See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245231122

Modeling and Backstepping-based Nonlinear Control Strategy for a 6 DOF

Quadrotor Helicopter

Article  in  Chinese Journal of Aeronautics · June 2008

DOI: 10.1016/S1000-9361(08)60034-5

CITATIONS READS

214 2,247

2 authors, including:

Ashfaq Ahmad Mian

Nanjing University of Aeronautics & Astronautics
12 PUBLICATIONS   415 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Ashfaq Ahmad Mian on 09 April 2014.

The user has requested enhancement of the downloaded file.

 Chinese
Journal of
Chinese Journal of Aeronautics 21(2008) 261-268
Aeronautics
www.elsevier.com/locate/cja

Modeling and Backstepping-based Nonlinear Control

Strategy for a 6 DOF Quadrotor Helicopter
Ashfaq Ahmad Mian*, Wang Daobo
College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China

Received 19 October 2007; accepted 27 March 2008

Abstract

In this article, a nonlinear model of an underactuated six degrees of freedom (6 DOF) quadrotor helicopter is derived on the basis
of the Newton-Euler formalism. The derivation comprises determining equations of the motion of the quadrotor in three dimensions and
approximating the actuation forces through the modeling of aerodynamic coefficients and electric motor dynamics. The derived model
composed of translational and rotational subsystems is dynamically unstable, so a sequential nonlinear control strategy is used. The con-
trol strategy includes feedback linearization coupled with a PD controller for the translational subsystem and a backstepping-based PID
nonlinear controller for the rotational subsystem of the quadrotor. The performances of the nonlinear control method are evaluated by
nonlinear simulation and the results demonstrate the effectiveness of the proposed control strategy for the quadrotor helicopter in
quasi-stationary flights.

Keywords: underactuated systems; quadrotor helicopter; backstepping control

1 Introduction In Fig.1, l represents the distance between each

motor and the pivot center, I , T and \ represent
Helicopters exhibit a number of important
the Euler angles about the body axes x, y and z ,
physical effects such as aerodynamic effects, inertial
respectively, Ti (i 1, 2,3, 4) is the thrust force pro-
counter torques, gravity effect, gyroscopic effects,
duced by each propeller marked. The earth-fixed
and friction, etc., which makes it difficult to design
frame is denoted by E { X , Y , Z } , and the body-
a real-time control for them.
fixed frame by B {x, y, z} .
A quadrotor helicopter is a highly nonlinear,
multivariable, strongly coupled, and underactuated
system (six degrees of freedom (6 DOF) with only 4
actuators). The main forces and moments acting on
the quadrotor are produced by propellers. There are
two propellers in the system rotating in opposite
direction to balance the total torque of the system.
Fig.1 shows the free body diagram and axes of a
quadrotor helicopter.

*Corresponding author. Tel.: +86-13770928494. Fig.1 Forces and moments acting on a quadrotor helicopter.
E-mail address: [email protected]
Foundation item: Higher Education Commission, Government of Paki-
stan (1-3/PM-OVER/China/2005)
· 262 · Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268

Simultaneous increase or decrease in speed of copter with only two backstepping steps. Results
the four motors will generate vertical motion. When from the nonlinear simulation verify the effective-
the motor pair (3, 1) is allowed to operate inde- ness of the proposed control strategy for the
pendently, the pitch angle T about the y axis can quadrotor helicopter under near quasi-stationary
be controlled with the indirect control of motion conditions.
along the same axis. Similarly, the independent op-
eration of the motor pair (2, 4) could control the roll 2 Quadrotor Dynamics
angle I about the x axis with an indirect control of The aerodynamic forces and moments are
motion along the same axis. Finally, the counter- sought out by combining momentum with blade
clockwise rotation of the pair (3, 1) and the pair (2, element theory[7-8]. A quadrotor has four motors
4) can control the yaw angle \ about the z axis. In with propellers. The power applied to each motor
this manner, the quadrotor helicopter has 6 DOF. generates a net torque on the rotor shaft, Qi , which
Most recent studies on the theoretical analysis results in a thrust, Ti . If the rotor disk is rotating,
of a 6 DOF quadrotor helicopter were carried out by there is a difference in relative velocity between the
means of a commercially available four rotor aerial blade and the air as the rotor is moving on forward
robot (Draganflyer V Ti)[1]. A control strategy for and backward sweep and causing a net moment
the quadrotor was designed by using internal lin- about the roll axis, Ri . Forward velocity also
earization[2], whereas a quaternion-based feedback causes a drag force on the rotor that acts in opposi-
control scheme was proposed for exponential atti- tion to the direction of travel, Di . Thrust and drag
tude stabilization[3]. However, in this case, the can be defined with the aerodynamic coefficients,
problem to control an underactuated quadrotor was CT and CD as
degenerated to the one of controlling a fully actu- T CT U Ar 2: 2 (1)
ated one.
D CD U Ar 2: 2 (2)
In this article, a feedback linearization- and
where A is a blade area, U the density of air, r the
backstepping-based PID (BS-PID) control strategy
radius of the blade and : the angular velocity of
is designed for motion control of the underactuated
a propeller.
quadrotor. The main idea is to associate the robust-
In much the same way, the torque Q and the
ness against disturbances offered by backstepping
rolling moment R could be defined with the CQ
with robustness against model uncertainties by inte-
and CR as
gral action. The integrator action in backstepping
Q CQ U Ar 2 : 2 r (3)
proposed for linear systems[4-5] is obtained by add-
ing the integral of tracking error to the error found R CR U Ar 2 : 2 r (4)
in first step of backstepping procedure. The total force, f total , and the total moment,
Besides higher payload capacity and better W total , acting on the body frame of a quadrotor are
maneuverability, a quadrotor helicopter has impor- given by
tant advantages in having small rotors and being 1
f total  Cx, y , z Ac U ( x 2  y 2  z 2 )  mgZ 
enclosed, thereby able to be safer for indoor flights. 2
4 4
On the other side, it is rather energy consuming and (5)
¦ Ti z  ¦ Di ( x y)
is fairly big in size. i 1 i 1
4 4
This article is devoted towards deriving a com- W total (1)i ¦ Qi z  (1)i 1 ¦ Ri ( x y )  h<
plete dynamic model of a quadrotor helicopter on i 1 i 1
4
the basis of Ref.[6], and to design feedback lineari-
¦ Di ( y x)  (T4  T2 )lx  (T3  T1 )ly 
zation with PD control and backstepping-based PID i 1

control strategy for the nonlinear quadrotor heli- {[( D2  D4 )  ( D3  D1 )]l}z (6)
 Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268 · 263 ·

In Eq.(5), the first term represents the friction force Dx 4 ) and ( Dy 3  Dy1 ) the drag force unbalances
on the quadrotor body in horizontal motion with during forward and sideward flights, respectively.
Cx , y , z denoting longitudinal drag coefficients, Ac 2.2 Translational dynamics
is the fuselage area, x , y and z speeds in the x, y
By neglecting the effects of body moments on
and z direction, respectively, Z is the vertical axis in
the translational dynamics, from Eq.(5) and Eq.(7),
inertial coordinates, (x y) the direction of velocity, m
the translational dynamics governing the quadrotor
the total mass of quadrotor, and g the force because
are given by
of gravity. In Eq.(6), h is the vertical distance be-
tween propeller center and center of gravity (CG) of mX (cos I sin T cos \  sin I sin \ )<
4 4
quadrotor. 1
¦ Ti  ¦ Di  2 Cx Ac U x x (12)
Given the quadrotor being a single rigid body i 1 i 1

with 6 DOF, and assuming that the earth is flat ne- mY (cos I sin T sin \  sin I cos \ )<
4 4
glecting ground effect, the equations of motion for a 1
¦ Ti  ¦ Di  2 C y Ac U y y (13)
rigid body subjected to a body force, f b  R 3 , and i 1 i 1

a body moment, W b  R 3 , when applied at the cen- 1 4

ter of mass and expressed in Newton-Euler formal- mZ  mg  (cos I cos T )¦ Ti  C z Ac U z z (14)
i 1 2
ism[6], are given by
ª mI 0 º ª v b º ª Ȧb u mv b º ª f bº 3 Engine Model
«0 « »« » « b» (7)
¬ J »¼ ¬« Ȧ b ¼» ¬« Ȧb u JȦb ¼» ¬« IJ ¼» Let a current I a at a driving voltage Va flow
b 3 b
where v  R is the body velocity vector, Z  through a DC motor with inductance Lm , resistance
R 3 the body angular velocity vector, m  R the Rm , and back electro motive force (EMF) voltage
total mass, I  R 3u3 an identity matrix, and J  Vemf ; then,
R 3u3 an inertial matrix. dI a
Va  Vemf Lm  Rm I a (15)
dt
2.1 Rotational dynamics
The motor converts the current into a me-
Assuming that in a symmetric design of quad-
chanical torque applied to the shaft, Tm K tm I a .
rotor, the inertia tensor is diagonal, the moment
The torque, which produces angular velocity Zm
equation governing the quadrotor is given by
according to inertia J m and motor load Tl , is de-
IJb Ȧb u JȦb  IJ total (8) scribed by
From Eq.(6) and Eq.(8), rotational dynamics of dZ
Tm J m m  Tl (16)
the quadrotor in body axis are given by dt
J I T\
x
  ( J  J )  l T  T  
y z 4 2
By defining Vemf K eZm , neglecting the in-
4 4 ductance of the motor because of its small size and
¦ (1)i1 Rxi  h¦ Dyi (9) introducing propeller and gearbox models, from Eq.
i 1 i 1
(15) and Eq.(16) it can be obtained
J yT I\
  ( J  J )  l (T  T ) 
z x 3 1
K tm K e d K
4 4 Z m  Zm  3 Zm2  tm Va (17)
Rm J m K rg J m Rm J m
¦ (1)i1 Ryi  h¦ Dxi (10)
i 1 i 1
where K is the gear box efficiency, d the drag fac-
4
 ( J  J )  (1)i Q  tor, and rg the gear reduction ratio.
J z\ IT x y ¦ i
i 1
[( Dx 2  Dx 4 )  ( Dy 3  Dy1 )]l (11) 4 Control Strategy
where Rx and Ry represent the rolling moments, A nonlinear control strategy is proposed to sta-
h( Dx ) and h( D y ) the drag moments, and ( Dx 2  bilize the quadrotor under near quasi-stationary
· 264 · Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268

conditions, i.e., hovering or near hovering. Since (13). Let xd and y d be the desired speed in x and
only the hovering is considered, the terms in trans- y direction, respectively; then the errors at desired
lational and rotational dynamics associated with and actual speed are separately given by
vehicular velocity become zero. Thus, the drag ex xd  x (23)
forces and rolling moments because of the velocity ey y d  y (24)
are neglected, and the thrust and the torque coeffi-
The desired roll and pitch angles in terms of
cients are supposed to be constant, i.e.,
errors between actual and desired speeds are, thus,
Ti b: i2 ½° separately given by
¾ (18)
Qi d : i2 °¿ Id arcsin(ue sin \  ue cos \ ) (25)
x y

where b and d are thrust and drag factors, respec- uex

§ sin I sin \ ·
tively. Td arcsin ¨  ¸ (26)
© cos I cos \ cos I cos \ ¹
The inputs to the quadrotor, namely, the verti-
where ue and ue are
cal force input u1 , the roll actuator input u2 , the x y

K x ex m K y ey m
pitch actuator input u3 and the yaw moment input uex , ue y
u1 u1
u4 are defined as
where K x and K y are the positive constants and
u1 b(: 12  : 22  : 32  : 24 ) ½
° u1 is the desired vertical force input by the altitude
u2 b(: 24  : 22 ) °
¾ (19) control.
2 2
u3 b (: 3  : 1 ) °
2 2 2 2 ° 4.3 Rotational control
u4 d (: 2  : 4  : 1  : 3 ) ¿
The backstepping-based PID control technique
The control strategy is so designed that the al-
is designed for rotational subsystem, in which the
titude of the quadrotor is stabilized by using the
control inputs u2 , u3 and u4 control the quadrotor
vertical force input u1 . The desired roll and pitch
during hovering.
angles are formed on the rotation controller by the
Let the roll tracking error be defined as
position subsystem. The rotation controller is used
e I  Id (27)
to stabilize the quadrotor under near quasi-station-
ary conditions with control inputs u2 , u3 and u4 . The first error considered in designing the
backstepping is
4.1 Altitude control
z1 K1e  K 2 ³ edt (28)
The altitude subsystem Eq.(14) containing ver-
where K1 and K 2 are positive tuning parameters,
tical force input u1 is given by
and ³ edt the integral of roll error.
mZ mg  u1 cos I cos T (20)
Lyapunov theory is used while using the
which can be linearized by selecting u1 as
mg Q Lyapunov function z1 as a positive definite and its
u1  (21) time derivative as a negative semi definite,
cos I cos T cos I cos T
1 2
The necessary condition for Eq.(21) is cos I ˜ V1 z1 (29)
2
cos T z 0 , where Q , a PD controller, is given by
Its derivative is given by
Q  K d z  K p ( z  zd ) (22)
V1 z1 z1 z1 ( K1I  K1Id  K 2 e) (30)
where K p and K d are the proportional and the
There is no control input in Eq.(30). By letting
derivative positive gains and zd the desired alti-
I be the virtual control, the desired virtual control
tude.
(I) is defined as
d
4.2 Position control K 2e c1 z1
(I)d Id   (31)
Position subsystem is given by Eq.(12) and Eq. K1 K1
 Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268 · 265 ·

where c1 is a positive constant for increasing the c1c2 K 2

I  K1K 2
convergence speed of the roll tracking loop. K1

Now, the virtual control I is the roll rate of a K2

D c2   c1
quadrotor with its own error K1
1 Considering the control law given by Eq.(38)
z2 I  (I)d ( z1  c1 z1 ) (32)
K1 and the characteristic equation of regulation dy-
The augmented Lyapunov function for the se- namics, because the rotational subsystem is both
cond step is given by observable and controllable, the pole placement
1 2 1 2 technique is used to place the poles at desired loca-
V2 z1  z2 (33)
2 2 tion to find the roots of the characteristic equation.
The derivative of Eq.(33) is given by Selecting larger values for c1 and c2 makes the
V z z  z z2 1 1 2 2 (34) derivative of the Lyapunov function more negative,
thus, making the regulation dynamics faster.
By putting z1 and z2 in Eq.(34), the fol-
Similar to the roll subsystem, the backstepping-
lowing can be obtained
based PID control is designed for pitch and yaw
ª § cK ·
V2 z2 « e ¨ K12  1 2 ¸  ³ edt  K1K 2   subsystem to obtain u3 and u4 as follows:
¬ © K1 ¹
J y ª § c4 K 4 cK ·
§K ·  § Jy  Jz · lu2  º u3 « e ¨  c3c4  K32  3 4 ¸ 
e ¨ 2  c1 ¸  T\¨ ¸  Id » l «¬ © K 3 K3 ¹
© K1 ¹ © Jx ¹ Jx ¼
§ c3c4 K 4 · § K ·
 z1[c1K1e  c1K 2 ³ edt ] (35) ³ edt ¨©  K3 K 4 ¸  e ¨ c4  4  c3 ¸ 
K3 ¹ © K3 ¹
The desirable dynamics are § J z  J x ·º
c Td  I\
 ¨ ¸» (39)
¨ J y ¸»
V2 c2 z2  2 ( z1  c1 z1 ) (36) © ¹¼
K1
Jz ª § c6 K 6 cK ·
where c2 is a positive tuning parameter. By put- u4 « e ¨  c5c6  K52  5 6 ¸ 
l ¬« © 5 K K5 ¹
ting z1 and z1 in Eq.(35), the desirable dynamics
§ c5c6 K 6 ·
are given by ³ edt ¨©  K5 K 6 ¸ 
K5 ¹
§c K · §cc K ·
V2 e(c2 )  e ¨ 2 2  c1c2 ¸  ³ edt ¨ 1 2 2 ¸ (37) § K · º
© K1 ¹ © K1 ¹ e ¨ c6  6  c5 ¸  \d » (40)
© K5 ¹ »¼
The desirable dynamics ensure negative defi-
niteness of position tracking error, its integration, 5 Results and Discussion
and velocity tracking error.
The angles and their time derivatives of rota-
Eq.(35) will be negative, if u2 is given by
tional subsystem do not depend on translation com-
Jx ª § c2 K 2 cK ·
u2 « e ¨  c1c2  K12  1 2 ¸  ponents, as the 6 DOF equations governing the
l ¬ © 1 K K1 ¹
quadrotor helicopter have shown. However, the
§ c1c2 K 2 · § K · translations depend on the angles. Ideally, it can be
³ edt ¨©  K1K 2 ¸  e ¨ c2  2  c1 ¸ 
K1 ¹ © K1 ¹ thought as two subsystems: the one of angular rota-
§ J y  J z ·º tions and the other one of linear translations.
Id  T\
 ¨ ¸» (38) Rotational control keeps the 3D orientation of
© J x ¹ ¼»
the quadrotor helicopter to the desired state. Roll
As a rule, Eq.(38) is a PID, where the gains of
and pitch angles are usually made to be zero to re-
each mode are given by
alize hovering. The rotational controller is responsi-
c2 K 2 cK
P  c1c2  K12  1 2 ble for compensating the initial errors, stabilizing
K1 K1
· 266 · Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268

roll, pitch, and yaw angles and maintaining them at from the conventional optimized one. From Fig.3, it
zero. This is accomplished by way of the backstep- is evident that the backstepping-based PID control-
ping-based nonlinear control law. ler presents higher robustness and better transient
Table 1 summarizes different system parame- performances than the traditional PID version.
ters of the prototype quadrotor helicopter.
Table 1 Physical parameters of quadrotor

Parameter Value

l /m 0.305 0

J x /(kg ˜ m 2 ) 0.015 4

J y /(kg ˜ m 2 ) 0.015 4
2
J z /(kg ˜ m ) 0.030 9

m /kg 0.615 0

Next, a closed loop system with nonlinear con-

trol algorithm is simulated. The initial conditions
are I T \ 0.524 rad and zd 1 m . The ref-
erence inputs to the controller are xd y d 0 m/s ,
zd 1 m , and \ d 0 rad .
Fig.2 shows the response of the nonlinear con-
troller to stabilize the quadrotor during hovering.
The simulation results in Fig.2 are acquired with a
model inclusive of actuators’ dynamics. Although Fig.2 Attitude control of a quadrotor helicopter.
the initial conditions are very strict, it can be seen
from Fig.2 that the controller succeeded in control-
ling the roll, pitch, and yaw angles of the quadrotor
in less than 8 s.
The nonlinear simulation results obtained with
a backstepping-based PID controller are compared
with a conventional one, where the motor dynamics
are included in the dynamic model and omitting the
gyroscopic effects thus removing the cross coupling.
An optimization algorithm is used to find the best
possible set of PID parameters. In Ref.[9], in order
to obtain the parameters for the roll, the pitch, and
the yaw, the objective function in the optimization
algorithm was to minimize the integral of the abso-
lute error (IAE). The IAE is a performance criterion
that considers the difference between the set point
and the output that exists when a system is excited
by a step input. The optimization toolbox of MAT-
LAB was used to obtain the controller’s gain for the
PID controller.
Fig.3 Comparison of backstepping-based PID with con-
Fig.3 shows the comparison of the results from
ventional optimized PID for rotational subsystem.
the backstepping-based PID controller with those
 Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268 · 267 ·

The altitude rate and position rate response of

the quadrotor helicopter are shown in Fig.4.
Results from Fig.4 indicate that the position
controller effectively makes the attitude controller
keep the quadrotor helicopter at a given point. The
integral term in the backstepping control helps in
eliminating the steady state error.
Fig.5 shows the rotor speed response of a
quadrotor during hovering.
From Fig.5, for the two pairs of propellers (1, 3)
and (2, 4) rotating in opposite direction, as is shown
Fig.5 Control response of a quadrotor helicopter.
that the rotor speed is able to produce sufficient lift
to overcome the weight of the quadrotor helicopter Fig.6 shows the simulation results of the alti-
and enable it to hover at a given point. tude subsystem.

Fig.6 Altitude control of a quadrotor helicopter.

As is indicated in Fig.6, by properly valuing
the proportional and derivative terms like K d
1.96, K p 3.98 , the controller is able to stabilize the
attitude angles and enable the quadrotor helicopter
to hover at a given point.

6 Conclusions

Fig.4 Altitude and position rate response of a quadrotor

A stabilization nonlinear control method for a
helicopter.
quadrotor helicopter is presented. The modeling of
the quadrotor is on the basis of Newton-Euler for-
malism. A novel control strategy is applied to rota-
tional subsystem of the quadrotor helicopter, where-
in the integral of the tracking error is considered in
the first step of the backstepping procedure, The
control law derived for the nonlinear quadrotor is
thus a backstepping-based PID for regulation dy-
namics. The stabilization ability of the nonlinear
controller is examined through nonlinear simulation
and the results indicate effectiveness of the pro-
posed control strategy for the quadrotor helicopter.
 · 268 · Ashfaq Ahmad Mian et al. / Chinese Journal of Aeronautics 21(2008) 261-268

References html.

[7] Prouty R W. Helicopter performance, stability and control. Florida:

[1] McKerrow P. Modelling the Draganflyer four-rotor helicopter.
Krieger Publishing Company, 1996.
Proceedings of the IEEE International Conference on Robotics
[8] Seddon J. Basic helicopter aerodynamics. Oxford: BSP Profes-
and Automation. 2004; 3596-3601.
sional Books, 1990.
[2] Pounds P, Mahony R, Hynes P, et al. Design of a four-rotor aerial
[9] Smith C, Corripio A. Principles and practice of automatic process
robot. Proceedings of Australian Conference on Robotics and
control. 2rd ed. New York: John Wiley & Sons, 1997.
Automation. 2002; 145-150.

[3] Tayebi A, McGilvray S. Attitude stabilization of a four-rotor aerial Biography:

robot. 43rd IEEE Conference on Decision and Control. 2004; Ashfaq Ahmad Mian Born in 1973, he received B.S. from
1216-1221. University of Engineering and Technology and M.S. from
[4] Kanellakopoulos I, Krein P. Integral-action nonlinear control of Quaid-e-Azam University, Pakistan, in 1997 and 2000 re-
induction motors. Proceedings of the 12th IFAC World Congress. spectively. Since then, he joined the R&D organization,
1993; 251-254. NESCOM. Now he is a Ph.D. candidate in Nanjing Univer-
[5] Krstic M, Kanellakopoulos I, Kokotovic P. Nonlinear and adaptive sity of Aeronautics and Astronautics. His main research in-
control design. New York: John Wiley & Sons, 1995. terests are autopilot design, nonlinear control of MIMO sys-
[6] Koo T J, Ma Y, Sastry S. Nonlinear control of a helicopter based tems, controller fragility and robust control.
unmanned aerial vehicle model. http://citeseer.ist.psu.edu/417459. E-mail: [email protected]

View publication stats