1056 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO.

2, APRIL 2020

USDE-Based Sliding Mode Control for Servo

Mechanisms With Unknown System Dynamics
Shubo Wang , Member, IEEE, Liang Tao , Qiang Chen , Jing Na , Member, IEEE,
and Xuemei Ren

Abstract—This article proposes an unknown system dy- id d-axis stator current.

namics estimator (USDE) based sliding mode control for iq q-axis stator current.
servo mechanisms with unknown dynamics and modeling ud d-axis stator voltage.
uncertainties. An invariant manifold is first constructed by
introducing an auxiliary variable based on a first-order low- uq q-axis stator voltage.
pass filter. This is used to design a USDE with only one np Number of pole pairs.
tuning parameter (i.e., time constant for the filter) and a Tl Load torque.
simpler structure than other estimators. The USDE is used KT Torque time constant.
to compensate for the effect of the lumped unknown system Td Load disturbance.
dynamics since it can be easily incorporated into control
synthesis. Moreover, to avoid the chattering phenomenon Tf Friction torque.
in the conventional sliding mode control methods, a novel ψf Rotor flux linkage.
reaching law is designed based on hyperbolic functions q Angular position.
to guarantee that the sliding mode variable infinitely ap- q̇ Angular speed.
proaches to the equilibrium point instead of crossing it.
Consequently, the fast convergence and chattering-free
property can be achieved simultaneously. Simulations and
I. INTRODUCTION
experiments are provided to validate the effectiveness and ERVO mechanisms have been widely applied in many
superior performance of the proposed method.
Index Terms—Chattering-free property, servo mecha-
S industrial fields [1]–[4]. However, the unknown system
dynamics (e.g., nonlinear frictions, disturbances, and parameter
nisms, sliding mode control (SMC), unknown system dy-
namics estimator (USDE).
uncertainties) introduced by the transmission devices can reduce
the system performance. To overcome this issue, PID control
NOMENCLATURE was adopted to control servo mechanisms [5]. However, PID
J Motor inertia. control with fixed gains cannot guarantee satisfactory control
performance due to the nonlinear dynamics in the servo mecha-
nisms. With the development of control theories and techniques,
Manuscript received September 17, 2018; revised May 14, 2019 various advanced control methods have been developed to deal
and August 29, 2019; accepted January 24, 2020. Date of publica-
tion February 6, 2020; date of current version April 15, 2020. Rec- with the nonlinear dynamics in the systems, such as adaptive
ommended by Technical Editor M. O. Efe. This work was supported control [6]–[10], dynamic surface control [11], [12], and sliding
in part by the National Natural Science Foundation of China under mode control (SMC) [13]–[16]. Moreover, neural network [17]–
Grant 61803216, Grant 61973274, and Grant 61922037, in part by the
China Post-Doctoral Science Foundation under Grant 2019M650159, [24] and fuzzy logic system [25]–[29] have also been utilized
in part by the Natural Science Foundation of Shandong Province under to approximate the unknown nonlinearities due to their learning
Grant ZR2018BF022, and in part by Zhejiang Provincial Natural Science and approximation ability.
Foundation under Grant LY17F030018. (Corresponding author: Qiang
Chen.) Among these control methods, SMC has been widely used for
S. Wang is with the School of Automation, Qingdao University, Qing- nonlinear control systems due to its simple algorithm and high
dao 266071, China (e-mail: wangshubo1130@126.com). robustness to disturbances [30]–[33]. However, the chattering
L. Tao is with the College of Information Engineering, Zhejiang Univer-
sity of Technology, Hangzhou 310023, China, and also with the College phenomenon in the conventional SMC cannot be completely
of Electrical Engineering, Anhui Polytechnic University, Wuhu 241000, eliminated. Recently, by using a new-type sliding surface design,
China (e-mail: taoliang@ahpu.edu.cn). the nonsingular terminal SMC (NTSMC) has been applied to de-
Q. Chen is with the College of Information Engineering, Zhe-
jiang University of Technology, Hangzhou 310023, China (e-mail: rive some new-type SMC methods to achieve faster convergence
sdnjchq@zjut.edu.cn). time [34]–[36], i.e., nonsingular fast TSMC (NFTSMC). The
J. Na is with the Faculty of Mechanical and Electrical Engineering, main feature is that the NFTSMC can retain the advantages of
Kunming University of Science and Technology, Kunming 650500, China
(e-mail: najing25@163.com). the NTSMC and provide a faster control convergence than that
X. Ren is with the School of Automation, Beijing Institute of Technol- of the NTSMC. However, in practical industry applications, the
ogy, Beijing 100081, China (e-mail: xmren@bit.edu.cn). unknown disturbance and nonlinearities may be large, and thus,
Color versions of one or more of the figures in this article are available
online at https://ieeexplore.ieee.org. the SMC solely may not effectively compensate such external
Digital Object Identifier 10.1109/TMECH.2020.2971541 disturbances, due to the serious chattering problem induced in

1083-4435 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: USDE-BASED SLIDING MODE CONTROL FOR SERVO MECHANISMS WITH UNKNOWN SYSTEM DYNAMICS 1057

the control. Therefore, it is necessary to design an observer to

estimate and compensate the unknown disturbance to reduce the
chattering issue in the SMC schemes.
The disturbance observer (DOB) techniques have also been
used to address the unknown disturbances. Following this idea,
many DOBs-based control approaches have been applied in
some control systems, such as servo mechanisms [37] and
robotic systems [38]. Among these results, various DOBs have
been proposed in the literatures [39]–[43]. Specifically, a new
idea called extended state observer (ESO) that has a similar Fig. 1. Schematic diagram of motor drive servo mechanisms.
property as DOB was developed in [44]. This technique has
also been used in some control systems to handle unknown
dynamics [45]–[49]. Recently, a new uncertainty and distur- II. PROBLEM FORMULATION
bance estimator (UDE) [50] was also proposed to estimate the In this article, we consider the servo mechanism driven by
lumped unknown system dynamics. Subsequently, the UDE- a permanent magnet synchronous motor (PMSM) as shown in
based control methods have been successful applied in some Fig. 1, which can be modeled as
fields, such as inverters control [51], rotary servo system [52],
⎛ ⎞ ⎛ −R ⎞⎛ ⎞
and two-degrees-of-freedom nature [53], which also show the i̇d L −np q̇ 0 id
satisfactory performance. ⎜ ⎟ ⎜ np ψf ⎟ ⎜ ⎟
⎝ i̇q ⎠ = ⎝ −np q̇ −R ⎠ ⎝ iq ⎠
In this article, we provide an alternative control scheme for L L
nonlinear servo mechanisms with unknown system dynamics, q̈ 0 KT
0 q̇
J
which can specifically handle the chattering issue of the SMC ⎛ u
⎞
d
by incorporating a new estimator into the control design. For this L
⎜ uq ⎟
purpose, an unknown system dynamics estimator (USDE)-based +⎝ L ⎠. (1)
−Tf −Tl −Td
SMC is proposed. First, a USDE is developed by introducing J
auxiliary filtered variables to estimate the lumped unknown
system dynamics, where it has only one parameter (i.e., time In practice, the d- axis reference current i∗d is set to be i∗d = 0,
constant of the low-pass filters) to be tuned. The estimated in order to approximately eliminate the couplings between the
variable can be incorporated into the control as a feedforward angular velocity and current. If the controllers for the two current
term to compensate for the effect of these unknown system loops work well, the output satisfies id = i∗d = 0, and then (1)
dynamics. Moreover, to eliminate the chattering problem in the can be reduced to the following form [54]:
traditional SMC schemes, a new structural reaching law (RL) is  −R KE  uq
i̇q iq
designed by using hyperbolic functions. The stability of the con- = KT L L
+ −Tf −T L
l −Td
(2)
trol system and the convergence of control and estimation errors q̈ J 0 q̇ J
are all guaranteed. The control performance of the developed
where KE = np ψf .
control method is verified via both comparative simulations and
Choose a state variable as x = [x1 , x2 ]T = [q, q̇]T . Then, (2)
experiments.
can be written as
The main contributions of this article are listed as follows.
1) A USDE is proposed by defining auxiliary filtered vari- ẋ1 = x2
(3)
ables to estimate the unknown dynamics (e.g., nonlinear ẋ2 = J1 (K1 u + K2 x2 − Tl − Td − Tf )
friction, uncertainties, and external disturbances) of servo
mechanisms. The USDE is derived by constructing an where u = uq , K1 = KT /R and K2 = KT KE /R.
ideal invariant manifold with simple filter operations, so To facilitate the design of estimator, we rewrite (3) as
that only the filter gain needs to be selected, while the
ẋ1 = x2
convergence of the estimation error to a small compact (4)
set can be guaranteed. ẋ2 = b1 [u − F (x1 , x2 )]
2) A new structural RL is designed with hyperbolic functions where b = J/K1 , and F (x1 , x2 ) = (Tf + K2 x2 + Tl + Td )/
to guarantee fast convergence and chattering elimination K1 denotes the unknown system dynamics.
of the sliding mode variable. This is achieved by making Assumption 1 [56]: The unknown disturbance Td , nonlinear
the sliding mode variable infinitely approach to zero friction Tf , and load torque Tl are bounded, and the derivative
instead of crossing it. of F is bounded by supt≥0 |Ḟ | ≤ , where  > 0 is a constant.
This article is organized as follows. Section II presents the
problem formulation. Sections III and IV provide the design of
USDE and the controller, respectively. Sections V and VI show III. USDE DESIGN
the comparative simulations and experiments. Section VII gives In this section, an USDE is developed to estimate the lumped
some conclusions. unknown system dynamic F . For this purpose, the filtered

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
1058 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO. 2, APRIL 2020

variables x2f , uf of x2 , u are defined as which can be presented in the time domain (assuming zero initial
condition) as
κẋ2f + x2f = x2 , x2f (0) = 0
(5) 1
κu̇f + uf = u, uf (0) = 0 ėF = − ef + Ḟ . (14)
κ
where κ > 0 is the time constant of the filter shown in (5), which
From (13), one can find that the error eF can be small by
can be set as a small constant.
setting parameter κ small. Then, we can obtain the following
In the following, an invariance manifold [55] is employed to
lemma.
design an USDE. Then, we have the following lemma.
Lemma 2 [56]: For servo mechanism (4) with unknown sys-
Lemma 1 [56]: Consider (4) and (5), an auxiliary variable γ
tem dynamic estimator (10), the estimation error eF is bounded
is defined as
by
(x2 − x2f ) (uf − F ) 
γ= − (6)
κ b |eF (t)| ≤ e2F (0)e−t/k + k 2 2 (15)
is bounded and decreases exponentially for any finite κ > 0, and
 and thus, F → F̂ holds for k → 0 or  → 0.
(x2 − x2f ) (uf − F )
lim lim − =0 (7) Proof: Consider a Lyapunov function as VF = e2F /2, then
κ→0 t→∞ κ b the derivative of VF is calculated based on (14)
it means (x2 − x2f )/κ − (uf − F )/b = 0 is an invariant man- 1
ifold for κ > 0. V̇F = − e2F + eF Ḟ
k
Proof: The time derivative of γ is calculated as 
1 k
≤ − V F + 2 . (16)
(x2 − x2f ) 1  k 2
γ̇ = − u̇f − Ḟ √
κ b From (16), we can conclude that |eF (t)| ≤ e2F (0)+k2 2 , which
implies that eF (t) → 0 holds for k → 0 or  → 0. 
1 κ  As shown in Lemma 1, the USDE is derived by constructing
=− γ − Ḟ . (8)
κ b an ideal invariant manifold with simple filter operations, while
Select a Lyapunov function as Vγ = γ 2 /2, then the convergence of the estimation error can be guaranteed as
other estimators (e.g., DOB, ESO). Specifically, we only need
1 1 1 1 2 κ
V̇γ = γ γ̇ = − γ 2 + γ Ḟ ≤ − γ 2 + γ + 2 Ḟ 2 to set the filter gain κ in the USDE.
κ b κ 2κ 2b
1 κ IV. CONTROLLER DESIGN
≤ − V γ + 2 2 . (9)
κ 2b
The conventional SMC suffers the chattering problem, which
By solving (9), we can verify that Vγ (t) ≤ e−t/k Vγ (0) + can reduce the control accuracy. To eliminate the chattering phe-
κ  /2b2 . Therefore, the variable γ will converge exponentially
2 2
nomenon, we will develop a new structural RL with hyperbolic
√
to a small region as |γ(t)| ≤ Vγ2 (0)+κ2 2 /2b2 , and its ultimate functions, which can be used to design a control to achieve
bounded depends on the parameters κ and . motion tracking.
It is shown that the ideal invariant manifold can provide a
mapping from (x2f , uf ) to the unknown system dynamic F A. New Structural RL
without knowing any information of ẋ2 . Hence, based on (7),
the USDE is designed as The RL is constructed as

b(x2 − x2f ) ṡ = −k1 tanh(αs) − k2 |s| · asinh(βsp ) (17)

F̂ = uf − . (10)
κ where s denotes the sliding mode variable; k1 , k2 , α, and β
In order to analyze the convergence of estimator (10), a low- represent the positive constants; p > 0 denotes an odd inte-
 s
−e−s
pass filter κ+1 is applied on (4), such that we have ger; tanh(s) = ees +e −s is the hyperbolic tangent function, and
√
 1 1 1 1 asinh(s) = ln(s + s2 +1) is the inverse hyperbolic function.
[x2 ] = [u] − [F ]. (11) When the sliding mode variable s is far away from equi-
κ + 1 b κ + 1 b κ + 1
librium point, the convergence of s is mainly controlled by
Based on (5), one has −k2 |s| · asinh(βsp ), which makes s have a large convergence
x2 − x2f u f − Ff rate, since the value of −k2 |s| · asinh(βsp ) is much larger than
ẋ2f = = (12) −k1 tanh(αs). When the sliding mode variable s approaches to
κ b
equilibrium point, the term −k2 |s| · asinh(βsp ) is almost zero,
where Ff denotes the filtered variable of F , which is given by
and thus, s is mainly controlled by −k1 tanh(αs). The most
κḞf + Ff = F . Then it follows from (8) and (10) that F̂ = Ff .
remarkable feature of this RL is that the variation of s in the unit
Consequently, the estimation error eF is
 sampling time is less than the value of s when s enters a small
1 κ neighborhood of equilibrium point, i.e., s is close to equilibrium
eF = F − F̂ = 1 − F = [F ] (13)
κ + 1 κ + 1 point at each sampling time, but cannot reach equilibrium point.

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: USDE-BASED SLIDING MODE CONTROL FOR SERVO MECHANISMS WITH UNKNOWN SYSTEM DYNAMICS 1059

Fig. 3. Diagram of the proposed control system.

Fig. 2. Convergent example of the proposed RL.

Substituting (4) into (20), one can obtain that

Thus, the chattering issue encountered in the conventional RLs
1
can be effectively remedied. ṡ =(K1 u − F (x1 , x2 )) − ẍd + λ1 ė. (21)
To illustrate the convergence of the developed RL, an example J
is given. The parameters are selected as k1 = 40, k2 = 20, α = Then the controller u can be designed as
5, β = 10, and p = 3; the initial value of sliding mode variable J 
s is chosen as s0 = 50. Fig. 2 shows the simulation result. From u= −k1 tanh(αs) − k2 |s| · asinh(βsq ) − k3 s
K1
Fig. 2, one can clearly find that s converges to a very small value 
8.6 × 10−4 after only 0.06 s, and the chattering is avoided in the + F̂ + J ẍd − Jλ1 ė (22)
steady-state performance.
where k1 , k2 , and k3 are all positive constants set by the design-
Remark 1: The proposed RL (17) makes the sliding mode
ers.
variable s infinitely approach to equilibrium point but cannot
There are two kinds of parameters to be set by the designers.
cross it. Therefore, once s reaches an ideal neighborhood [0, δ],
For the proposed estimator, only the filter coefficient κ needs
s can be considered as practically convergent, and the conver-
sinh(a) to be set as explained in Section III, which can be selected as a
gent time is approximated by t = k11α ln[ sinh(αtanh(k s /k1 ))
]+
small positive constant to tradeoff the convergence rate and the
ln(plogb s0 +1)
k2 p , where ks
is the slope of δ. From Fig. 2, it is known robustness. For the proposed control (22), the feedback gains
that both the high precision (i.e., δ that is small enough) and fast k1 , k2 should be set to achieve fast tracking error convergence
convergence are guaranteed [57]. as explained in Remark 1.
Remark 2: In the boundary layer-based RL, the chattering The practical implementation of the proposed control algo-
problem still exists in the systems with fast varying unmod- rithm can be summarized as follows:
eled dynamics, which may generate the unpredictable low-and
medium-frequency oscillations. Moreover, the finite time stabil- The Implementation of the Proposed Control Algorithm.
ity may not be guaranteed due to the asymptotic stability within 1: Start procedure
the boundary layer [34], [58]. 2: Set the filter constant κ, sliding mode surface
parameters λ1 ,and control parameters k1 , k2 , k3 , α and
B. Controller Design β;
4: Compute the sliding mode surface s from (19);
In this section, a SMC based on the proposed RL is designed
5: Compute the control action u via (22) for integration
for the studied servo mechanisms to achieve motion tracking
interval t ∈ [ti , ti+1 ], i ∈ N ;
control. The overall control system structure including the pro-
6: Calculate the filter variables x2f and uf based on (5);
posed USDE and the SMC is shown in Fig. 3. To design the
SMC, we define the tracking error e as 7: Compute the USDE F̂ given in (10);
8: Continuation: let i := i + 1;
e = x1 − xd (18) 9: End procedure

where xd is the rotation trajectory to be tracked.

The sliding mode surface is defined as
C. Stability Analysis
s = ė + λ1 e (19)
This section will analyze the stability of the closed-loop con-
where λ1 > 0 is a positive constant. The derivative of (19) is trol system by using Lyapunov stability theory. The following
obtained as main result is given.
Theorem 1: Consider the control system (3) consisting of
ṡ = ë + λ1 ė = ẋ2 − ẍd + λ1 ė. (20) the USDE (10), and control law (22), then all signals in the

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
1060 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO. 2, APRIL 2020

closed-loop system are bounded, and the tracking error e and

the estimation error eF will converge to a small set around zero.
Proof: Consider a Lyapunov function as
1 2 1 2
V = s + eF . (23)
2 2

The time derivative of V is obtained as
V̇ = sṡ + eF ėF

1
=s (K1 u − F (x1 , x2 )) − ẍd + λ1 ė
J

1
+ eF − eF + Ḟ . (24)
κ
Substituting the control u in (22) and the estimation error (14)
into (24), it yields
V̇ = s (−k1 tanh(as) − k2 |s| · asinh(βsp ) − k3 s − eF )
 Fig. 4. Simulation results for square wave—position tracking, tracking
1 error, control law, and disturbance estimation.
+ eF − eF + Ḟ
κ
= − k1 s · tanh(αs) − k2 s|s| · asinh(βsp ) − k3 s2 − seF The friction dynamics are given by Tf = γ1 (tanh(β1 x2 ) −
1 2 tanh(β2 x2 )) + γ2 tanh(β3 x2 ) + γ3 x2 with γ1 = 0.07, γ2 =
− e + eF Ḟ 0.08, γ3 = 0.075, β1 = 100, β2 = 1, and β3 = 1000. The ex-
κ F
 ternal disturbance is Td = 0.05 sin(t). The controller parame-
1
= − k1 |s|tanh(α|s|)−k2 |s|2 · asinh(β|s|p ) − k3 − s2 ters are selected as k1 = k2 = k3 = 20, λ1 = λ2 = 10, α = 5,
2
 β = 10, and p = 3. The filter time constant is κ = 0.01. To show
1 1 the chattering elimination of the developed RL, the traditional
− − 1 eF 2 + Ḟ 2
κ 2 SMC is used for comparison. In the conventional SMC, the
sliding mode surface is designed as s∗ = ė + λ2 e, and controller
2 is u∗ (t) = b[−k3 sign(s∗ ) − ab x2 + ẍd − λ2 ė].
≤ −k1 |s|tanh(α|s|) − k2 |s|2 · asinh(β|s|p ) + (25)
2 Case1: A square wave with amplitude 2 rad and period 7.5 s
for any control gain k3 > 0.5, κ < 1. is employed as the reference signal. The simulation results are
According to the boundedness theorem and the analysis shown in Fig. 4. Fig. 4 shows that the system output trajectory
given in [59], the inequality (25) implies that s and eF are with the developed controller can track the reference signal accu-
all bounded, and the sliding mode variable s can eventually rately, and the tracking error is smaller than the SMC. Especially,
2
converge to a small neighborhood limt→∞ s(t) ∈ (|s| ≤ 3k 1
)∩ the developed RL (17) can avoid the chattering phenomenon,
1
while the SMC still leads to significant chattering in the control
(|s| ≤ ( 3k2 b ) p+2 ). Then, in combination with the design of the
2

signal. Moreover, the suggested USDE can precisely estimate

sliding mode surface (19), the tracking errors s, e all converge
the unknown dynamics.
to a small bounded set. This completes the proof.
Case 2: To illustrate the effectiveness of the proposed con-
Remark 3: The filter time constant κ determines the estima-
trol scheme, a composite reference signal xd = 0.4 sin(2t) +
tion performance of the USDE. A too small κ will produce the
0.2 cos(t) is used as the desired trajectory. The simulation results
oscillation, and a large κ may lead to phase delay. A very large
are shown in Fig. 5. Fig. 5 describes the tracking performance,
k2 in (22) could increase the convergence speed of sliding mode
tracking error, control action, and unknown dynamics estimation
variables s, but may force the initial value of the controller
responses. From Fig. 5, it is concluded that the proposed control
to exceed the system limit. Besides, the steady-state tracking
scheme produces better tracking performance than the SMC,
accuracy could be improved by increasing k1 under the condition
and the control signal does not suffer the chattering issue. In
that (k1 a + k2 p) · ts < 1 according to [59].
addition, the unknown dynamics can be accurately estimated by
using the USDE.
V. SIMULATION Case 3: To further validate the effectiveness of the developed
Numerical simulations are utilized to confirm the validity control and show the advantages of the proposed USDE, an
of the developed USDE and hyperbolic RL-based sliding con- ESO [44] is combined with the proposed new RL-based SMC
trol control (USDE-HRLSMC). The system parameters in (4) (ESO-HRLSMC) for comparison, showing the response of the
are given as J = 0.1 kg/m2 , KE = 0.2 V/(rad/s), KT = 5 N · proposed USDE and ESO. In the design of ESO, the unknown
m/A, Ra = 5Ω, Tl = 0.1 N · m, and thus K1 = KT /Ra = 1. dynamics are defined as an extra system state x3 = (Td + Tf +

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: USDE-BASED SLIDING MODE CONTROL FOR SERVO MECHANISMS WITH UNKNOWN SYSTEM DYNAMICS 1061

Fig. 7. Simulation results for parameter variations.

Fig. 5. Simulation results for xd = 0.4 sin(2t) + 0.2 cos(t)—position

tracking, tracking error, control law, and disturbance estimation.

Fig. 8. Simulation results for parameter uncertainties on dynamics

response of servo mechanism.

jumps at 5 s
0.1, 0 ≤ t < 5
J= (27)
0.2, 5 ≤ t ≤ 15.
Fig. 6. Simulation results for different control method USDE-HRLSMC
and ESO-HRLSMC. Fig. 7 shows the tracking performance and estimation perfor-
mance of servo mechanism with such parameter changes. It is
clearly seen that the tracking error produces a small fluctuation
Tl )/J. Then, an ESO can be designed as [60] and the estimation F̂ produces an oscillation at t = 5 s, which
⎧ can be recovered quickly. This indicates that the proposed esti-
˙
⎨ x̂1 = x̂2 − 3ω0 (x̂1 − x1 )
⎪ mate and control can handle the parameter variations effectively.
x̂˙ 2 = KJ1 u + x̂3 − 3ω02 (x̂1 − x1 ) (26) Case 5: This case will study the effect of parameter un-
⎪
⎩˙ certainties on the dynamic response of servo mechanisms. In
x̂3 = −ω03 (x̂1 − x1 )
this case, we use Δ(t) to denote the parameter uncertainties
where x̂ = [x̂1 , x̂2 , x̂3 ]T is the observe state, and ω0 > 0 is the in the turntable servomechanism system, so that the lumped
observer gain. In this simulation, ω0 is set as ω0 = 80. The unknown dynamics F can be written as F = (Tf + Tl + Td +
desired trajectory is selected as xd = 0.4 sin (2t) + 0.2 cos (t). Δ(t))/K1 . In this simulation, the parameter uncertainties are
The simulation result is given in Fig. 6, from which we can given as Δ(t) = 0.2 sin(t). A square wave with amplitude 2 rad
see that the tracking performance of the developed USDE- and period 7.5 s is employed as the reference signal. The simu-
HRLSMC is slightly better than the ESO-HRLSMC. This is due lation results are shown in Fig. 8, where the position tracking,
to the developed USDE that can estimate the unknown dynamics tracking error, and unknown system dynamics estimation are
with exponential convergence. given. From Fig. 8, it can be seen that the servo mechanism has a
Case 4: In this case, we will study the effect of parameter satisfactory dynamic response even in the presence of parameter
variations on the tracking error. The system inertia parameter J uncertainties. This is due to the parameter uncertainties that

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
1062 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO. 2, APRIL 2020

TABLE II
COMPARISON RESULTS OF PERFORMANCE INDICES

Fig. 9. Diagram of the laboratory test rig.

TABLE I
PARAMETERS OF TURNTABLE AND MOTOR

parameter is given as λ1 = 10. The filter time constant is κ =

0.1.
2) TSMC [61]: The control law u = F̂ + k1 s + k2 |s|γ
sgn(s) + (δ1 + δ2 )sgn(s), where the tracking error is e = x1 −
xd and s = ė + λ1 e + λ2 β(e), where β(e) = |e|γ sgn(e), s =
0 or s = 0, |e| ≥ μ; or β(e) = 1 e + 2 |e|2 sgn(e), s = 0, |e| ≤
μ, with 1 = (2 − γ)μγ−1 , 2 = (γ − 1)μγ−2 , μ > 0 being a
small positive constant. The control parameters are γ = 9/11,
λ1 = 5, λ2 = 1, k1 = 0.5, k2 = 0.1, r = 9/11, and δ1 = δ2 =
can be lumped into the unknown system dynamics, and then
0.01.
estimated by using the USDE and compensated in the control
3) ANC [62]: An ANC with a linear sliding mode term was
design.
proposed in [62]. The control law is u = ks + Ŵ T Φ + u1 ,
where u1 = σs/|s| for s = 0 or u1 = 0 for s = 0 is a slid-
VI. EXPERIMENTAL RESULTS
ing mode term, and s = Λe1 + e2 is the filtered error with
In this section, we will give experimental results to verify the ˙
e1 = xd − x1 and e2 = ẋd − x2 . The adaptive law is Ŵ = ΓsΦ,
effectiveness of the suggested method based on a turntable servo and the control parameters are Λ = 15, k = 1, Γ = 0.5, and
system. σ = 0.005.
4) PID Control: This is a standard PID control, where the
A. Experimental Setup control parameters are given as kp = 30, ki = 0.05, kd = 10.
To show the performance of the developed control method, To illustrate the control performance, the following four per-
a turntable servo mechanism is used as the experimental plat- formance indexes are introduced in thisarticle:
form (see Fig. 9). The experimental setup is comprised of a 1) integrated absolute error IAE = |e(t)|dt;

PMSM (HC-UFS13), an encoder with the pulsewidth modu- 2) integrated square error ISDE = (e(t) − e(0))2 dt,
lation (PWM) amplifier located in the driving card (MR-J2S- where e(0) represents the average error;

10 A), a digital signal processor (DSP, TMS3202812) perform- 3) integrated absolute control IAU = |u(t)|dt;

ing as the controller. A Dell Computer with Intel(R) Pentium 4) integrated square control ISDU = (u(t) − u(0))2 dt
G3250 T CPU 2.8 GHz is adopted in the experiments. The where u(0) denotes the average control.
control algorithm is implemented by using C++ program in The performance indexes under different references are listed
CSS 3.0 in the DSP environment. The sampling time is 0.01 s. in Table II .
The output position is measured by means of an encoder with
a resolution of 800 divisions. A gear transmission system with C. Experimental Results
a gear ratio of 80 is included; then, the encoder output signals
To verify the control performance of the proposed method,
have a resolution of 64 000 per rotation. The system parameters
sinusoidal trajectories with different periods are employed in
are shown in Table I.
this section.
Case 1: A slowly-varying sinusoid reference signal xd =
B. Controller Implementation
0.4 sin(2πt/4) is first adopted to test the control performance.
To compare the control performance, four control methods are The experimental results are shown in Figs. 10 and 11. Fig. 10
implemented for comparison—the presented control method, shows the tracking performance [see Fig. 10(a)], tracking error
the terminal SMC (TSMC) [61], adaptive neural control (ANC) [see Fig. 10(b)], and the control actions [see Fig. 10(c)] of
and PID control. different control schemes. From these results, one can see that
1) USDE-HRLSMC: This proposed control scheme is given the control performance of the developed control scheme is
in Section III. The controller parameters are selected as k1 = 2, better than the other three control methods (i.e., TSMC, ANC,
k2 = 0.5, p = 3, α = 0.5, β = 1, and the sliding mode surface and PID). The control action of the proposed control scheme

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: USDE-BASED SLIDING MODE CONTROL FOR SERVO MECHANISMS WITH UNKNOWN SYSTEM DYNAMICS 1063

Fig. 12. Unknown dynamics estimation. (a) With input disturbance

−2x1 . (b) Normal case.

To further test the effectiveness of the presented USDE, an

extra disturbance signal d = −2x1 is added in this case. The
experiment results are shown in Fig. 12. Fig. 12 shows the
estimated disturbance in the normal case and with the added
signal d , respectively. From Fig. 12, it is clearly seen that the
USDE can accurately estimate the unknown disturbance even if
there exists a manually added disturbance d .
Case 2: In this case, a sinusoidal signal with larger amplitude
xd = 0.8 sin (2πt/4) is used as the desired trajectory. The output
tracking and control signals of different controllers are shown
in Fig. 13, respectively. From these results, one can again find
that the tracking error of the proposed control method is smaller
than the TSMC, ANC, and PID schemes. In addition, the control
signal of the devised control is smoother than the TSMC, which
suffers the chattering issue. This is mainly because the USDE
F̂ is incorporated into the controller design and the new RL is
adopted.
The performance indexes are summarized in Table II. From
these results, we can find that the designed control with new RL
Fig. 10. Experiment results for xd = 0.4 sin(2πt/4). (a) Position track-
ing. (b) Tracking errors. (c) Control signals. and USDE produces the smaller IAE and ISDE than the other
three control algorithms. By comparing the USDE-HRLSMC
and TSMC, the IAU of USDE-HRLSMC is larger than TSMC,
while the ISDU is smaller than TSMC. Among these four
controller, ANC and PID control produce similar control per-
formances.
Case 3: To test the transient performance of the proposed
control method with a fast varying trajectory, a setpoint xd = 0.8
is also used as the reference signal in this case. The position
tracking and tracking error are shown in Fig. 14. Among the con-
trollers, the proposed USDE-HRLSMC produces better tracking
Fig. 11. Unknown dynamics estimation. performance and there is no overshoot. This is because the
unknown system dynamics are estimated by using the developed
is also smoother than TSMC, ANC, and PID [see Fig. 10(c)]. USDE, which is incorporated into the control design to compen-
Moreover, the unknown system dynamics estimation is shown in sate their effect on the control response. The rising time of ANC
Fig. 11. From Fig. 11, one can see that the proposed USDE can and PID control method are faster than the USDE-HRLSMC and
effectively estimate the unknown dynamics (e.g., nonlinear fric- TSMC, while these two methods produce overshoot. In addition,
tion and external disturbance). From these experiment results, the TSMC method does not result in overshoot, but the rising
it can be concluded that the developed control produces better time is the smallest among the four control methods.
control performance when the estimator F̂ is incorporated into Case 4: In this case, we employed a triangular wave reference
the control to compensate the unknown dynamics. to test the control performance of the USDE-HRLSMC and

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
1064 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO. 2, APRIL 2020

Fig. 15. Experimental results for triangular wave.

are consistent with the simulation results given in Section V.

Moreover, the unknown dynamics could be accurately estimated
by employing the presented USDE.

VII. CONCLUSION
In this article, an USDE-based SMC scheme is developed for
servo mechanisms with unknown system dynamics. A USDE
with only one tuning parameter is developed by introducing
auxiliary filtered variables derived based on low-pass filters only.
Then a new structural RL is designed by using hyperbolic func-
tions, which can remedy the chattering phenomenon encoun-
tered in the convectional SMC schemes. Finally, the USDE and
the new RL are incorporated into the SMC synthesis to achieve
Fig. 13. Experiment results for xd = 0.8 sin(2πt/4). (a) Position track- satisfactory motion control performance for the servo systems.
ing. (b) Tracking errors. (c) Control signals. Simulation and experimental results validate the effectiveness
of the proposed algorithm. Since the proposed USDE has only
one tuning parameter, which is easy to be set by the designers,
future studies can be carried out to extend this USDE to the
control design of robotic manipulators, hydraulic systems, and
other nonlinear systems with unknown dynamics.

REFERENCES
[1] S. Wang, X. Ren, J. Na, and T. Zeng, “Extended-state-observer-based
funnel control for nonlinear servomechanisms with prescribed tracking
performance,” IEEE Trans. Autom. Sci. Eng., vol. 14, no. 1, pp. 98–108,
Jan. 2017.
[2] J. Na, Q. Chen, X. Ren, and Y. Guo, “Adaptive prescribed performance
motion control of servo mechanisms with friction compensation,” IEEE
Trans. Ind. Electron., vol. 61, no. 1, pp. 486–494, Jan. 2014.
[3] J. Na, Y. Li, Y. Huang, G. Gao, and Q. Chen, “Output feedback control of
uncertain hydraulic servo systems,” IEEE Trans. Ind. Electron., vol. 67,
no. 1, pp. 490–500, Jan. 2020.
Fig. 14. Setpoint response.
[4] M. Yang, C. Wang, D. Xu, W. Zheng, and X. Lang, “Shaft torque limiting
control using shaft torque compensator for two-inertia elastic system with
backlash,” IEEE/ASME Trans. Mechatronics, vol. 21, no. 6, pp. 2902–
USDE-TSMC. The experimental results of the tracking perfor- 2911, Dec. 2016.
mance, tracking error, control action, and unknown dynamics [5] G. Zhang, “Speed control of two-inertia system by PI/PID control,” IEEE
estimation responses are shown in Fig. 15. From Fig. 15, it can Trans. Ind. Electron., vol. 47, no. 3, pp. 603–609, Jun. 2000.
[6] B. Brahmi, M. Saad, C. Ochoa-Luna, M. H. Rahman, and A. Brahmi,
be concluded that the proposed USDE-HRLSMC can achieve “Adaptive tracking control of an exoskeleton robot with uncertain dynam-
better tracking performance than USDE-TSMC, i.e., it has ics based on estimated time-delay control,” IEEE/ASME Trans. Mecha-
relatively smaller tracking error and control chattering, which tronics, vol. 23, no. 2, pp. 575–585, Apr. 2018.

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: USDE-BASED SLIDING MODE CONTROL FOR SERVO MECHANISMS WITH UNKNOWN SYSTEM DYNAMICS 1065

[7] D. Meng, Y. Jia, J. Du, and S. Yuan, “Robust discrete-time iterative learning [31] Q. Xu, “Piezoelectric nanopositioning control using second-order discrete-
control for nonlinear systems with varying initial state shifts,” IEEE Trans. time terminal sliding-mode strategy,” IEEE Trans. Ind. Electron., vol. 62,
Autom. Control, vol. 54, no. 11, pp. 2626–2631, Nov. 2009. no. 12, pp. 7738–7748, Dec. 2015.
[8] Y. J. Liu, S. Liu, S. Tong, X. Chen, C. L. P. Chen, and D. J. Li, “Adaptive [32] J. Qin, Q. Ma, H. Gao, and W. X. Zheng, “Fault-tolerant cooperative
control-based barrier Lyapunov functions for a class of stochastic nonlinear tracking control via integral sliding mode control technique,” IEEE/ASME
systems with full state constraints,” Automatica, vol. 87, pp. 83–93, 2018. Trans. Mechatronics, vol. 23, no. 1, pp. 342–351, Feb. 2018.
[9] H. Li, S. Zhao, W. He, and R. Lu, “Adaptive finite-time tracking control [33] Q. Chen, S. Xie, M. Sun, and X. He, “Adaptive non-singular fixed-time at-
of full state constrained nonlinear systems with dead-zone,” Automatica, titude stabilization of uncertain spacecraft,” IEEE Trans. Aerosp. Electron.
vol. 100, pp. 99–107, 2019. Syst., vol. 54, no. 6, pp. 2937–2950, Dec. 2018.
[10] S. Wang, J. Na, and Y. Xing, “Adaptive optimal parameter estimation and [34] S. Yu, X. Yu, B. Shirinzadeh, and Z. Man, “Continuous finite-time control
control of servo mechanisms: Theory and experiments,” IEEE Trans. Ind. for robotic manipulators with terminal sliding mode,” Automatica, vol. 41,
Electron., doi: 10.1109/TIE.2019.2962445. no. 11, pp. 1957–1964, 2005.
[11] G. Sun, D. Li, and X. Ren, “Modified neural dynamic surface approach to [35] L. Yang and J. Yang, “Nonsingular fast terminal sliding-mode control for
output feedback of MIMO nonlinear systems,” IEEE Trans. Neural Netw. nonlinear dynamical systems,” Int. J. Robust Nonlinear Control, vol. 21,
Learn. Syst., vol. 26, no. 2, pp. 224–236, Feb. 2015. no. 16, pp. 1865–1879, 2011.
[12] S. Wang, H. Yu, J. Yu, J. Na, and X. Ren, “Neural-network-based adaptive [36] S. Y. Chen and F. J. Lin, “Robust nonsingular terminal sliding mode
funnel control for servo mechanisms with unknown dead-zone,” IEEE control for nonlinear magnetic bearing system,” IEEE Trans. Control Syst.
Trans. Cybern., doi: 10.1109/TCYB.2018.2875134. Technol., vol. 19, no. 3, pp. 636–643, May 2011.
[13] S. Ding, L. Liu, and W. X. Zheng, “Sliding mode direct yaw-moment [37] K.-H. Kim and M.-J. Youn, “A nonlinear speed control for a PM syn-
control design for in-wheel electric vehicles,” IEEE Trans. Ind. Electron., chronous motor using a simple disturbance estimation technique,” IEEE
vol. 64, no. 8, pp. 6752–6762, Apr. 2017. Trans. Ind. Electron., vol. 49, no. 3, pp. 524–535, Jul. 2002.
[14] S. Ding and S. Li, “Second-order sliding mode controller design subject [38] W.-H. Chen, D. Ballance, P. Gawthrop, and J. O’Reilly, “A nonlinear
to mismatched term” Automatica, vol. 77, pp. 388–392, 2017. disturbance observer for robotic manipulators,” IEEE Trans. Ind. Electron.,
[15] G. P. Incremona, M. Rubagotti, and A. Ferrara, “Sliding mode control vol. 47, no. 4, pp. 932–938, Sep. 2000.
of constrained nonlinear systems,” IEEE Trans. Autom. Control, vol. 62, [39] E. Kim, “A fuzzy disturbance observer and its application to control,”
no. 6, pp. 2965–2972, Jun. 2017. IEEE Trans. Fuzzy Syst., vol. 10, no. 1, pp. 77–84, Feb. 2002.
[16] J. J. Rath, M. Defoort, H. R. Karimi, and K. C. Veluvolu, “Output feedback [40] M. Chen and S. S. Ge, “Adaptive neural output feedback control of
active suspension control with higher order terminal sliding mode,” IEEE uncertain nonlinear systems with unknown hysteresis using disturbance
Trans. Ind. Electron., vol. 64, no. 2, pp. 1392–1403, Feb. 2017. observer,” IEEE Trans. Ind. Electron., vol. 62, no. 12, pp. 7706–7716,
[17] L. Liu, Y. Liu, and S. Tong, “Neural networks-based adaptive finite-time Dec. 2015.
fault-tolerant control for a class of strict-feedback switched nonlinear [41] M. White, M. Tomizuka, and C. Smith, “Improved track following in
systems,” IEEE Trans. Cybern., vol. 49, no. 7, pp. 2536–2545, Jul. 2019. magnetic disk drives using a disturbance observer,” IEEE/ASME Trans.
[18] Q. Chen, H. Shi, and M. Sun, “Echo state network based back- Mechatronics, vol. 5, no. 1, pp. 3–11, Mar. 2000.
stepping adaptive iterative learning control for strict-feedback sys- [42] Y. Yan, J. Yang, Z. Sun, C. Zhang, S. Li, and H. Yu, “Robust speed regu-
tems: An error-tracking approach,” IEEE Trans. Cybern., doi: lation for PMSM servo system with multiple sources of disturbances via
10.1109/TCYB.2019.2931877. an augmented disturbance observer,” IEEE/ASME Trans. Mechatronics,
[19] M. Chen, S. Shao, and B. Jiang, “Adaptive neural control of uncertain vol. 23, no. 2, pp. 769–780, Apr. 2018.
nonlinear systems using disturbance observer,” IEEE Trans. Cybern., [43] M. Wu, F. Gao, P. Yu, J. She, and W. Cao, “Improve disturbance-rejection
vol. 47, no. 10, pp. 3110–3123, Oct. 2017. performance for an equivalent-input-disturbance-based control system by
[20] B. Xu, D. Yang, Z. Shi, Y. Pan, B. Chen, and F. Sun, “Online recorded data- incorporating a proportional-integral observer,” IEEE Trans. Ind. Elec-
based composite neural control of strict-feedback systems with application tron., vol. 67, no. 2, pp. 1254–1260, 2020.
to hypersonic flight dynamics,” IEEE Trans. Neural Netw. Learn. Syst., [44] J. Han, “From PID to active disturbance rejection control,” IEEE Trans.
vol. 29, no. 8, pp. 3839–3849, Aug. 2018. Ind. Electron., vol. 56, no. 3, pp. 900–906, Mar. 2009.
[21] C. Yang, X. Wang, Z. Li, Y. Li, and C. Su, “Teleoperation control based [45] W. Xue, W. Bai, S. Yang, K. Song, Y. Huang, and H. Xie, “Adrc with
on combination of wave variable and neural networks,” IEEE Trans. Syst., adaptive extended state observer and its application to air– fuel ratio control
Man, Cybern.: Syst., vol. 47, no. 8, pp. 2125–2136, Aug. 2017. in gasoline engines,” IEEE Trans. Ind. Electron., vol. 62, no. 9, pp. 5847–
[22] Z. Peng, J. Wang, and D. Wang, “Containment maneuvering of marine 5857, Sep. 2015.
surface vehicles with multiple parameterized paths via spatial-temporal [46] R. Cui, L. Chen, C. Yang, and M. Chen, “Extended state observer-based
decoupling,” IEEE/ASME Trans. Mechatronics, vol. 22, no. 2, pp. 1026– integral sliding mode control for an underwater robot with unknown
1036, Apr. 2017. disturbances and uncertain nonlinearities,” IEEE Trans. Ind. Electron.,
[23] Q. Chen, L. Shi, J. Na, R. Xuemei, and Y. Nan, “Adaptive echo state vol. 64, no. 8, pp. 6785–6795, Aug. 2017.
network control for a class of pure-feedback systems with input and output [47] H. Li, S. Zheng, and X. Ning, “Precise control for gimbal system of double
constraints,” Neurocomputing, vol. 275, pp. 1370–1382, 2017. gimbal control moment gyro based on cascade extended state observer,”
[24] Y. Wei, J. H. Park, H. R. Karimi, Y. Tian, and H. Jung, “Improved stability IEEE Trans. Ind. Electron., vol. 64, no. 6, pp. 4653–4661, Jun. 2017.
and stabilization results for stochastic synchronization of continuous-time [48] J. Na, B. Jing, Y. Huang, G. Gao, and C. Zhang, “Unknown system
semi-Markovian jump neural networks with time-varying delay,” IEEE dynamics estimator for motion control of nonlinear robotic systems,” IEEE
Trans. Neural Netw. Learn. Syst., vol. 29, no. 6, pp. 2488–2501, Jun. 2018. Trans. Ind. Electron., vol. 67, no. 5, pp. 3850–3859, May 2020.
[25] Q. Zhou, H. Li, C. Wu, L. Wang, and C. K. Ahn, “Adaptive fuzzy control [49] J. Liu, S. Vazquez, L. Wu, A. Marquez, H. Gao, and L. G. Franquelo, “Ex-
of nonlinear systems with unmodeled dynamics and input saturation using tended state observer-based sliding-mode control for three-phase power
small-gain approach,” IEEE Trans. Syst., Man, Cybern.: Syst., vol. 47, converters,” IEEE Trans. Ind. Electron., vol. 64, no. 1, pp. 22–31, Jan. 2017.
no. 8, pp. 1979–1989, Aug. 2017. [50] Q. Zhong and D. Rees, “Control of uncertain LTI systems based on
[26] L. Cao, H. Li, N. Wang, and Q. Zhou, “Observer-based event-triggered an uncertainty and disturbance estimator,” ASME J. Dyn. Syst., Meas.,
adaptive decentralized fuzzy control for nonlinear large-scale systems,” Control, vol. 126, no. 4, pp. 905–910, 2004.
IEEE Trans. Fuzzy Syst., vol. 27, no. 6, pp. 1201–1214, Jun. 2019. [51] S. Y. Gadelovits, Q. Zhong, V. Kadirkamanathan, and A. Kuperman,
[27] H. Li, Z. Chen, L. Wu, and H. Lam, “Event-triggered control for nonlinear ‘‘UDE-based controller equipped with a multi-band-stop filter to improve
systems under unreliable communication links,” IEEE Trans. Fuzzy Syst., the voltage quality of inverters,” IEEE Trans. Ind. Electron., vol. 64, no. 9,
vol. 25, no. 4, pp. 813–824, Aug. 2017. pp. 7433–7443, Sep. 2017.
[28] H. Li, J. Wang, H. Du, and H. R. Karimi, “Adaptive sliding mode control [52] B. Ren, Q. Zhong, and J. Dai, “Asymptotic reference tracking and distur-
for Takagi– Sugeno fuzzy systems and its applications,” IEEE Trans. Fuzzy bance rejection of ude-based robust control,” IEEE Trans. Ind. Electron.,
Syst., vol. 26, no. 2, pp. 531–542, Apr. 2018. vol. 64, no. 4, pp. 3166–3176, Apr. 2017.
[29] Z. Zhang, H. Li, C. Wu, and Q. Zhou, “Finite frequency fuzzy H∞ control [53] Q.-C. Zhong, A. Kuperman, and R. K. Stobart, “Design of UDE-based
for uncertain active suspension systems with sensor failure,” IEEE/CAA controllers from their two-degree-of-freedom nature,” Int. J. Robust Non-
J. Automatica Sinica, vol. 5, no. 4, pp. 777–786, 2018. linear Control, vol. 21, no. 17, pp. 1994–2008, 2011.
[30] M. Tao, Q. Chen, X. He, and M. Sun, “Adaptive fixed-time fault-tolerant [54] S. Li and Z. Liu, “Adaptive speed control for permanent-magnet syn-
control for rigid spacecraft using a double power reaching law,” Int. J. chronous motor system with variations of load inertia,” IEEE Trans. Ind.
Robust Nonlinear Control, vol. 29, no. 12, pp. 4022–4040, 2019. Electron., vol. 56, no. 8, pp. 3050–3059, Aug. 2009.

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.
1066 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 25, NO. 2, APRIL 2020

[55] A. Astolfi and R. Ortega, “Immersion and invariance: A new tool for Qiang Chen received the B.S. degree in mea-
stabilization and adaptive control of nonlinear systems,” IEEE Trans. surement and control technology and instru-
Autom. Control, vol. 48, no. 4, pp. 590–606, Apr. 2003. mentation from Hebei Agricultural University,
[56] J. Na, A. S. Chen, G. Herrmann, R. Burke, and C. Brace, “Vehicle engine Baoding, China, and the Ph.D. degree in control
torque estimation via unknown input observer and adaptive parameter science and engineering from the Beijing Insti-
estimation,” IEEE Trans. Veh. Technol., vol. 67, no. 1, pp. 409–422, tute of Technology, Beijing, China, in 2006 and
Jan. 2018. 2012, respectively.
[57] L. Tao, Q. Chen, Y. Nan, and C. Wu, “Double hyperbolic reaching law with He is currently an Associate Professor
chattering-free and fast convergence,” IEEE Access, vol. 6, pp. 27 717– with the College of Information Engineering,
27 725, 2018. Zhejiang University of Technology, Hangzhou,
[58] G. Bartolini, A. Ferrara, E. Usai, and V. I. Utkin, “On multi-input China. His research interests include iterative
chattering-free second-order sliding mode control,” IEEE Trans. Autom. learning control, neural networks, and adaptive control with applications
Control, vol. 45, no. 9, pp. 1711–1717, Sep. 2000. to motion control systems.
[59] L. Tao, Q. Chen, and Y. Nan, “Disturbance-observer based adaptive control
for second-order nonlinear systems using chattering-free reaching law,”
Int. J. Control, Autom. Syst., vol. 17, no. 2, pp. 356–369, 2019.
[60] Q. Zheng, L. Q. Gaol, and Z. Gao, “On stability analysis of active distur-
bance rejection control for nonlinear time-varying plants with unknown
dynamics,” in Proc. IEEE Conf. Decis. Control, 2007, pp. 3501–3506.
[61] Q. Chen, X. Ren, J. Na, and D. Zheng, “Adaptive robust finite-time neural
control of uncertain PMSM servo system with nonlinear dead zone,” Jing Na (Member, IEEE) received the B.Eng.
Neural Comput. Appl., vol. 28, no. 12, pp. 3725–3736, 2017. and Ph.D. degrees from the School of Au-
[62] L. Wang, T. Chai, and L. Zhai, “Neural-network-based terminal sliding- tomation, Beijing Institute of Technology, Bei-
mode control of robotic manipulators including actuator dynamics,” IEEE jing, China, in 2004 and 2010, respectively.
Trans. Ind. Electron., vol. 56, no. 9, pp. 3296–3304, Sep. 2009. From 2011 to 2013, he was a Monaco/ITER
Postdoctoral Fellow with the ITER Organization,
Saint-Paul-les-Durance, France. From 2015 to
2017, he was a Marie Curie Intra-European Fel-
low with the Department of Mechanical Engi-
neering, University of Bristol, Bristol, U.K. Since
2010, he has been with the Faculty of Mechan-
Shubo Wang (Member, IEEE) received the
ical and Electrical Engineering, Kunming University of Science and
B.S. degree in physics from Binzhou Univer- Technology, Kunming, China, where he became a Professor in 2013.
sity, Shandong, China, in 2008, the M.S de-
He has coauthored one monograph and more than 100 international
gree in control science and engineering from the
journal and conference papers. His current research interests include
School of Information Science and Engineer-
intelligent control, adaptive parameter estimation, nonlinear control and
ing, Central South University, Hunan, China, in applications for robotics, vehicle systems, and wave energy convertors.
2011, and the Ph.D. degree in control science
Dr. Na is currently an Associate Editor of the IEEE TRANSACTIONS ON
and engineering from the Beijing Institute of
INDUSTRIAL ELECTRONICS and Neurocomputing. He was the recipient of
Technology, Beijing, China, in 2017.
the Best Application Paper Award of the 3rd IFAC International Confer-
Since 2017, he has been with the School
ence on Intelligent Control and Automation Science (IFAC ICONS 2013),
of Automation, Qingdao University, Qingdao,
and the 2017 Hsue-shen Tsien Paper Award.
China, where he became an Associate Professor in 2019. He has
coauthored one monograph and more than 30 international journal
and conference papers. His current research interests include adaptive
control, adaptive parameter estimation, neural network, motor control,
nonlinear control, and applications.

Xuemei Ren received the B.S. degree from

Shandong University, Shandong, China, in
1989, and the M.S. and Ph.D. degrees in con-
trol engineering from the Beijing University of
Aeronautics and Astronautics, Beijing, China, in
1992 and 1995, respectively.
Liang Tao received the B.Eng. degree in From 2002, she was a Professor with the
electronic information engineering from Anhui School of Automation, Beijing Institute of Tech-
Agricultural University, Hefei, China, and the nology, Beijing. From September 2001 to March
Ph.D. degree in control science and engineer- 2002, and April 2005 to October 2005, she vis-
ing from the Zhejiang University of Technology, ited the Department of Electrical Engineering,
Hangzhou, China, in 2011 and 2019, respec- The Hong Kong Polytechnic University, Hong Kong. From July 2006 to
tively. July 2007, she was a Visiting Scholar with the Automation and Robotics
He is currently a Lecturer with the College of Research Institute, University of Texas at Arlington, Arlington, TX, USA.
Electrical Engineering, Anhui Polytechnic Uni- She has authored or coauthored more than 100 academic papers. Her
versity, Wuhu, China. His research interests in- research interests include nonlinear systems, intelligent control, neural
clude servo system control, adaptive control, network control, adaptive control, multidrive servo systems, and time
and parameter identification. delay systems.

Authorized licensed use limited to: Northwestern University. Downloaded on May 03,2020 at 07:20:32 UTC from IEEE Xplore. Restrictions apply.