K. Shafique, B. A. Khawaja, F. Sabir, S. Qazi, and M. Mustaqim, “Internet
[24] W. Yu and J. Rosen, “A novel linear PID controller for an upper limb of things (IoT) for next-generation smart systems: A review of current
exoskeleton,” Proc. IEEE Conf. Decis. Control, pp. 3548–3553, 2010. challenges, future trends and prospects for emerging 5G-IoT Scenarios,”
[25] Y. Oktarina, F. Septiarini, T. Dewi, P. Risma, and M. Nawawi, “Fuzzy-PID IEEE Access, vol. 8, pp. 23022–23040, 2020.
Controller Design of 4 DOF Industrial Arm Robot Manipulator,” vol. 8, no. [50] G. A. Akpakwu, B. J. Silva, G. P. Hancke, and A. M. Abu-Mahfouz, “A
2, pp. 123–136, 2019. Survey on 5G Networks for the Internet of Things: Communication
[26] John J. Craig, “Introduction to robotics.,” 2008. Technologies and Challenges,” IEEE Access, vol. 6, pp. 3619–3647, 2017.