Proceedings of the 8th International Conference on WeCC.
5
Systems and Control, Marrakech, Morocco, October
23-25, 2019
Sliding mode differentiator-based robust backstepping control of
nonlinear uncertain system*
Fatma Massaoudi, Dorsaf Elleuch, Hafedh Abid, Tarak Damak
This approach presents a compromise between robustness
Abstract— A new robust technique control is developed in
this paper for a nonlinear uncertain system. This technique is against uncertainties and precision that limites derivators
based on backstepping method integrating sliding mode number use by many fields [9]. The virtual controller
differentiator used to estimate the derivative of tracking error. deivative is determinated by a higher-order sliding mode
This technique has many objects especially minimizing sensors differentiator [11].
number and ameliorating control performance. The efficiency
of this technique is illustrated by simulation results. The I. BACKSTEPPING CONTROL
differentiator used is robust in presence of external
disturbance. A. Definition
I. INTRODUCTION The backstepping technique is based on breaking the
Many new technologies that are related to internal As a problem of design on the global system down to a sequence
robust approach to control nonlinear systems (NLS), of sub-problems on the lower order systems, and recursively
backstepping control (BSC) is widely used. The backstepping use virtual controls to have the intermediate control laws
controller design method offers an effecient tool of designing based on the Control Lyapunov Function. Beginning from
controllers for nonlinear uncertain systems having triangular lower order system. [1]
structure [1]. Backstepping technique is a performing design
It is a systematic design approach used especially by
strategy dealing with nonlinear system problems. In order to strict feedback Systems. BSC method is based on defining
develop the BSC algorithm, the suitable Lyapunov function
the error variable. Then defining the Lyapunov function.
for nonlinear system is chosen and the virtual BSC laws are
After that, derivating this Lyapunov function and
constructed recursively in some steps then, a stable control
determinating a virtual variable. Finally, defining the next
system is obtained. Thanks to this approach, the closed-loop
integrator.
system global stability is reached [2]. The advantages of
backstepping control include the guaranteed global of Thus, the most crucial object of backstepping control is
regional stability, the stress on robustness, and computable to obtain loop systems equivalent to one order system that
transient performance [1]. The backstepping method has guarantees the lyapunov stability.
received a great deal of interest since its proposition, and has
been widely applied to the control problems arising from the The backstepping strategy has been used also in aviation
aerospace engineering [3][4][5], mechanical engineering technology, to reduce the stabilization and trajectory tracking
[6]….Yet, satisfactory performance is not easly reached. In problems of quadrotor .However, it is not easy to obtain
fact, the BSC parameters are chosen arbitrarily. Selecting performance owing to the arbitrary choice of BSC
proper parameters is important to have a satisfying response. parameters.
Indeed, an inappropriate parameters selection can cause B. Backstepping control law
system instability . Sensors are widely used in experiments
and applications. However, they are costly and not available.
The NL system has the following form:
Thus, differentiators are integrated minimizing sensors
number . Constracting a robust sliding mode differentiator
consists a classical methodology for signal processing x1 x2
x2 f x h x u
strategy [7]. This strategy is based on estimating the system
state output or variables in order to minimize observation
error. This procedure does not need the unknown input to
become uniformly limited [8]. However, these numerical With u is the BSC control, x1 ; x2 the state vector, f(x)
differentiators sometimes lead to gains adjustment problems. and h(x) are uncertain nonlinear functions.
The tracking-error has the following expression:
*Research is supported by the LAB-STA, ENIS, Sfax, Tunisia
Fatma Massaoudi is with LAB-STA, ENIS, University of Sfax,3038,
Sfax, Tunisia; e-mail: [Link]@[Link] .
x1 x1d
Dorsaf Elleuch is with LAB-STA, ENIS, University of Sfax,3038, Sfax,
Tunisia; e-mail: dorsafellouch@[Link] .
Hafedh Abid is with LAB-STA, ENIS, University of Sfax,3038, Sfax, x1d: The desired state vector.
Tunisia; e-mail: abidhafedh@[Link] The backstepping error variable is defined by the following
Tarak Damak, is with LAB-STA, ENIS, University of Sfax,3038, Sfax,
Tunisia; e-mail: [Link]@[Link] expression:
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 235
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
z1 with is the derivative of and is a u f , , x1 , x1d , x1d , x1d , z1 , z2
positive constant chosen in order to guarantee the system
stability.
z1 x1 x1d x1 x1d
z1 derivative is described by this equation: Thus, the BSC control law depends on differentiator
z1 x1 x1d x1 x1d variables. For that, a sliding mode differentiator estimates the
tracking error derivatives.
The lyapunov function is chosen as follows :
The BSC block diagram is illustred as follows :
1 2
v1 z1
2
So, the derivative of lyapunov function is expressed by:
v1 z1 z1
v1 z1 ( x2 x1d x2 x1d )
v1 c1 z12 z1 (c1 z1 x2 x1d x2 x1d )
The system stability condition is v1 0
v1 c1 z12 z1 z2
So z2 c1 z1 x2 x1d x2 x1d
z2 x2 x1d x2 x1d 1 Figure1. BSC block diagram
with 1 c1 z1 c1 ( x1 x1d x1 x1d )
II. SLIDING MODE DIFFERENTIATOR
z2 x2 x1d x2 x1d 1
z2 x2 x1d x2 x1d c1 x1 x1d x1 x1d Most of applications integrate differentiators to estimate
z2 f hu x1d x2 x1d c1 z2 1 outputs or state variables.
1 2 A various types of differentiators can be studied like :
Considering v2 z2 v1
2 Prescribed convergence law;
The derivative of v2 is:
Twisting Algorithm (TWG) ;
v2 v1 z2 z2
SuperTwisting Algorithm .
The system stability is verified when v2 0
The SM numerical derivator used has the following
v2 c1 z12 z1 z2 z2 z2
structure :
Then
v2 c1 z12 z1 z2 z2 z1 z1 c2 z2 c2 z2 f hu x1d x2 x1d c1 z2 1
Z 0 Z1 0 L1/ n 1 Z 0 f (t ) n / n 1 sign( Z 0 f (t ))
So, ( n 1)/ n
v2 c1 z12 c2 z22 Z1 Z 2 1 L1/ n Z1 Z 0 sign( Z1 Z 0 )
Then, Z Z L1/ ( n 1i ) Z Z ( n i )/ ( n 1 i )
sign( Z i Z i 1 )
c2 z2 z1 f hu x1d x2 x1d c1 z2 1 0 i i 1 i i i 1
Z n 1 n 1 [Link]( Z n 1 Z n 2 )
The BS expression is given by this equation:
1
u
h
c z
2 2 z1 f x1d x2 x1d c1 z2 1 Where Z0 , Z1...Z n present the estimate expressions of
f , f ,..., f ( n) , the chosen coefficients are
Where c1 and c2 are the SBC parameters. 1 n .... 1 0 and f(t) is the signal to be
x1 x1d , x1 x1d , x1 x1d and x1 x1d differentiated. [8] .
Then After that, a fourth-order differentiator is integrated.
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 236
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
Z 0 Z1 0 Z 0 3/ 4 sign( Z 0 ) R
c2 z2 z1 2 x1d
2/3
u mR 2 R
Z1 Z 2 1 Z1 Z 0 sign( Z1 Z 0 )
g
sin x x c z
1
Z Z Z Z 1/ 2 sign( Z Z )
1 1d 1 2
R
2 3 2 2 1 2 1
Z 3 3 sign( Z 3 Z 2 ) R 0.8 0.1sin8t 0.3cos4t , m=1, g=9.81
The desired position trajectory is x1d 0.5sin 0.5t 0.5cos t
The constant is chosen as 6
Where Z 0 is the estimation of The fourth-order differentiator parameters are the following:
Z1 is the estimation of 0 5 , 1 14 , 2 6 , 3 0.2
The BSC parameters are chosen in order to guarantee system
Z 2 is the estimation of convergence as follows:
Z3 is the estimation of c1 40 and c2 40
III. SIMULATION RESULTS
A. Results without external disturbance
a) Without SM differentiator
BSC tracking results before and after adding external Fig.3 and Fig.4 illustrate respectively the tracking
disturbance are shown and results are comparing without and trajectory and the torque variation using the BSC.
with differentiator.
Simulation results prove that the BSC is performing and
A pendulum is considered. This dynamic system has effective and the system convergence is reached.
motions which are restricted to some vertical plane. It is
shown in the following figure [10].
Figure 2. Pendulum
Figure 3. Tracking trajectory
The pendulum dynamic model in state space has the
following equation [10]:
x1 x2
R g 1
x2 2 x1 sin x1 u
R R mR 2
With :
m : the mass moving along the pendulum Rod ;
R(t) : the distance swept by the mass from O ;
g : the gravitational constant ;
x1 x : the angular position, u: the torque.
The SBC law is expressed as follows:
Figure 4. Backstepping control
b) With SM differentiator
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 237
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
A fourth-order differentiator is integrated for estimating
the successives derivates of the tracking error sigma.
Fig.5 and Fig.6 show respectively the tracking trajectory
and the torque variation using the SBC after integrating SM
differentiator. The system convergence can be concluded and
the results with differentiator are more performing than those
without differentiator. In fact the trajectory position and the
desired position are similar .
Fig.7, Fig.8, Fig.9 and Fig.10 show sigma variation and
its derivatives .The obtained results prove that the estimated
derivatives are more effective than those without using
differentiator. Thus, differentiator robustness is concluded.
Integrating numerical derivator has improved system
performance and stability.
Figure 8. Derivative of sigma variation
Figure 5. Tracking trajectory with differentiator
Figure 9. Second derivative of sigma
Figure 6. Backstepping control with differentiator
Figure 10. Third derivative of sigma
B. Results with external disturbance
The external disturbance used is chosen “Random Signal”
added to the angular position x which value is between -0.03
and 0.03.
a) Without SM differentiator
Fig.11 and Fig.12 represent respectively the tracking
Figure 7. Sigma variation trajectory and the torque variation using the backstepping
controller with external disturbance. The system stability and
convergence are reached despite the perturbation.
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 238
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
Figure 11. Tracking trajectory with uncertainty Figure 14. Backstepping control with differentiator and external
disturbance
Figure 12. Backstesping control with uncertainty
b) With SM differentiator and external disturbance Figure 15. Sigma variation
Fig.13 and Fig.14 describe respectively the tracking
trajectory and the torque variation using the SBC after
integrating SM differentiator and external disturbance. The
obtained results prove that system convergence is reached in
the presence of perturbation.
Fig.15, Fig.16, Fig.17 and Fig.18 represent sigma
variation and its derivatives .Simulation results prove that the
estimated derivatives are more performing than those without
using differentiator. So, the differentiator is robust against
external disturbance.
Figure 16. Derivative of sigma variation
Figure13. Tracking trajectory with differentiator and external disturbance
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 239
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
V. REFERENCES
[1] Z. Sheng, [Link]-qi ,”Dynamic backstepping control for pure-
feedback nonlinear systems”, Computational Aerodynamics
Institution,pp.1-20,2017.
[2] J. Zhang, Ch. Sun, R. Zhang, Ch. Qian, L. Xue, “Adaptive
Backstepping Controller Design for Reentry Attitude of Near Space
Hypersonic Vehicle”, International Conference on Intelligent Science
and Big Data Engineering, vol. 8261, pp.642-648, 2013.
[3] Lee T, and Kim Y. “Nonlinear adaptive flight control using
backstepping and neural networks controller”, Journal of Guidance,
Control and Dynamics,vol. 24(4),pp.675-682,2001.
[4] Farrell J, Sharma M, and Polycarpou M. Backstepping “based flight
control with adaptive function approximation”, Journal of Guidance,
Control and Dynamics, vol .28(6), pp.1089-1102,2005.
Figure 17. Second derivative of sigma variation [5] Sonneveldt L, Chu Q P, and Mulder J A. “Nonlinear flight control
design using constrained adaptive backstepping”, Journal of
Guidance, Control and Dynamics, vol. 30(2),pp. 322-336,2007.
[6] Bridges M M, Dawson D M, and Abdallah C T. “Control of rigid-link,
flexible-joint robots: a survey of backstepping approaches” ,Journal of
Robotic Systems , vol. 12(3), pp.199-216,1995.
[7] M. A. M. Basri, K. A. Danapalasingam, A. R. Husain, “design and
optimization of backstepping controller for an underactuated
autonomous quadrotor unnamed aerial vehicle”, TRANSACTIONS
OF FAMENA XXXVIII-3 ,pp. 27-44,2014.
[8] [Link], « Robust exact differentiation via sliding mode technique »,
Automatica, vol. 34, pp. 373-384, 1998.
[9] L. Sidhom, « Etude et contribution aux techniques de
differentiations », INSA de Lyon, 2011.
[10] [Link] , “Principles of 2-Sliding mode design”,Preprint submitted
to Automatica, vol.43(4), pp. 576-586,2007.
[11] [Link], [Link] and [Link] , “Sliding-Mode-Based differentiation
and its application”,Elsevier,vol.50(1),pp.1699-1704,2017
Figure 18. Third derivative of sigma variation
IV. CONCLUSION
A robust technique is used in order to control a nonlinear
dynamic system. A backstepping control based on SM
differentiator is developed for this uncertain nonlinear
system. In fact the differentiator is used to the tracking error
derivatives estimation. Simulation results confirm the
performance and efficiency of this approach. This numerical
derivator is robust against external disturbance. In order to
ameliorate robustness of uncertain nonlinear system,
backstepping control and sliding mode control combination
can be applied.
978-1-7281-1938-0/19/$31.00 ©2019 IEEE 240
Authorized licensed use limited to: UNIVERSITE DE SFAX. Downloaded on June 19,2020 at [Link] UTC from IEEE Xplore. Restrictions apply.
