International Journal of Automation and Computing 12(2), April 2015, 134-141

DOI: 10.1007/s11633-015-0880-3

Second-order Sliding Mode Approaches for the Control of

a Class of Underactuated Systems
Sonia Mahjoub1 Faiçal Mnif1,2 Nabil Derbel1
1
Advanced Control and Energy Management Laboratory, University of Sfax,
Sfax Engineering School, Sfax 3038, Tunisia
2
Department of Electrical and Computer Engineering, Sultan Qaboos University, Muscat, Oman

Abstract: In this paper, first-order and second-order sliding mode controllers for underactuated manipulators are proposed. Sliding
mode control (SMC) is considered as an effective tool in different studies for control systems. However, the associated chattering
phenomenon degrades the system performance. To overcome this phenomenon and track a desired trajectory, a twisting, a super-
twisting and a modified super-twisting algorithms are presented respectively. The stability analysis is performed using a Lyapunov
function for the proposed controllers. Further, the four different controllers are compared with each other. As an illustration, an example
of an inverted pendulum is considered. Simulation results are given to demonstrate the effectiveness of the proposed approaches.

Keywords: Underactuated manipulator, sliding mode control, twisting algorithm, super-twisting algorithm, inverted pendulum.

1 Introduction The high frequency components of the control propagat-

ing on the system can excite the unmodeled fast dynamics
Underactuated mechanical systems (UMS) are increas- and therefore cause undesired oscillations. In fact, this can
ingly present in the robotic field. They have less actuators degrade the system performance or may even lead to insta-
than their degrees of freedom. In these systems, we find bility. In the literature, three main approaches have been
manipulators, vehicles and humanoids with several passive presented which help to reduce the chattering effects. The
joints. Underactuations arise by deliberate design for the class of methods consist in the use of the saturation control
purpose of reducing the weight of the manipulator or might instead of the discontinuous one. It ensures the convergence
be caused by actuator failures. The difficulty in control- to a boundary layer of the sliding manifold. Moreover, a
ling underactuated mechanisms is due to the fact that tech- switching function, inside the boundary layer of the sliding
niques developed for fully actuated systems cannot be di- manifold, was approximated by a linear feedback gain[3, 4] .
rectly used. These systems are not feedback linearizable, However, the accuracy and the robustness of the sliding
yet they exhibit nonholonomic constraints and nonmini- mode were partially lost.
mum phase characteristics[1] . Moreover, it has been shown The second class of methods consist in the use of a system
that it is not difficult to stabilize this class of systems by observer-based approach[5] . It can reduce the problem of
continuous controllers. Because of this, the class of under- robust control to the problem of exact robust estimation.
actuated mechanical systems presents challenging control This can lead to the deterioration of the robustness with
problems. One of the common methods used to control un- respect to the plant uncertainties or disturbances.
deractuated systems is the SMC based on Lyapunov design. Using high-order sliding mode controllers given by
The SMC has always been considered as an efficient ap- Levant[6, 7] as a way to reduce the chartering phenomenon
proach in control systems, due to its high accuracy and and keep the main advantages of the original approach of
robustness against internal and external disturbances. The the SMC is another way to eliminate chattering. On the
SMC approach consists of two steps. The first is to choose a other hand, the second-order sliding mode control is rel-
manifold in the state space that forces the state trajectories atively simple to implement and it gives good robustness
to remain along it. The second is to design a discontinu- to external disturbances. The second-order sliding mode
ous state-feedback capable of forcing the system to reach control (SSMC) approach can reduce the number of dif-
the state on the manifold in a finite time. However, the ferentiator stages in the controller. However, the stability
drawback of the SMC is the presence of the chattering ef- proofs are based usually on a geometrical or homogeneity
fect, caused by the switching frequency of the control[2] . method since the Lyapunov function is a difficult task to
define[7] . The stability and the convergence using SSMC is
Regular paper challenging and several trials were made to deal with those
Special Issue on Advances in Nonlinear Dynamics and Control
Manuscript received March 23, 2014; accepted September 25, 2014. difficulties. Recently, Moreno and Osorio[8] constructed a
Recommended by Guest Editor Fernando Tadeo Lyapunov function that provides a finite time convergence,
c Institute of Automation, Chinese Academy of Science and

Springer-Verlag Berlin Heidelberg 2015 a robustness and an estimate of the convergence time for
 S. Mahjoub et al. / Second-order Sliding Mode Approaches for the Control of a Class of Underactuated Systems 135

super-twisting algorithm. In [9], a multivariable super- 2 Dynamic model and control approach
twisting structure was proposed, which analyzed the sta- for an inverted pendulum
bility using the ideas of Lyapunov function given in [8].
The inverted pendulum system shown in Fig. 1 is a 2.1 Dynamic model
typical benchmark of non-linear underactuated mechanical
systems[10] . For this system, the control input is the force u The dynamical behavior of an inverted pendulum can be
that moves the cart horizontally and the output is the an- described by the differential equations as[12]
gular position of the pendulum θ. Therefore, the inverted
(m + M )ÿ + ml(θ̈ cos θ − θ̇2 sin θ) = τ
pendulum has been a popular candidate to illustrate dif-
ferent control methods. However, despite of its simple me- ÿ cos θ + lθ̈ + g sin θ = 0 (1)
chanical structure, this prototype is not easy to control and
where l is the length of the pendulum, m is the pendulum
requires sufficiently sophisticated control designs. Indeed,
mass, M is the cart mass, τ is the horizontal force action,
it is proven that the system is not feedback linearizable and
θ is the angular deviation, and y is the position of the cart
has no corresponding constant relative degree[11] . More-
which is moving horizontally.
over, Zhao and Spong[12] have shown that several geometric
Let x1 = y, x2 = ẏ, x3 = θ and x4 = θ̇. According to the
properties of the system are lost when the pendulum moves
canonical form of a class of underactuated systems, we can
through horizontal positions. The objective consists of mov-
transform (1) into the following state space representation
ing the cart from an initial position to a desired one and
keeping the pendulum in the vertical position with a mini- ẋ1 = x2
mum of its oscillations around this position. In general, the
ẋ2 = f1 + b1 τ
main difficulty is to swing up the pendulum from the down-
ward vertical position and to keep the cart stable. Numer- ẋ3 = x4
ous control techniques have been employed to stabilize the ẋ4 = f2 + b2 τ (2)
inverted pendulum such as proportional-integral-derivative
(PID) controllers where the control gains are adjustable and where x = [x1 , x2 , x3 , x4 ]T is the state variable vector, τ
updated online with a stable adaptation mechanism[13] . is the control input, f1 , f2 , b1 and b2 are nominal nonlinear
functions, expressed as

mlx24 sin x3 − mg sin x3 cos x3

f1 =
M + m sin2 x3
(m + M )g sin x3 − mlx24 cos x3
f2 =
l(M + m sin2 x3 )
1
b1 =
M + m sin2 x3
− cos x3
b2 = . (3)
l(M + m sin2 x3 )

Let

τ = M + m sin2 x3 u − (mlx24 sin x3 − mg sin x3 cos x3 ).

(4)
Fig. 1 Inverted pendulum
Then, (2) becomes
The objective of this paper is to develop a robust position
tracking controller based on the first-order and the second-
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
x˙1 x2 0
order sliding mode approaches applied to an inverted pen- ⎢ x˙2 ⎥ ⎢ 0 ⎥ ⎢ 1 ⎥
Ẋ = ⎢ ⎥= ⎥ +⎢ ⎥u
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
dulum. Stability of the closed loop system is carried out ⎣ x˙3 ⎦
⎢
⎣ x4 ⎦ ⎣ 0 ⎦
using candidate Lyapunov functions for the proposed con- x˙4 g sin x3 − cos x3
l l
trollers. The contributions of this paper are presenting the
stability analysis of twisting and super-twisting controllers. Ẋ = f (x) + g(x)u. (5)
Further, a modified super-twisting algorithm with simple
And (5) can be expressed by
stability analysis is proposed. This controller has almost
the same propriety as the super-twisting algorithm. The pa- ⎡
x˙1
⎤ ⎡
0 1 0 0
⎤⎡
x1
⎤ ⎡
0
⎤
per is organized as follows. Section 2 describes the model of ⎢ x˙2 ⎥ ⎢ 0 0 0 0 ⎥⎢ x2 ⎥ ⎢ 1 ⎥
the inverted pendulum and the first sliding mode controller. ⎥=⎢ ⎥ +⎢ ⎥ u.
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ ⎥⎢
⎣ x˙3 ⎦ ⎣ 0 0 0 1 ⎦⎣ x3 ⎦ ⎣ 0 ⎦
Section 3 deals with the sliding mode controllers and the g sin x3 − cos x3
x˙4 0 0 0 x4
design of second-order sliding mode controllers. Section 4 lx3 l

discusses the simulation results of the proposed controllers. (6)

136 International Journal of Automation and Computing 12(2), April 2015

Voytsekhovsky et al.[14, 15] proposed a method that can where

approximate the original system with an input-output lin- 
6 sin x3 sin x3 6g sin x3 2
earizable control system in new coordinates. This stabi- fe (x) = − l 4 x4 + x4 +
cos4 x3 cos2 x3 cos3 x3
lization method of nonlinear system using sliding mode 
2g sin x3 g sin x3
control is based on coordinate transformation by mapping − 3x24 +
T : x → ξ defined by cos3 x3 cos x3

3g g sin x3
− 2g
ξi = Li−1 h(x), i ∈ {1, 2, 3, 4} (7) cos2 x3 l
f
−6x24 3g 2g cos x3
T
ge (x) = + 3x24 − + . (13)
with ξ = (ξ1 ξ2 ξ3 ξ4 ) . T is defined as a local diffeomor- cos2 x3 l cos x3 l
phism with T (0) = 0.
By neglecting 2x4 tan(x3 ) because this term is o(x3 , x4 )2 ,
The dynamical system in the new coordinates can be
we obtain a feedback linearizable nonlinear system in the
approximated by the system model
state ξ with
ξ̇1 = ξ2 ξ̇1 = ξ2
ξ̇2 = ξ3 ξ̇2 = ξ3
ξ̇3 = ξ4 ξ̇3 = ξ4
ξ̇4 = L4f (T −1 (ξ)) + Lg L3f h(T −1 (ξ))u (8) ξ̇4 = fe (ξ) + ge (ξ)u
z = ξ1 . (14)
where Lf h(x) is the Lie derivative of h(x) along the vector
f (x).
2.2 First-order sliding mode controller
Consider the output system function defined by[15]
 Define the surface, s = {ξ ∈ R4 | s(ξ) = 0} , for λ > 0,
1 + sin x3
z = h(x) = x1 + l ln (9)  3
cos x3 d
s(ξ) = +λ (z − zd ). (15)
dt
with ξ = T (x) and T1 (x) = h(x) = ξ1 . Define the transfor-
mation T : x → ξ by We choose zd = [2 0 0 0]T . The time derivative of s along
the system trajectory ξ is equal to
⎡ ⎤ ⎡ ⎤
h(x) ξ1 = T1 (x)
ṡ(ξ) = ξ (4) + 3λξ (3) + 3λ2 ξ (2) + λ3 ξ (1) =
⎢ Lf h(x) ⎥ ⎢ ξ2 = T2 (x) ⎥
T (x) = ⎢ ⎥=⎢ ⎥ = ξ. (10)
⎢ ⎥ ⎢ ⎥
⎣ L2f h(x) ⎦ ⎣ ξ3 = T3 (x) ⎦ fe (ξ) + ge (ξ)u + 3λz (3) + 3λ2 z (2) + λ3 z (1) . (16)
L3f h(x) ξ4 = T4 (x)
The sliding mode control is expressed by
Then,
u = ueq + usw (17)
⎡ ⎤
1+sin x3
x1 + l ln cos x3 where usw is the switching control, ueq is the equivalent
control yielded from ṡ(ξ) = 0 , and
⎢ lx4
⎥
⎢ x2 + cos x3
⎥
T (x) = ⎢ ⎥.
⎢ ⎥
lx4
tan x3 g + cos
fe (z) + 3λz (3) + 3λz (3) + λ3 z
⎢ x3
⎥
ueq = −
⎣ ⎦
3g
2
cos3 x3
− 1
cos x3
lx34 + cos2 x3
− 2g x4 ge (ξ)
(11) usw = ηsgn(s) + ks (18)

Differentiating ξ, we obtain where η and K are positive constants.

It is notable that for small deviations, we have ge (ξ) <
−3 − gl < 0. Choosing the Lyapunov candidate as

lx4
ξ˙1 = z (1) = x2 + ln
cos x3
 1 2
˙ (2) lx4 V = s (19)
ξ2 = z = tan x3 g + 2
cos x3
and differentiating V along the trajectories of (14) yields

2 1
ξ˙3 = z (3) = − lx34 +
cos3 x3 cos x3
3g V̇ = sṡ = −η|s| − ks2 ≤ 0. (20)
( − 2g)x4 − 2x4 tan x3 u
cos2 x3
Then, the system is stable and the convergence of the
ξ˙4 = z (4) = fe (x) + ge (x)u (12) sliding mode is guaranteed.
 S. Mahjoub et al. / Second-order Sliding Mode Approaches for the Control of a Class of Underactuated Systems 137

3 Second-order sliding mode controller The Lyapunov function can be chosen as

The drawback of the first-order sliding mode control is 1 1

V1 = λ2 s1 2 + ṡ21 . (28)
the chattering phenomenon. As a solution to resolve this 2 2
problem, a higher-order sliding mode (HOS) is proposed. Differentiating (28) yields
In fact, HOS appears as an effective application to counter-
act the chattering phenomenon and the switching control V̇1 = λ2 s1 ṡ1 + ṡ1 s̈1 =
signals, with higher relative degrees in a finite time[8, 16] .
λ2 s1 ṡ1 + ṡ1 (Ψ(ξ) + ge (ξ)u) =
The HOS has been introduced in [6], with the goal to
get a finite time on the sliding set of order r defined by ṡ1 [λ2 s1 − Ks1 − Kβsgn(s˙1 )] =
...
s = ṡ = s̈ = s = · · · = s(r−1) = 0. s defines the sliding ṡ1 sgn(s˙1 )[λ2 s1 sgn(s˙1 ) − Ks1 sgn(s˙1 ) − Kβ] =
variable with the r-th order sliding and with its (r − 1)-th |s1 | [(λ2 |s1 | − K|s1 |)sgn(s˙1 s1 ) − Kβ)] ≤
first time derivatives depending only on the state x. The
|ṡ1 | [(λ2 − K)|s1 | − Kβ] ≤ 0. (29)
first-order sliding mode tries to keep s = 0. In the case
of second-order sliding mode control, which only needs its Therefore, the system is stable if λ2 − K < 0.
measurement or evaluation of s, the relation should be ver-
ified as 3.2 Super twisting controller
s(x) = ṡ(x) = 0. (21) SSMC controllers require the of values of the derivatives
except for the super twisting algorithm (STW). The STW is
In the following, a twisting algorithm, a super-twisting algo- a continuous sliding mode algorithm ensuring main proper-
rithm and a modified super-twisting algorithm with a pre- ties of the first-order sliding mode control for systems with
scribed convergence law are used. Lipschitz continuous matched uncertainties or disturbances
3.1 Twisting controller with bounded gradients[7] . It has been developed to control
systems with a relative degree equal to one in order to avoid
3.1.1 Controller approach chattering.
Consider the sliding surface Trajectories on the two sliding planes are characterized
 2 by twisting around the origin, but the continuous control
d
s1 = + λ1 ξ. (22) law u(t) is constituted by two terms. The first is defined
dt
by the discontinuous time derivative and the second is a
Differentiating (22) twice gives continuous function of the available sliding variable[2] .
3.2.1 Controller approach
s̈1 = fe (ξ) + ge (ξ)u + 2λ1 z (3) + λ21 z (2)
The derivative of the sliding surface is given as
s̈1 = Ψ(ξ) + ϕu (23)
ṡ = fe (ξ) + ge (ξ)u + 3λz (3) + 3λ2 z (2) + λ3 z (1) (30)
where Ψ(ξ) = fe (ξ) + 2λz (3) + λ2 z (2) and ϕ(ξ) = ge (ξ) .
We assume that functions Ψ and ϕ are bounded such which can be expressed as
that
ṡ = Ψ1 (ξ) + ϕ(ξ)u (31)
|Ψ| ≤ Ψd , 0 < ϕm ≤ ϕ ≤ ϕM (24)

where Ψd , ϕm , ϕM and ϕd are positive scalars. Then, we where Ψ1 (ξ) = fe (ξ)+3λz (3) +3λ2 z (2) +λ3 z (1) . The control
have law can be expressed by[16]

Ψ Ψd u1 − Ψ1 (ξ)
| |< . (25) u= (32)
ϕ ϕM ϕ(ξ)
By imposing s̈1 = 0, the equivalent control can be ex- where the super twisting controller is
pressed as ueq = − Ψ
ϕ
.
1
3.1.2 Stability analysis u1 = −k1 sgn(s)|s| 2 − k2 s + σ. (33)
The dynamic control law using the twisting algorithm is
given by[8] Variations of the term σ are described by
K
usw = (s1 + βsgn(ṡ1 )) (26) σ̇ = −k3 sgn(s) − k4 s (34)
ge (ξ)

with β > 0, 0 < K ≤ KM and KM > 1 Ψd

. The total where k1 , · · · , k4 are positive scalars.
1−β ϕM
control is defined by Substituting (32) and (33) into (31) gives
1
u = ueq + usw . (27) ṡ = −k1 sgn(s)|s| 2 − k2 s + σ. (35)
138 International Journal of Automation and Computing 12(2), April 2015

3.2.2 Stability analysis where

We want to prove the stability of the system with the use ⎡
Ψ11 0 Ψ13
⎤
of Lyapunov function candidate define by Ψ=⎣ 0 Ψ22 Ψ23 ⎦
⎢ ⎥
k5 2 Ψ31 Ψ23 Ψ33
V2 (s, z) = 2k3 |s| + k4 s2 + σ + γ2 (36)
2 ⎡ ⎤
Υ11 0 Υ13
where k5 is a positive scalar and Υ=⎣ 0
⎢
Υ22 Υ23 ⎦
⎥
(43)
γ = k1 sgn(s)|s| + k2 s − σ.
1
2 (37) Υ31 Υ23 Υ33
with
We introduced the positive scalar k5 for more flexible stabil-
ity conditions and for more generalization of the expression Ψ11 = k13
of V2 . Substituting (37) into (36) gives Ψ22 = 5k1 k22

2 s k1 k5
k5 s Ψ13 = − k12
V2 (s, σ) = 2k3 |s| + k4 s2 + σ 2 + k1 + k2 s − σ = |s| 2
2 |s|
Ψ23 = −2k1 k2
1
(4k3 |s|2 + 2k4 s2 |s| + k5 σ 2 |s| + 2k12 s2 + Ψ31 = Ψ13
2|s|
4k1 k2 s 2
|s| − 4k1 sσ |s| + 2k22 s2 |s|− Ψ32 = Ψ23

4k2 s|s|σ + 2σ 2 |s|). (38) Ψ33 = k1 (44)

and
Define the subspace as
Υ11 = 4k12 k2
κ = {(s, σ) ∈ R2 /s = 0}. (39)
Υ22 = 2k23
Differentiating (38) with respect to time gives s
Υ13 = −k1 k2
|s|
∂V 2 ds ∂V 2 dσ 1
V̇2 (s, σ) = + = Υ23 = k4 k5 − 2k22
∂s dt ∂σ dt 2
∂V 2 ∂V 2
ṡ + σ̇ = Υ32 = Υ23
∂s ∂σ
−ṡ  5 5 Υ33 = 2k2 . (45)
5 − 2k3 |s| 2 sgn(s) − 2k4 s|s| 2 −
|s| 2 Matrices Ψ and Υ are positives definite if
3
2k12 s|s| 2 − 4k1 k2 s|s|2 + k1 k2 s2 |s|sgn(s)+ 5
k12 = k3 k 5
5 8
2k1 σ|s|2 − k1 σ|s|2 s − 2k22 s|s| + 2
2
k2 = 2k4 k5 . (46)
5
2k2 |s| 2 + k12 s2 sgn(s) |s| −

So matrices Ψ and Υ are positive definite and consciously
σ̇ 
V̇2 (s, σ) ≤ 0. Thus, we can conclude that the system is

− k5 σ|s| + 2k1 s |s| + 2k2 s|s| − 2σ|s| . (40)
|s|
stable.
Substituting (34) and (35) into (40) yields
3.3 A modified super twisting controller
−6k12 k2 s2 4k12 σs k13 s4 6k1 k22 s2 To simplify the stability analysis, we propose a modified
V̇2 (s, σ) = + + 7 − +
|s| |s| |s| 2 |s| super twisting controller (MSTW).
2k13 s2 2k1 σ 2 3.3.1 Controller approach
4k22 sσ − 3 − − 2k23 s2 −
|s| 2 |s| ṡ can be expressed as
2k12 k2 s4 2k2 s3 σ 8k1 k2 sσ ṡ = Ψ1 (ξ) + ϕ(ξ)u = −k6 s + σ1 . (47)
2k2 σ 2 + 3
− 13 + −
|s| |s| |s|
The total controller can be expressed by
k3 k5 sσ 2k1 k2 s3 σ
− k5 k4 sσ − 5 + u = ueq + ∆u (48)
|s| |s| 2
k1 k22 s4 k1 s 2 σ 2 where
5 + 5 . (41)
|s| 2 |s| 2
∆u = ϕ(ξ)−1 (k6 s + σ1 ). (49)
1 T
Define X = (|s| 2 s σ) . Then, it is easy to show that Variations of σ1 are given by

1 σ˙1 = −k7 (s − k8 σ1 − k9 sgnσ1 ). (50)

V̇2 (s, σ) ≤ − 1 X T ΨX − X T ΥX (42)
|s| 2
Parameters k6 , k7 , k8 and k9 are positive scalars.
 S. Mahjoub et al. / Second-order Sliding Mode Approaches for the Control of a Class of Underactuated Systems 139

3.3.2 Stability analysis

For the stability proof, the Lyapunov function candidate
is given by
1 2 1 2
V3 (s, σ1 ) = s + σ1 . (51)
2 2k7
Differentiating (51) with respect to time gives
1
V̇3 (s, σ1 ) = sṡ + σ1 σ˙1 =
k7
1
s(−k6 s + σ1 ) + σ1 (s + σ˙1 ). (52)
k7
Substituting (50) into (52) gives
Fig. 5 Evolution of the sliding surface and the control using TW
V̇2 (s, σ1 ) = −k6 s2 − k8 σ12 − k9 |σ| < 0. (53)

4 Simulation results and discussions

Parameters of the inverted pendulum system are set as
M = 20 kg, m0 = 1.8 kg, l = 0.3 m, g = 9.8 N/kg. The
initial conditions of the cart pendulum are (y0 , y˙0 ) = (0, 0),
(θ0 , θ˙0 ) = (0.1, 0) and the desired position is set as yd = 2,
θd = 0 and y˙d = θ˙d = 0.

Fig. 6 Evolution of the sliding surface and the control using

STW

Fig. 2 Evolution of the position of θ for the uncertain system

Fig. 3 Evolution of the position of y for the uncertain system Fig. 7 Evolution of the sliding surface and the control using
MSTW

Simulations are done using λ = 1 and k = 20 for the

SMC, k1 = 40 and k2 = 90 for the twisting controller.
In Figs. 2 and 3, the simulation results for the four con-
trollers are shown. The convergence of state variables has
been established for all controllers. Furthermore, the state
variables for STW and MSTW controllers converge faster
than those of TW and SMC. 20% of mass uncertainties,
have been considered for the pendulum and cart. We can
notice the robust behaviors of the controllers with respect
to parametric uncertainties. Figs. 4–7 show that the pro-
posed SSMC is able to compensate effectively the chatter-
Fig. 4 Evolution of sliding surface and control by SMC
140 International Journal of Automation and Computing 12(2), April 2015

ing phenomenon better than the first-order sliding mode. [6] A. Levant. Higher-order sliding modes, differentiation and
Moreover, with the super-twisting controller and modified output-feedback control. International Journal of Control,
super-twisting controller, the chattering is eliminated. vol. 76, no. 9, pp. 924–941, 2003.
For more comparisons between these approaches, we con-
sider the following criteria [7] A. Levant. Chattering analysis. IEEE Transactions on Au-
tomatic Control, vol. 55, no. 6, pp. 1380–1389, 2010.
 20
J= u2 dt
0 [8] J. A. Moreno, M. Osorio. Strict Lyapunov functions for the
super twisting algorithm. IEEE Transactions on Automatic
which is proportional to the energy delivered to the system. Control, vol. 57, no. 4, pp. 1035–1040, 2012.
It is clear from Table 1 that STW gives the least delivered
energy, while the MSTW delivers the second compared to
[9] I. Nagesha, C. Edwardsb. A multivariable super-twisting
other approaches. sliding mode approach. Automatica , vol. 50, no. 3, pp. 984–
988, 2014.
Table 1 Comparison of the energy criterion (×104 J)
[10] I. Fantoni, R. Lozano. Non-linear Control for Underactu-
SMC TW STW MSTW
ated Mechanical Systems, London, UK: Springer, pp. 21–
3.6954 1.5798 1.2624 1.3056
42, 2002.

5 Conclusions [11] B. Jakubczyk, W. Respondek. On the linearization of con-

trol systems. Bult. Acad. Polon. Sei. Math., vol. 28, pp. 517–
In this paper, a second-order sliding mode controller has 522, 1980.
been designed for underactuated manipulators. This con-
troller keep the main advantages of the original sliding mode [12] J. Zhao, M. W. Spong. Hybrid control for global stabi-
approach and removes the chattering caused by the slid- lization of the cart pendulum system. Automatica, vol. 37,
no. 12, pp. 1941–1951, 2001.
ing mode approach. Simulation results of the twisting, the
super-twisting and the modified super-twisting controllers
show that these controllers give better performance com- [13] W. D. Chang, R. Hwang, J. G. Chsieh. A self-tuning PID
control for a class of nonlinear systems based on the Lya-
pared to the first-order sliding mode controller. It has been
punov approach. Journal of Process Control, vol. 12, no. 2,
shown that the proposed modified super-twisting controller pp. 233–242, 2002.
have almost the same properties of the super-twisting with
a simple stability analysis. Moreover, the second-order
[14] D. Voytsekhovsky, R. M. Hirschorn. Stabilization of single-
sliding-mode controller is an effective tool for the control
input nonlinear systems using higher-order term compen-
of uncertain nonlinear systems since it overcomes the main sating sliding mode control. International Journal of Robust
drawback of the classical sliding-mode control approach. and Nonlinear Control, vol. 18, no. 4–5, pp. 468–480, 2008.

References [15] C. Aguilar, R. Hirschorn. Stabilization of an Inverted Pen-

dulum, Report on Summer Natural Sciences and Engineer-
[1] A. Isidori. Nonlinear Control Systems, Berlin, Germany: ing Research Council of Canada, Research Project, Queens
Springer Verlag, 1995. University at Kingston, Department of Mathematics and
Statistics, 2003.

[2] W. Perruquetti, J. P. Barbot. Sliding Mode Control in En-

gineering, FL, USA: CRC Press, 2002. [16] S. V. Emel′ yanov, S. V. Korovin, L. V. Levantovsky.
Higher-order sliding modes in control systems. Differential
Equations, vol. 29, no. 11, pp. 1627–1647, 1993.
[3] J. A. Burton, A. S. I. Zinober. Continuous approximation of
variable structure control. International Journal of Systems
Science, vol. 17, no. 6, pp. 875–885, 1986. Sonia Mahjoub received the M. Sc. de-
gree in real time control and the Ph. D. de-
gree in automatic control, from the Sfax
[4] H. Lee, V. I. Utkin. Chattering suppression methods in Engineering School, Tunisia in 2005 and
sliding mode control systems. Annual Reviews in Control, 2013, respectively.
vol. 31, no. 2, pp. 179–188, 2007. Her research interests include applica-
tion of nonlinear control to underactuated
mechanical systems.
[5] K. C. Veluvolu, Y. C. Soh, W. Cao. Robust observer with E-mail: mahjoub sonia@yahoo.fr (Cor-
sliding mode estimation for nonlinear uncertain systems. responding author)
IET Control Theory Applications, vol. 1, no. 5, pp. 1533– ORCID iD: 0000-0002-3025-1754
1540, 2007.
 S. Mahjoub et al. / Second-order Sliding Mode Approaches for the Control of a Class of Underactuated Systems 141

Faiçal Mnif received the B. Eng. degree Nabil Derbel received the Engineer-
in electrical engineering from Sfax Engi- ing Diploma from the Sfax Engineering
neering School, Tunisia in 1988, the M. Sc. School, Tunisia in 1986, the M. Sc. de-
and the Ph. D. degrees in control engineer- gree in automatic control from the Poly-
ing from the Polytechnique School of Mon- technique School of Montreal in 1986, the
treal in 1991 and 1996, respectively. From Ph. D. degree from the Laboratory of Au-
1996 to 2000, he was with the Superior tomation and Analysis Systems of Toulouse
Technology in Montreal, Canada. In 2000, in 1989, and the State Doctorate degree
he joined the National Institute of Applied from the Sfax Engineering School, Tunisia.
Sciences and Technology, where he is currently an associate pro- He joined the Tunisian University in 1989, where he held differ-
fessor. Since 2001, he has been on leave to Sultan Qaboos Univer- ent positions in research and education. Currently, he is a full
sity in Oman, where he is an associate professor in Department professor of automatic control at the Sfax Engineering School.
of Electrical and Complex Mechanical Systems. He is an IEEE senior member.
His research interests include nonlinear control theory and ap- His research interests include optimal control, complex sys-
plications, with an emphasis on the control of complex mechan- tems, fuzzy logic, neural networks, and genetic algorithm.
ical systems. E-mail: n.derbel@enis.rnu.tn
E-mail: mnif@squ.edu.om