International Journal of Control, Automation and Systems 17(1) (2019) 203-211 ISSN:1598-6446 eISSN:2005-4092

http://dx.doi.org/10.1007/s12555-017-0544-x http://www.springer.com/12555

Disturbance Observer-based Trajectory Following Control of Robot Ma-

nipulators
Mohamadreza Homayounzade* and Amir Khademhosseini

Abstract: Robotic manipulators are usually subject to different types of disturbances. If the effects of such dis-
turbances are not taken into account, it can lead to unsatisfactory tracking performance of the robot and may even
destabilize the robot control system. In this paper a novel nonlinear disturbance observer-based controller is de-
signed for robotic manipulators. Previous disturbance observer-based controllers which are designed for robotic
systems undergo the restricting assumption that the external disturbance is constant. In this paper a novel two-stage
procedure is proposed to design a disturbance observer that improves disturbance attenuation ability of current dis-
turbance observer-based controllers. The proposed method can cope with non-constant disturbance. Furthermore,
it is shown that even for the fast time varying disturbance, the controller achieves satisfactory tracking performance.
The proposed controller guarantees semi-global asymptotic position and velocity tracking and moreover, removes
restrictions of previous studies on the number of degrees of freedom (DOFs), joint types, and manipulator configu-
ration. The effectiveness of the proposed method is verified against different types of external disturbance applied
on the robot manipulator and the results are compared with the results of previous methods. Furthermore, the results
support the theoretical conclusions.

Keywords: Disturbance observer, robotic manipulators, stability analysis, time varying disturbance, trajectory fol-
lowing.

1. INTRODUCTION bance observer, and afterward eliminating the disturbance

effect utilizing feed-forward compensation in the control
In recent years with the growing interest in high- system. Due to feed-forward nature of this compensation,
precision control of robotic systems, a significant number disturbance observers generally provide satisfactory con-
of researches have focused on improving the tracking per- trol performance without utlilizing large feedback gains
formance of robotic systems without the higher cost asso- [1]. Disturbance observers have been used in a variety
ciated with such improvements. of fields such as independent joint control [2–4], friction
The principal challenge in achieving this goal is to estimation and compensation [5, 6], uncertainties and dis-
overcome the difficulties imposed by the disturbances. turbance attenuation [7–10], and contact force estimation
Robotic manipulators are mostly subject to different [11–15]. A survey of nonlinear disturbance observers ap-
sources that exert disturbances to the system such as joint plied to Euler-Lagrange systems is presented in [16].
frictions, system uncertainties, end-effector unknown pay- In [17] a disturbance observer is developed to estimate
load. External disturbances when unaccounted for can ad- the disturbance generated by an exogenous system based
versely affect the robot performance and may even desta- on solving linear matrix inequalities and afterward a ro-
bilize the control system. bust controller is designed for the class of time delay un-
Hence, it seems crucial to use some kind of disturbance certain systems.
suppression or attenuation to achieve the desired perfor- In [18] a new method is proposed to improve the stabil-
mance. Disturbance observer-based control method is the ity robustness of disturbance observers. In [19], a fuzzy
most efficient method utilized to quell the disturbance ef- disturbance observer is developed and its application to
fect. The basic concept behind utilizing disturbance ob- the control of a nonlinear system under the internal and
server is to aggregate all the internal and external distur- external disturbances is presented.
bances exerting on the manipulator into a single distur- In spite of the fact that the disturbance observer method
bance term, estimating the lumped term utilizing a distur- has been extensively used in the robot manipulator cont-

Manuscript received September 5, 2017; revised June 2, 2018; accepted August 8, 2018. Recommended by Associate Editor Nam H. Jo
under the direction of Editor Hyun-Seok Yang.

Mohamadreza Homayounzade is with the Department of Mechanical Engineering, Fasa University, Fasa, Iran (e-mail: homayounzade.
[email protected]). Amir Khademhosseini is with the Department of Computer Engineering, Fasa University, Fasa, Iran (e-mail: khadem_h
@fasau.ac.ir).
* Corresponding author.

⃝ICROS,
c KIEE and Springer 2019
204 Mohamadreza Homayounzade and Amir Khademhosseini

trol, in the majority of these cases, the analysis or design is ulator subject to different types of disturbance and the re-
based on linearized models and linear control techniques sults are compared with previous results.
[20, 21]. Since the robot manipulator dynamic is highly The rest of the paper is organized as follows: In Sec-
nonlinear and coupled, the validity of utilizing linear tech- tion 2 the dynamic modelling of the robot dynamic is pre-
niques for robotic systems may be in doubt. sented. In Section 3 the controller along with disturbance
In order to resolve the limitations and shortcomings estimator is dersigned. The error dynamics of the system
of utilizing linear disturbance observers for nonlinear dy- is presented in Section 4. The stability of the system is an-
namic of robot manipulators, Chen et al. [22] proposed alyzed by the Lyapunov theroem in Section 5. Simulation
a nonlinear disturbance observer for a specific class of results are presented in Section 6 and Section 7 concludes
robotic manipulators without the need for the acceleration the paper.
measurements. Their method depends on finding a proper
observer gain matrix that stabilizes the control system. 2. DYNAMIC MODELLING OF ROBOTIC
However, Chen et al. could find such a proper gain matrix SYSTEMS
for a 2-link planar manipulator with revolute joints. After-
wards, Nikoobin et al. [23], generalized Chen’s method to Dynamic equation of an n-degrees-of-freedom serial
be applicable to n-link planar manipulators with revolute rigid manipulator is given by
joints by exploiting the explicit relation for this particular M(q)q̈ +C(q, q̇)q̇ + G(q) + τd = τ , (1)
class of robots.
The method proposed in [22] has been applied to other where q, q̇, q̈ represent respectively the n × 1 vectors of
classes of robotic systems such as observer-based control joint positions, velocities and accelerations; M is the n × n
of 6-DOF parallel manipulators [24] and dynamic surface inertia matrix; C(q, q̇) is the n × n centrifugal and Coriolis
control of mobile wheeled inverted pendulums [25]. Al- matrix; τ is the n×1 vector of applied joint torques; and τd
though the method designed in [23] show promising re- is the n × 1 disturbance vector which lumps the effects of
sults in disturbance estimation, their method is restricted torque due to disturbance, end-effector payloads, external
to the planar, serial manipulators with revolute joints. force, friction force, and unmodeled dynamics.
However, industrial robots are generally non-planar and Property 1: The inertia matrix is positive definite
may have prismatic joints in addition to revolute joints. (P.D.) and there exists positive real numbers Mm and MM
Furthermore, in [22, 23] the external disturbance is sup- such that
posed to be constant.
Mm ∥x∥2 ≤ xT M(q)x ≤ MM ∥x∥2 ∀x ∈ ℜn , (2)
The above limitations serve as the stimulation to look
for a more general design method. In [9], the authors de- where throughout this paper ∥·∥ represents the 2 induced
signed a disturbance observer-based controller for robotic matrix norm for matrices and Euclidean norm for vectors.
manipulators which removes previous restrictions on the Property 2: The centripetal-Coriolis matrix can be de-
number of DOFs, joint types, or manipulator configura- signed such that it possesses the following skew symmet-
tion. However, their method similar to previous researches ric property
suffers from the restricting assumption that the external [ ]
1
disturbance is required to be constant. In [26] an integral xT Ṁ −C(q, q̇) x = 0 ∀ x ∈ ℜn . (3)
2
sliding disturbance observer is designed to compensate for
unknown disturbances for a class of nonlinear systems ex- Definition 1: For any arbitrary P.D. or N.D. (negative
posed to disturbance with bounded first time derevative. definite) symmetric matrix such as P, throughout this pa-
The above limitations serve as the motivation for this per by Pm and PM , we mean the minimum and maximum
paper to design a general method that guarantees stability eigenvalues of the matrix P. Hence, for any arbitrary vec-
of the controlled system exposed to non-constant distur- tor such as ξ we can state
bance. Pm ∥ξ ∥2 ≤ ξ T Pξ ≤ PM ∥ξ ∥2 . (4)
In this paper we design a nonlinear disturbance
observer-based controller for robotic systems. The pro- Let us define position tracking error, i.e., z1 , by
posed controller guarantees asymptotic trajectory tracking
z1 = q − qd , (5)
and removes the restrictions of the previous methods that
require the external disturbance to be constant or its first where the vector qd represents the desired angular position
time derevative to be bounded. The proposed method can of joints. Moreover, let the filtered velocity tracking error
be utilized for non-constant external disturbance and it of the links (i.e., z2 ) be defined by
requires the measurement of joint positions and velocities
z2 = q̇ − q̇d + Λ(q − qd ), (6)
but not accelerations.
In order to illustrate the effectiveness of the proposed where Λ is an n × n P.D. matrix and the vector q̇d repre-
method, simulations are done using a 2-link robot manip- sents the desired joint angular velocity.
 Disturbance Observer-based Trajectory Following Control of Robot Manipulators 205

3. CONTROL OBJECTIVE

The control objective is to achieve semi-global asymp-

totic position/velocity tracking in the presence of external
disturbance. A disturbance observer is designed to esti-
mate the external disturbance. In fact, the disturbance ob-
server composed of two individual estimators: the first es-
timates the external disturbance (i.e., τd ); and the second
estimates the time derivative of disturbance with respect
to time (i.e., τ̇d ).
The schematic of the disturbance observer-based con-
troller is shown in Fig. 1. The control output is the torque
exerted on the robot joints. As shown in Fig. 1, the con- Fig. 1. Schematic of disturbance observer-based control
troller torque is constructed by three individual variables system.
fed-back to the controller as an input, i.e., z1 which rep-
resents the joints positions, z2 represents the joints veloci-
ties, and τ̂d represents an estimation of the external distur-
bance.
Remark 1: The variables z1 , z2 which represent respec-
tively the joint positions and velocities are the conven-
tional variables required for feedback controller design.
However, since the external disturbance is unknown, τ̂d is
utilized in the controller as an estimation of the external
disturbance.

3.1. Nonlinear disturbance observer design

As shown in Fig. 1, the disturbance observer estimates Fig. 2. Schematic diagram of disturbance observer.
the external disturbance by the measurement of joint posi-
tions and velocities. The output of disturbance estimator,
i.e., τ̂d , is then utilized as feedback to the controlled sys- 3.2. Controller design
tem to compensate the effect of external disturbance. Let’s initially define the function fd by
Let’s initially describe the functions f1 (z2 ) and f2 (z2 )
by fd = −C(q, q̇)q̇d +C(q, q̇)Λz1 − G(q)
γ − M(q)q̈d + MΛ(z2 − Λz1 ). (13)
f1 (z2 ) = − z2 − α LM −1Cz2 − α LM −1 K1 z2 , (7)
α
f2 (z2 ) = −α M −1Cz2 − α M −1 K1 z2 . (8) The following controller is designed:
Consider the variable S to be calculated by the following
τ = −K1 z2 + τ̂d − fd , (14)
differential equation
Ṡ = f2 (z2 ). (9) where τ̂d represents estimation of disturbance calculated
by (11) and K1 is an n × n P.D. gain matrix.
Moreover, lets the vector β̂ be calculated by Remark 2: The controller constitutes three individual
β̂ = S − α z2 . (10) terms: the first term (i.e., −K1 z2 ) represents as the feed-
back of system states and designed to stabilize the closed-
Consequently the estimate of external disturbance (i.e., τ̂d ) loop system; the second term (i.e., τ̂d ) is designed to sup-
is calculated by press the unknown disturbance effect; the third term (i.e.,
fd ) is designed to eliminate the known nonlinear terms in
τ̂d = R − α Lz2 , (11) robot dynamic equation.
where the vector R in (11) is calculated from the following
differential equation: 4. ERROR DYNAMICS
Ṙ = f1 (z2 ) + β̂ . (12)
Error dynamics of the closed-loop system are developed
The block diagram schematic of (9) to (12) is shown in in this section and they will be utilized in the next section
Fig. 2. to prove the stability of the controlled system.
206 Mohamadreza Homayounzade and Amir Khademhosseini

4.1. Error dynamics of the trajectory tracking 5. STABILITY ANALYSIS

The robot dynamic equation can be represented in terms
In this section initially the main theorem regarding this
of z1 and z2 by
paper will be presented and henceforth; the stability of the
system will be analyzed utilizing the Lyapunov theorem.
ż1 = z2 − Λz1 , (15)
Let us define the augmented state vector E and the block
M(q)ż2 = −C(q, q̇)z2 + fd − τd + τ . (16) diagonal matrix B by
( )
where fd is defined previously by (13). E = [z1 T , z2 T , τ̃d T , η T ]T , B = diag ε I, γ I, α M −1 , I ,
Substituting the control input calculated by (14) in (16), (26)
the system tacking error dynamic is calculated by
where ε , α , γ are all positive
{ constants. Furthermore,
} let
M(q)ż2 = −C(q, q̇)z2 − K1 z2 − τ̃d , (17) us define the set Ω1 = E(t) | ∥E(t)∥2 ≤ ℓ , where ℓ rep-
resents the upper bound of this region.
where τ̃d represents the disturbance estimation error and Theorem 1: The control input calculated by (14) en-
defined by sures semi-globally asymptotic stability of the system in
trajectory tracking and disturbance estimation provided
τ̃d = τd − τ̂d . (18) that the system states{ are initialized within}any bounded
region such as Ω1 = E(t) | ∥E(t)∥2 ≤ BBMm ℓ and the gain
4.2. Error dynamics of disturbance estimation matrices Λ, K1 , L satisfy the following conditions.
Differentiating (11) with respect to time, one obtains
1
γ Λm > , (27a)
τ̂˙d = − z2 − α LM −1Cz2 − α LM −1 K1 z2 2
α CM MM 1 MM
K1m > + ε , (27b)
+ β̂ − α L M −1 [−Cz2 − K1 z2 − τ̃d ] . (19) Mm 2 γ
1
Lm > δ MM
2
. (27c)
Eliminating equal terms and simplifying (19), one obtains 2α
γ Proof: Consider the following Lyapunov function
τ̂˙d = − z2 + β̂ + α LM −1 τ̃d . (20)
α
1
V = E T BE. (28)
Consequently, one obtains 2
Differentiating the above equation with respect to time,
τ̃˙d =τ̇d − τ̂˙d yields
γ
=η + z2 − α L M −1 τ̃d . (21)
α V̇ =ε z1 T ż1 + γ z2 T ż2 + α τ̃d T M −1 τ̃˙d
where the variable η is defined by 1 d
+ α τ̃d T (M −1 )τ̃d + η T η̇ . (29)
2 dt
η = τ̇d − β̂ . (22) Substituting the error equations (15), (17), (21) and (25)
in (29), results in
Differentiating (10) with respect to time, one obtains
V̇ =ε z1 T [z2 − Λz1 ]
˙
β̂ = − α M −1 C z2 − α M −1 K1 z2 + γ z2 T M −1 [−Cz2 − K1 z2 − τ̃d ]
[ γ ]
− α M −1 [−Cz2 − K1 z2 − τ̃d ] . (23)
+ α τ̃d T M −1 η + z2 − α LM −1 τ̃d
α
Simplifying the above equation results T d −1
+ 2 α τ̃d
1
(M )τ̃d
[ dt ]
˙ + η T τ̈d − α M −1 τ̃d .
β̂ = α M −1 τ̃d . (24) (30)
[ ] [ ]
Differentiating (22) with respect to time, one obtains Considering
[ that
] τ̃d T[ α M −1 η
] − η T α M −1 τ̃d = 0,
z2 T −γ M −1 τ̃d + τ̃d T γ M −1 z2 = 0, and assuming that
η̇ = τ̈d − α M −1 τ̃d . (25) τ̈d = 0, equation (30) can be simplified to

In the next section it will be shown that the controlled sys- V̇ =ε z1 T [z2 − Λz1 ]
[ ]
tem is asymptotically stable. + z2 T −γ M −1Cz2 − γ M −1 K1 z2
 Disturbance Observer-based Trajectory Following Control of Robot Manipulators 207

[ ] ∫ ]
+ τ̃d T −α 2 LM −2 τ̃d
t
+ λ3 ∥τ̃d ∥ dt .
2
(38)
1 d 0
+ α τ̃d T (M −1 )τ̃d . (31)
2 dt Equation (38) implies that E(t) ∈ L2 ∩ L∞ , and moreover
Remark 3: The observer-based controller is designed from Barballat conditions it can be concluded that
to cope with the ramp external disturbance, i.e., τ̈d = 0.
Moreover, it is shown numerically that even if for fast time lim (q−qd ) = 0, lim (q̇−q̇d ) = 0, lim (τd −τ̂d ) = 0.
t→∞ t→∞ t→∞
varying disturbance the controller works well. (39)
Remark 4: For any arbitrary vector such as a and b we
can state that 6. NUMERICAL SIMULATIONS
aT b ≤ ∥a∥ × ∥b∥ , In this section the effectiveness of the proposed con-
1( 2 )
trol method is evaluated numerically by performing sim-
aT b ≤ ∥a∥ + ∥b∥2 . (32)
2 ulation on a two link robot manipulator and a 4 DOF
Considering Remark 4, one can simplify (31) to obtain SCARA (selective compliance assembly robot arm) and
the results are compared with the results of methods pro-
1 ( )
posed in [10, 31].
V̇ ≤ ε ∥z1 ∥2 + ∥z2 ∥2 − ε Λm ∥z1 ∥2
2
1 6.1. Comparison with [31]
+ γ M −1 ∥C∥ ∥z2 ∥2 − γ K1m ∥z2 ∥2
MM In [31], a disturbance observer-based controller is de-

1 1 d −1
signed for robotic systems exposed to eternal disturbance
− α Lm
2
∥τ̃ d ∥2
+ α (M ) ∥τ̃d ∥ .
2
MM 2 2 dt which is supposed to be generated by a linear exogenous
(33) system
In any bounded region the inertia matrix and the ξ̇ = Aξ , d = Cξ , (40)
matrix can be bounded by ∥C∥ ≤
Corriollis-centripedal

CM and dtd (M −1 ) ≤ δ [27–30]. Hence, where d ∈ R represents an external disturbance and ξ ∈
( ) Rm . Afterward, a disturbance estimator is designed to es-
V̇ ≤ ∥z1 ∥2 12 ε − ε Λm timate ξ while the matrices A, C are assumed to be certain
( )
γ CM γ K1m and these matrices have been utilized to construct the dis-
+ ∥z2 ∥2 21 ε + − turbance observer. In [31] the effectiveness of the distur-
Mm MM
( ) bance observer-based controller is evaluated by control-
α Lm 1
2
+ ∥τ̃d ∥2 − + αδ . (34) ling a two-link robotic manipulator, each link of which is
MM 2 2 directly driven by a DC motor. Moreover, it is assumed
By designing the matrices Λ, K1 , L and the constants γ , α , that there is a periodic disturbance torque acting on the
ε such that the conditions (27a) to (27c) are held, it can be end of the second link, given by d(t) = 0.5 sin(2t + 1)
stated that N.m. However, the amplitude and frequency of the ex-
ternal disturbance is required to be known. It is worth
V̇ ≤ −λ1 ∥z1 ∥2 − λ2 ∥z2 ∥2 − λ3 ∥τ̃d ∥2 , (35) mentioning that our method does not require the knowl-
edge of the amplitude and frequency of disturbance, and
where λ1 to λ3 are all positive constants and clearly de-
these parameters are not utilized in the controller design.
fined from (34). From (35), it can be deduced that V̇ ≤ 0,
The simulation results of the method presented in [31]
and consequently
and those of the traditional computed torque controller
1 1 (CTC) without utilizing any disturbance compensation are
Bm ∥E(t)∥2 ≤ V ≤ V (0) ≤ BM ∥E(0)∥2 . (36)
2 2 shown in Fig. 3. It is worth mentioning that Fig. 3 is ex-
Hence, actly the same as presented in [31].
We simulate the proposed method in the same situation
BM as performed in [31] and the results are shown in Fig. 4.
∥E(t)∥2 ≤ ∥E(0)∥2 , ∀t > 0. (37)
Bm Comparing Fig. 4 with Fig. 3, it can be justified that the
Thus, any trajectory initiated within any bounded region same control performance is obtained utilizing our method
{ }
while compared to [31], here the amplitude and frequency
such as Ω2 = E | ∥E∥2 ≤ BBMm ℓ will remain in Ω1 .
of the disturbance is not required to be known.
Integrating (35) with respect to time and considering
(36), we obtain 6.2. Comparison with [10]
[ ∫t ∫ t
2 In this section we perform simulation on a 4 DOF
∥E(t)∥2 ≤ℓ − λ1 ∥z1 ∥2 dt + λ2 ∥z2 ∥2 dt SCARA manipulator and the results are compared with
Bm 0 0
208 Mohamadreza Homayounzade and Amir Khademhosseini

Fig. 4. Performance of disturbance observer-based con-

troller.

Fig. 3. Performance of disturbance observer-based con-

troller [31].

the results presented at [10]. SCARA is an industrial 4-

DOF robotic arm, which is widely used in the assembly of
electronic circuits and devices.
The first two joints of the arm, which are used to gen-
erate motion in a horizontal plane, are revolute and have
parallel axes of rotation. The third joint of the arm is a
prismatic joint, which controls the vertical motion (z-axis)
of the end-effector. Finally, the last joint is revolute and is
used to orient the gripper about the z-axis. Fig. 5 depicts a
schematic diagram of this manipulator.
The dynamics of the SCARA in the form of (1) is cal- Fig. 5. SCARA manipulator.
culated by [10]:
   
p1 + p2 c2 p3 + 0.5p2 c2 0 −p5 0
 p3 + 0.5p2 c2 p3 0 −p5   0 
M(q) =  
, G(q) =  
 −p4 g  , (41)
0 0 p4 0 
−p5 −p5 0 −p5 0
 
−p2 s2 q̇2 −0.5p2 s2 q̇2 0 0
 0.5p2 s2 q̇1 0 0 0  where the SCARA robot parameters are defined as
C(q, q̇) = 

,
0 0 0 0 
0 0 0 0 p1 = m1 x1 2 + m2 (x2 2 + a1 2 ) + (m3 + m4 )(a1 2 + a2 2 )
 Disturbance Observer-based Trajectory Following Control of Robot Manipulators 209

Table 1. System parameters.

Parameter Value Parameter Value
m1 15 kg I1 0.02087m1 kg.m2
m2 12 kg I2 0.08m2 kg.m2
m3 3 kg I3 0.05 kg.m2
m4 3 kg I4 0.02m4 kg.m2
x1 0.25 m x2 0.2 m

4
+ ∑ Ii ,
i=1
p2 = 2a1 [x2 m2 + a2 (m3 + m4 )] ,
4
p3 = m2 x2 2 + a2 2 (m3 + m4 ) + ∑ Ii ,
i=2
p4 = m3 + m4 , p5 = I4 , (42)

where Ii is the moment of inertia around the centroid, mi

is the mass, xi is the mass centre, and ai is the length for
link i. For the simulation purpose the system parameters,
similar to [10], are considered as Table 1.
In [10] the authors proposed a disturbance observer-
based controller for robotic systems exposed to distur-
bance, assuming that the time derivative of disturbance
be equal to zero, i.e., τ̇d = 0. Moreover, it is shown
through simulations that for time-varying disturbance the
controller works desirably. In this paper we try to re-
solve the assumption made in [10], such that the pro-
posed method is proved to be asymptotically stable as-
suming τ̈d = 0. To show the effectiveness of the proposed
method over previous ones, our results are compared with
[10] assuming that the SCARA manipulator is exposed to
unbounded time varying disturbance i.e., τid = 1 + 0.5t
∀i = 1 : 4. This scenario may happen as in the case that
the robot has an unexpected contact with the environment
in which the contact force may be considered as time in-
creasing external disturbance.
The SCARA robot is simulated by the proposed control
method and the method proposed in [10], and the results
are shown in Fig. 6. From the comparison it can be ob-
served that the performance of the controller proposed in
[10] degraded exposed to this type of disturbance; how-
ever, the proposed method can handle this mission.

7. CONCLUSIONS

In this paper, a novel disturbance observer-based con-

troller is formulated to stabilize robot manipulators sub-
jected to external disturbance. A disturbance observer is
designed to estimate the external disturbance exerted on
the robot manipulator. It is shown mathematically and nu-
merically that by a judicious selection of controller and Fig. 6. Performance of disturbance observer-based con-
disturbance observer gains, the estimation and tracking troller.
errors will asymptotically get vanished. It is shown that
210 Mohamadreza Homayounzade and Amir Khademhosseini

the proposed control algorithm can cope with even time [13] S. Katsura, Y. Matsumoto, and K. Ohnishi, “Analysis and
varying external disturbance. The effectiveness of the pro- experimental validation of force bandwidth for force con-
posed method is compared with previous methods in dif- trol,” IEEE Trans. on Industrial Electronics, vol. 53, no. 3,
ferent scenarios. pp. 922-928, June 2006.
[14] Y. Ohba, M. Sazawa, K. Ohishi, T. Asai, K. Majima, Y.
REFERENCES Yoshizawa, and K. Kageyama, “Sensorless force control
for injection molding machine using reaction torque ob-
[1] S. Li, J. Yang, W. H. Chen, and X. Chen, Distur- server considering torsion phenomenon,” IEEE Trans. on
bance Observer-Based Control: Methods and Applica- Industrial Electronics, vol. 56, no. 8, pp. 2955-2960, Au-
tions, CRC, Boca Raton, FL, April 2014. gust 2009.

[2] C. H. Choi and N. Kwak, “Robust control of robot [15] E. Sariyildiz and K. Ohnishi, “An adaptive reaction force
manipulator by model-based disturbance attenuation,” observer design,” IEEE/ASME Trans. on Mechatronics,
IEEE/ASME Trans. on Mechatronics, vol. 8, no. 4, pp. 511- vol. 20, no. 2, pp. 750-760, April 2015.
513, December 2003. [16] A. Mohammadi, H. J. Marquez, and M. Tavakoli, “Non-
[3] C. Zhongyi, S. Fuchun, and C. Jing, “Disturbance linear disturbance observers: design and applications to
observer-based robust control of free-floating space manip- Euler-Lagrange systems,” IEEE Control Systems, vol. 37,
ulators,” IEEE Systems Journal, vol. 2, no. 1, pp.114-119, no. 4, pp. 50-72, August 2017.
March 2008.
[17] M. Chen and W. H. Chen, “Disturbance-observer-based ro-
[4] J. H. She, M. Fang, Y. Ohyama, H. Hashimoto, and M. bust control for time delay uncertain systems,” Interna-
Wu, “Improving disturbance-rejection performance based tional Journal of Control, Automation, and Systems, vol.
on an equivalent-input-disturbance approach,” IEEE Trans. 8, no. 2, pp. 445, April 2010.
on Industrial Electronics, vol. 55, no. 1, pp. 380-389, Jan-
[18] K. Kong and M. Tomizuka, “Nominal model manipula-
uary 2008.
tion for enhancement of stability robustness for disturbance
[5] B. Bona and M. Indri, “Friction compensation in robotics: observer-based control systems,” International Journal of
an overview,” Proc. of 44th IEEE Conf. on Decision and Control, Automation and Systems, vol. 11, no. 1, pp. 12-20,
Control, pp. 4360-4367, 2005. February 2013.
[6] U. Sawut, N. Umeda, K. B. Park, T. Hanamoto, and T. [19] E. Kim, “A fuzzy disturbance observer and its application
Tsuji, “Frictionless control of robot arm with sliding mode to control,” IEEE Transactions on Fuzzy Systems, vol. 10,
observer,” Proc. of the IEEE International Vehicle Elec- no. 1, pp.77-84, February 2002.
tronics Conf., pp. 61-66, 2001.
[20] B. K. Kim and W. K. Chung, “Advanced disturbance ob-
[7] Z. J. Yang, Y. Fukushima, and P. Qin, “Decentralized adap- server design for mechanical positioning systems,” IEEE
tive robust control of robot manipulators using disturbance Trans. on Industrial Electronics, vol. 50, no. 6, pp. 1207-
observers,” IEEE Trans. on Control Systems Technology, 1216, December 2003.
vol. 20, no. 5, pp.1357-1365, September 2012.
[21] S. Komada, N. Machii, and T. Hori, “Control of redundant
[8] Y. X. Su, B. Y. Duan, C. H. Zheng, Y. F. Zhang, G. D. Chen,
manipulators considering order of disturbance observer,”
and J. W. Mi, “Disturbance-rejection high-precision mo-
IEEE Trans. on Industrial Electronics, vol. 47, no. 2, pp.
tion control of a Stewart platform,” IEEE Trans. on Con-
413-420, April 2000.
trol Systems Technology, vol. 12, no. 3, pp. 364-374, May
2004. [22] W. H. Chen, D. J. Ballance, P. J. Gawthrop, and J. O’Reilly,
[9] A. Mohammadi, H. J. Marquez, and M. Tavakoli, “Distur- “A nonlinear disturbance observer for robotic manipula-
bance observer-based trajectory following control of non- tors,” IEEE Trans. on Industrial Electronics, vol. 47, no.
linear robotic manipulators,” Proc. of the 23rd CANCAM, 4, pp. 932-938, August 2000.
pp. 5-9, 2011. [23] A. Nikoobin and R. Haghighi, “Lyapunov-based nonlinear
[10] A. Mohammadi, M. Tavakoli, H. J. Marquez, and F. disturbance observer for serial n-link robot manipulators,”
Hashemzadeh, “Nonlinear disturbance observer design for Journal of Intelligent & Robotic Systems, vol. 55, no. 2, pp.
robotic manipulators,” Control Engineering Practice, vol. 135-153, July 2009.
21, no. 3, pp. 253-267, March 2013. [24] R. Hao, J. Wang, J. Zhao, and S. Wang, “Observer-based
[11] S. Katsura, Y. Matsumoto, and K. Ohnishi, “Modeling of robust control of 6-DOF parallel electrical manipulator
force sensing and validation of disturbance observer for with fast friction estimation,” IEEE Trans. on Automation.
force control,” IEEE Trans. on Industrial Electronics, vol. Science Engineering, vol. 13, no. 3, pp. 1399-1408, July
54, no. 1, pp. 530-538, February 2007. 2016.
[12] M. Rakotondrabe, C. Clévy, K. Rabenorosoa, and K. Ncir, [25] J. Huang, S. Ri, L. Liu, Y. Wang, J. Kim, and G. Pak, “Non-
“Force estimation and control of an instrumented platform linear disturbance observer-based dynamic surface control
dedicated to automated micromanipulation tasks,” Proc. of of mobile wheeled inverted pendulum,” IEEE Trans. Con-
IEEE Conf. on Automation Science and Engineering, pp. trol System Technology, vol. 23, no. 6, pp. 2400-2407,
722-727, 2010. November 2015.
 Disturbance Observer-based Trajectory Following Control of Robot Manipulators 211

[26] Y. S. Lu and C. W. Chiu, “A stability-guaranteed integral Mohamadreza Homayounzade received

sliding disturbance observer for systems suffering from the B.S. degree in Mechanical engineering
disturbances with bounded first time derivatives,” Interna- in 2007 from Shiraz University and M.S.
tional Journal of Control, Automation and Systems, vol. 9, and Ph.D. degrees in Mechanical engineer-
no. 2, pp. 402-409, April 2011. ing, in 2009 and 2014, respectively, from
[27] T. Burg, D. Dawson, J. Hu, and M. De Queiroz, “An adap- Isfahan University of Technology (IUT),
tive partial state-feedback controller for RLED robot ma- Isfahan, Iran. Since 2014, he has been with
nipulators,” IEEE Trans. on Automatic Control, vol. 41, no. the Department of Mechanical Engineer-
7, pp. 1024-1030, July 1996. ing at Fasa University, Fasa, Iran. His re-
search interests are robotics (analysis, design, and manufactur-
[28] T. C. Lee, B. S. Chen, and Y. C. Chang, “Adaptive con- ing), control of dynamical systems, vibrations, and mechatron-
trol of robots by linear time-varying dynamic position feed- ics.
back,” International Journal of Adaptive Control and Sig-
nal Processing, vol. 10, no. 6, pp. 649-671, November
1996. Amir Khademhosseini received the B.S.
degree in Computer engineering in 2005
[29] M. Homayounzade, M. Keshmiri, and M. Keshmiri,
from Shiraz University and his M.S. de-
“Adaptive control of electrically-driven robot manipulators
gree in Computer engineering in 2009,
without velocity/current measurements,” Proc. of ASME
from Isfahan University, Isfahan, Iran.
International Mechanical Engineering Congress and Ex-
Since 2012, he has been with the De-
position, 2014.
partment of Computer Engineering at Fasa
[30] M. Homayounzade and M. Keshmiri, “Noncertainty equiv- University, Fasa, Iran. His research inter-
alent adaptive control of robot manipulators without veloc- ests are computer architecture, embedded
ity measurements,” Advanced Robotics, vol. 28, no. 14, pp. systems, FPGA, IoT, and mechatronics.
983-996, July 2014.
[31] W. H. Chen, “Disturbance observer based control for non-
linear systems,” IEEE/ASME Transactions on Mechatron-
ics, vol. 9, no. 4, pp.706-710, December 2004.