You are on page 1of 7

Robust Control Design of 6-DOF Robot for

Nuclear Power Plant Dismantling

Hamza Khan Saad Jamshed Abbasi Karam Dad Kallu


School of Mechanical Engineering School of Mechanical Engineering School of Mechanical Engineering
Pusan National University Pusan National University Pusan National University
Busan, South Korea Busan, South Korea Busan, South Korea
hamzakhan.0496@gmail.com saadjamshed93@gmail.com karamdadkallu@gmail.com

Min Cheol Lee*


School of Mechanical Engineering
Pusan National University
Busan, South Korea
mclee@pusan.ac.kr

Abstract— Controller design for robotic manipulator in a implementing control algorithms in real-time, the system are
nuclear environment has been a challenging task for more than effected by disturbance and uncertainties which results in
a decade. In this paper three different control techniques were such a way that mismatches system parameters. Therefore, to
proposed which include PID control, sliding mode control overcome this problem robust control is needed.
(SMC) and sliding mode control with sliding perturbation
observer (SMCSPO). Dynamic modeling of the system was done
using the Lagrange approach. The trajectory was planned for In recent years, control of robotic manipulator movement
manipulator’s end-effector, then trajectory for each joint was has gained researchers attentions. Different control
computed. Then control algorithms were designed and algorithms have been proposed and implemented on the
implemented in MATLAB/Simulink. The simulation results robotic manipulators, with each algorithm having their own
illustrated that PID is a linear control, therefore, when advantages and disadvantages. Conventional PID control is
controlling the system with uncertainties the performance of the linear control and wide usage for robotic manipulator has
PID is not good. Sliding mode control is a robust non-linear long history in industry which led to the development of an
control and have good performance characteristics but industrial PID control paradigm[5, 6]. But the problem is that
information of upper bound of perturbation is necessary. To
the PID is a linear control and can’t control the non-linear
overcome this problem SMCSPO a robust control with
disturbance rejection logic has been implemented and result systems or when system is being disturbed by uncertainties,
shows that SMCSPO performance is better than SMC and PID therefore, Sliding Mode Control (SMC) generally known as
in the presence of uncertainties and external disturbance due to a non-linear robust control is used. SMC has gained
disturbance compensation. researcher attention because of its robust nature [7, 8]. SMC
consist of a reaching phase and sliding phase. In the reaching
Keywords—6-DOF Robotic arm, Trajectory planning, PID, phase, controller tends to bring the system on sliding surface
SMC, SPO, SMCSPO and then sliding phase starts which helps the system to remain
I. INTRODUCTION on sliding surface. However, with conventional SMC there is
a problem that when system reaches sliding surface it is
The scope of industrial robot applications were difficult to remain on the sliding surface because of high gain
established from the conventional handling, assembly and switching control. The system with high gain move back and
welding task leading to a wide range of production forth around the sliding surface which results in chattering
applications [1]. Robots and their precise operation are [9].
getting popular. Usage of robot has been increased in a large
number of applications like handling task that might be tough Now to remove the chattering a perturbation compensator
or risky for human beings [2]. In industrial applications of is needed where the perturbation includes system
robot the main objective of replacing human with robots to uncertainties, response error and external disturbance. In [10]
provide efficiency and accuracy. For nuclear applications a sliding perturbation observer (SPO) is proposed. SPO is the
thermal and radioactive environment of nuclear plant requires combination of perturbation observer and sliding observer.
human-less automation [3]. After the Fukushima nuclear Now sliding mode control with sliding perturbation observer
power plant incident in Japan, nuclear power plants (SMCSPO) is the combination of SMC and SPO, which
dismantling gained attention all over the world [4]. utilizes only partial state feedback to estimate the system
states and perturbation with reasonable accuracy [4]. This
Humans provide guidance to the robot to achieve the estimated perturbation is useful in reduction of chattering by
desired task, which includes the robot end-effector trajectory using perturbation compensation technique.
planning, joint motion control and calculation and producing
joint torque to accurately track these trajectories. For position In the current study, PID, SMC and SMCSPO were
and velocity control of the robot end effector we must know implemented on a 6-Degree of Freedoms (DOF) industrial
the current position and velocity of the robot. When manipulator IRB-120 used in the dismantling tasks of nuclear

978-1-7281-3058-3/19/$31.00 © 2019 IEEE

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
power plants reactor vessel internals. For simulation and B. Dynamic Modeling
testing of designed control algorithm MATLAB/Simulink is Table. 2. IRb-120 Model Parameters
used and then the results were compared. The main
contribution of the current study is dynamic equations Link Link Link Centre of Mass expressed
mass/kg length/m in base coordinate frame
calculation of all six joints and controller implementation and (x,y,z)/m
then focusing on the comparison of the controller on the basis 1 3.36 .29 0, 0, -.05
of system responses. 2 6.8 .27 .124, 0, 0
3 6.22 .07 .058, .024, 0
II. SYSTEM DESIGN AND DYNAMIC MODEL 4 2 .347 0, 0, -.09
5 1.3 .07 0, .06, 0
In October 2009, ABB Company presented a smallest 6 .11 .002 0, 0, -.09
multipurpose industrial robot IRB-120 [11]. This section
presents the basic mechanical design, kinematics and
dynamic model of IRB-120.
A. System Design and Kinematics
This robotic manipulator has 6-DOF. IRB-120 was
chosen because it resembles with the actual existing system
being used in the nuclear decommissioning operation. The
CAD model of IRB-120 is presented in Fig.1. The axis
configuration of the IRB-120 is shown in Fig.2.

For kinematic the Denavit-Hartenberg (DH) Parameter of


robot are given in Table. 1. Which will be used for inverse
kinematics of the robot for trajectory planning.
Fig. 3. Schemetic of 3-link manipulator
Table. 1. DH Parameters of IRB-120

Joint 𝛉 (°) 𝐝 (𝐦𝐦) 𝐚 (𝐦𝐦) 𝛂 (°) The first step toward the control implementation is the
1 𝜃1 * 290 0 -90 dynamic modeling. Robot motion is observed by robot
2 𝜃2* 0 270 0 dynamics related to the forces and torque which creates these
motions. For dynamic equations of robot were obtained by
3 𝜃3* 0 70 90
Lagrange approach [12]. For dynamic equations the link
4 𝜃4 * 374 0 -90
length, link mass and center of masses are shown in Table. 2.
5 𝜃5* 0 0 90
6 𝜃6* 75 0 0 In [11] the three links of manipulator was resolved for
dynamic model of the robot.

The dynamic equation can be expressed as (1) [13].

𝑀(𝜃)𝜃̈ + 𝐶(𝜃, 𝜃̇)𝜃̇ + 𝑁(𝜃, 𝜃̇ ) = 𝜏 (1)

Where 𝜏 , 𝑀(𝜃) , 𝐶(𝜃, 𝜃̇) and 𝑁(𝜃, 𝜃̇ ) are actuator torque


vector, mass matrix, coriolis matrix and gravity term
respectively. And the equation of motion derived for first
three links are given in (2) to (4)

Fig. 1. IRB-120 CAD model 𝜏1 = 𝑀11 𝜃1̈ + 𝐶11 𝜃1̇ + 𝐶12 𝜃2̇ + 𝐶13 𝜃3̇ + 𝑁1 (2)

𝜏2 = 𝑀22 𝜃2̈ + 𝑀23 𝜃3̈ + 𝐶21 𝜃1̇ + 𝐶22 𝜃2̇ + 𝐶13 𝜃3̇ + 𝑁2 (3)

𝜏3 = 𝑀32 𝜃2̈ + 𝑀33 𝜃3̈ + 𝐶31 𝜃1̇ + 𝐶32 𝜃2̇ + 𝐶13 𝜃3̇ + 𝑁3 (4)

The M components are given in (5) to (9)

𝑀11 = 𝐼𝑦2 𝑠22 + 𝐼𝑦3 𝑠23


2
+ 𝐼𝑧1 + 𝐼𝑧2 𝑐22 + 𝐼𝑧3 𝑐23
2
+ 𝑚2 𝑟12 𝑐22 +
2
𝑚3 (𝑙1 𝑐2 + 𝑟2 𝑐23 ) (5)

𝑀22 = 𝐼𝑥2 + 𝐼𝑥3 + 𝑚3 𝑙12 + 𝑚2 𝑟12 + 𝑚3 𝑟22 + 2𝑚3 𝑙1 𝑟2 𝑐3 (6)

𝑀23 = 𝑀32 = 𝐼𝑥3 + 𝑚2 𝑟22 + 𝑚3 𝑙1 𝑟2 𝑐3 (7)

Fig. 2. Kinematic model of IRB-120

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
𝑀33 = 𝐼𝑥3 + 𝑚2 𝑟22 (8)

And

𝑀12 = 𝑀13 = 𝑀21 = 𝑀31 = 0 (9)

The C terms are computed from the given formula Fig. 4. Virtual simulation of robotic arm in nuclear decommissioning

1 𝜕𝑀𝑖𝑗 𝜕𝑀𝑖𝑘 𝜕𝑀𝑘𝑗 In our case we want to follow a circular trajectory and then
𝐶𝑖𝑗 (𝜃, 𝜃̇) = ∑𝑛𝑘=1 Τ𝑖𝑗𝑘 𝜃𝑘̇ = ∑𝑛𝑘=1( + + )𝜃𝑘̇ control the robotic manipulator joints to follow that circular
2 𝜕𝜃𝑘 𝜕𝜃𝑗 𝜕𝜃𝑖
(10) trajectory. The origin of circle trajectory was taken at
(450,350,300) and with radius of 150mm. The circle was
And the gravity term 𝑁𝑖 is given by divided into 20 points and for each point the joint angle was
calculated using inverse kinematics of robot. The inverse
0 kinematics was done by inverse jacobian method. The robotic
𝑁(𝜃, 𝜃̇) = [−(𝑚2 𝑔𝑟1 + 𝑚3 𝑔𝑙1 )𝐶2 − 𝑚3 𝑟2 𝐶23 ] (11) arm following trajectory is shown in the Fig. 5. And the
−𝑚3 𝑟2 𝐶23 trajectory for each joint to move end effector in a circular is
shown in Fig. 6.
The (2) to (4) are the three link manipulator equation motion
but the link 5 also have some effect of gravity while motion.
So deriving the equation for link 5.

𝜏5 = (𝑚5 𝑙52 + 𝑚5 𝑙3 𝑙5 𝐶5 )𝜃̈5 + 𝑚5 𝑙52 𝜃̈3 − 𝑚5 𝑙3 𝑙5 𝑆5 𝜃̇3 𝜃̇5 +


𝑚5 𝑔𝑙5 𝑆35 (12)

The (12) presents the equation of motion for link 5. Link 4


and link 6 are roll joints and they have no effect of gravitation.
Therefore, the equation for link 4 and 6 is given by

𝜏4 = 𝑚5 𝜃̈ + 𝑚4 𝑙4 𝑟3 𝜃̇ (13)

𝜏6 = 𝑚6 𝜃̈ + 𝑚6 𝑙6 𝑟4 𝜃̇ (14)

Where

𝐶𝑖 = 𝐶𝑜𝑠𝜃𝑖 𝑎𝑛𝑑 𝐶𝑖𝑗 = 𝐶𝑜𝑠(𝜃𝑖 + 𝜃𝑗 ) (15)


Fig. 5. Circular trajectory of end-effector of robotic arm
Similarly,

𝑆𝑖 = 𝑆𝑖𝑛𝜃𝑖 𝑎𝑛𝑑 𝑆𝑖𝑗 = 𝑆𝑖𝑛(𝜃𝑖 + 𝜃𝑗 ) (16)

Where I is the moment of inertia, l is the link length, 𝑟𝑖


represents the distance to the joints from center of masses of
each link and m is the mass of each link. The dynamic
equations used in Simulink for the system control design.
III. TRAJECTORY PLANNING
For a required task it is difficult to directly control the
manipulator, therefore, to achieve optimization the simpler
approach is to plan a trajectory path for robot and the
manipulator should be controlled to track that trajectory [14].
There are many different operations being conducted by the
robotic manipulator in which one is the surface plate cutting.
In Fig .4. a virtual simulation of the robotic arm is shown
which is controlled using joystick and try to cut the plate in a
circular shape during nuclear decommissioning process. The
left side of the Fig. 4. shows that a circular trajectory being
created for the robotic manipulator for surface cutting and the
right side shows the plate after cutting. Fig. 6. Joint trajectory of each link

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
Fig. 6. Shows the trajectory against each link when the Where 𝐾 and 𝜓 are switching gain and perturbation
robot arm end effector is trying to follow the circular respectively. The gain K should be greater than perturbation
trajectory. Now this trajectory will be set to the desired path to verify (20) 𝑠𝑎𝑡(𝑠) is the saturation function which is
for each joint when controlling the arm. defined as
IV. CONTROL DESIGN s⁄ if |s| > ε
|s| c
In this section, 3 different types of controller are discussed sat(s) = { s (21)
and compared including PID, SMC and SMCSPO. SMCSPO ⁄εc if |s| ≤ εc
is most robust controller which estimates the perturbation and
then compensate it. Where s and εc is the sliding surface and the boundary
layer respectively. When the system reach sliding surface the
A. PID Controller sliding phase begins here to remain the system on sliding
PID controller may be called as the most popular control surface. Now during sliding phase, the transfer function
technique that is still being used by most of the industry as between the sliding surface and perturbation is shown in (22).
well as researchers in their work. PID control has proven its
effectiveness in almost every domain. It is used in various 𝑠 1
= (22)
𝜓 𝑝+𝐾⁄𝜀𝑐
configuration and forms to suit the user and the system needs.
In some cases, it is used as a standalone controller while in
others it may be part of a hierarchical system or a distributed If the perturbation in system is greater a larger value of K
control system. Its applicability in every condition has made is needed which cause the chattering in the response and this
it suitable for application in almost all of fields of study. The gain K also increase the breaking frequency of the transfer
PID controller is generally known as three terms controller function which will result in more effect of perturbation on
based upon the name i.e. P (Proportional), I (integral) and D the system. 𝑝 is s-domain variable.
(Derivative). The general form of control can be written as C. Sliding ModeControl with Sliding Perturbation
(17) [15]. Observer
𝑑𝑒 SMCSPO is a robust control with disturbance rejection
𝑢 = 𝑘𝑝 𝑒 + 𝑘𝑖 ∫ 𝑒 (𝜏) 𝑑𝜏 + 𝑘𝑑 (17) technique. Basically SMCSPO in an integration of SMC and
𝑑𝜏
Sliding perturbation observer (SPO). The block diagram of
Where u and e are the control signal and the error, 𝑘𝑝 , 𝑘𝑖 and SMCSPO is presented in Fig. 7.
𝑘𝑑 are the gains for proportional, integral and derivative term
of PID controller, respectively. i. Sliding Perturbation Observer
B. Sliding Mode Control SPO is non-linear observer that uses only partial feedback
Sliding mode control has been widely used for non-linear (position) to estimate velocity and perturbation [12]. The
system control. The important difficulties in SMC are the state estimation equations and estimation error as shown
stable reaching condition and phenomena of chattering. For below
known mathematical model of the system, SMC can be used
for reference signal tracking. However, for the systems 𝑥̂1̇ = 𝑥̂2 − 𝑘1 . 𝑠𝑎𝑡(𝑥̃1 ) − 𝛼1 . 𝑥̃1 (23)
having uncertainties with disturbance an additional
disturbance compensator is required then SMC is used to 𝑥̂2̇ = −𝑘2 . 𝑠𝑎𝑡(𝑥̃1 ) + 𝛼3 . 𝑢 − 𝛼2 . 𝑥̃1 (24)
generate control input. SMC belongs to a class of control
systems identified as Variable Structure Control (VSC) 𝑥̂3̇ = 𝛼32 . (𝛼3 . 𝑥̂2 + 𝑢 − 𝑥̂3 ) (25)
which arises from the theory of variable structure systems
such as relay systems [16]. 𝜓̂ = 𝛼3 (𝛼3 ∙ 𝑥̂2 − 𝑥̂3 ) (26)
The first main objective is to design the sliding surface which Where 𝑘1 , 𝑘2 , 𝛼3 , 𝛼1 , 𝛼2 are constants with positive values.
is shown in (18).

𝑠 = 𝑒̇ + 𝑐 ∙ 𝑒 (18)

Where 𝑠 , 𝑐 and 𝑒 are sliding surface, constant and error


respectively. Now to verify the sliding condition during
reaching phase the following condition must be satisfied must
be satisfied.

𝑠 ∙ 𝑠̇ ≤ 0 (19)
Fig. 7. SMCSPO block diagram
Where

𝑠̇ = −𝐾 ∙ 𝑠𝑎𝑡(𝑠) + 𝜓 (20)

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
ii. SMC Integration with SPO (SMCSPO) iii. Design Procedure
In this section the SMC is integrated with SPO. The estimated A general systematic design procedure for SMCSPO is
sliding surface is given as presented in this section. After achieving the condition |𝑠̂ | ≤
𝜀𝑐 and |𝑥̃1 | ≤ 𝜀𝑜 . It is assumed that 𝑐 = 𝐾⁄𝜀𝑐 then the
𝑠̂ = 𝑒̂̇ + 𝑐 ∙ 𝑒̂ (27) controller observer and dynamics takes the following form

Where the estimated error is 𝑒̂ = 𝑥̂ − 𝑥𝑑 and the actual 𝑥̃̇1


sliding surface is given by 𝑥̃̇2 =
𝑥̃̇3
𝑠 = 𝑒̇ + 𝑐 ∙ 𝑒 (28) [ 𝑠̇ ]
−𝑘1⁄
𝜀𝑜 1 0 0
Where in (28) 𝑒 = 𝑥 − 𝑥𝑑 is the actual error. Where the −𝑘2⁄ −𝛼3 0
estimated error of sliding surface is given by 𝜀𝑜 𝛼32
0 𝛼33 −𝛼32 0
𝑠̃ = 𝑠̂ − 𝑠 (29) 𝑘2⁄ 𝐾 𝑘1 2 𝐾 −𝐾
𝜀 + ( − ) − (2 + 𝛼32 ) 𝛼3
[ 𝑜 𝜀𝑐 𝜀𝑜 𝜀𝑜 𝜀𝑐 ]
This estimated sliding surface error can be written as
𝑥̃1 0
𝑥̃2 𝜓̇
𝑠̃ = 𝑥̃̇1 + 𝑐 ∙ 𝑥̃1 (30) × [ ] + [0] (38)
𝑥̃3 1 𝛼3
𝑠 0
Now again to verify sliding condition during reaching phase,
𝑠̂ ∙ 𝑠̂̇ ≤ 0 must be satisfied. And to satisfy this condition then The parameters are set in such way that pole placement
𝑠̂̇ should be method is used and compared with desired poles.

𝑠̂̇ = −𝐾 ∙ 𝑠𝑎𝑡(𝑠̂ ) (31)


Where [𝜆 + 𝑐] [𝜆3 + (𝑘1⁄𝜀𝑜 ) 𝜆2 + (𝑘2⁄𝜀𝑜 ) 𝜆 + 𝛼32 (𝑘2⁄𝜀𝑜 )]
=0 (39)
𝑠̂⁄ if |𝑠̂ | > 𝜀𝑜
|𝑠̂ | After coefficient comparing of characteristics equations the
𝑠𝑎𝑡(𝑠̂ ) = { (32)
𝑠̂⁄ if |𝑠̂ | ≤ 𝜀𝑜 following relation is obtained
𝜀𝑜
𝑘1 𝑘2 𝜆𝑑 𝐾
= 3𝜆𝑑 , = 𝜆𝑑 , 𝛼3= √ , = 𝜆𝑑 (40)
By using previous results, computing 𝑠̂̇ 𝜀𝑜 𝜀𝑜 3 𝜀𝑐

𝑘2 𝑐∙𝑘1 𝑘12 𝑘
𝑠̂̇ = 𝛼3 ∙ 𝑢 − 𝑥̃1 { + − } − { 1 } 𝑥̃2 + 𝑐{𝑥̂2 − 𝑥̇ 𝑑 } − V. SIMULATION AND RESULTS
𝜀𝑜 𝜀𝑜 𝜀𝑜 2 𝜀𝑜
𝑥̈ 𝑑 + 𝜓̂ (33) MATLAB/Simulink is used to simulate and verify the
controller results. All three controller were applied on the
Now solving for control law and it can be obtained as system. In the simulation an external disturbance was
introduced to check controller’s performance. Link 4, link5
1 𝑘2 𝑐∙𝑘1 𝑘2
and link 6 were controlled using the proposed scheme. The
𝑢= [−𝐾𝑠𝑎𝑡(𝑠̂ ) + 𝑥̃1 { + − 12} − 𝑐{𝑥̂2 − 𝑥̇ 𝑑 } − results of each controller are as follow.
𝛼3 𝜀𝑜 𝜀𝑜 𝜀𝑜
𝜓̂ + 𝑥̈ 𝑑 ] (34) A. PID Controller
The gain parameters of PID were set in the PID controller
The resulting 𝑠̂ when substituting (34) in (33) is
equation to see the response of each link. Following Table. 3.
𝑘1 shows the system parameters. The parameters were chosen
𝑠̂̇ = −𝐾 ∙ 𝑠̂ − ∙ 𝑥̃2 (35) randomly after observing the system response, PID is
𝜀𝑜
manually tuned using Ziegler Nichols closed loop method.
From above equation it can be seen that 𝑥̃2 effect the resulting And the system trajectory error according to set parameters
𝑠̂̇ -dynamics. Therefore, to satisfy sliding condition the robust are shown in the Fig. 8.
control gains should follow
Table. 3. PID Controller Parameters
𝑘12
𝐾> (36) Gains
𝜀𝑜 Joints
𝒌𝒑 𝒌𝒊 𝒌𝒅
After reaching phase when the system reach sliding surface, 4 400 300 50
then the actual s dynamics within boundary layer becomes 5 1500 1200 100

𝐾 𝑘2 𝑘1 𝐾 𝑘1 𝐾 6 250 200 35
𝑠̇ + 𝑠 = 𝑥̃1 { −( − ) (𝑐 − )} − 𝑥̃2 { + 𝑐} − 𝜓̃
𝜀𝑐 𝜀𝑜 𝜀𝑜 𝜀𝑐 𝜀𝑜 𝜀𝑐
(37)

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
Table. 5. SMCSPO Parameters

Gains
Joints
𝝀 𝜺𝒐 𝒌𝟏 𝒌𝟐 𝜶𝟑
4 40 1.5 180 7200 3.65

5 48 1.3 187 8985 4

6 30 1 90 2700 3.163

Fig. 8. Trajectory tracking error for PID

The simulation result seems to be controlled but the gain


and control input was very high and the results are not
realistic. The results shows the system is highly non-linear.
The simulation failed when using lower values of PID gains.
B. SMC
SMC was implemented on the system to check the system
response. Table. 4. shows the system parameters for SMC.
According to the system parameters in the system, trajectory Fig. 10. Trajectory tracking error for SMCSPO
error is shown in the Fig. 9.

The system result showed that the link-5 is still having


oscillations. When an external disturbance was introduced
that system failed to simulate because the gain K was then
lower than the perturbation.
C. SMCSPO
SMCSPO is the robust controller. The system parameters
for the links are shown in the Table. 5. System parameters are
selected on the basis of value of 𝜆𝑑 which depends upon the
physical limitations of the system. In this research value of
𝜆𝑑 has been selected after manual tuning.

Table. 4. SMC Parameters


Fig. 11. Actual and Estimated Perturbation

Gain
Joints The error in trajectory of link 4, link 5 and link 6 is shown in
K c
the Fig. 10.
4 120 2 The results of SMCSPO show no oscillation in link 5. The
5 150 .8 system is well controlled. All the non-linearities were
compensated with SPO. When external disturbance was
6 100 .8 introduced there were no need to increase the gain because
SPO estimated the perturbation and then compensated from
the system response and refined results were obtained.
With this the estimated perturbation of the system when some
external disturbance were introduced is shown in the Fig. 11.
VI. CONCLUSION
In this research a robust control scheme SMCSPO has
been proposed for tracking trajectory. Control performance
of conventional PID control, SMC and SMCSPO were
compared through simulations and it has been observed that
the output tracking error of SMCSPO is less than other
controllers. When external disturbance was introduced to the
system, both PID and SMC control parameters were needed
to be tuned. In SMC when external disturbance was
Fig. 9. Trajectory tracking error for SMC introduced, initially the value of gain K was less than
perturbation which was then increased to settle the response.

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.
When increasing value of K the breaking frequency is boundary layer method," Journal of Central South
increased which make the system sensitive to more University, vol. 25, pp. 2219-2244, 2018.
perturbation. SMCSPO with external disturbance estimation, [12] B. Siciliano, L. Sciavicco, L. Villani, and G. Oriolo,
estimated the disturbance and compensated it from the system Robotics: modelling, planning and control: Springer
response in order to provide better results. Science & Business Media, 2010.
[13] R. M. Murray, A mathematical introduction to
ACKNOWLEDGMENT robotic manipulation: CRC press, 2017.
This research was supported by the nuclear research and [14] A. Piazzi and A. Visioli, "Global minimum-jerk
development program through the National Research trajectory planning of robot manipulators," IEEE
Foundation of Korea(NRF) funded by the Ministry of Science transactions on industrial electronics, vol. 47, pp.
and ICT(MSIT, Korea). [NRF-2019M2C9A1057807] 140-149, 2000.
[15] K. J. Aström and R. M. Murray, Feedback systems:
This research was funded by the Technology Innovation an introduction for scientists and engineers:
Program (10073147, Development of Robot Manipulation Princeton university press, 2010.
Technology by Using Artificial Intelligence) funded By the [16] F. Moldoveanu, "Sliding mode controller design for
Ministry of Trade, Industry & Energy(MOTIE, Korea). robot manipulators," Bulletin of the Transilvania
University of Brasov. Engineering Sciences. Series
REFERENCES
I, vol. 7, p. 97, 2014.

[1] E. Abele, M. Weigold, and S. Rothenbücher,


"Modeling and identification of an industrial robot
for machining applications," CIRP annals, vol. 56,
pp. 387-390, 2007.
[2] W. G. Hao, Y. Y. Leck, and L. C. Hun, "6-DOF PC-
Based Robotic Arm (PC-ROBOARM) with
efficient trajectory planning and speed control," in
2011 4th International Conference on Mechatronics
(ICOM), 2011, pp. 1-7.
[3] T. Moore, "Robots for nuclear power plants," IAEA
Bulletin, vol. 27, pp. 31-38, 1985.
[4] S. J. Abbasi, K. D. Kallu, and M. C. Lee, "Efficient
Control of a Non-Linear System Using a Modified
Sliding Mode Control," Applied Sciences, vol. 9, p.
1284, 2019.
[5] M. A. Johnson and M. H. Moradi, PID control:
Springer, 2005.
[6] E. M. Jafarov, M. A. Parlakci, and Y. Istefanopulos,
"A new variable structure PID-controller design for
robot manipulators," IEEE Transactions on Control
Systems Technology, vol. 13, pp. 122-130, 2004.
[7] K. D. Young and U. Ozguner, "Sliding mode:
Control engineering in practice," in Proceedings of
the 1999 American Control Conference (Cat. No.
99CH36251), 1999, pp. 150-162.
[8] X.-G. Yan, S. K. Spurgeon, and C. Edwards,
Variable structure control of complex systems:
Springer, 2017.
[9] X. Liu, W. Jiang, and X.-C. Dong, "Nonlinear
adaptive control for dynamic and dead-zone
uncertainties in robotic systems," International
Journal of Control, Automation and Systems, vol.
15, pp. 875-882, 2017.
[10] J. T. Moura, H. Elmali, and N. Olgac, "Sliding mode
control with sliding perturbation observer," Journal
of Dynamic Systems, Measurement, and Control,
vol. 119, pp. 657-665, 1997.
[11] M. H. Barhaghtalab, V. Meigoli, M. R. G. Haghighi,
S. A. Nayeri, and A. Ebrahimi, "Dynamic analysis,
simulation, and control of a 6-DOF IRB-120 robot
manipulator using sliding mode control and

Authorized licensed use limited to: City, University of London. Downloaded on July 12,2020 at 15:10:22 UTC from IEEE Xplore. Restrictions apply.

You might also like