Ajwad 2018
Ajwad 2018
Ajwad 2018
An International Journal
Syed Ali Ajwad, Jamshed Iqbal, Raza Ul Islam, Ahmed Alsheikhy, Abdullah
Almeshal & Adeel Mehmood
To cite this article: Syed Ali Ajwad, Jamshed Iqbal, Raza Ul Islam, Ahmed Alsheikhy,
Abdullah Almeshal & Adeel Mehmood (2018): Optimal and Robust Control of Multi DOF
Robotic Manipulator: Design and Hardware Realization, Cybernetics and Systems, DOI:
10.1080/01969722.2017.1412905
none defined
ABSTRACT KEYWORDS
Robots have become an integral part of industrial automation. AUTAREP manipulator;
Their ultimate role and contribution in this sector is essentially a autonomous mechatronics
function of the associated control strategy to ensure precision, process; industrial robotic
manipulator; robotic arm
repeatability, and reliability, particularly in an environment
control
polluted with disturbances and uncertainties. This research aims
to present a design of the modern control strategies for a 6 degree
of freedom robotic manipulator. Based on derived kinematic and
dynamic models of the robot, optimal and robust control
strategies are simulated and practically realized on a custom
developed pseudo-industrial framework named as AUTonomous
Articulated Robotic Educational Platform. Results of the experi-
mental trials in terms of trajectory tracking demonstrate efficiency
and usefulness of the presented control approaches.
Introduction
Thanks to recent advances in various domains of technology, the use of robots
can fulfill competitive requirements of the global market. Robots have become
very popular in many areas like medical (Iqbal, Tsagarakis, and Caldwell
2015), automation industries (Iqbal et al. 2016), space exploration (Iqbal et al.
2014), nuclear power plants (Iqbal et al. 2012), and so on. Currently, robots
are the key part of industrial production plants. They have reduced the pro-
duction time and increased the product quality significantly (Mitsi et al. 2005).
The job of robots in industry mostly involves welding, automated assembly, wafer
handling, sorting, and performing pick and place tasks. Robots can work in sce-
narios where human interaction is harmful because of environmental challenges.
Articulated robotic arms are the most common types of robots deployed in
industrial processes (Manzoor et al. 2014). These robots usually resemble
CONTACT Jamshed Iqbal [email protected] Department of Electrical Engineering, FAST National
University of Computer and Emerging Sciences, A.K. Brohi Road, H-11/4, Islamabad, Pakistan.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/ucbs.
© 2017 Taylor & Francis Group, LLC
2 S. A. AJWAD ET AL.
(CTC) has been demonstrated in Islam, Iqbal, and Khan (2014). Through
simulation, it has been shown that SMC outperforms CTC in terms of
transient response and disturbance rejection. In Girirajkumar, Kumar, and
Anantharaman (2010), adaptive proportional derivative (PD) fuzzy SMC
has been implemented to control the position of a robotic manipulator.
The system parameters are estimated through fuzzy logic algorithm. Tahir
and Jaimoukha (2013b) presented a robust control law based on model
predictive control for linear discrete-time systems. The novelty of the
proposed strategy lies in incorporating a state-feedback structure in the outer
controller while considering feedback gains as decision variables in
optimization (Tahir and Jaimoukha 2013a).
Common industrial applications of multidegree of freedom (DOF) robotic
manipulator include welding, assembling, and painting. Table 1 presents a
comparative review of prominent robots for these applications. Despite the
attention of research community round the globe, the subject of manipulation
of multi-DOF robots has only seen a limited number of real-world implemen-
tations of modern control strategies. Most of the reported works are either
restricted to PID control or simulation of modern control strategies. In con-
trast with most of other reported works, scope of the present paper is not just
limited to simulation environment. LQR and SMC strategies have also been
implemented on a 6 DOF serial link pseudo-industrial robotic arm which is
a central part of custom developed AUTonomous Articulated Robotic
Educational Platform (AUTAREP) (Iqbal et al. 2014). The proposed
framework has been developed for in-house testing and validation of
advanced control schemes and motion planning algorithms. The designed
control laws are strongly backed by the mathematical models of the
manipulator, thus enriching theoretical foundations as well.
This article primarily focuses on the precise and robust control of a robotic
manipulator which cannot be achieved through a linear control law. Although
trivial control strategies are still the main workhorse in the industry, however,
stringent demands of precision, repeatability and reliability highlight the role
of nonlinear and robust control techniques in near future. Consequently, the
present work, implemented on a pseudo-industrial arm, finds potential in
robotics based industrial applications.
Figure 1. AUTAREP robotic manipulator: a comparison with human arm. Note: AUTAREP,
AUTonomous Articulated Robotic Educational Platform.
CYBERNETICS AND SYSTEMS: AN INTERNATIONAL JOURNAL 5
Mathematical Model
The design of a control algorithm requires the knowledge of derived
mathematical model of a robot (Islam et al. 2012). Modeling plays a vital
role in investigating the relationship between motion and forces required to
accomplish a given task and subsequently designing a control strategy. A
complete mathematical model of a robot can be formulated by considering
its kinematic and dynamic motion.
Kinematic Model
Kinematic model describes the time-based states of the system (position,
velocity, and acceleration) and the resulting orientation of end-effector. It
involves derivation of forward kinematics (FK) and inverse kinematics (IK).
FK can be computed using Denavit–Hartenberg (DH) parameters, geometri-
cal representation, screw theory-based representation and Hayati–Roberts
representation. The DH parameters based model of the robot has been
computed in Iqbal, Islam, and Khan (2012). The frame assignment of each
joint of the robot is shown in Figure 2.
where
A ¼ l2c2 þ l2c23 − l4s234
B ¼ l1 − l2s2 − l2s23 − l4c234
Using (1) and information about ROM of various joints of the arm, 3D
work-envelope of the manipulator is illustrated in Figure 3. The workspace
indicates the manipulators ability to operate the objects inside spherical radius
of 580 mm.
Inverse kinematics model of the robot has been formulated using algebraic
and geometric techniques. The derived joint angles of the robotic arm are
given by (2)–(5).
�
h1 ¼ Atan2 py ; px ð2Þ
1 �
c3 ¼ ðc1 px þ s1 py þ l4 s234 Þ2 þ ðpz l1 þ l4 c234 Þ2 l22 l32
2l2 l3
qffiffiffiffiffiffiffiffiffiffiffiffi
s3 ¼ � 1 c23
h3 ¼ Atan2ðs3 ; c3 Þ ð3Þ
1 � �
c2 ¼ c1 px þ s1 py þ l4 s234 ðc3 l3 þ l2 Þ ðpz l1 þ l4 c234 Þs3 l3
ðc3 l3 þ l2 Þ2 þs23 l32
1 � �
s2 ¼ c1 px þ s1 py þ l4 s234 s3 l3 þ ðpz l1 þ l4 c234 Þðc3 l3 þ l2 Þ
ðc3 l3 þ l2 Þ2 þs23 l32
h2 ¼ Atan2ðs2 ; c2 Þ ð4Þ
h4 ¼ h234 ðh2 þ h3 Þ ð5Þ
where px, py, and pz denote the object distance from robot base center in x, y
and z axis respectively.
Dynamic Model
Dynamic model describes the relation between time-based state of a system
under consideration and forces/torques that are causing the change in the
states. Scientific community has proposed various methods for the formu-
lation of dynamic model. Dynamic model of AUTAREP arm has been
developed using Euler–Lagrange scheme which is most commonly used
energy based technique. The Lagrangian of the complete system has been
computed by difference of potential and kinetic energies. The resulted
dynamic model is given by (6).
s ¼ MðqÞ€q þ V ðq; q_ Þ þ GðqÞ ð6Þ
where τ denotes the input torque and q represents the link’s angular position.
MðqÞ; V ðq; q_ Þ, and G(q) represent matrices corresponding to inertia, force
(including Corollis and centrifugal forces) and gravity respectively. The
derived dynamic model is given in Islam, Iqbal, and Khan (2014).
Controller Design
The overall research problem consists of derivation of kinematic and dynamic
models of the arm followed by the design of the control algorithm. The
control law based on the dynamic model is presented below.
where u(t) and x(t) represent matrices corresponding to input and current
state, respectively. R(t) is a real symmetric positive definite matrix while Q
(t) is a positive semidefinite matrix. The matrices R(t) and Q(t) are chosen
to individually define the contribution of states and control inputs for optimal
operation (Khan, Islam, and Iqbal 2012). Acceleration of the robot links can
be written as,
K ¼ R 1 BT P ð10Þ
where P is symmetric positive definite matrix and it is a solution of Riccati
equation. The equation may have more than one solutions but only one
solution satisfies positive semidefinite condition and is used to compute K.
ueq forces the system to move to the sliding surface from its initial state and
keeps S ¼ 0 but disturbance or uncertainty may cause S ≠ 0. To nullify this
10 S. A. AJWAD ET AL.
where λ and I are positive constants. The controlled torque has been
computed by following the design procedure and consequently, can be written
as in (20).
s ¼ MðqÞ½kðe_ Þ þ I ðeÞ þ €qd � þ V ðq; q_ Þ þ GðqÞ kSignðSÞ fS ð20Þ
The term ζ S has been used for strong reachability purpose. The input
torque ensures the sliding mode and keeps the robot dynamics on sliding
surface S for all future values of time.
Simulation Results
The simulations for LQR and SMC have been performed in MATLAB R2009a
using derived dynamic model. The designed controller has been subjected to
CYBERNETICS AND SYSTEMS: AN INTERNATIONAL JOURNAL 11
various test inputs, such as step, ramp, and sine along with different initial
conditions. Simulation results indicate that both control laws are capable of
successfully tracking almost every input. Figures 5–8 present tracking
response of LQR and SMC for various joints of the robotic arm.
Robustness of the control strategies against variable payload has also been
tested by varying pay-load from 0 to 250 g. The change in amplitude of the
variable load has been selected between 10 and 200 g. It is clear from
Figure 9 that LQR’s performance degrades for variable payload while SMC
remains stable. In case of elbow joint (Figure 9a), LQR results in position vari-
ation up to 27.8%, which reduces as we go from shoulder joint (Figure 9b) to
base joint (result not shown) due to impact of gravity. On the other hand,
Figure 9. Robustness against variable load, (a) elbow joint (b) shoulder joint.
Hardware Realization
The proposed control algorithms have also been implemented on various joints
of actual robotic arm through embedded hardware of AUTAREP (Iqbal et al.
2014). The electronic hardware is centered on a 16-bit digital microcontroller
dsPIC33FJ256GP710A. The position and direction information of each link is
provided to microcontroller using optical encoder mounted on each motor.
With more than 100 Kernel command, the designed graphic user interface oper-
ates the controller through Universal Asynchronous Receiver/Transmitter
(UART) protocol. Teaching pendent, which is interfaced with controller through
digital input ports, can also be used to operate the system. Overall architecture,
showing the interface of a single motor segment with microcontroller is
illustrated in Figure 10a while fabricated hardware is depicted in Figure 10b.
The power rating of customized H-bridges for driving motors is 50 V/5 A.
Desired speed of the motor is achieved by controlling the voltage through
pulse width modulation of respective duty cycle. To avoid any mechanical
damage to the arm, current sensing resistor has been added to limit the
current supply in case of motor stall. Each motor draws the current of 0.65
and 1.5 A in normal and stall conditions respectively.
CYBERNETICS AND SYSTEMS: AN INTERNATIONAL JOURNAL 13
Figure 10. AUTAREP electronic hardware (a) block diagram and (b) developed control board.
Note: AUTAREP, AUTonomous Articulated Robotic Educational Platform.
The differential equations of the control laws have been transformed into
difference equation, for computation inside microcontroller, using Euler’s
technique. By definition, time derivative can be described as:
dq
€q ¼ limd!1 ð21Þ
dt
where δ represents a small change. In digital domain, (21) can be
approximated as
q ð kÞ qðk 1Þ
q_ ffi ð22Þ
T
where T denotes the sampling time which is set as 5 ms. The terms q(k) and
q(k − 1) represent present and previous position respectively. Likewise,
integration can also be approximated in microcontroller using trapezoidal,
14 S. A. AJWAD ET AL.
The experimental results of base, shoulder and elbow joints have been
presented in Figure 11.
Considering the response of base joint, although the rise time in case of
LQR is less as compared to SMC but SMC is showing dominancy in other
parameters like overshoot and settling time. Comparing the responses of both
strategies for shoulder and elbow joints, it can be seen clearly that SMC is
responding more effectively and expeditiously in all aspects.
Coupling effect in the hardware has also been investigated. It has been
observed that in the real system, the coupling effect does not diminish
totally and influences the stability of the robot. The robot joints have been
moved individually as well as simultaneously to study this effect. In both
cases, significant difference in the response has been observed as illustrated
in Figure 12. The trajectory corresponding to “individual joint” represents
the response of a particular joint while the other joints are kept static.
Whereas, trajectory with legend “multiple joints” represents the response of
all joints stimulated at same time. There are many reasons responsible for this
effect, which include mechanical play, backlash, and sharing same power
source.
Figure 11. Coupling effect in trajectories corresponding to joints, (a) base, (b) shoulder, and
(c) elbow.
CYBERNETICS AND SYSTEMS: AN INTERNATIONAL JOURNAL 15
Figure 12. Coupling effect in trajectories corresponding to joints, (a) base, (b) shoulder, and
(c) elbow.
Conclusion
This research article is focused on control of multi-DOF robotic arm. LQR
and SMC control techniques have been implemented on a custom developed
platform. Both strategies have been tested and compared for various inputs.
Through simulation results, it has been illustrated that SMC shows better
performance for all kinds of reference trajectories and robustness against vari-
able payload. The digitized algorithms have also been implemented on base,
shoulder and elbow joints of actual hardware to validate simulation results
in a real environment. It is envisaged in near future to implement a control
strategy on wrist joint and finally to incorporate trajectory planning and
collision avoidance algorithms so as to offer a fully integrated and multi-
dimensional tool for researchers, industrial engineers, interns and students.
References
Ajwad, S. A., M. I. Ullah, K. Baizid, and J. Iqbal. 2014. A comprehensive state-of-the-art
on control of industrial articulated robots. Journal of the Balkan Tribological Association
20 (4):499–521.
Ajwad, S. A., J. Iqbal, M. I. Ullah, and A. Mehmood. 2015. A systematic review of current
and emergent manipulator control approaches. Frontiers of Mechanical Engineering 10 (2):
198–210. doi:10.1007/s11465-015-0335-0(Higher Education Press).
Ajwad, S. A., U. Iqbal, and J. Iqbal. 2015. Hardware realization and PID control of
multi-degree of freedom articulated robotic arm. Mehran University Research Journal of
Engineering and Technology 34(S1): 1–12. ISSN: 0254-7821.
Arisoy, A., M. K. Bayrakceken, S. Basturk, M. Gokasan, and O. S. Bogosyan. 2011. High order
sliding mode control of a space robot manipulator. Proceedings of 5th international
16 S. A. AJWAD ET AL.
Mahyuddin, M. N., S. G. Khan, and G. Herrmann. 2014. A novel robust adaptive control
algorithm with finite-time online parameter estimation of a humanoid robot arm. Robotics
and Autonomous Systems 62 (3):294–305. doi:10.1016/j.robot.2013.09.013.
Manzoor, S., R. U. Islam, A. Khalid, A. Samad, and J. Iqbal. 2014. An open-source multi-DOF
articulated robotic educational platform for autonomous object manipulation. Robotics and
Computer-Integrated Manufacturing 30 (3):351–62. doi:10.1016/j.rcim.2013.11.003.
Mitsi, S., K.-D. Bouzakis, G. Mansour, D. Sagris, and G. Maliaris. 2005. Off-line programming
of an industrial robot for manufacturing. The International Journal of Advanced Manufac-
turing Technology 26 (3):262–67. doi:10.1007/s00170-003-1728-5 (Springer-Verlag).
Piltan, F., A. Salehi, A. Jalali, and A. Zare. 2011. Design sliding mode controller for robot
manipulator with artificial tuneable gain. Canadian Journal of Pure and Applied Science
5 (2):1573–80. doi:10.1049/ip-cta:19990518.
Saini, S. C., Y. Sharma, M. Bhandari, and U. Satija. 2012. Comparison of pole placement and
LQR applied to single link flexible manipulator. 2012 International conference on com-
munication systems and network technologies, 843–47. IEEE, Rajkot, India, 11-13 May
2012. doi:10.1109/CSNT.2012.183.
Simmons, G., and Y. Demiris. 2005. Optimal robot arm control using the minimum variance
model. Journal of Robotic Systems 22 (11):677–90. doi:10.1002/rob.20092(Wiley
Subscription Services, Inc., A Wiley Company).
Tahir, F., and I. M. Jaimoukha. 2013a. Causal state-feedback parameterizations in robust
model predictive control. Automatica 49 (9):2675–82. doi:10.1016/j.automatica.2013.
06.015.
Tahir, F., and I. M. Jaimoukha. 2013b. Robust feedback model predictive control of
constrained uncertain systems. Journal of Process Control 23 (2):189–200. doi:10.1016/j.
jprocont.2012.08.003.
Wai, R.-J., and R. Muthusamy. 2013. Fuzzy-neural-network inherited sliding-mode control for
robot manipulator including actuator dynamics. IEEE Transactions on Neural Networks and
Learning Systems 24 (2):274–87. doi:10.1109/TNNLS.2012.2228230.
Xie, S. Q., Z. C. Duan, A. Shaw, and Y. L. Tu. 2002. A fuzzy integral sliding mode control
algorithm for high-speed laser beam focus tracking control. The International Journal of
Advanced Manufacturing Technology 20 (4):296–302. doi:10.1007/s001700200155
(Springer-Verlag London Limited).