electronics

Article
Output Feedback Control via Linear Extended State
Observer for an Uncertain Manipulator with Output
Constraints and Input Dead-Zone
Duc Thien Tran 1,2 , Hoang Vu Dao 2 , Truong Quang Dinh 3 and Kyoung Kwan Ahn 2, *
1 Automatic Control Department, Ho Chi Minh City University of Technology and Education,
Ho Chi Minh City 700000, Vietnam; thientd@hcmute.edu.vn
2 School of Mechanical Engineering, University of Ulsan, Ulsan 44610, Korea; hoangvudaocsp@gmail.com
3 Warwick Manufacturing Group (WMG), University of Warwick, Coventry CV4 7AL, UK;
T.Dinh@warwick.ac.uk
* Correspondence: kkahn@ulsan.ac.kr; Tel.: +82-52-259-2282




Received: 21 July 2020; Accepted: 14 August 2020; Published: 20 August 2020


Abstract: This paper proposes an output feedback controller with a linear extended state observer
(LESO) for an n-degree-of-freedom (n-DOF) manipulator under the presence of external disturbance,
an input dead-zone, and time-varying output constraints. First, these issues are derived in mathematical
equations accompanying an n-DOF manipulator. The proposed control is designed based on the
backstepping technique with the barrier Lyapunov function (BLF) and a LESO. The LESO is used for
estimating both the unmeasured states and the lumped uncertainties including the unknown frictions,
external disturbances, and input dead-zone, in order to enhance the accuracy of the robotic manipulator.
Additionally, the BLF helps to avoid violation of the output constraints. The stability and the output
constraint satisfaction of the controlled manipulator are theoretically analyzed and proven by the
Lyapunov theorem with a barrier Lyapunov function. Some comparative simulations are carried out on
a 3-DOF planar manipulator. The simulation results prove the significant performance improvement of
the proposed control over the previous methods.

Keywords: backstepping control; extended state observer; Lyapunov theorem; barrier Lyapunov
function; time-varying output constraint; robotic manipulator

1. Introduction
In recent years, robots have attracted the interest of many researchers in institutes, universities,
and technology companies around the world [1]. Challenges such as highly nonlinear dynamics, modeling
error, and external disturbances can degrade the control performance of the robotic manipulator. In order to
improve the accuracy and reliability of the robotic manipulator, researchers have developed controller
approaches to handle these problems. Some well-known robotic controllers such as computed torque
control [2], backstepping control [3–5], and sliding mode control [3,5–8], etc. have been widely applied in
robotic applications.
Backstepping control is one of the most useful techniques for controlling nonlinear systems [9],
regardless of the mismatched and matched uncertainties. In order to improve the effectiveness of
the backstepping control, some advanced approximators, such as fuzzy logic systems (FLSs) [10–12],
neural networks [13–16], and extended state observers [17,18], were applied to the backstepping
control to compensate for the uncertainties. In Reference [10], an FLS was used in an advanced
backstepping control to approximate the unknown nonlinearities of a manipulator. In Reference [19],
Wang et al. designed an adaptive fuzzy backstepping control for an underwater vehicle manipulator
system; the FLS was used to estimate the system parameters. The results in these papers proved the

Electronics 2020, 9, 1355; doi:10.3390/electronics9091355 www.mdpi.com/journal/electronics

Electronics 2020, 9, 1355 2 of 20

effectiveness of the FLS. However, it is difficult to analyze the stability of an FLS to control the whole
system [20] in which it is embedded because of the complexity in selecting the membership functions
and fuzzy rules of the fuzzy system. In Reference [13], neural network estimators were provided to
approximate the unknown disturbance and unknown dead-zone in a robotic manipulator. Although the
results demonstrated its advantages in approximating the unknown nonlinear functions, the selecting
neural network structure required the expert’s experience. It is hard to implement the neural network
without the results of the expert’s experiments. In the 1990s, Jingquing Han firstly proposed a linear
extended state observer (LESO) for estimating the uncertainties and unmeasured states. The observer is
well-known as a simple structure observer and it can work well under the inaccuracy of mathematical
models and strong nonlinearities to approximate the uncertainties [21]. The LESOs and backstepping
technique were applied together in many systems such as hydraulic systems [17,22,23], spacecraft [24],
inertia wheel pendulum [25] and mass-spring mechanical systems [26].
In addition to the above uncertainties, the constraints from inputs, outputs, and state variables
are other challenges encountered in practice. They arise in the application of robots when the robot
and human co-operate or collaborate in manufacturing processes and daily life. Transgression of the
constraints may produce not only decay of the system performance [27], but also unsafe operation for
both the robot and the human. By designing the advanced controllers with a BLF whose output is
infinite at corresponding limits, these approaches guarantee that the barriers will not be broken [28,29].
Consequently, the constraints are ensured to be valid all the time. In Reference [30], an adaptive neural
network control was proposed for a robotic manipulator under the presence of an input dead-zone
and output constraint. While the input dead-zone and modeling error were approximated by a neural
network, the output constraint was overcome by the BLF. In Reference [31], a BLF was combined with
an adaptive neural network to design an advanced control for a two-DOF hydraulic robot with output
constraints. The neural network (NN) was provided to estimate the unknown model of the robot.
In Reference [32], a fuzzy logic system was employed with the BLF to approximate unknown nonlinear
functions and to tackle the output constraint in a class of a nonstrict-feedback system. In Reference [33],
an adaptive fuzzy backstepping surface control was designed based on a time-varying BLF for uncertain
strict-feedback nonlinear systems. The fuzzy logic system approximated the unknown nonlinear
functions and the BLF helped to overcome the asymmetric time-varying output constraints.
From the aforementioned above, this paper proposes an advanced output feedback control via
a linear extended state observer for an n-DOF manipulator, regardless of the uncertainties and the
time-varying output constraint. The uncertainties such as unknown frictions, external disturbances,
and input dead-zone are taken into account in this study. In order to handle these issues, they are
firstly described with n-DOF manipulator dynamics. The proposed control is designed based on
the backstepping technique with an LESO and BLF. While the LESO approximates the uncertainties
and estimates the unmeasured states, the BLF helps to guarantee the satisfaction of the output
responses with the constraints. Compared with a neural network and a fuzzy logic system, the LESO
possesses a simple structure and does not require the experience of the designer. The stability and the
constrained satisfaction are analyzed by the Lyapunov approach with the barrier Lyapunov function.
Comparative simulations are implemented on a 3-DOF planar manipulator, and the simulation results
prove that the proposed method significantly improves the performance over previous approaches.
The rest of this paper is organized as follows: We provide the robotic manipulator dynamics and
problem formulations in Section 2. Section 3 describes the control design, which includes the linear
extended state observer design, the proposed control design, and the proof of stability. The effectiveness
of the proposed control is exhibited by some simulation results in Section 4. Finally, some conclusions
are presented in Section 5.
Electronics 2020, 9, x FOR PEER REVIEW 3 of 22

2. Robotic Manipulator Dynamics

Electronics 9, 1355 we consider an n-DOF manipulator under the presence of unknown friction, input
2020,paper,
In this 3 of 20

dead-zone, and external disturbance. Its dynamics in the joint space are expressed by [2]
2. Robotic Manipulator Dynamics
 ( ,q)q +G(q) +JT ( q) fext + τfric = H( τ) ,
( )q+Cq
Mq (1)
In this paper, we consider an n-DOF manipulator under the presence of unknown friction,
n ×1
input q , q , q
wheredead-zone,  ∈ R and present angular
external position,
disturbance. Its velocity,
dynamicsand acceleration
in the joint spacevectors in the joint
are expressed space of
by [2]
n×n n× n
the manipulator, respectively; M ( q ) ∈ R presents the inertia matrix; C (q , q ) ∈ R expresses the
.. . .
M(q)q + C(q, q)q + G(qn)×1 + JT (q)fext + τ f ric = H (τ),
Coriolis and centrifugal term matrix; G (q ) ∈ R derives the gravity vector; J q is a nonsingular () (1)
. ..
where q, q,
Jacobian q ∈ Rn×1
matrix; f ext present angular
presents position,
the external velocity, and
disturbance vector; τ fr ic is the
acceleration vectors in thefriction
unknown joint space of
vector;
n×n presents the inertia matrix; C(q, q . n×n expresses the
the manipulator, respectively; M ( q ) ∈ R
and H ( τ ) is the torque vector acting on joints within the dead-zone. ) ∈ R
Coriolis and centrifugal term matrix; G(q) ∈ Rn×1 derives the gravity vector; J(q) is a nonsingular
The friction model, τ fric , is exhibited as
Jacobian matrix; fext presents the external disturbance vector; τ f ric is the unknown friction vector;
and H (τ) is the torque vector acting on joints within the dead-zone.
 q 
τ = b q + c tanh   , (2)
The friction model, τ f ric , is exhibited fric
as ψ 

where b ∈ R n×n and c ∈ R n×n are positive diagonal .

. !
matrices;
q ψ is a positive constant; and
τ f ric = bq + ctanh , (2)
 q    q 1   q n 
T ψ
n×1
tanh   =  tanh   ,..., tanh   ∈ R .
where  ψb ∈ Rn×n  ψ and c ∈ ψ Rn×n  are positive diagonal matrices; ψ is a positive constant;
.  .   . T
q q1 q
and tanh ψ = tanh ψ , . . . , tanh ψn ∈ Rn×1 .
Property 1 [34]. M ( q ) is a positive definite symmetric matrix and the condition
Property 1 [34]. M(q) is a positive definite symmetric matrix and the condition 0 < λmin (M(q)) ≤ kM(q)k ≤
λmax (M(q)) ≤ σ0 holds where σ0 is a positive constant, and λmin (M) and λmax (M) are the minimum and
0 < λ m in ( M ( q ) ) ≤ M ( q ) ≤ λ m ax ( M ( q ) ) ≤ σ 0 holds where σ 0 is a positive constant, and λ m in ( M ) and λ m ax ( M )
maximum eigenvalues of a matrix M.
are the minimum. and maximum
 . eigenvalues of a matrix M . h .  . i
Property 2 [2]. M(q) − 2C q, q is a skew-symmetric matrix, that is provided as xT M(q) − 2C q, q x = 0.
Assumption 1 [13]. The input dead-zone nonlinearity presented in Figure 1 can be derived as follows:
Property 2 [2]. M ( ) ( )
 q −2C q,q is a skew-symmetric matrix, that is provided as


hr (τ − τr ) τ ≥ τr
x M( q) − 2C( q, q )  x = 0 .


T

H (τ) = 

0 τl < τ < τ r , (3)



hl (τ − τl ) τ ≤ τl



Assumption 1 [13]. The input dead-zone nonlinearity presented in Figure 1 can be derived as follows:
where τr and τl are unknown constants for which: τr > 0, τl < 0, and H (τ) = [H (τ1 ), . . . , H (τn )]T .

H (τ )
hr (τ − τ r )

0
τl τr τ

hl (τ − τ l )

Figure 1. Dead-zone model.

Figure 1. Dead-zone model.
Assumption 2. The dead-zone nonlinear functions are smooth functions and their derivatives are bounded by
unknown positive constants such that
.
0 < Hr1 < hr (τ − τr ) < Hr2 , [τr , +∞)
. . (4)
0 < Hl1 < hl (τ − τl ) < Hl2 , (−∞, τl ]
Electronics 2020, 9, 1355 4 of 20

Based on Equations (3) and (4), the dead-zone functions can be represented as follows:

H (τ) = τ + ξ(t), (5)

where ξ(t) is an unknown vector whose elements are smooth functions. Additionally, their derivatives
are bounded.
.
Let x1 = q ∈ Rn and x2 ∈ q ∈ Rn ; then the robotic dynamics (1) can be rewritten as follows:
.
x1 = x2
. , (6)
x2 = M−1 (x1 )(u − C(x1 , x2 )x2 −G(x1 ) − ∆)

where xi = [xi1 , xi2 , . . . , xin ]T , (i = 1, 2); ∆ = JT (q)fext + τ f ric + ξ(t) derives a lumped disturbance
including the unknown friction, external disturbance, and an unknown vector of the input dead-zone,
and u depicts the torque control signal.
In this paper, we design an advanced control to guarantee that the output responses track a reference
xd = [xd1 , xd2 , . . . , xdn ]T and satisfy the condition, kci (t) < x1i (t) < kci (t), where kc (t) = [kc1 (t), . . . , kcn (t)]T
T
and kc (t) = [kc1 (t), . . . , kcn (t)] are time-varying functions.
Assumption 3. In this study, we suppose that the manipulator operates in a bounded workspace. It means that
the reference signals, xdi , i = 1, 2, . . . , n, are bounded and known, |xdi | ≤ Xd0 , and Xd0 is a positive constant.

3. Control Design
Before we design the LESO to estimate the lumped disturbance and the unmeasured states in the
manipulator, an extra state x3 ∈ Rn×1 is added in the manipulator dynamics (6) to present the lumped
T
disturbance, M−1 (x1 )∆. The system state is presented as x = [xT1 , xT2 , xT3 ] ∈ R3n×1 . Note that the extra
state x3 is continuously differentiable and bounded as in [35,36].
Assumption 4. The difference of the state x3 (t) is assumed to be bounded, i.e., kδ(t)k∞ ≤ δ, where δ is a
positive constant.
The manipulator dynamics (6) is represented as follows:
.
x1 = x2
.
x2 = F(x1 , x2 ) + H(x1 )u + x3 , (7)
.
x3 = δ(t)

where δ(t) presents the derivative of the state x3 (t); x3 defines −M−1 (x1 )∆; H(x1 ) derives M−1 (x1 );
and F(x1 , x2 ) describes −M−1 (x1 )(C(x1 , x2 )x2 + G(x1 )).
Assumption 5. We suppose that the functions F(x1 , x2 ) are locally Lipschitz for x2 in its practical range.

3.1. Linear Extended State Observer Design

The LESO is designed to not only approximate the lumped disturbance, x3 , but also estimate the
unmeasured system state vector, x2 . Let x̂ express the estimated system state of x and let e
x present the
estimation error, (e
x = x − x̂). Now, we represent the robotic dynamics (7) as follows:
.
x = An x + Bn u + φ(x) + D(x)
, (8)
y = x1
Electronics 2020, 9, 1355 5 of 20

     
 0n×n In×n 0n×n   0n×n   0n×1 
where An =  0n×n 0n×n In×n  ∈ R3n×3n ; Bn =  H(x1 )  ∈ R3n×n ; φ(x) =  F(x)  ∈ R3n×1 ;
     
     
0n×n 0n×n 0n×n 0n×n 0n×1
  
 
 0n×1 
and D(x) =  0n×1  ∈ R3n×1 .
 
 
δ
The extended state observer is presented as follows:
.
x̂ = An x̂ + Bn u + φ(x̂) + κ(x1 − x̂1 ), (9)

T T
where φ(x̂) = [ 0n×n F(x1 , x̂2 ) 0n×n ] , κ = [ 3κ0 In×n 3κ20 In×n κ30 In×n ] ∈ R3n×n presents the
observer gain matrix; and κ0 > 0 is adjusted to enhance the observer performance.
From Equations (8) and (9), the estimation error dynamics is computed as
.
x + φ(x) − φ(x̂) − κe
x = Ane
e x1 + D(x), (10)

T xi ∆
where D(x) = [ 0n×n 0n×n δ(t) ] . Now, we define ζi = ∈ Rn×1 (i = 1, 2, 3), φ
e= φ(x) − φ(x̂),
e
κ0 i−1
and then the LESO (10) is represented as

. φ
e D(x)
ζ = κ0 An1 ζ + + 2 , (11)
κ0 κ0
 
 −3In×n In×n 0n×n 
where An1 =  −3In×n 0n×n In×n  ∈ R3n×3n is a Hurwitz matrix.
 
 
−In×n 0n×n 0n×n
From Assumption 5, the below inequality can be obtained:

|φ|
e = |φ(x2 ) − φ(x̂2 )| ≤ c|ζ2 |. (12)

Theorem 1. When the LESO (9) is used to estimate the lumped disturbance of the unmeasured states, and the
inequality (12) is guaranteed, then the estimation errors are bounded with the appropriate constant.

Proof of Theorem 1. A Lyapunov function is taken into account as follows:

1 T
V0 = ζ Pζ, (13)
2
where P derives a positive definite matrix. It is chosen as a solution of the following Lyapunov equation:

ATn1 P + PAn1 = −2I3n×3n . (14)

From (11), the differential Lyapunov function is presented as

. T
φ φ
e e 
D(x) D(x)
 
V0 = 12 κ0 ζT ATn1 P + PAn1 ζ + 12 κ0 + κ20
Pζ + 21 ζT P κ0 +
κ20
T
φ
e
D(x) δ
 
= −κ0 ζT ζ + κ0 + κ2 Pζ ≤ − κ0 − c
kζk22 + λ (P)kζk2 . (15)
0
κ0 κ20 max
  
= − κ0 − κ0 kζk2 + κδ2 λmax (P) kζk2

c
0
 T
 φ D ( x )   c  2 δ
= −κ 0 ζ ζ +  +  Pζ ≤ −  κ 0 −  ζ 2 + 2 λmax ( P ) ζ
T
. (15)
 κ0 κ0  κ0  κ0
2 2

  c  δ 
=  −  κ 0 −  ζ 2 + 2 λmax ( P )  ζ 2
 κ0  κ0 
Electronics 2020, 9, 1355    6 of 20

The differential Lyapunov Function (15) is negative when −  κ 0 − c  ζ + δ 2 λmax ( P ) ≤ 0 ; that

κ  2 κ
The differential Lyapunov Function (15) is negative when − κ0 − κ0c0 kζk2 +0 δ2 λmax (P) ≤ 0;

κ
δ 0
ζ 2 ≥kζk2 ≥2 δλmax ( P) . The estimation errors in LESO are reducing and the stability of the
implies λ
κ0 (κ0 κ−0 (cκ)20 −c) max
that implies (P). The estimation errors in LESO are reducing and the stability of
the ESO is guaranteed [37] when the bandwidth, κ0 , increases. 
ESO is guaranteed [37] when the bandwidth, κ 0 , increases. □

3.2. Proposed Control Design

Figure 22 presents
presents the
thestructure
structureofofthe
theproposed
proposed control with
control an an
with n-DOFn-DOF manipulator
manipulator under the
under
presence
the of unknown
presence of unknown external disturbance,
external friction,
disturbance, inputinput
friction, dead-zone,
dead-zone, and output
and output constraints. The
constraints.
proposed
The control
proposed consists
control of aoffull
consists state
a full feedback
state feedback control
control based
basedononthetheBLF,
BLF,and andaa linear
linear extended
observer. The backstepping control is designed with the barrier Lyapunov function to avoid the
state observer.
violation of the output constraint. Because all states, such as as position
position and
and velocity,
velocity, of the manipulator
used to
are used todesign
designthethebackstepping
backstepping control,
control, thisthis control
control is named
is named as a as
fulla state
full state feedback
feedback control-
control-based
based barrier
barrier LyapunovLyapunov function.
function. The extended
The extended state observer
state observer was employed
was employed to approximate
to approximate the
the lumped
lumped disturbance
disturbance in the manipulator
in the manipulator dynamics dynamics and unmeasured
and unmeasured variable, variable,
x̂2 . x̂2 .

External disturbance

Proposed control n-DOF Manipulator

xd
Unknown frictions
Reference Full state u Output
feedback control Nominal dynamics
constraints
based BLF
x1 Input deadzone
x̂2
x̂3 x1
Linear extended
state observer

Figure 2. Structure of the proposed control.

Figure BLF = barrier
2. Structure of theLyapunov control. n-DOF = n-degree-of-freedom.
proposedfunction;

The tracking errors, , (i = Lyapunov

BLF e=i barrier 1, 2), are defined
function;asn-DOF = n-degree-of-freedom.

n×1
The tracking errors, ei , ( i =1,2) , are
e1 = x1 − xd ∈ R
defined as n×1 , (16)
e2 = x2 − α1 ∈ R

where xd ∈ Rn×1 is the reference signal.

The time-varying upper and lower boundary errors of e1 are computed as:

kai (t) = kci (t) − xdi (t)

, (17)
kbi (t) = kci (t) − xdi (t)

where kci (t) ≤ x1i (t) ≤ kci (t), kai is the time-varying upper boundary error, and kbi is the time-varying
lower boundary error.
The virtual control, α1 is computed as
.
α1 = −K1 e1 − λ1 e1 + xd , (18)

where λ1 ∈ Rn×n is a positive diagonal matrix; λ1 = diag([λ11 , . . . , λ1n ]) ∈ Rn×n expresses a positive
diagonal matrix. The elements in the matrix λ1 are stated as
v
t .
u 2  . 2
 kai (t)   kbi (t) 
λ1i =   +   . (19)
kai (t)  kbi (t) 
   
Electronics 2020, 9, 1355 7 of 20

The control input is calculated as follows:

.
u = −κξ − K2 e2 − M(x1 )x̂3 + C(x1 , x̂2 )α1 + G(x1 ) + M(x1 )α1 , (20)
    
h(e11 ) 1−h(e11 ) h(e1n ) 1−h(e1n )
where κξ = 2 −e2 + 2 −e2 e11 , . . . , 2 −e2 + 2 −e2 e1n ∈ Rn×1 , h(e1i ) is derived as follows
ka1 11
kb1 11
kan 1n
kbn 1n

(
0, e1i ≤ 0
h(e1i ) = , i = 1, . . . , n, (21)
1, otherwise

and K2 ∈ Rn×n is a positive diagonal matrix.

3.3. Stability Analysis

Lemma 1 [22]. The following inequality holds for any positive constant k ∈ R and x ∈ R so that |x| < k:

k2 x2
log ≤ . (22)
k2 − x2 k 2 − x2

Theorem 2. The control law in (18) and (20), which utilizes the estimation value of the unmeasurable
state, and the lumped disturbance from the ESO in (9) guarantee the ultimately uniformly bounded tracking
performance and satisfaction of the output constraint in (17) of the manipulator described by (1), under unknown
friction, external disturbance, and an unknown vector of the input dead-zone.

Proof of Theorem 2. Step 1. We take the time derivative of the position error e1 shown as
. . .
e1 = x2 − xd = e2 + α1 − xd . (23)

From (18), substituting the virtual control signal, α1 , into (23), the result is represented as
.
e1 = e2 − K1 e1 − λ1 e1 . (24)

Based on Yu et al. [38], to guarantee the constrained performance of the joint angles, we can select
a barrier Lyapunov function as follows:

n   2   2
kbi

1 X   kai   
V1 = h(e1i ) log 2  + (1 − h(e1i )) log . (25)
2 k − e2 2 − e2
kbi
  
i=1 ai 1i 1i

In order to simplify the BLF (25), we state variables by

e1i e
ξai = , ξbi = 1i , ξi = h(e1i )ξai + (1 − h(e1i )ξbi ). (26)
kai kbi

As a result, the Lyapunov Function (25) is rewritten as follows:

n
X 1 1
V1 = log . (27)
2
i=1
1 − ξ2 i

The differential barrier Lyapunov function is calculated by

n 
 . . 
 h(e1i )ξai  . ξ
  
. X k (
e1i + a1 e1i  + 
 1 − h ( e 1i )) bi  . k 
e1i + b1 e1i .

V1 =     (28)
kai  kbi 
i = 1 1 − ξai kai 1 − ξ2bi kbi
 2  
Electronics 2020, 9, 1355 8 of 20

Substituting (24) into (28), the result presents as

. .  !
. n      
P h(e1i )ξai ka1 (1−h(e1i ))ξbi . kb1
V1 = e2i − kai + k1i + e1i + e − kbi + k1i + e1i
2
i = 1 1−ξai )kai
( kai (1−ξ2bi )kbi 1i kbi
n
. (29)
P ξ2
≤− k1i 1−ξ 2 + κξ e2
i=1

Step 2. Based on (6), the time derivative of error, e2 is expressed as follows

. . . .
e2 = x2 − α1 = M(x1 )−1 [u − C(x1 , x2 )x2 − G(x1 ) − ∆(t)] − α1 . (30)
.
From (30) and (20), the time derivative of error, e2 is represented as
.
e2 = M(x1 )−1 [−C(x1 , x2 )x2 − G(x1 ) − ∆(t)
. i .
−κξ − K2 ê2 − M(x1 )x̂3 + C(x1 , x̂2 )α1 + G(x1 ) + M(x1 )α1 − α1 . (31)
−1
= M(x1 ) [−κξ − K2 (e2 −e x2 ) − ∆(t) − M(x1 )x̂3 − C(x1 , x2 )e2 + ∆C(x1 ,e
x2 )α1 ]

To consider the stability of the dynamics system (1), including not only the position tracking
performance at Step 1, but also the speed tracking performance (31), the following Lyapunov function
is investigated as
1
V2 = V1 + eT2 M(x1 )e2 . (32)
2
Next, the time derivative of the Lyapunov function (32) is calculated as

. . . 1 .
V 2 = V 1 + eT2 M(x1 )e2 + eT2 M(x1 )e2 . (33)
2
Replacing (29) and (31) into (33) with property 2, the result is expressed by

. n
ξ 2
2 − e2 K2 e2 + e2 c0 |ζ2 | − e2 ( ∆ (t) + M(x1 )x̂3 )
T T T
P
V2 = − k1i 1−ξ
i=1 . (34)
x2 + ∆C(x1 ,e
+eT2 K2e x2 )α1

In order to demonstrate the stability of the entire closed-loop system, including the estimation
performance of the ESO, we choose a Lyapunov function via (15) and (32) as follows:
n
1 T 1X 1 1
V = V0 + V2 = ζ Pζ + log + eT2 M(x1 )e2 . (35)
2 2
i=1
1 − ξ 2
i
2

The time derivative of the Lyapunov Function (35) is computed as

. n
ξ2 δ
 
2 − κ0 −
c
kζk22 + λ − eT2 K2 e2
P
V =− k1i 1−ξ κ0 κ20 max
(P)kζk2
i=1 . (36)
−eT2 (∆(t) − M(x1 )x̂3 ) + eT2 K2e
x2 + ∆C(x1 ,e
x2 )α1

∆
We define ε = ∆(t) − M(x1 )x̂3 ∈ Rn×1 as the disturbance estimation error. These inequalities
T
hold: −eT2 ε ≤ 12 eT2 e2 + 21 εT ε and eT2 K2e
x2 ≤ 21 eT2 K2 e2 + 12e x2 . Thus, when Lemma 1 is investigated,
x2 K2e
Equation (36) is rewritten as below:

.   n
V ≤ − κ0 − c
kζk22 − K1
log 1 2 − 12 eT2 (K2 − In×n )e2
P
κ0 1−ξi
i=1 , (37)
δ 1eT
+ 12 εT ε + λ (P)kζk2 + 2 x2 K2ex2 + ∆C(x1 ,e
x2 )α1 ≤ −co V + D2
κ2 max
0
Electronics 2020, 9, 1355 9 of 20

δ
     
where c0 = min λmin κ0 − c
P−1 , 2λmin (K1 ), λmin (K2 − I)M−1 and D = k 12 εT ε + λ (P)kζk2
κ0 κ20 max
T
+ 21e
x2 K2ex2 + ∆C(x1 ,e
x2 )α1 k∞ .
From Reference [39], we can state that when the system is controlled by the proposed control,
it is ultimately uniformly bounded under the presence of unknown frictions, external disturbances,
and input dead-zone. From (37), we can find the inequation as follows:

0 ≤ V (t) ≤ µ + (V (0) − µ)e−c0 t ≤ µ + V (0), (38)

D
where µ = c02 .
From (25), (26), (35) and (38), we achieve the results as

1 1
log ≤ µ + V (0), (39)
2 1 − ξ2 i
 2 
kai

1
log ≤ µ + V (0), e1i > 0



 2 k 2 −e2
 ai 1i
. (40)

  2 
 kbi
 2 log 2 2 ≤ µ + V (0), e1i ≤ 0
1



kbi −e1i

After we take the exponentials on both sides of (40), the results are given as
 k2

 ai
≤ e2(µ+V (0)) , e1i > 0
 kai2 −e21i


 2
kbi
. (41)
2(µ+V (0)) , e ≤ 0

2 ≤ e

1i

 k2 −e
bi 1i

The following inequality can then be achieved:

 p
e1i ≤ kai p

 1 − e−2(µ+V (0)) , e1i > 0
. (42)
−2(µ+V (0)) ,

e1i ≥ −k
bi 1 − e e1i ≤ 0


As a result, we can conclude that the output constraints are guaranteed. The proof is complete. 

4. Numerical Simulations

4.1. Simulation Descriptions

Some simulations were conducted on MATLAB Simulink with a 3-DOF planar manipulator to
illustrate the superiorities of the proposed control. The MATLAB Simulink was configured with a
sampling time of 0.001 s; the solver type was ODE3. Additionally, the simulation time was 30 s.
The 3-DOF planar manipulator presented in Figure 3 is a planar robot with 3 rotary actuators.
The parameters of the manipulator are presented in Table 1. Additionally, all mass exists as a point mass
at the distal end of each link, and the center of mass in each link is presented by i PC = li Xi ,(i = 1,2,3) .
By using the Newton iteration method in [2], the dynamics of the manipulator are presented in the
Appendix A.

Table 1. Parameters of the 3-DOF planar manipulator.

Symbol Description Symbol Description

l1 = 0.35 m Length of 1st link m1 = 0.23 kg Mass of 1st link
l2 = 0.3 m Length of 2nd link m2 = 0.2 kg Mass of 2nd link
l3 = 0.15 m Length of 3rd link m3 = 0.1 kg Mass of 3rd link
g = 9.81 ms−2 Gravity constant - -
 The 3-DOF planar manipulator presented in Figure 3 is a planar robot with 3 rotary actuators.
The parameters of the manipulator are presented in Table 1. Additionally, all mass exists as a point
i
mass at the distal end of each link, and the center of mass in each link is presented by PC = li Xi ,(i=1,2,3)
. By using the Newton iteration method in [2], the dynamics of the manipulator are presented in the
Electronics 2020,
Appendix A. 9, 1355 10 of 20

y3 x3

y2 l3 q3

x2
y1 l2
g q2
y0
l1 x1
q1
x0
Figure 3.
Figure Structure of
3. Structure of the
the 3-DOF
3-DOF manipulator.
manipulator.

The friction model vector

Tableincludes the viscous
1. Parameters of the and coulomb
3-DOF frictions, which is presented as follows:
planar manipulator.
. !
Symbol Description . qSymbol Description
τ f ric = bq + c tanh ∈ R3 , (43)
l1 = 0.35 m Length of 1st link mψ1 = 0.23 kg Mass of 1st link
l2 = 0.3 m Length of 2nd link m2 = 0.2 kg Mass of 2nd link
where bl3 ==0.15 m ([1, 1, 1])(
0.5diag Nms/rad
Length of 3rd), clink
= 0.2diag([1, m31,=10.1 kg ), and ψ =
])(Nm 10.ofThe
Mass 3rddead-zone
link
functions ms−2 as hr (τ −Gravity
are defined
g = 9.81 τr ) = (constant
τ − τr ); hr (τ − τl ) = 1.2-(τ − τl ), where τr = 0.2 and - τl = −0.2.
During the simulation period, an external disturbance along the x-axis of the original coordinate system
is applied after the
The friction 20th vector
model as f = the
second,includes ( N). The
−40viscous andtrajectory
coulomb signals, xd and
frictions, zd , in
which is the Cartesian
presented as
coordinate, are sine waves, xd = 0.4 + 0.2 cos(2π f f re t)(m), yd = 0(m), and zd = 0.2 sin(2π f f re t)(m),
follows:
where f f re is the frequency of the trajectory. Additionally, the rotary angle around the z-axis is zero.

4.2. Simulation Results

The merits of the proposed controller are illustrated through comparisons with two other controllers:

• The backstepping control (BC):

.
u = −K2 e2 + C(x1 , x2 )α1 (t) + G(x1 ) + M(x1 )α1 (t) − e1 (44)

α1 = x2d − K1 e1 (45)

• The linear extended state observer via backstepping control (LESOBC):

.
u = −K2 e2 + C(x1 , x̂2 )α1 (t) + G(x1 ) + M(x1 )α1 (t) − e1 − M(x1 )x̂3 , (46)

α1 = x2d − K1 e1 (47)

where x̂3 is the estimated lumped disturbance. This estimated lumped disturbance is approximated
by the LESO in (9).

The parameters of these controllers are exhibited in Table 2. In order to ensure equality in comparisons
between the controllers, parameters of the backstepping are firstly selected. Next, some parameters of the
LESOBC are inherited from the BC, and others, the observer gains, are adjusted. Finally, the proposed
control copies the parameters in the LESOBC and uses the upper and lower boundaries.
The simulations are divided into two cases. In the first case, we conduct simulations with the
low-frequency references which are set to 0.1 Hz. In the second case, the frequency of the reference is
increased to 0.5 Hz.
Electronics 2020, 9, 1355 11 of 20

Table 2. Parameters of the three controllers.

Controllers Parameters
Backstepping control K1 = 10diag([2, 2, 1.5]), K2 = 30diag([1.5, 1.5, 0.5]);
LESOBC K1 = 10diag([2, 2, 1.5]), K2 = 30diag([1.5, 1.5, 0.5]); κ0 = 200
K1 = 10diag([2, 2, 1.5]), K2 = 30diag([1.5, 1.5, 0.5]); κ0 = 200,
Proposed control  h iT  h iT
ka1 (t) = 0.5e−0.6t + 0.02 1 1 1 , kb1 (t) = − 0.6e−0.6t + 0.04 1 1 1

4.2.1. The First Simulation Case

Trajectories in the Cartesian coordinate with frequency at 0.1 Hz are generated for the 3-DOF
planar manipulator. These references in the joint space are computed through the inverse-kinematic
equations of the manipulator, which are presented in Figure 4. The references of joint 1, joint 2,
and joint 3 are exhibited by a dashed black line, a dashed red line, and a dashed blue line, respectively.
Figure 5 presents the output responses of the planar robot at the joint spaces. The results of the
reference, backstepping control, LESOBC, and the proposed control are exhibited by dashed black lines,
black lines, dashed dot black lines, and red lines, respectively. The boundaries of the output responses
are presented by dashed dot black lines. From these results, we easily realize that the output responses
of the backstepping control transgressed the output constraints. In order to exhibit the effectiveness of
the proposed method, the differences between the references and output responses are presented in
Figure 6. These results prove that the LESOBC and proposed control guarantee the satisfaction of the
output responses with the output constraints under the presence of an unknown friction and external
disturbance which arises at the 15th second.
The responses of the lumped uncertainties, x3 , at three joints are illustrated in Figure 7a with a
dashed blue line, dashed dot black line, and dashed red line for joint 1, joint 2, and joint 3, respectively.
In the first 15 s, the lumped uncertainties are unknown frictions at the joint space. In the last 15 s,
an external
Electronics force
2020, is applied
9, x FOR at the end-effector along the x-axis. As a result, the uncertainties in
PEER REVIEW 12each
of 22
joint include not only the unknown friction but also the external disturbance. Therefore, they increased
significantly and reduced
lines, black lines, dashed thedotaccuracy
black lines,of the
andbackstepping control. Figure
red lines, respectively. The 7b presents the
boundaries estimated
of the output
lumped
responses disturbance,
are presented x̂3 , which is thedot
by dashed result of the
black LESO.
lines. From Comparing to the
these results, welumped disturbance
easily realize in
that the
Figure
output7a, the results
responses of the
of the estimated control
backstepping lumpedtransgressed
disturbance thein Figure
output7b proved that
constraints. In the LESO
order has a
to exhibit
limited bandwidth,of
the effectiveness which
the means it cannot
proposed approximate
method, the disturbance
the differences betweenat high
the frequency.
referencesTheandaccuracy
output
of the LESOBC
responses and proposed
are presented control6.is These
in Figure improved significantly
results prove that by the
using LESO toand
LESOBC estimate the lumped
proposed control
disturbance.
guarantee theThe estimated
satisfaction ofresults of the
the output LESO are
responses presented
with in Figure
the output 8. The
constraints estimated
under errors in
the presence of
position,
an unknown velocity, andand
friction lumped disturbance
external are bounded.
disturbance which arises at the 15th second.

Figure 4. References
Figure 4. References of
of the
the 3-DOF
3-DOF planar
planar manipulator
manipulator in
in joint
jointspace
spaceat
at0.1
0.1Hz.
Hz.

100

80

60
t 1 (Deg.)

t 2 (Deg.)

40

20
 Figure 4. References of the 3-DOF planar manipulator in joint space at 0.1 Hz.
Electronics 2020, 9, 1355 12 of 20

(a) (b)

Electronics 2020, 9, x FOR PEER REVIEW (c) 13 of 22

Figure
Figure 5. Output
5. Output responses
responses of the
of the threecontrollers
three controllers in
in (a)
(a) joint
joint1;1;(b)
(b)joint
joint2; 2;
and (c)(c)
and joint 3 at30.1
joint at Hz.
0.1 Hz.

(a) (b)

(c)
Figure
Figure 6. Error
6. Error responses
responses of the
of the threecontrollers
three controllers in
in (a)
(a) joint
joint1;1;(b)
(b)joint
joint2; 2;
and (c)(c)
and joint 3 at30.1
joint at Hz.
0.1 Hz.

The responses of the lumped uncertainties, x 3 , at three joints are illustrated in Figure 7a with a
dashed blue line, dashed dot black line, and dashed red line for joint 1, joint 2, and joint 3,
respectively. In the first 15 s, the lumped uncertainties are unknown frictions at the joint space. In the
last 15 s, an external force is applied at the end-effector along the x-axis. As a result, the uncertainties
Electronics 2020, 9, x FOR PEER REVIEW 14 of 22

Electronics 2020, 9, x FOR PEER REVIEW 14 of 22

Electronics 2020, 9, 1355 13 of 20

(a) (b)
Figure 7. Lumped disturbance response of the planar manipulator at 0.1 Hz: (a) lumped disturbance
(a) (b)
and (b) estimated lumped disturbance.
Figure7.7.Lumped
Figure Lumpeddisturbance
disturbanceresponse
responseofofthe
theplanar
planarmanipulator
manipulatoratat0.1
0.1Hz:
Hz:(a)
(a)lumped
lumpeddisturbance
disturbance
and
and(b)
(b)estimated
estimatedlumped
lumpeddisturbance.
disturbance.

(a) (b)

(a) (b)

(c)
Figure8.8. Estimated
Figure Estimatederror
errorresults
resultsof
ofthe
theextended
extendedstate
stateobserver
observerat
at0.1
0.1Hz:
Hz: (a)
(a) position
position error,
error,(b)
(b)velocity
velocity
(c)
error,
error,and
and(c)
(c)lumped
lumpeduncertainties
uncertaintieserrors.
errors.
Figure 8. Estimated error results of the extended state observer at 0.1 Hz: (a) position error, (b) velocity
Figure
Figure 99 respectively
error, and respectively presents
presents the
(c) lumped uncertainties the torque
torque responses
errors. responses ofof the
the BC,
BC, LESOBC,
LESOBC, and and the
the proposed
proposed
control.
control.The
Thecontrol
controlresponses
responsesofofthe BCBC
the areare
presented in Figure
presented 9a when
in Figure 9a whenthe the
input dead-zones
input dead-zonesare not
are
compensated in the BC. In the LESOBC and proposed control, the torque responses are
Figure 9 respectively presents the torque responses of the BC, LESOBC, and the proposeddifferent from the
BC because
control. thecontrol
The input dead-zones
responses ofinthe
twoBCcontrollers are compensated
are presented in Figure 9aby the LESO.
when Finally,
the input in order are
dead-zones to
 Electronics 2020, 9, x FOR PEER REVIEW 15 of 22

Electronics 2020, 9, 1355 14 of 20

not compensated in the BC. In the LESOBC and proposed control, the torque responses are different
from the BC because the input dead-zones in two controllers are compensated by the LESO. Finally,
in order
evaluate to evaluate the
the effectiveness of effectiveness
the proposedofcontrol
the proposed control
in detail, in detail,
the root meanthe root mean
square errorsquare
(RMSE)error
is used
(RMSE) is used for evaluating the responses of the three controllers. The results
for evaluating the responses of the three controllers. The results are shown in Table 3. are shown in Table 3.

(a) (b)

(c)

FigureFigure 9. Control
9. Control signals
signals ofofthe
theproposed
proposed control
controlatat
0.10.1
Hz:Hz:
(a) backstepping control;control;
(a) backstepping (b) LESOBC; and
(b) LESOBC;
(c) proposed control.
and (c) proposed control.

3. The
TableTable rootroot
3. The mean square
mean error
square (RMSE)
error (RMSE)for
forthe
the tracking errorsofofthe
tracking errors the manipulator
manipulator at Hz.
at 0.1 0.1 Hz.

Controllers 1st(Deg)
1st Joint Joint 2nd Joint
2nd Joint (Deg) 3rd Joint
3rd Joint (Deg)
Controllers
(Deg) (Deg) (Deg)
Backstepping control 24.5205 47.3142 43.4415
Backstepping control 24.5205 47.3142 43.4415
LESOBC 0.0055 0.008 0.0038
LESOBC
Proposed control 0.0055
0.0001 0.008
0.0001 0.0038
0.0001
Proposed control 0.0001 0.0001 0.0001

4.2.2. 4.2.2.
The Second Simulation
The Second Case
Simulation Case
The trajectories in Cartesian
The trajectories coordinate
in Cartesian withwith
coordinate frequency at 0.5
frequency Hz Hz
at 0.5 are are
applied forfor
applied thethe
3-DOF
3-DOFplanar
planar manipulator.
manipulator. Similarly, compared
Similarly, compared to the previous
to the previous simulation,
simulation, the trajectories
the trajectories in the
in the joint
joint space
space of the
of the manipulator are calculated by using the inverse-kinematic equations. Figure 10 presents
manipulator are calculated by using the inverse-kinematic equations. Figure 10 presents the output error the
output
responses of error
three responses
controllersofatthree
threecontrollers
joints. Theat output
three joints. The outputoferror
error responses responses of the
the backstepping control,
backstepping control, LESOBC, and proposed control are respectively presented by black lines,
LESOBC, and proposed control are respectively presented by black lines, dashed dot blue lines, and red
dashed dot blue lines, and red lines. Additionally, the upper and lower error boundaries are plotted
lines. Additionally, the upper and lower error boundaries are plotted by the dashed dot black lines.
by the dashed dot black lines. The results in this figure show that the backstepping control still
The results in this figure show that the backstepping control still transgresses the constraints and the
LESOBC begins breaking the output constraints because the accuracy of the LESOBC is improved by the
LESO and it does not depend on the output boundaries. The proposed control is developed by integrating
the output constraints into the control design. Therefore, its output responses are still guaranteed, although
the working frequency increases. With this simulation, we see the effectiveness of the BLF in tackling the
 Electronics 2020, 9, x FOR PEER REVIEW 16 of 22

transgresses the constraints and the LESOBC begins breaking the output constraints because the
accuracy of the LESOBC is improved by the LESO and it does not depend on the output boundaries.
The proposed control is developed by integrating the output constraints into the control design.
Electronics 2020, 9, 1355 15 of 20
Therefore, its output responses are still guaranteed, although the working frequency increases. With
this simulation, we see the effectiveness of the BLF in tackling the output constraints. The lumped
uncertainties, in this case, are the unknown frictions in the first 15 s and a combination of the
output constraints. The lumped uncertainties, in this case, are the unknown frictions in the first 15 s and a
unknown frictions and external disturbances in the last 15 s.
combination of the unknown frictions and external disturbances in the last 15 s.

(a) (b)

(c)
Figure
Figure 10. Error
10. Error responses
responses of three
of three controllersin:
controllers in: (a)
(a) joint
joint 1;
1;(b)
(b)joint
joint2;2;and (c)(c)
and joint 3 at30.5
joint at Hz.
0.5 Hz.

The lumped
The lumped disturbance
disturbance responses,x3x, 3are
responses, presented
, are presentedin
in Figure 11a.Because
Figure 11a. Becausethethe frequency
frequency of of
the trajectory is increased from 0.1 Hz to 0.5 Hz and the lumped uncertainties concern
the trajectory is increased from 0.1 Hz to 0.5 Hz and the lumped uncertainties concern the velocity,the velocity,
Electronics 2020, 9, x FOR PEER REVIEW 17 of 22
the amplitude
the amplitudeof the
of lumped disturbances
the lumped disturbancesincreased
increased if we compare
if we compareit itwith
withthethe previous
previous simulation.
simulation.
Figure 11b derives the estimated lumped disturbance, x̂3x̂, of the LESO.
Figure 11b derives the estimated lumped disturbance, 3 , of the LESO.

(a) (b)
Figure 11.Lumped
Figure11. Lumpeddisturbance
disturbanceresponse
responseofofthe
theplanar
planarmanipulator
manipulatoratat0.5
0.5Hz:
Hz:(a)
(a)lumped
lumpeddisturbance
disturbance
and (b) estimated lumped disturbance.
and (b) estimated lumped disturbance.

The effectiveness of the LESO is presented in Figure 12, with the estimated error responses of
the position, the velocity, and the lumped disturbance. The torque signals of the BC, LESOBC, and
proposed control are exhibited in Figure 13. In the LESOBC and the proposed control, the control
responses are also different from the BC because the input dead-zones are overcome by the LESO.
 Electronics 2020, 9, 1355 16 of 20

The effectiveness of the LESO is presented in Figure 12, with the estimated error responses of the
(a)
position, the velocity, and the lumped disturbance. The torque signals of the(b) BC, LESOBC, and proposed
control are11.
Figure exhibited
Lumpedin Figure 13.response
disturbance In the LESOBC andmanipulator
of the planar the proposed control,
at 0.5 Hz: (a)the control
lumped responses are
disturbance
alsoand
different from the BC because
(b) estimated lumped disturbance. the input dead-zones are overcome by the LESO. Finally, in order to
evaluate the effectiveness of the proposed control in detail, the root mean square error (RMSE) is used for
evaluating the responses
The effectiveness of of
thethe threeiscontrollers.
LESO presented The results12,
in Figure arewith
shown theinestimated
Table 4. error responses of
In summary,
the position, the proposed
the velocity, and thecontroller that integrates
lumped disturbance. Thethe BLF-backstepping
torque signals of the BC,control and ESO
LESOBC, andis
effective to
proposed control
control aretheexhibited
manipulator at different
in Figure 13. Inmotion frequencies
the LESOBC (0.1proposed
and the Hz and 0.5 Hz) under
control, a bunch
the control
of problems
responses areasalso
unknown
different friction, external
from the disturbance,
BC because and unknown
the input dead-zones input
are dead-zone.
overcome by The proposed
the LESO.
controller outperforms the ESOBC and BC in terms of the RMSE as mentioned
Finally, in order to evaluate the effectiveness of the proposed control in detail, the root mean square in Tables 3 and 4,
and especially
error (RMSE) is entirely
used fordominates
evaluatingthe theESOBC
responsesat low frequency
of the (0.1 Hz) and
three controllers. Thethe BC atare
results both low and
shown in
high 4.
Table frequency, with respect to the prescribed-constraint performance as shown in Figures 6 and 10.

Electronics 2020, 9, x FOR PEER REVIEW 18 of 22

(a) (b)

(c)
12. Estimated
Figure 12. Estimatederror
errorresults
results of the
of the extended
extended statestate observer
observer at 0.5 at
Hz:0.5(a)Hz: (a) position
position error, (b)error, (b)
velocity
velocity
error, anderror, and (c)uncertainties
(c) lumped lumped uncertainties
errors. errors.

Table 4. RMSE for the tracking errors of the manipulator at 0.5 Hz.

Controllers 1st Joint (Deg) 2nd Joint (Deg) 3rd Joint (Deg)
Backstepping control 30.4138 52.7901 115.0904
ESOBC 0.2315 0.1260 0.6836
Proposed control 0.0159 0.0179 0.0197
 (c)
Figure 12. Estimated error results of the extended state observer at 0.5 Hz: (a) position error, (b)
velocity error, and (c) lumped uncertainties errors.
Electronics 2020, 9, 1355 17 of 20

(a) (b)

(c)
Figure 13.Control
Figure13. Controlsignals
signalsof
ofthe
theproposed
proposedcontrol
controlatat0.5
0.5Hz:
Hz:(a)
(a)backstepping
backsteppingcontrol;
control;(b)
(b)LESOBC;
LESOBC;
and
and(c)
(c)proposed
proposedcontrol.
control.

5. Conclusions
This paper proposed an output feedback control via an extended state observer for an n-DOF
robotic manipulator under the presence of unknown friction, external disturbances, input dead-zone,
and the time-varying output constraints. These challenges are presented with n-DOF manipulator
dynamics in mathematical equations. The proposed controller was developed from the LESO and
the barrier Lyapunov function with the backstepping framework. The LESO estimated both the
lumped disturbance and the unmeasured states in the robotic manipulator. Additionally, the BLF
guaranteed that the output responses avoid violation of the constraints. Next, the Lyapunov approach
was theoretically conducted to analyze the stability and robustness of the proposed control of the
manipulator. Some simulations were conducted on the 3-DOF planar manipulator. The comparative
results between the proposed control and the other controllers, such as backstepping control and
the LESOBC, prove the superiority of the proposed control in improving accuracy against the
lumped disturbances.
In future works, some advanced control can be developed from this algorithm to deal with other
challenges such as finite-time convergence, the chattering effect, and input constraints, besides the
output constraints. Some adaptive approximations can be investigated in this algorithm when the
LESO is used as a fault detector or a force estimator for the manipulator.
Electronics 2020, 9, 1355 18 of 20

Author Contributions: Conceptualization, D.T.T., and K.K.A.; methodology, D.T.T.; software, H.V.D. and
D.T.T.; validation, D.T.T. and K.K.A.; formal analysis, D.V.H. and D.T.T.; investigation, D.T.T.; resources, D.T.T.;
data curation, D.T.T.; writing—original draft preparation, D.T.T.; writing—review and editing, D.T.T., D.V.H.,
and T.Q.D.; visualization, D.T.T.; supervision, K.K.A.; project administration, K.K.A.; funding acquisition, K.K.A.
All authors have read and agreed to the published version of the manuscript.
Funding: This research was supported by the Basic Science Program through the National Research Foundation
of Korea (NRF), funded by the Ministry of Science and ICT, South Korea (NRF-2020R1A2B5B03001480) and by
the Ministry of Trade, Industry & Energy (MOTIE, Korea) under the Industrial Technology Innovation Program
(No.10067184).
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A
The inertia matrix, the Coriolis and centrifugal term matrix, and the gravity vector of the 3-DOF
planar manipulator are presented as follows:
     
 M11 M12 M13     C11 C12 C13   G1 
, C q, q. =  C21
     
M(q) =  M21 M22 M23   C22 C23 , G(q) =  G2 
   (A1)

M31 M32 M33 C31 C32 C33 G3
    

M11 = l2 (m3 s3 (s3 (l2 + l1 c2 ) + l1 c3 s2 ) +m3 c3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 )) + l1 (s2 (l1 m2 s2 + m3 c3 (s3 (l2 + l1 c2 ) + l1 c3 s2 )
−m3 s3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 )) + c2 (m2 (l2 + l1 c2 ) + m3 s3 (s3 (l2 + l1 c2 ) + l1 c3 s2 ) +m3 c3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 ))) + l21 m1 (A2)
+l3 m3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 ) + l2 m2 (l2 + l1 c2 )

M12 = l2 (m3 s3 (s3 (l2 + l1 c2 ) + l1 c3 s2 ) +m3 c3 (l3 + c2 (l2 + l1 c2 ) − l1 s2 s3 )) + l3 m3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 ) + l2 m2 (l2 + l1 c2 ) (A3)
 
M13 = l3 m3 (l3 + c3 (l2 + l1 c2 ) − l1 s2 s3 ); M21 = M12 ; M22 = l22 m2 + l2 l2 m3 s23 + m3 c3 (l3 + l2 c3 ) + l3 m3 (l3 + l2 c3 ) (A4)

M23 = l3 m3 (l3 + l2 c3 ); M31 = M13 ; M32 = M23 ; M33 = l23 m3 (A5)

. . . .
C11 = −x2 l1 l2 m3 s23 − x3 l1 l3 m3 s23 − x2 l1 l2 m3 s2 − x3 l2 l3 m3 s3 (A6)
. . . . . .
C12 = −x1 l1 l3 m3 s23 − x2 l1 l3 m3 s23 − x3 l1 l3 m3 s23 − x1 l1 l2 (m2 + m3 )s2 − x2 l1 l2 (m2 + m3 )s2 − x3 l2 l3 m3 s3 (A7)
. . .  . . . .
C13 = −l3 m3 (l1 s23 + l2 s3 ) x1 + x2 + x3 ; C21 = x1 l1 l3 m3 s23 + x1 l1 l2 m2 s2 + x1 l1 l2 m3 s2 − x3 l2 l3 m3 s3 (A8)
. . . .  . . .  . . 
C22 = −x3 l2 l3 m3 s3 C23 = −l2 l3 m3 s3 x1 + x2 + x3 ; C31 = l3 m3 x1 l2 s3 + x2 l2 s3 + x1 l1 s23 ; C32 = l2 l3 m3 s3 x1 + x2 (A9)

C33 = 0; G1 = g(l1 m1 c1 + l1 m2 c1 + l1 m3 c1 + l3 m3 c123 + l2 m2 c12 + l2 m3 c12 ); G2 = g(l3 m3 c123 + l2 m2 c12 + l2 m3 c12 ) (A10)

G3 = gl3 m3 c123 (A11)

The Jacobian matrix of the robot is expressed by

 
 −l1 s1 − l2 s12 − l3 s123 −l2 s12 − l3 s123 −l3 s123 
 
J =  l1 c1 + l2 c12 + l3 c123 l2 c12 + l3 c123 l3 c123 ,
 (A12)

1 1 1


where    
si = sin(qi ), sij = sin qi + q j , sijk = sin qi + q j + qk
   
ci = cos(qi ), cij = cos qi + q j , cijk = cos qi + q j + qk (A13)
 
i, j, k = 1, 3

References
1. Baek, J.; Kwon, W.; Kim, B.; Han, S. A Widely adaptive time-delayed control and its application to robot
manipulators. IEEE Trans. Ind. Electron. 2018, 66, 5332–5342. [CrossRef]
2. Craig, J.J. Introduction to Robotics: Mechanics and Control; Pearson Prentice Hall: Upper Saddle River, NJ, USA,
2005; Volume 3.
Electronics 2020, 9, 1355 19 of 20

3. Tran, D.T.; Truong, H.V.A.; Ahn, K.K. Adaptive backstepping sliding mode control based rbfnn for a hydraulic
manipulator including actuator dynamics. Appl. Sci. 2019, 9, 1265. [CrossRef]
4. Truong, H.V.A.; Tran, D.T.; To, X.D.; Ahn, K.K.; Jin, M. Adaptive fuzzy backstepping sliding mode control for
a 3-DOF hydraulic manipulator with nonlinear disturbance observer for large payload variation. Appl. Sci.
2019, 9, 3290. [CrossRef]
5. Tran, D.T.; Ba, D.X.; Ahn, K.K.; Thien, T.D. Adaptive backstepping sliding mode control for equilibrium
position tracking of an electrohydraulic elastic manipulator. IEEE Trans. Ind. Electron. 2020, 67, 3860–3869.
[CrossRef]
6. Duc-Thien, T. Adaptive sliding mode control with backstepping technique for hydraulic manipulator.
In Proceedings of the 33rd Institute of Control Robotics and Systems, Buan, Korea, 17 May 2018.
7. Jung, S. Improvement of tracking control of a sliding mode controller for robot manipulators by a neural
network. Int. J. Control. Autom. Syst. 2018, 16, 937–943. [CrossRef]
8. Chen, K.Y. Robust Optimal Adaptive Sliding Mode Control with the Disturbance Observer for a Manipulator
Robot System. Int. J. Control. Autom. Syst. 2018, 16, 1701–1715. [CrossRef]
9. Nikdel, N.; Badamchizadeh, M.; Azimirad, V.; Nazari, M. Adaptive backstepping control for an n-degree
of freedom robotic manipulator based on combined state augmentation. Robot. Comput. Manuf. 2017, 44,
129–143. [CrossRef]
10. Chang, W.; Li, Y.; Tong, S. Adaptive fuzzy backstepping tracking control for flexible robotic manipulator.
IEEE/CAA J. Autom. Sin. 2018, 1–9. [CrossRef]
11. Park, S.; Lee, H.; Han, S.; Lee, J. Adaptive Fuzzy Super-twisting Backstepping Control Design for MIMO
Nonlinear Strict Feedback Systems. Int. J. Control. Autom. Syst. 2018, 16, 1165–1178. [CrossRef]
12. Lv, W.; Wang, F.; Zhang, L. Adaptive fuzzy finite-time control for uncertain nonlinear systems with dead-zone
input. Int. J. Control. Autom. Syst. 2018, 16, 2549–2558. [CrossRef]
13. He, W.; Huang, B.; Dong, Y.; Li, Z.; Su, C.Y. Adaptive neural network control for robotic manipulators with
unknown deadzone. IEEE Trans. Cybern. 2017, 48, 2670–2682. [CrossRef] [PubMed]
14. Lin, C.H. Nonlinear backstepping control design of LSM drive system using adaptive modified recurrent
laguerre orthogonal polynomial neural network. Int. J. Control. Autom. Syst. 2017, 42, 494–917. [CrossRef]
15. Lin, C.H.; Ting, J.C. Novel nonlinear backstepping control of synchronous reluctance motor drive system for
position tracking of periodic reference inputs with torque ripple consideration. Int. J. Control. Autom. Syst.
2019, 17, 1–17. [CrossRef]
16. Yi, G.; Mao, J.; Wang, Y.; Guo, S.; Miao, Z. Adaptive tracking control of nonholonomic mobile manipulators
using recurrent neural networks. Int. J. Control. Autom. Syst. 2018, 16, 1390–1403. [CrossRef]
17. Yao, J.; Jiao, Z.; Ma, D. Extended-state-observer-based output feedback nonlinear robust control of hydraulic
systems with backstepping. IEEE Trans. Ind. Electron. 2014, 61, 6285–6293. [CrossRef]
18. Liu, J.; Gai, W.; Zhang, J.; Li, Y. Nonlinear adaptive backstepping with ESO for the quadrotor trajectory
tracking control in the multiple disturbances. Int. J. Control. Autom. Syst. 2019, 17, 2754–2768. [CrossRef]
19. Wang, J.; Hung, J.Y. Adaptive Backstepping Control for an Underwater Vehicle Manipulator System Using
Fuzzy Logic. In Proceedings of the IECON 2018—44th Annual Conference of the IEEE Industrial Electronics
Society, Washington, DC, USA, 21–23 October 2018.
20. Wu, Y.; Huang, R.; Li, X.; Liu, S. Adaptive neural network control of uncertain robotic manipulators with
external disturbance and time-varying output constraints. Neurocomputing 2019, 323, 108–116. [CrossRef]
21. Han, J. From PID to active disturbance rejection control. IEEE Trans. Ind. Electron. 2009, 56, 900–906. [CrossRef]
22. Tran, D.T.; Do, T.C.; Ahn, K.K. Extended high gain observer-based sliding mode control for an electro-hydraulic
system with a variant payload. Int. J. Precis. Eng. Manuf. 2019, 20, 2089–2100. [CrossRef]
23. Jun, G.H.; Ahn, K.K. Extended-state-observer-based nonlinear servo control of an electro-hydrostatic actuator.
J. Drive Control 2017, 14, 61–70.
24. Chen, H.T.; Song, S.M.; Zhu, Z.B. Robust Finite-time Attitude Tracking Control of Rigid Spacecraft Under
Actuator Saturation. Int. J. Control. Autom. Syst. 2018, 16, 1–15. [CrossRef]
25. Mario, R.N.; Hebertt, S.R.; Rubén, G.M.; Alberto, L.J. Active Disturbance Rejection Control of the Inertia
Wheel Pendulum through a Tangent Linearization Approach. Int. J. Control. Autom. Syst. 2019, 17, 18–28.
[CrossRef]
26. Zhao, Y.; Yu, J.; Tian, J. Robust output tracking control for a class of uncertain nonlinear systems using
extended state observer. Int. J. Control. Autom. Syst. 2017, 15, 1227–1235. [CrossRef]
Electronics 2020, 9, 1355 20 of 20

27. Ren, B.; Ge, S.S.; Tee, K.P.; Lee, T.H. Adaptive neural control for output feedback nonlinear systems using a
barrier lyapunov function. IEEE Trans. Neural Netw. 2010, 21, 1339–1345. [CrossRef]
28. Li, Y.; Yang, C.G.; Yan, W.; Cui, R.; Annamalai, A. Admittance-based adaptive cooperative control for multiple
manipulators with output constraints. IEEE Trans. Neural Netw. Learn. Syst. 2019, 30, 3621–3632. [CrossRef]
29. Wang, C.; Wu, Y.; Yu, J. Barrier Lyapunov functions-based adaptive control for nonlinear pure-feedback
systems with time-varying full state constraints. Int. J. Control. Autom. Syst. 2017, 15, 2714–2722. [CrossRef]
30. He, W.; David, A.O.; Yin, Z.; Sun, C. Neural network control of a robotic manipulator with input deadzone
and output constraint. IEEE Trans. Syst. Man Cybern. Syst. 2016, 46, 759–770. [CrossRef]
31. Guo, Q.; Liu, Y.; Wang, Q.; Jiang, D. Adaptive neural network control of Two-DOF robotic arm driven by
electro-hydraulic actuator with output constraint. In Proceedings of the IET Conference, Guiyang, China,
19–22 June 2018; p. 7.
32. Zhou, Q.; Wang, L.; Wu, C.; Li, H.; Du, H. Adaptive fuzzy control for nonstrict-feedback systems with input
saturation and output constraint. IEEE Trans. Syst. Man Cybern. Syst. 2017, 47, 1–12. [CrossRef]
33. Edalati, L.; Sedigh, A.K.; Shooredeli, M.A.; Moarefianpour, A. Adaptive fuzzy dynamic surface control of
nonlinear systems with input saturation and time-varying output constraints. Mech. Syst. Signal. Process.
2018, 100, 311–329. [CrossRef]
34. Mien, V.; Mavrovouniotis, M.; Ge, S.S. An adaptive backstepping nonsingular fast terminal sliding mode
control for robust fault tolerant control of robot manipulators. IEEE Trans. Syst. Man Cybern. Syst. 2018, 49,
1448–1458. [CrossRef]
35. Yang, H.; Sun, J.; Xia, Y.; Zhao, L. Position control for magnetic rodless cylinders with strong static friction.
IEEE Trans. Ind. Electron. 2018, 65, 5806–5815. [CrossRef]
36. Tran, D.T.; Jin, M.; Ahn, K.K. Nonlinear extended state observer based on output feedback control for a
manipulator with time-varying output constraints and external disturbance. IEEE Access 2019, 7, 156860–156870.
[CrossRef]
37. Guo, Q.; Zhang, Y.; Celler, B.G.; Su, S. Backstepping control of electro-hydraulic system based on
extended-state-observer with plant dynamics largely unknown. IEEE Trans. Ind. Electron. 2016, 63, 6909–6920.
[CrossRef]
38. Yu, J.; Zhao, L.; Yu, H.; Lin, C. Barrier Lyapunov functions-based command filtered output feedback control
for full-state constrained nonlinear systems. Automatica 2019, 105, 71–79. [CrossRef]
39. Huang, A.C.; Chien, M.C. Adaptive Control of Robot Manipulators: A Unified Regressor-Free Approach; World Scientific:
Singapore, 2010.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).