Proceedings of the 36th Chinese Control Conference

July 26-28, 2017, Dalian, China

Observer-based Super Twisting Sliding Mode Control for Fixed Wing

Unmanned Aerial Vehicle
Qi Dong1, Qun Zong1, Ruilong Zhang1, Chaofan Zhang1, Jing Zong2
1. School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
E-mail: [email protected]
2. College of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072
E-mail: [email protected]

Abstract: Fixed-wing unmanned aerial vehicle (UAV) should be under control in the presence of model uncertainty and external
disturbance, which is considered as torque perturbation in this paper. Such a challenge imposes tight requirements to the
enhanced robustness and accuracy of the vehicle autopilot. The key of this paper is to introduce the model of attitude dynamics
and propose a novel finite-time observer-based super twisting sliding mode control scheme for fixed-wing UAV. The observer
designed in this paper can not only estimate the attitude rate but also the torque perturbation exactly in finite time. Then
combined with the observer, a continuous homogeneous second-order sliding mode controller based on the terminal sliding
mode and super twisting algorithm is designed to achieve a finite-time and exact attitude tracking with chattering attenuation.
Finally, some simulations are provided to verify the effectiveness and robustness of the integrated control scheme in attitude
tracking.
Key Words: Fixed-wing UAV, finite-time observer, super twisting algorithm, torque perturbation, attitude tracking

convex integrated design method for a small-scale

1 Introduction
helicopter in 2010. In 2005, backstepping technique was
The research on fixed-wing unmanned aerial vehicle designed for fixed-wing UVAs to achieve desired velocity
(UAV) has received much attention during the last few and roll angle by Ren [3]. In 2008, the backstepping control
decades because of the need for military and civilian method was proposed by to deal the problem of path
application. No matter on the cluster formation or single following control for a small fixed-wing UAV by Dongwon
vehicle control, fixed-wing UAV is subject to significant Jung and Panagiotis Tsiotras [4]. Using CMAC (Cerebellar
model uncertainties and severe torque perturbation due to Model Articulation Controller ) nonlinear approximator, a
the complex aerodynamic characteristics of the aviation. direct approximate adaptive control was presented for an
Such a challenge imposes tight requirements to the experimental prototype quadrotor helicopter in 2011 [5].
enhanced robustness and accuracy of the vehicle autopilot. The reference [6] derived an explicit transfer function from
Therefore, a more advanced controller, which is robust to the plant input to the output to guarantee the stability of the
the uncertainties and disturbances with high control controlled system by using nonlinear dynamic inversion
accuracy, would be a significant improvement over current process. By using dynamic model inversion, a trajectory
fixed-wing UAV flight technology. In summary, the urgent tracking controller for tiltrotor UAV was presented by B-M
task of fixed-wing UAV attitude control is to track the Kim in 2010 [7], furthermore, an online adaptive neural
attitude commands more accurately and rapidly while networks were employed to compensate the model inversion
owning sufficient robustness and adaptiveness to external errors. In [8], quaternion-based dynamic output feedback for
disturbances and inaccuracies of the model. the attitude tracking problem of a rigid body without
In the past few years, some researches have been made in velocity measurement was proposed by Tayebi in 2008. In
the UAV attitude control via various control methods, such 2013, [9] proposed a quaternion-based backstepping
as PID, dynamic inversion, adaptive control and controller which was able to track the trajectory and show
backstepping control. In [1], a PID controller with the uniformly asymptotically stability by using rotational
noninteger order was proposed for quadrotor UVA and the dynamics. Additionally, in 2012, a real-time attitude
coefficients of disturbance were approximated by neural stabilization control scheme for the efficient performance of
networks. In [2], Song proposed a controller by applying the a mini-UVA Quad-Rotor was proposed in [5]. These linear
and nonlinear control methods have got great development
*
This work is supported by National Natural Science Foundation (NNSF) in this respect for UAV attitude control. However, sliding
of China under Grant 61673294 and 61273092, Ministry of Education mode control (SMC) is still a more effective way
Equipment Development Fund under Grant No.6141A02033311, Nature
Science Foundation of Hebei Province under Grant F2015202150, Youth
Foundation of Hebei Educational Committee under Grant QN2015068.

6382
considering the UAV dynamics with serious uncertainty and The angular acceleration of fixed-wing UAV can be
external disturbances. written by using Euler’s moment equation as

Jωnb = - S  ωnb  Jωnb + τ d

More recently, compared to a decoupled single-variable
 (1)
structure [11], the multivariable super twisting sliding mode
7
control (STSMC) was proposed to obtain more elegant where ωnb ª¬Z p Zq Zr º¼  3u1 represents the angular
solutions [12]. As a matter of fact, the fixed wing UAV
velocity of the body frame, relative the NED frame
attitude model shows an obvious multivariable
characteristic. Therefore, the application of multivariable referenced in the body frame, J  3u3 is the inertia matrix
STSMC to fixed wing UAV is bound to bring a more and S ( ) is the symbol of the skew-symmetric operator
excellent and practical attitude control scheme. The key
ª 0 Zr Zq º
contributions are summarized as follows: « »
S (ω ) « Zr Z p »
b (2)
n 0
• A novel integrated multivariable finite-time observer
« Zq Zp 0 »¼
and second-order super twisting sliding mode controller ¬
design is proposed for fixed wing UAV to guarantee that the The aerodynamic moments acting on the fixed-wing
attitude tracking error converge to zero in finite time with UAV can be described as [9]
the control chattering elimination effectively.
qS >bCl cCm bCn @
T
• The multivariable finite-time observer designed in this  τd (3)
paper can not only estimate the attitude rate but also the
where q 0.5UVa 2 is the dynamical pressure; b is the span
torque perturbation exactly in finite time, which can deal
with the unpredictable attitude rate problem in the most of the wing; c is the mean aerodynamic chord and the S is
applications. the surface area of the wings. Cl , Cm , Cn are the
• The continuous finite-time and homogeneous sliding aerodynamic coefficients which are approximated as
mode controller designed in this paper combines super
b b
twisting and terminal sliding mode characteristics and can Cl Cl0  ClE E  Cl p Z p  Clr Zr  ClGa G a  ClGr G r
2 Va 2 Va
obtain superior rapidity while retaining the tracking
c
accuracy. Cm Cm0  CmD D  Cmq Zq  CmGe G e
2 Va
b b
2 The Attitude Dynamics of Fixed Wing UAV Cn Cn0  CnE E  Cn p Z p  Cnr Zr  CnGa G a  CnGr G r
2 Va 2 Va
In this section, the attitude dynamics model of fixed-wing
UAV is introduced. The UAV attitude and coordinate where G a , G e , G r are the deflection angels of the aileron,
systems are shown in Fig. 1. elevator and rudder respectively. Let τ d τ + d w , and we
can get

Jωnb = - S  ωnb  Jωnb + τ + d w (4)

D
E Fnx
where d w is described as
Fnz
E Fbx
Fbz ª § b b ·º
Fby « qSb ¨¨ Cl0  ClE E  Cl p p  Clr r ¸¸ »
Fny « © 2 V a 2 Va ¹»
« »
« § c · »
Fig. 1 UAV attitude and coordinate systems dw qSc ¨¨ Cm0  CmD D  Cmq q ¸¸
« 2 Va ¹ »
« © »
The rotational dynamics of UAV, the North-East-Down
« § b b ·»
(NED) frame Fn , the body frame Fb and the relationship « qSb ¨ Cn0  CnE E  Cn p p  Cnr r ¸¸ »
«¬ ¨
between these two frames are shown in Figure 1. The NED © 2 V a 2 Va ¹ »¼
frame has its Fnx-axis aligned towards the North, Fny is The unit quaternion qn ,b is defined as
pointing East and Fnz is pointing towards the center of the
earth. The body frame is suited to the rigid body, where Fnx  qn,b ªcos §¨ - ·¸ k T sin §¨ - ·¸º > q0 q1 q2 q3 @T (5)
« »
is pointing along the fuselage of the body, Fny is pointing ¬ ©2¹ © 2 ¹¼
through the right wing while Fnz is pointing downwards. D
is the angle of attack and E is the sideslip angle.

6383
which represents the rotation from the body frame to the ª qb1 qb 2 qb3 º
NED frame. - is the angle of rotation around the unit vector « qb 3 qb 2 »» b
1 « qb 0
k and q02  q12  q22  q32 1 is satisfied. qe
2 « qb 3 qb 0 qb1 » n
 ω  ωd  (11)

The quaternion kinematics of the fixed-wing UAV is « »

¬ qb 2 qb1 qb 0 ¼
given by
Define the error of attitude rate
ª q1 q2 q3 º
«q ωe ωnb  ωd
1« 0 q3 q2 »» b (6)
qn ,b ω
2 « q3 q0 q1 » n where ωd is desired angular velocity. Thus equation (4) and
« »
¬ q2 q1 q0 ¼ (11) can be translated into
The quaternion of attitude error between NED frame and ª qb1 qb 2 qb 3 º
« »
a desired frame is defined as 1 « qb 0 qb 3 qb 2 »
qe ω
qd ,n ª¬ qn,1 qn,2 qn,3 qn,4 º¼
T
(7) 2 « qb 3 qb 0 qb1 » e (12)
« »
¬ qb 2 qb1 qb 0 ¼
Thus the attitude error between the body frame and the
desired frame can be denoted by using quaternion product as Jωe = - S  ωe  ωd  J  ωe  ωd   Jωd + τ + d w

 qd ,b qd ,n … qn,b (8)
3 The Finite-time Integrated Control for Fixed
Remark 1: The symbol … represents the product of two Wing UAV
T
quaternions. Assuming q1 ª¬qo1 qx1 q y1 qz1 º¼ is one 3.1 The Finite-time Observer Design for UAV

T Let qev [qb,2 qb,3 qb,4 ]T and in order to describe the

quaternion and q2 ª¬qo 2 qx 2 q y 2 qz 2 º¼ is the other one,
flight process of UAV accurately as much as possible, the
we can obtain
external disturbances should be considered. Hence a more
ª qo1 qx1  q y1 qz1 º ª qo 2 º realistic model of UAV can be described as
«q qo1  q z1 q y1 »» « qx 2 »
q1 … q2 « x1 « » (9) 1
« q y1 qx1 » « q y 2 » R1ωe
q z1 qo1  qev 2 (13)
« »« »
¬« qz1  q y1 qx1 qo1 ¼» ¬ qz 2 ¼ ωe  J 1
1
SJ ˜ (ωe  ωd )  ωd  J 1τ  J 1'
The goal of the attitude quaternion is defined as where
> r1 0 0 0@
T
qd ª qb 0 qb3 qb 2 º
R1 «q qb 0 qb1 »»
where >1 0 0 0@ and > 1 0 0 0@ represent the « b3
T T

«¬ qb 2 qb1 qb 0 »¼
same physical state, although mathematically they are
and ' d w is the disturbance in the flight process of UAV,
different. Using qd to represent desired attitude quaternion
which is a Lipschitz continuous time signal with Lipschitz
of UAV has double-valueness, thus the quaternion error can
be shifted to the origin as constant G , i.e. ' d G almost everywhere.

ª1 qb ,00 º Let
« q »
ª1 Kd ,b º « b ,1 »
1
qe « ε »
(10) We R1ωe (14)
¬ d ,b ¼ « qb ,2 » 2
« »
¬ qb ,3 ¼ Then it can be obtained that

when qe 0 ˈit means that the flight reaches the goal qev We
1 1 1 (15)
attitude. We ( R1  R1 J 1SJ )ωe  R1 J 1SJωd
2 2 2
The quaternion kinematic of quaternion error is given by
1
  R1ωd  T  μ
2
1 1
> μ1 , μ2 , μ3 @
T
where μ R1 J 1' , T R1 J 1τ and μ .
2 2

6384
 The finite-time observer and controller are designed 1 1 1
τ  R1ωe  R1 J 1SJ ˜ (ωe  ωd )  R1ωd
based on oriented control model as Eqs. (15). For this 2 2 2
purpose, we design the following finite-time multivariable ª k11 IL1 (qev1 ,We1 ) sign(IL1 ) º
1/3
(23)
2 « »
observer to estimate We and P .
  μˆ  L3 « k12 IL 2 (qev 2 ,We 2 ) sign(IL 2 ) »  z
1/3

1 1 1
« 1/3 »
Wˆe ( R1  R1 J 1SJ )ωˆ e  R1 J 1SJωd «¬ k13 IL 3 (qev 3 , We3 ) sign(IL 3 ) »¼
2 2 2 (16)
1
  R1ωd  T  -
2
ª k21sign(IL1 ) º
where the output error injection signal
z  L «« k22 sign(IL 2 ) »»
V «¬ k23 sign(IL 3 ) »¼
- k1 1
 [  k2V
V 2 (17) where the parameters k1i , k2i (i 1, 2,3) and L are positive
V
[  k3  k4V gains to be designed.
V
Define the estimate error of observer as
where V We  Wˆe And define z [  P .We can obtained ef [ P1  Pˆ1 , P2  Pˆ 2 , P3  Pˆ3 ]T
(24)
that [e f 1 , e f 2 , e f 3 , ]T
1 1
V Ŵˆe
We  W R1 (ωe  ωˆ e )  R1 J 1SJ (ωˆ e  ωe ) Let
2 2
V (18) K z  ef
  μ  k1 1
 [  k2V (25)
V 2
Substituting (11) and (12) into (3), it can be derived that
V
z [  μ  k3  k4V  μ qev We
V
ª k11 IL1 (qev1 ,We1 ) 1/3 sign(IL1 ) º
and we let 2« » (26)
 L « k12 IL 2 (qev 2 ,We 2 ) sign(IL 2 ) »  K
3 1/3
We
J
1 1
R1 (ωe  ωˆ e )  R1 J 1SJ (ωˆ e  ωe )  μ « 1/3 »
2 2 (19) «¬ k13 IL 3 (qev 3 ,We3 ) sign(IL 3 ) »¼
I μ ª k21sign(IL1 ) º
and the terms J and I are assumed to satisfy Ki « »
 L « k222 sign(IL 2 ) » +e f

J d G2 V ¬« k23 sign(IL 3 ) ¼»
(20)
I d G3 From (6) and Theorm 1, it can be concluded that e fif d G1i

During the sliding motion V V 0 and z 0 , since with the finiteboundary G1i ! 0 
from (19), J (0) 0 . We can get z [  μˆ 0 . If the Lemma 1: Considering the third-order system (13) with a
estimate μ̂ is chosen as [12] uniformly bounded signal e f . Then, there exist positive

μˆ [ (21) values of the gains k1i , k2i (i 1, 2,3) and L such that the
variables qev , We and Ki converge to zero globally,
3.2 Finite-time continuous controller design
uniformly and in finite time, despite any bounded
Considering the second-order uncertain system (12), define
perturbation e fif d G1i . The detailed proof and gains selection
the continuous differentiable variable as
can be found in the reference [9].
 ILi (qevi ,Wei ) qevi  Di Li 1/2 «¬Wei »¼3/2 ˈ(i 1, 2,3) (22)
where We [We1 ,We 2 ,We3 ] . The finite-time continuous 4 Simulation Results
controller is designed as Some simulations are provided to verify the effectiveness
and robustness of the integrated control scheme in attitude
tracking. In this simulation, the UAV parameters are set as

6385
 c 0.76m , b 1.96m , S 1.37m2 , m 20.64kg , 1
qe0
2 2 qe1
I xx 1.6073kg m , I yy 7.5085kg m ,
0.5 qe2

I zz 7.1865kg m , J xz 2
0.56144kg m 2 qe3

e
The aerodynamic coefficients are set as 0

q
Cl0 0.007 , ClE 0.3666 , Cl p 1.6623 ,
-0.5
Clr 0.1947 , ClG 0.5053 , ClG 0.1091 ,
a r
-1
Cm0 0.0445, CmD 0.7067 , Cmq 1.7125 , 0 2 4
time(s)
6 8 10

CmG 0.5428 , Cn0 0.0023 , CnE 0.1098 , Fig. 3 UAV attitude quaternion tracking error curves
e

Cnp 0.1840 , Cnr 0.5745 , CnG 0.0669 , 1

a Zpe

CnG 0.1961 . 0.5

Zqe
r
Zre
The finite-time observer and controller parameters are set

Ze(rad/s)
0
in the Table 1.
-0.5
Table 1 Observer-Controller Parameters
-1
Title Value
k11 6 -1.5
0 2 4 6 8 10
time(s)
k12 6
Fig. 4 UAV attitude rate error curves
k13 6
k21 2 40
Wp
k22 2 Wq
20
Wr
k23 2
W (Nm)

0
L 1
k1 k2 10 -20
k3 k4 1
-40

In Fig. 2 and Fig. 3, the attitude quaternion and attitude -60

0 2 4 6 8 10
quaternion error tracking curves are provided respectively to time(s)
demonstrate the process of attitude tracking. From the Fig. 5 UAV attitude and coordinate systems
results, it can be obtained that the integrated control scheme
1
proposed in this paper can track the attitude commands with ef1
ef1
fast convergence and small tracking error after a finite time.
0.5 ef1

1.5
q
b0
ef

q 0
b1
1 q
b2
q -0.5
b3
b

0.5
q

-1
0 1 2 3 4 5
0 time(s)
Fig. 6 The estimate error of observer
-0.5
0 2 4 6 8 10 Fig. 4 shows the UAV attitude rate error curves and from
time(s)
Fig. 2 UAV attitude quaternion tracking curves
the simulation, we can draw a conclusion that the angular
rates change with fast convergence and low chattering and
are all less than 1rad/s which satisfies the engineering
application requirement. The aerodynamic moments curves
are presented in Fig. 5. From that, it can be observed that the

6386
aerodynamic moments changes continuously with lower [2] Song B, Mills J K, Liu Y. Nonlinear dynamic modeling and
control of a small-scale helicopter, International Journal of
chattering. In Fig. 6, the estimate errors of observer are
Control, Automation and Systems, 8(3): 534-543, 2010.
shown to demonstrate the accuracy and the rapidity of the [3] AIAA. Nonlinear Trajectory Tracking for Fixed Wing UAVs
designed multivariable observer. It can be seen from the via Backstepping and Parameter Adaptation, Aiaa Journal,
2005.
results that the estimation errors converge to the equilibrium [4] Jung D, Tsiotras P. Bank-To-Turn Control for a Small UAV
point in a short time, which can compensate for the Using Backstepping and Parameter Adaptation, World
controller more effectively. Congress, 41(2): 4406-4411, 2008.
[5] Nicol C, Macnab C J B, Ramirez-Serrano A. Robust adaptive
control of a quadrotor helicopter, Mechatronics, 21(6):
5 Conclusions
927-938, 2011.
A novel finite-time integrated control scheme including a [6] Ducard G, Geering H P. Stability analysis of a dynamic
inversion based pitch rate controller for an unmanned aircraft,
multivariable observer and second-order super twisting in International Conference on Intelligent Robots and
sliding mode controller is proposed for fixed wing UAV to Systems, 2008: 360-366.
guarantee that the attitude tracking error converge to zero in [7] Kim B M, Kim B S, Kim N W, et al. Trajectory tracking
controller design using neural networks for a tiltrotor
finite time with the control chattering elimination effectively. unmanned aerial vehicle, Proceedings of the Institution of
The multivariable finite-time observer designed in this paper Mechanical Engineers Part G Journal of Aerospace
Engineering, 224(8): 881-896, 2010.
can estimate torque perturbation exactly in finite time, and
[8] Tayebi A. Unit Quaternion-Based Output Feedback for the
the continuous finite-time and homogeneous sliding mode Attitude Tracking Problem, IEEE Transactions on Automatic
controller designed in this paper combines super twisting Control, 53(6): 1516-1520, 2008.
[9] Oland E, Kristiansen R. Quaternion-based backstepping
and terminal sliding mode characteristics and can obtain control of a fixed-wing unmanned aerial vehicle, in IEEE
superior rapidity while retaining the tracking accuracy. The Aerospace Conference, 2013:1-7.
simulations are provided to verify the effectiveness and [10] Gonz, Lez I, Salazar S, et al. Real-Time Attitude Stabilization
of a Mini-UAV Quad-rotor Using Motor Speed Feedback,
robustness of the integrated control scheme in attitude Journal of Intelligent & Robotic Systems, 70(1): 93-106,
tracking. 2013.
[11] Moreno J A, Osorio M. Strict Lyapunov functions for the
super-twisting algorithm, IEEE Transactions on Automatic
References Control, 57(4): 1035-1040, 2012.
[12] Nagesh I, Edwards C. A multivariable super-twisting sliding
[1] Efe M Ö. Neural Network Assisted Computationally Simple
mode approach, Automatica, 50(3): 984-988, 2014.
PI, D, Control of a Quadrotor UAV, IEEE Transactions on
[13] Kamal S, Moreno J A, Chalanga A, et al. Continuous terminal
Industrial Informatics, 7(2): 354-361, 2011.
sliding-mode controller, Automatica, 69: 308-314, 2016.

6387