Special Issue Article

Advances in Mechanical Engineering

2016, Vol. 8(10) 1–10
Ó The Author(s) 2016
Finite time motion control for a DOI: 10.1177/1687814016674673
aime.sagepub.com
launching platform based on sliding
mode disturbance observer

Yang Qiu1, Jian Hu1, Hui Lu1 and Long Liu2

Abstract
A kind of launching platform driven by two permanent magnet synchronous motors which is used to launch kinetic load
to hit the target always faces strong external disturbance such as the air current impulsion which would degrade their
tracking accuracy greatly. In this article, the position control problem of a launching platform with strong disturbance is
studied based on the finite time control technique. A sliding mode disturbance observer is proposed to observe the dis-
turbance, based on which a finite time control technique is proposed for position loop. Rigorous mathematical analysis is
given for the closed-loop system performance in the presence of disturbances. The experimental results show that this
method not only makes the position tracking error of the closed-loop system converge faster but also makes the bound-
ary of steady-state error smaller than that of the conventional proportional–integral–derivative control scheme and slid-
ing mode control scheme, which means the closed-loop system has stronger anti-disturbance property.

Keywords
Launching platform, motion control, finite time control, anti-disturbance, observer

Date received: 8 December 2015; accepted: 29 August 2016

Academic Editor: Nasim Ullah

Introduction produced would exert strong external transient distur-

bance on the system and (2) when the azimuth rotary
The launching platform is a typical mechatronic servo mechanism and the pitching rotary mechanism rotate
system, usually composed of a sensor (such as the photo- around their respective rotation axis, a gyroscopic torque
electric encoder), an actuator (such as an AC servo would be exerted on one mechanism by another mechan-
motor), a mechanical transmission mechanism (such as a ism as a disturbance due to the gyroscopic effect. All
turret, a rack, and a reducer.), load, and a controller. these factors would bring great difficulty to improve the
Figure 1 shows a typical launching platform, which is accuracy of the system. To solve the problem of system
also a typical multi-variable strong-coupling nonlinear model uncertainty including the parametric uncertain
system. It is mainly used to launch kinetic load to hit the
target, and it is applied in military and civilian fields
widely, such as for anti-missile air defense to combat 1
School of Mechanical Engineering, Nanjing University of Science and
invading target and artificial rainfall to hit the target Technology, Nanjing, China
cumulonimbus. How to improve the tracking accuracy 2
No.650 Institute, AVIC Hongdu Aviation Industry Group, Nanchang,
and robustness of launching platform is always a hot China
topic of scholars at home and abroad. The improvement
Corresponding author:
of the comprehensive performance of the launching plat- Jian Hu, School of Mechanical Engineering, Nanjing University of Science
form mainly faces the challenges as follows: (1) when the and Technology, Nanjing 210094, China.
kinetic load is launched, a strong air current impulsion Email: hujiannjust@163.com

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016
2 Advances in Mechanical Engineering

Figure 1. The setup diagram of the small launching platform.

and the other uncertain nonlinearities such as the exter- using different kinds of finite time control methods is
nal disturbance and improve the tracking performance presented in Wang et al.12 Integral terminal sliding
of electro-mechanical servo system just like the launching mode controller and continuous finite time controller
platform, many advanced nonlinear controllers have are designed for the speed loop and compared with the
been investigated,1,2 such as adaptive robust control corresponding control method of asymptotical stabi-
(ARC),3–6 adaptive backstepping control,7–9 active dis- lity; the controller based on the finite time control can
turbance rejection control,10 and neural network–based make the output of system track the desired speed ref-
intelligent control.11 erence signal in finite time and obtain a better dynamic
From the point of view of optimization, the control response and anti-disturbance performance. However,
method of the finite time convergence is the optimal a large chattering phenomenon is caused by high
control method with respect to the time. In the control switching gains. Thus, a composite integral terminal
system, the convergence performance is a key perfor- sliding mode control method based on disturbance
mance index. The finite time convergence of the system observer is proposed to reduce chattering. Through
means that the state of the system reaches equilibrium feed-forward compensation based on disturbance esti-
in a finite time and then always stays at the equilibrium mation, the proposed controller can take a smaller
point. Studies have shown that to control an uncertain switching gain, which could reduce chattering and
system with interference, finite time control technology simultaneously achieve a good disturbance rejection
not only has a faster convergence in a neighborhood of performance. In this article, by referring to the sliding
the origin but also has a better robustness and anti- mode technology in previous studies12–16 and integrat-
interference compared to an infinite time control tech- ing the forward compensation method in Krstic et al.,17
nology (exponential convergence or general other a novel finite time controller is proposed for a launch-
asymptotically convergence). Therefore, finite time con- ing platform, in which a sliding mode disturbance
trol technology’s advantages attract us to put it into observer is proposed to observe the external distur-
the design of a nonlinear controller for the launching bance, and the final control law is derived based on the
platform. Generally speaking, the design method of the estimated disturbance produced by the observer.
finite time controller is classified into three types: the Compared with the control method in Wang et al.,12
open-loop control method, the discontinuous state the control value of the proposal is much smoother due
feedback control method, and the continuous state to the special design and thus the chattering is reduced
feedback control method. Open-loop control method obviously.
lacks anti-interference ability and robustness, which This article is organized as follows. Dynamic model
leads to its limit in practical application. As the repre- of a small launching platform is presented in section
sentative of the discontinuous feedback control ‘‘Dynamics of a small launching platform.’’ Section
method, bang–bang control is not easy to achieve and ‘‘Design of finite time controller’’ gives the sliding mode
is easy to produce jitter and other issues. Therefore, the disturbance observer and the finite time controller
continuous feedback control method based on the sys- design procedure and theoretical results. Simulation
tem state is very worthy of study. and results are presented in section ‘‘Semi-physical
A finite time control problem for a permanent mag- simulation research.’’ Section ‘‘Conclusion’’ contains
net synchronous motor (PMSM) speed servo system the conclusion.

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

Qiu et al. 3

Dynamics of a small launching platform x_ 1 = x2

The launching platform is composed of an azimuth- x_ 2 = u1 u  u2 x2 + d(x, t) ð3Þ
axis servo subsystem and the pitch-axis servo subsys- y = x1
tem, and each subsystem is composed of a controller, a
servo driver, a servo motor, a speed reducer, a rotating where u1 = ku =Jequ and u2 = Bequ =Jequ are the known
mechanism, sensors, and so on. The specific system dia- nominal parameter values, and d(x, t) = f (t, y, y_ )=Jequ is
gram is shown in Figure 1. In order to improve the the transformed external disturbance which is needed
control performance of the system, when building the to be estimated and compensated.
system’s model, the effect of many nonlinear factors,
such as the viscous friction and the coupling distur- Design of finite time controller
bance torque between the rotary mechanisms caused by
gyroscopic effect, should be considered. In addition, Sliding mode disturbance observer design
only the mechanical dynamical characteristics are con- In this section, the design process of the sliding mode
sidered while the current’s dynamic characteristics are observer will be given. To design a sliding mode distur-
neglected when modeling since the electrical response bance observer with finite time convergence property,
speed is much higher than the mechanical response the following sliding mode surface of the observer is
speed. introduced
Since the mathematic model of azimuth-axis servo
subsystem is almost the same as that of the pitch-axis s1 = z1  x 2 ð4Þ
servo system, here the azimuth-axis servo subsystem is
taken as the research object first. The coupling distur- where z1 is the internal dynamics of the observer and it
bance torque applied on one servo subsystem by is designed as
another servo subsystem caused by gyroscopic effect is p =q1
regarded as a time-varying disturbance in the servo sub- z_ 1 =  k1 s1  b1 sign(s1 )  e1 s11  ju2 x2 jsign(s1 ) + u1 u
system. Thus, the controller design for the pitch-axis ð5Þ
servo subsystem is the same as that for the azimuth-axis
servo subsystem, so we just show how to design the where k1, b1, e1, p1, and q1 are the coefficients of the dis-
controller for the azimuth-axis servo subsystem and turbance observer, p1 \ q1 and they are all the positive
show the control performance of azimuth-axis servo odd numbers, k1, b1, and e1 are all the positive numbers
subsystem. Considering all the above factors, the math- and b1  D  |d(x,t)|
ematical model of the azimuth-axis servo system can be 
expressed as 1 when s1 .0
sign(s1 ) =
f1g when s1 \0 ð6Þ
Jequ€y = ku u  Bequ y_  dn  c1 v  c2 v_  t(t, y, y_ ) ð1Þ sign(0) 2 ½1, 1
where Jequ is the equivalent inertia of the PMSM’s rotor Differentiating equation (4) and substituting the sec-
and load; y is the output position of azimuth axis of the ond item of equations (3) and (5) into it, we could
launching platform; ku is the voltage-torque coefficient; obtain
Bequ is the equivalent viscous friction coefficient;
v and v_ are the angular speed and angular acceleration s_ 1 = z_ 1  x_ 2
of pitch axis of the launching platform, respectively; c1 p =q1
=  k1 s1  b1 sign(s1 )  e1 s11 ð7Þ
and c2 are the coupling disturbance coefficients between
two frames; dn is the constant disturbance; and t(t, y, y_ )  ju2 x2 jsign(s1 ) + u2 x2  d(x, t)
is the other uncompensated disturbance and modeling
error. To simplify the controller design, we could just Then, design the estimation of d(x, t) as
regard the constant disturbance, coupling disturbance,
^ t) =  k1 s1  b1 sign(s1 )
d(x,
and all the other uncompensated disturbance and mod- ð8Þ
p =q1
eling errors as a total external disturbance f. Then, the  e1 s11  ju2 x2 jsign(s1 ) + u2 x2
system model can be transformed as

Jequ€y = ku u  Bequ y_ + f (t, y, y_ ) ð2Þ

Lemma 1(Man and Yu18). Assume that there exists a con-
To design the controller, we need to build the sys- tinuous positive definite function V0(t) which satisfies
tem’s state equation. Denote x = ½x1 , x2 T = ½y, y_ T as the following inequality
systematic states, and the systematic state space model
after parametric linearization is as follows V_ 0 (t) + aV0 (t) + lV0g (t)  0, 8t.t0 ð9Þ

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

4 Advances in Mechanical Engineering

then V0(t) converges to the equilibrium point in finite could make the expression s2 + l1 s + l0 Hurwitz,
time ts where s is the differential operator. a1 and a2 satisfy
a1 = a2/(22a2), a22 (0,1).
1 aV 1g (t0 ) + l The time derivative of s2 is given by
ts  t0 + ln 0 ð10Þ
a(1 + g) l
s_ 2 = e_ 1 (t) + l0 sign(e0 (t))je0 (t)ja1 + l1 sign(e1 (t))je1 (t)ja2
where a . 0, l . 0, and 0 \ g \ 1.
= u1 u  u2 x2 + d(x, t)  €xd + l0 sign(e0 (t))je0 (t)ja1
+ l1 sign(e1 (t))je1 (t)ja2 + s_ 1
Theorem 1. Considering the nonlinear system with
model uncertainty (equation (3)), the sliding mode dis- ð14Þ
turbance observer is designed according to equations Design the control law u as
(4), (5), and (8). Then, the disturbance approximation
error of the proposed sliding mode disturbance obser- 1h ^ t)  €xd + l0 sign(e0 (t))je0 (t)ja1
ver is convergent in the finite time, when the output is u=  u2 x2 + d(x,
u1
measured at discrete sampling times with a sufficiently + l1 sign(e1 (t))je1 (t)ja2 + l2 s2 + l3 sign(s2 )js2 ja 
small sampling interval without measurement errors.
ð15Þ

Proof. See Appendix 1. where l2, l3, and a are the parameters of the controller
and l2 . 0, l3 . 0, and 0 \ a \ 1.
Substituting equation (15) into equation (14), we
Remark 1. The proposed sliding mode disturbance
could have
observer still requires that the external disturbance
d(x,t) has a known upper boundary, that is, b1  D, ^ + s_ 1  l2 s2  l3 sign(s2 )js2 ja
s_ 2 = d(t)  d(t)
just like most of the existing sliding mode disturbance ð16Þ
observers. Comparing with the existing results, the pro- =  l2 s2  l3 sign(s2 )js2 ja
posed sliding mode disturbance observer can guarantee
the disturbance estimate error convergent to zero in the Main results
finite time. However, the premise is that there’s no
Lemma 2 (Man and Yu18). Let k1, ., kn . 0 be such that
noise and sampling interval is small enough. In addi- the polynomial sn + kn sn21 + k2s + k1 is Hurwitz
tion, the disturbance approximation error converges and consider the system
only in the sense of equivalent values, since the pro-
posed disturbance estimation law (equation (8)) con- x_ 1 = x2
tains switching terms.
..
. ð17Þ
Finite time controller design x_ n1 = xn
In this section, we develop the tracking control scheme x_ n = u
for the case where all states are available using sliding
There exists b 2 (0, 1) such that, for every a 2
mode control approach. To develop the sliding mode
(1 2 b, 1), the origin is a globally finite time-stable
tracking control, we define
equilibrium for the system (equation (17)) under the
e0 (t) = x1  xd (t) ð11Þ feedback

e1 (t) = x2  x_ d (t) + s1 ð12Þ u = x a (x1 , . . . , xn )

ð18Þ
where xd(t) is the desired position signal and xd(t) is =  k1 sign(x1 )jx1 ja1      kn sign(xn )jxn jan
continuous of order 2, and xd (t), x_ d (t), and €xd (t) all have where a1, ., an satisfy
a upper boundary.
A sliding mode surface of controller is given by ai ai + 1
ai1 = i = 2, . . . , n ð19Þ
2ai + 1  ai
s2 = e1 (t)
ð with an + 1 = 1 and an = a.
+ ðl0 sign(e0 (t))je0 (t)ja1 + l1 sign(e1 (t))je1 (t)ja2 Þdt According to the controller designed above, we
could get the theorem as given below.
ð13Þ
where l0, l1, a1, and a2 are the parameters of the con- Theorem 2. Considering the nonlinear system with the
troller and all of them are bigger than zero. l0 and l1 external disturbance (equation (3)) and assuming full

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

Qiu et al. 5

state information available, the tracking error of the The following three controllers are compared:
closed-loop system is convergent in the finite time with
the proposed sliding mode disturbance observer from 1. Proportional–integral–derivative (PID). This is
equations (4), (5), and (8) and sliding mode tracking the traditional position-velocity-current three-
control law from equations (13) and (15). loop PID controller. For this controller imple-
mentation, the motor driver is set at the velocity
Proof. See Appendix 1. loop. The position-loop PID controller is imple-
mented in the real-time control software, and its
gains tuned carefully via an error-and-try
Remark 2. Although the external disturbance is required method kp = 3, ki = 0.15, and kd = 0, which
to have an upper boundary to make the estimation denote the proportional gain, integral gain, and
error of the disturbance observer convergent to zero in derivative gain, respectively.
the finite time, the boundary of the external disturbance 2. Sliding mode controller (SMC). This is a classic
is not directly used to design the tracking control law. sliding mode controller proposed in Rubagotti
Thus, the control law is more applicable in reality. In et al.19 The guideline for choosing the control-
addition, the proposed disturbance observer–based slid- ler’s parameters c1, c2, and k is that c1 and c2
ing mode tracking control can explore the information should be chosen to make €e0 (t) + c1 e_ 0 (t) + c2
about the characteristic of disturbances, since the out- e0 (t) Hurwitz, and k is larger than the up bound
put of the sliding mode disturbance observer is used in of d(x,t). According to this guideline, the con-
the design of the sliding mode tracking control law troller’s parameters have been chosen as fol-
(equation (15)). It is worth to point out that s2 in equa- lows: c1 = 12, c2 = 48, and k = 5.0.
tion (13) has an additional term s1. The reason is that 3. Sliding mode observer based finite time controller
the disturbance observer’s sliding mode surface s1 (SMFTC). The finite time controller (equation
should be considered in the controller’s sliding mode (15)) with sliding mode disturbance observer
surface s2 to analyze the stability of the whole closed- (equation (8)) is proposed in this article and
loop system. described in section ‘‘Design of finite time con-
troller.’’ According to Lemma 1, l2 , l3 , and a0
Semi-physical simulation research determine the convergent velocity of the sliding
mode surface, and the guideline for these para-
To verify the feasibility of the proposed sliding mode meters is as follows: l2 .0, l3 .0, and 0\a2 \1.
observer–based finite time controller (SMFTC), a semi- According to Lemma 2, l0 , l1 , a1 , and a2 deter-
physical simulation platform has been built, which is mine the convergent velocity of the system
composed of a servo motor, a reducer, a load, a sensor, error, and the guideline for these parameters is
and so on and each part just simulates the correspond- as follows: l0 .0, l1 .0, a1 = a2 =(2  a2 ), and
ing part of a real launching platform. Extensive experi- 0\a2 \1. In addition, both the SMC and the
ments have been carried out on it. The parameters of proposed laws have been designed with the same
the semi-physical simulation platform are given as fol- bounds on disturbance and parametric uncer-
lows (Figure 2): tainties in our article. According to these guide-
lines, the controller parameters have been
Equivalent inertia: Jequ = 0.00138 kg m2; chosen as l0 = 12, l1 = 48, l2 = 2, and l3 = 5
Equivalent viscous friction coefficient: Beq = and a1 = 0.25, a2 = 0.4, and a = 0.3. The
0.4 N m s/rad; guideline for the observer parameters is p1 \ q1
Voltage-torque coefficient: ku = 1.36 N m/V. and they are all the positive odd numbers, k1,
b1, and e1 are all the positive numbers and b1 
|d(x,t)|. According to this guideline, the observ-
er’s parameters have been chosen as k1 = 50,
b1 = 4.0, e1 = 0.05, p1 = 3, and q1 = 7.

The three controllers are first tested for a sinusoidal-

like motion trajectory (x1d = 300 3 sin(3.14 t))
(12exp(20.01t)) (°) and two cases are tested for nor-
mal motion trajectory:

1. Position disturbance case. The position distur-

bance is implemented by applying u 2 0.22x1 to
Figure 2. Semi-physical simulation platform. the physical motion system. This type of

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

6 Advances in Mechanical Engineering

Figure 3. Tracking errors of the three controllers in position disturbance case.

disturbance will change d(x,t) greatly, and the Figure 4, from which we could see that the settling time
system will be operated under heavy unstruc- of disturbance observation is about 4 s, and the control
tured uncertainties. input value is shown in Figure 5). Thus, the total uncer-
2. Position-velocity-input disturbance case. This dis- tainties are greatly reduced and the tracking perfor-
turbance is implemented by applying 0.5u 2 0.15x2 mance has been greatly improved. In addition, the
2 0.28x1 to the physical motion system. finite time control technique helps it converge to its
Similarly, this type of disturbance will greatly equilibrium state faster and more accurately.
change the parameters u1 and u2, together with Case 2. To further test the robustness of the pro-
d(x,t). The dynamics of the system are almost posed algorithm to the external disturbance, the posi-
completely changed and can be thought as the tion disturbance is added, which will be the dominant
most exhaustive verification. factor affecting the tracking performance. The track-
ing performance of the three controllers is shown in
In addition, a simulated coupling disturbance is Figure 6. The disturbance estimation of SMFTC is
applied to the physical system in both cases and the shown in Figure 7 from which we could see that the
coupling coefficients c1 = 0.16 N m s/rad and settling time of disturbance observation is still about
c2 = 0.18 N m s/rad. What’s more, when t = 3 s, a step- 4 s and the control input value is shown in Figure 8.
disturbance of 1.3*(1 2 exp(220(t23)3)) N m is applied As seen, even with strong unstructured disturbances,
to the physical system in both cases. the proposed SMFTC and SMC are able to attenuate
unexpected effects. Contrarily, the PID controller
Case 1. The desired motion trajectory and corre- exhibits bad performance. However, the SMFTC has
sponding tracking performance under the three con- a faster convergence rate and a smaller steady track-
trollers are shown in Figure 3. As seen, the SMC ing error about 0.01° than SMC has. The reason is
controllers have better performance than PID control- that the power item in the finite time controller could
ler in terms of both transient and final tracking errors make the control value bigger and thus inhibit the
since it employed varying structure control method to influence of the disturbance faster when the error is
undermine the influence of the external disturbance. small.
However, the SMFTC has the best tracking perfor-
mance in all three controllers with a maximal steady
Conclusion
tracking error about 0.01°. The reason is that the slid-
ing mode–based observer could capture the distur- In this article, a finite time controller with sliding mode
bance, which could be compensated in the finite time disturbance observer has been proposed for the motion
controller (the disturbance estimation is shown in control of a launching platform driven by two PMSM

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

Qiu et al. 7

Figure 4. The disturbance observed by the sliding mode observer and its error.

Figure 5. The control value of SMFTC and SMC in position disturbance case.

motors. The stability of the closed-loop system is better dynamic response and anti-disturbance perfor-
ensured via Lyapunov’s method, which shows that the mance. Through feed-forward compensation based on
proposed controller could guarantee that the tracking disturbance estimation and special design, the proposed
error of the closed-loop system is convergent in the controller produces a smoother control value, which
finite time. Compared with the general control method could reduce chattering and simultaneously achieve a
of asymptotical stability, the controller based on the good disturbance rejection performance. Comparative
finite time control can make the output of system track experimental results are obtained to illustrate the effec-
the desired reference signal in a finite time and obtain a tiveness of the proposed scheme.

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

8 Advances in Mechanical Engineering

Figure 6. Tracking errors of the three controllers in position-velocity-input disturbance case.

Figure 7. The disturbance observed by the sliding mode observer and its error.

Declaration of conflicting interests Funding

The author(s) declared no potential conflicts of interest with The author(s) disclosed receipt of the following financial sup-
respect to the research, authorship, and/or publication of this port for the research, authorship, and/or publication of this
article. article: This work was supported by the National Natural

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

Qiu et al. 9

Figure 8. The control value of SMFTC and SMC in position-velocity-input disturbance case.

Science Foundation of China under Grant 51505224, the 10. Zheng Q, Gao LQ and Gao Z. On validation of extended
Natural Science Foundation of Jiangsu Province in China state observer through analysis and experimentation. J
under Grant BK20150776, and the National Natural Science Dyn Syst: T ASME 2012; 134: 024505-1–024505-6.
Foundation of China under Grant 51305203. 11. El-Sousy FFM. Intelligent optimal recurrent wavelet
Elman neural network control system for permanent-
magnet synchronous motor servo drive. IEEE T Ind
References
Inform 2013; 9: 1986–2003.
1. Iwasaki M, Seki K and Maeda Y. High-precision motion 12. Wang J, Li S and Li Q. Finite-time control for permanent
control techniques: a promising approach to improving magnet synchronous motor speed servo system via a dis-
motion performance. IEEE Ind Electron Mag 2012; 6: turbance observer. In: 40th annual conference of the IEEE
32–40. industrial electronics society, Dallas, TX, 29 October–1
2. Huang W-S, Liu C-W, Hsu P-L, et al. Precision control November 2014. New York: IEEE.
and compensation of servomotors and machine tools via 13. Fridman L, Levant A and Davila J. Observation of linear
the disturbance observer. IEEE T Ind Electron 2010; 57: systems with unknown inputs via high-order sliding-
420–429. modes. Int J Syst Sci 2007; 38: 773–791.
3. Yao B, Hu C, Lu L, et al. Adaptive robust precision 14. Lu YS. Sliding-mode disturbance observer with
motion control of a high-speed industrial gantry with switching-gain adaptation and its application to
cogging force compensations. IEEE T Contr Syst T 2011; optical disk drives. IEEE T Ind Electron 2009; 56:
19: 1149–1159. 3743–3750.
4. Yao J, Jiao Z and Ma D. Adaptive robust control of DC 15. Hall CE and Shtessel YB. Sliding mode disturbance
motors with extended state observer. IEEE T Ind Elec- observer-based control for a reusable launch vehicle. In:
tron 2014; 61: 3630–3637. AIAA guidance, navigation, and control conference and
5. Yao J, Jiao Z, Ma D, et al. High-accuracy tracking con- exhibit, San Francisco, CA, 15–18 August 2005, pp.1–26.
trol of hydraulic rotary actuators with modeling uncer- Reston, VA: AIAA.
tainties. IEEE/ASME T Mech 2014; 19: 633–641. 16. Nasimullah MI, Khan KW and Wang S. Fractional
6. Yao J, Deng W and Jiao Z. Adaptive control of hydrau- order adaptive fuzzy sliding mode controller for a
lic actuators with LuGre model based friction compensa- position servo system subjected to aerodynamics
tion. IEEE T Ind Electron 2015; 62: 6469–6477. loading and nonlinearities. Aerosp Sci Technol 2015; 43:
7. Ting CS, Chang YN, Shi BW, et al. Adaptive backstep- 381–387.
ping control for permanent magnet linear synchronous 17. Krstic M, Kanellakopoulos I and Kokotovic PV. Non-
motor servo drive. IET Electr Power App 2015; 9: linear and adaptive control design. New York: Wiley,
265–279. 1995.
8. Sun W, Gao H and Kaynak O. Adaptive backstepping 18. Man ZH and Yu XH. Terminal sliding mode control of
control for active suspension systems with hard con- MIMO linear systems. IEEE T Circuits: I 1997; 44:
straints. IEEE/ASME T Mech 2013; 18: 1072–1079. 1065–1070.
9. Yao J, Jiao Z and Ma D. A practical nonlinear 19. Rubagotti M, Estrada A, Castaños F, et al. Integral slid-
adaptive control of hydraulic servomechanisms with ing mode control for nonlinear systems with matched and
periodic-like disturbances. IEEE/ASME T Mech 2015; unmatched perturbations. IEEE T Automat Contr 2011;
20: 2752–2760. 56: 2699–2704.

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016

10 Advances in Mechanical Engineering

Appendix 1 V_ 2 (t) = s2 s_ 2
= s2 ðl2 s2  l3 sign(s2 )js2 ja Þ ð24Þ
Proof of theorem 1
(a + 1)=2 (a + 1)=2
Consider the following Lyapunov’s function candidate =  2l2 V2 (t)  2 l3 V2 (t)
According to Lemma 1, we could conclude that V2(t)
1 2
V1 (t) = s ð20Þ would converge to equilibrium in finite time, which
2 1 means that s2 would converge to zero and s_ 2 would also
Since b1  D, we could get that converge to zero in finite time. At this time, according
to equation (13), we could obtain that
V_ 1 (t) = s1 s_ 1
h i
p =q
= s1 k1 s1  b1 sign(s1 )  e1 s11 1  ju2 x2 jsign(s1 ) + u2 x2  d(x, t)
(p + q1 )=q1 ð21Þ
=  k1 s21  b1 s1 sign(s1 )  e1 s1 1  ju2 x2 jjs1 j + u2 x2 s1  d(x, t)s1
(p + q )=q
  k1 s21  b1 js1 j  e1 s1 1 1 1  d(x, t)s1
(p + q )=q (p + q )=2q
  k1 s21  e1 s1 1 1 1 =  2k1 V1 (t)  2(p1 + q1 )=2q1 e1 V1 1 1 1 (t)

According to Lemma 1, we could conclude that V1(t)

would converge to equilibrium in finite time, which
means that s1 would converge to zero and s_ 1 would also
converge to zero in finite time. e_ 0 (t) = e1 (t)  s1
Substituting equation (8) into equation (7), we could e_ 1 (t) =  l0 sign(e0 (t))je0 (t)ja1  l1 sign(e1 (t))je1 (t)ja2
obtain that ð25Þ
^ t)  d(x, t) = d(x,
s_ 1 = d(x, ~ t) ð22Þ Denote t1 as the time when s2 equals to zero and
denote t2 as the time when s1 equals to zero, then there
which implies that the disturbance estimation error exists a time t3 = max{t1,t2}, after t3
~ t) would converge to zero in finite time, that is,
d(x,
^ t) = d(x, t).
after finite time, d(x, e_ 0 (t) = e1 (t)
e_ 1 (t) =  l0 sign(e0 (t))je0 (t)ja1  l1 sign(e1 (t))je1 (t)ja2
Proof of theorem 2 ð26Þ
Consider the following Lyapunov’s function candidate According to Lemma 2, we could conclude that e0(t)
and e1(t) would converge to equilibrium in finite time,
1 2
V2 (t) = s ð23Þ which means that the tracking error of the system
2 2 would converge to zero in finite time.
The time derivative of V2(t) is given by

Downloaded from ade.sagepub.com at Universite Paris-Nanterre on October 25, 2016