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

net/publication/312691176

Stabilization and control of a UAV flight attitude angles using the backstepping
method

Article · January 2012

CITATIONS READS

17 413

1 author:

Mihai Lungu
University of Craiova & University Politehnica of Bucharest
147 PUBLICATIONS   879 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Control moment gyros View project

Special Issue - Advanced Observer-Based Control for Benchmark Control Problems: From Mathematical Modeling to Control Design View project

All content following this page was uploaded by Mihai Lungu on 28 August 2017.

The user has requested enhancement of the downloaded file.

 Stabilization and Control of a UAV Flight
Attitude Angles Using the Backstepping Method
Mihai Lungu


Abstract— The paper presents the design of a mini-UAV attitude This method disadvantage is the existence of nonlinearities
controller using the backstepping method. Starting from the nonlinear and their derivatives. This is a serious problem because the
dynamic equations of the mini-UAV, by using the backstepping aerodynamic forces and moments, which describe the system,
method, the author of this paper obtained the expressions of the
elevator, rudder and aileron deflections, which stabilize the UAV, at can not be modeled at a high precision degree [5]. An
each moment, to the desired values of the attitude angles. The attitude alternative to the dynamic inversion is the backstepping
controller controls the attitude angles, the angular rates, the angular method; it is a simpler method and its most important
accelerations and other variables that describe the UAV longitudinal advantage is that the nonlinearities must not be necessary
and lateral motions. To design the nonlinear controller, by using the cancelled in the control law.
backstepping technique, the nonlinear equations and the Lyapunov
In this paper the attitude controller is projected for the
analysis have been directly used. The designed controller has been
implemented in Matlab/Simulink environment and its effectiveness Sekwa UAV (Fig. 1) [6] by means of the backstepping method.
has been tested with a campaign of numerical simulations using data This UAV is characterized by a mobile centre mass in order to
from the UAV flight tests. The obtained results are very good and minimize the drag force. The design of the attitude angles
they are better than the ones found in previous works. controller (autopilot) means the determination of the three
deflections (elevator, ailerons and rudder deflections) so that
Keywords— Attitude angles, Backstepping, Controller, UAV. the angles characterizing the UAV attitude tend to their
imposed values. The Defense, Peace, Safety and Security
I. INTRODUCTION (DPSS) branch of the South African Council for Scientific and

A LL over the world, Unmanned Aerial Vehicles have a

board range of civil and military applications. A lot of
scientists design UAVs that have to meet some expectations
Industrial Research (CSIR) launched, some years ago, the
Sekwa program; the main objective of this research program
was to demonstrate the advantages of UAVs with variable
regarding their reliability, low cost, small size and so on and so centre mass [6]. The main characteristics of this UAV are the
forth. In the domain of small-size UAVs, the most systems are lack of the vertical empennage and the blended wing; these
still deployed as prototypes because of their lack of reliability. things lead to small drag forces and superior aerodynamic
Their reliability may be increased by improving their modeling performances. The only disadvantage may be the stability of
and their flight control systems [1], [2]. Modeling, simulation, the UAV; as a consequence, the designed autopilot must be
analysis and flight testing of full-size aircrafts or large UAVs very efficient. The purpose of the flight control system was to
are very well presented in lots of scientific works over the past increase the natural UAV stability so that it is characterized by
few decades [3]. nominal static stability and all its motion variables are
stabilized toward their desired values.
The main sensors that are used on UAVs are: three
accelerometers (for the measurement of the linear accelerations
u, v, w , which, by integrating, lead to velocities u, v, w ) three
gyrometers (for the measurement of the angular velocities
p, q, r ), sensors for static and dynamic pressure (the first for
the determination of the barometric altitude and the both
Fig. 1 The model of the Sekwa UAV
sensors for the determination of the flight velocity), a radio-
altimeter or other system for the measurement of the UAV’s
In this paper, the equations of the flying object motion are
height with respect to the ground and so on [7]-[9].
nonlinear [2], the backstepping method being used here for the
In contrast with [6], in this paper, the UAV flight control is
control of the UAV attitude angles. There are a lot of
made by using the backstepping method and the efficiency of
techniques for the control of flying objects when their flight is
this method will be demonstrated by a campaign of numerical
described by nonlinear equations: dynamic inversion, non-
simulations using data from the UAV flight tests. Most UAV
linear predictive control or techniques that use neuronal
autopilots use classical proportional-integral-derivative (PID)
networks [4]. The system nonlinearities are cancelled and a
controllers and ad-hoc methods to tune the controller gains
closed loop linear system is obtained by using the first method.
during the flight. This methodology is not the best one because
it has high risks and because there are a lot of limitations in the
M. Lungu is with the University of Craiova, Faculty of Electrical
Engineering, Avionics Division, 107 Decebal Street, 200440 Craiova, Dolj, UAVs performances and robustness. It is important to obtain
Romania (email: [email protected], [email protected]).
an integrated framework that enables it to rapidly synthesize,
implement, analyze and validate a controller configuration by
using iterative development cycles [10]-[14].
The mathematical procedure of using the backstepping
method to the stabilization and control of a mini UAV, the
determination of the control laws (deflections of the elevator,
rudder and ailerons), which assure the control of mini UAV,
and the new controller software implementation represent the
originality elements of this paper. Fig. 3 The command surfaces and the actuators of the Sekwa UAV
The paper is organized as follows: the dynamic equations of
the Sekwa UAV is given in section II, the controller design for The UAV has 6 actuators associated to the elevator, ailerons
the pitch angle channel is presented in section III; in section and rudder deflections 1 ,  2 ,  3 ,  4 ,  5 ,  6 (Fig. 3), an
IV and V the author determines the control laws for the actuator for the retractable landing gear  G  , an actuator for
stabilization and control of the roll and yaw angles. In section
VI, the designed controller is implemented in Matlab/Simulink
the steerable nose wheel  S  and an actuator for the
environment and its effectiveness is tested with a campaign of command of the trust (propulsion) force T  . Because in the
numerical simulations using data from the UAV flight tests; specialty literature one uses the classical commands of the
finally, some conclusions are shared in section VII. command surfaces (  e  the elevator deflection,  a  the
ailerons deflection and  r  the rudder deflection), it is
II. DYNAMIC EQUATIONS OF THE SEKWA MINI-UAV
necessary to be established some relationships between the
There are a lot of methods to define the model of the UAVs classical deflections and the deflections of the Sekwa UAV.
[15]-[17]. As one can see in [18], the control system architec- The deflection of a control surface is defined in radians with a
ture can be greatly simplified by judiciously expressing the positive deflection causing a negative moment. It is desired to
aircraft dynamics and carefully selecting the variables that work with one actuator per wing for control purposes,
need to be controlled. Thus, the complexity of the automatic therefore  2 and  3 should move together and  4 and  5
pilot is greatly reduced and most design techniques become
should move together, too; thus, 1   6 ,  2   3 ,  4   5 [6].
very efficient [6].
The relationships between the classical deflections  e ,  a ,  r 
and UAV Sekwa deflections 1 ,  2 ,  3 ,  4 ,  5 ,  6  are [6]:

 e   2   3   4   5  / 4 ,
 a    2   3   4   5  / 4 , (1)
 r  1   6  / 2 .

To obtain the UAV dynamics, the UAV is considered to be

Fig. 2 The inertial reference system
a rigid body with six degree of freedom [6], [18]. To simplify
the derivation of the UAV model, a static centre of mass
In order to obtain the mathematical model of the Sekwa
position is assumed. The dynamics of the UAV is considered
UAV, one needs to define three axis systems. The first one is
to be the resultant between the kinematics and the kinetics of
the inertial reference system and the Newton’s laws can be
the UAV. Because the purpose of this paper is not to deduce
successfully applied with respect to this frame. To use this
the dynamic equations of the Sekwa UAV, I will concentrate
frame, it is assumed that the surface of the Earth is flat and
on the UAV control design using the backstepping method; the
non-rotating. This frame [19] defines a plane that is tangential
Sekwa UAV dynamics is presented in detail in [6] and [18].
to the Earth surface (Fig. 2); the system centre is conveniently
Thus, the kinematics of the UAV is [6], [18]:
chosen as a point from the runway. The second frame (OXYZ)
  p  tan q sin   r cos  ,
is the body axis system with its origin in the centre mass of the
flying object. The OX axis is the longitudinal axis of the UAV,
the OY is the lateral axis, pointing out the aircraft right   q cos   r sin  , (2)
starboard wing, while OZ axis is the vertical axis and it is q sin   r cos 

  .
defined to reside within the UAV’s plane of symmetry. The cos 
third frame is the well known aerodynamic frame, its origin
being the same with the origin of the inertial frame; the OX The expressions of the roll, pitch and yaw moments
axis has, this time, the direction of the UAV total velocity. L, M , N  are [6], [18]:
This frame is very important because the determination of the
stability derivates is made with respect to this axis system.
 
 L  p d S C  C b  C l  C  S  c  C m  S  b  C n ,  technique. Thus, some nonlinearities may be maintained, while


M  p d S S  b  C l  C  c  C m ,  (3)
others, which are not necessary, are to be cancelled [5], [20];
the control law is simpler. In order to use the backstepping


 N  p d S S  C b  C l  S  S  c  C m  C  b  C n ,  method, the system must be written under the following form:

where p d is the dynamic pressure, S  the wing surface, x1  f1 x1 , x2  , x 2  f 2 x1 , x2 , x3  , x 3  f 3 x1 , x2 , x3 , u  , (7)
,   attack and sideslip angles, b  the wing span, c 
where f1 , f 2 , f 3 are nonlinear functions. x 2 plays the role of a
mean aerodynamic chord and C  cos , C  cos ,
virtual control for the command of the variable x1 toward
S  sin , S  sin  . In (3), the aerodynamic moment
zero, without taking into account the dynamics of x3 . Then,
coefficients Cl , Cm , Cn (roll, pitch and yaw, respectively)
x 3 is used as a virtual control for the convergence to zero of
have the expressions [6]:
the variable x 2 . Finally, the input variable u is the one that

 
 b stabilizes the variable x3 .
Cl  Cl   Clr  r  Cla  a  2V Cl p p  Clr r ,
 III. CONTROLLER DESIGN FOR THE PITCH ANGLE CHANNEL
 c
Cm  Cm 0  Cm   Cme  e  Cmq q , (4)
To stabilize the UAV pitch angle, the paper author considers
 2V


b

Cn  Cn   Cn r  r  Cn a  a  2V Cn p p  Cnr r .    the imposed (desired) value of the pitch angle  and
defines the error e     [5]. It is evident that e   and,
 

taking into account (2), it yields:

In (4) C m0 is the static moment coefficient and it has values
close to zero; that is why, in simulations, C m0 will be   e  r sin  
e    e   q     cos  , (8)
considered null. The equations describing the projections of  cos  
the UAV angular velocities along OXYZ frame axes are [6]:
with    positive constant. The design aim is e  0 ; as a

 p 
1
  
qr I y  I z  p d SC1 ,  result, it is desirable that:
 I x
   e  r sin   e  rsin
1
 pr I z  I x   p d SC 2  , q  0 q  . (9)
q  (5)
cos cos
 I y

r 

1
Iz
  
pq I x  I y  p d SC 3 ,  The equation of the error e , using (9), becomes:

e    e ; (10)
with

C1  C Cb  Cl  C S c  Cm  S  b  Cn ,
in order to demonstrate that (10) corresponds to a stable
system, the Lyapunov function V e   e2 has been
1
C2  Sb  Cl  C c  Cm , (6)
2
C3  S  Cb  Cl  S  S c  Cm  C b  Cn ; chosen [5] and, by computing V e  , it follows:


I x , I y , I z are the axial inertia moments. V e   e  e    e2  0,     0 . (11)
For the design of the Sekwa UAV controller, for the
stabilization and control of the attitude angle , ,   , the The Lyapunov function is negative and, as a consequence,
kinematics equations of the UAV, the equations of the the system (10) is a stable one and it converges to zero
aerodynamic moment coefficients Cl , Cm , Cn and the e  0 . The second error variable is now chosen as
equations describing the projections of the UAV angular eq  q  q , with q having the form (9), i.e.:
velocities along the OXYZ frame (equations (2), (4) and (5))
are needed.
  e  r sin
One of the best controller design procedures is the q  . (12)
backstepping method. The backstepping technique is based on cos
Lyapunov theory and it offers multiple possibilities for the use
of nonlinearities with respect to the dynamic inversion As a result, the following equation system is obtained:
 a1 e  a2  a  a3  r  A
e    e  eq cos  ,
 (20)
 
(13) with
eq  q  q .


1 1
By the substitution of q , with the form (5), in the second a1  p d SC c C m , a 2  p d SS  bC l ,
Iy e
Iy a

equation (13), it follows:

1
a3  p d SS  bC l ,
Iy r

eq 
1

pr I z  I x   pd S Sb  Cl  C c  Cm   q ;  (14)
pr I z  I x  
Iy 1
A  e cos    q eq  (21)
Iy
taking into account the expression of C l and C m from (4), pd S  b b 
 S  b Cl   Cl p p  C lr r  
(14) becomes: Iy  2V 2V 
pd S  c 
 C c  C m 0  C m   C mq q   q .
eq 
1
Iy
p S
Iy r

pr I z  I x   d S  b Cl   Cl  r   Iy  2V 


pd S
Iy

S  b Cl  a 
 a
b
2V
 
C l p p  C lr r  

 (15) IV. CONTROLLER DESIGN FOR THE ROLL ANGLE CHANNEL
The procedure is similar to the one presented in section III;
p S  c  the author of the paper considers   the desired (imposed)
 d C c  C m 0  C m   C m  e  C mq q   q .
Iy  e
2V  value of the roll angle and he defines the error [5] e     .
For   constant, it follows e   ; taking into account the
I want to stabilize the error e q to zero, as well. For this, the expression of  , it results:
following Lyapunov function

e    e  p    e  qtan  sin  r tan  cos . (22)
V e , e  
1 2 1 2
q  q e  e q (16)
2 2
For the convergence of the error e to zero, (22) must have


is chosen. By computing Vq e , eq , it follows:  the form e    e , with    positive constant. As a
consequence, the roll angular velocity has the expression [5]:
 
Vq e , eq  e e  eq eq 
p    e  q tan  sin   r tan  cos  . (23)
 
(17)
  e   q eq2  eq e cos    q eq  eq ,
2
 

Choosing the Lyapunov function V e  

1 2
e , one gets:
with  q  positive constant. 2
The first two terms in the right side of (17) are negative and,
for stability reasons, it is desirable that: V e   e  e   e2  0,     0 . (24)


eq e cos   q eq  eq  0  (18) Thus, the system is a stable one and convergent to zero
e  0    . The second error variable is now
or chosen: e p  p  p ; p has the form:

p S
pr I z  I x   d S  bC l  
1
e  cos    q e q  p    e  q tan  sin  r tan  cos . (25)
Iy Iy


pd S
Iy

S  b C l  r  C l  a 
 r a
b
2V
 
C l p p  C lr r  

 (19) Under these conditions, I obtain the following system:

pd S  c  e    e  e p ,

 C  c  C m 0  C m   C m  e  C mq q   q  0 ;
Iy  e
2V   
(26)
e p  p  p


thus, the first relationship between the deflections of the

and replacing p , with the form in (5), into the second
elevator, ailerons and rudder is:
equation (26), it follows:
e p 
1
Ix
  p S
qr I y  I z  p  d C  C b  Cl 
Ix
where

 
(27) 1 1

pd S
Ix
C S c  Cm  S  b  Cn ; b1  
Ix
p d SC S  c C m , b2 
e
Ix
p d Sb C C Cl  S  C n ,
a a

b3 
1
Ix

p d Sb C C Cl  S  C n ,
r r

taking into account the expressions of Cl , Cm and C n , it
results: B  e   p e p 
1
Ix
 p S

qr I y  I z  d C C bCl  
Ix (34)

   
2
pd S b p S
e p 
1
Ix
  p S
qr I y  I z  d C Cb Cl   Cl  r 
Ix r

Ix
C  C
2V
C l p p  C lr r  d C  S  c C m 0 
Ix
pd S  c  p S

pd S
Ix

C Cb Cl  a 
 a
b
2V
 
Cl p p  Clr r   p 

 
Ix
C S  c  C m  
 2V
C mq q   d S  bC n  
 Ix

p S  c 
(28)

pd S
S
b2

C n p p  C nr r  p . 
 d C S c  Cm0  Cm   Cm  e  Cmq q   Ix 2V
Ix  e
2V 


pd S
Ix

S  b Cn   Cn   r  Cn   a 
 r a
b
2V
 
Cn p p  Cnr r  .

 V. CONTROLLER DESIGN FOR THE YAW ANGLE CHANNEL
For the yaw angle stabilization, the approach is the same
with the one used in sections III and IV. The error variables
The error e p must be zero. For this, the Lyapunov function
e     , e r  r  r (35)

 
V p e , e p 
1 2 1 2
2
e  e p
2
(29)
are used here;  is the imposed (desired) value of the yaw
angle  , while r has the form [5]
 
is chosen. By computing V p e , e p , it results:
cosθ sin
r    e  q. (36)
  
Vp e , e p  e2   p e 2p  e p e   p e p  e p ,  (30) cos cos

where  p is a positive constant. The first two terms in the The Lyapunov functions, which have been used, are the
following ones:
right side of (30) are negative and, because it is desirable that
Vp e , e p   0 , it yields:
V e   e , Vr e , er   e2  er2 .
1 2 1 1
(37)
2 2 2
e   p e p  e p  0 . (31)
I obtain, in this case too, a relationship (the third one) between
Replacing e p , having the form (28), in the (31), I obtain: the deflections of the elevator, ailerons and rudder:

c1 e  c2  a  c3  r  C (38)
e   p e p 
1
Ix
 
p S
qr I y  I z  d C  C bCl  
Ix
with
p S 
 d C  C b Cl  r  Cl  a 
b
 
C l p p  C lr r   
Ix  r a
2V 
c1  
1
p d SS  S  c C m , c 2 
1

p d Sb S  C Cl  C  C n ,
  Iz Iz
e a a
pd S c (32)
 C  S  c  C m 0  C m   C m  e  C mq q  
Ix  e
2V  c3 
1

p d Sb S  C Cl  C  C n , 
  Iz r r
p S
 d S  b C n    C n   r  C n   a  p 
Ix r a
C  e   r e r 
1
 p S

pq I x  I y  d S  C bCl  
 
p S b2 Iz Iz (39)
 d S C n p p  C nr r  0 ;
Ix 2V p S
 d S  C
Iz
b2
2V
 p S

C l p p  C lr r  d S  S  c C m 0 
Iz
thus, the second relationship between the deflections of the pd S  c  p S
 S  S  c  C m   C mq q   d C  bC n   
elevator, ailerons and rudder is obtained as follows: Iz  2V  Iz

b1e  b2a  b3r  B , (33)


pd S
Iz
C
b2
2V

C n p p  C nr r  r . 
 The determination of the elevator, ailerons and rudder 
deflections e , a , r  is made by solving the system formed  e 
1
b2 c3  c2 b3 A  c2 a3  a2 c3 B  a2b3  a3b2 C ,
 det T
by equations (20), (33) and (38). Thus, it results the system: 
 a 
1
c1b3  c3b1 A  a1c3  a3c1 B  a3b1  a1b3 C , (43)
 det T
a1 e  a2  a  a3 r  A , 
  r 
1
b1c2  b2 c1 A  a2 c1  a1c2 B  a1b2  a2b1 C ,
b1 e  b2  a  b3 r  B , (40)  det T
c   c   c   C ,
1 e 2 a 3 r
where
with ai , bi , ci , A, B, C, i  1, 3 being described by (21), (34)
det T  a1b2 c3  a2b3c1  b1c2 a3  a3b2 c1  b3c2 a1  a2b1c3 . (44)
and (39). The equation system (40) has the matriceal form:

a1 a2 a3    e   A
 b b b       B  , (41)
 1 2 3  a  
 c1 c2 c3   r  C 

T

with the solution

1
 e  a1 a2 a3   A
    b b b    B  (42)
 a  1 2 3
 r   c1 c2 c3  C 
   
T 1

or

Fig. 4 Attitude controller for the control of the Sekwa UAV flight

Fig. 5 Matlab/Simulink model of the UAV attitude controller using the backstepping method
 In conclusion, the paper author has successfully implemen- u  18 m/s, p  q  r  0 deg/s ,   2 grd,
(46)
ted the nonlinear attitude controller on the nonlinear UAV   -2 deg,   5 deg ,   1.24 deg,   0.1 deg ;
aircraft model. The attitude controller controls the attitude
angles, the angular rates, the angular accelerations and other the desired values of the attitude angles have been chosen as:
variables that describe the dynamics of UAV. Putting together
  5 deg,   2 deg,   3 deg .
equations (2), (4), (5), (12), (21), (34), (36), (39), (43), (44)
In Fig. 6, I present the time evolutions of the variables
and the expression of the six errors e , eq , e , e p , e , er , the
t  , t  , t  , pt , qt , r t ,  a t ,  e t ,  r t  . One remarks the
block diagram for the constructing of the attitude controller
proper function of the backstepping method, because all the
has been obtained (Fig. 4).
angles, which define the UAV attitude (the roll angle, the pitch
angle and the yaw angle), track their imposed values.
VI. SIMULATION RESULTS
2
In this section, I present the obtained results after perfor-    0.4,  p  0.4
1
ming the numeric simulations. The designed control system    0.4,  p  1.4

Roll angle of the UAV [deg]

0
has been implemented in a Matlab/Simulink environment and    0 .7 ,  p  0 .7
-1
its effectiveness has been tested with a campaign of numerical    0.8,  p  1.4
-2
simulations. The Matlab/Simulink model of the structure in    1.2,  p  0.5
-3    1.4,  p  1.4
Fig. 4 is presented in Fig. 5.
The UAV has small dimensions: wing span b  1.7 m , wing -4

-5
surface S  0.39 m 2 , mean aerodynamic chord c  0.248 m
-6
and mass m  3.2 kg. The simulation data have been obtained -7

from literature; thus, the values of the aerodynamic -8

0 2 4 6 8 10 12 14 16 18
coefficients and of the inertia moments are [6]: Time [s]
Fig. 7 Time variation of the UAV roll angle for
Cm0  0, Cmq  5.3338  10 6  33.094  10 4  1.694 , different values of   and  p between 0.4 and 1.4
Cm  34  10  4  0.1287, Cm  12.828  10  4  0.458, 4
e

Cl p  0.484, Clr  0.17, Cl  0.35, Cl  0.105,

a r (45)
Pitch angle of the UAV [deg]

3
Cl  0.23809, Cn p  0.002061, Cnr  0.035424,
Cn  0.001833, Cn  0.04778, Cn  0.06581,
2
a r

I x  0.19 kg  m 2 , I y  0.05 kg  m 2 , I z  0.25 kg  m 2 . 1    0.4,  q  0.4

   0.4,  q  1.4
0    0.7,  q  0.7
2 3 5
   0.8,  q  1.4
2 4.5
   1.2,  q  0.5
0
-1
 [deg]

θ [deg]

1 4
ψ [deg ]

   1.4,  q  1.4
-2
0 3.5
-4
-1 3 -2
-6 -2 2.5
0 2 4 6 8 10 12 14 16 18
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time [s] Time [s] Time [s] Time [s]
2 3 0.5

0 2 0
Fig. 8 Time variation of the UAV pitch angle for
different values of   and  q between 0.4 and 1.4
p [deg/s]

r [deg/s]
q [deg/s]

-2 1 -0.5

-4 0 -1
5
   0.4,  r  0.4
-6 -1 -1.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time [s] Time [s] Time [s]
4.5    0.4,  r  1.4
Yaw angle of the UAV [deg]

3 3.5

2
3    0.7,  r  0.7
δ r [deg ]

-3.5
δ a [deg ]

   0.8,  r  1.4
δ e [deg ]

2.5
1
4
2
0    1.2,  r  0.5
-4 1.5

-1 1
3.5    1.4,  r  1.4
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time [s] Time [s] Time [s]
Fig. 6 Time variations of the variables 3

t , t , t , pt , qt , r t ,  a t , e t , r t 

2.5

The trim conditions are very important because they are the 2
0 2 4 6 8 10 12 14 16 18
starting values of the variables; the trim conditions are: Time [s]
Fig. 9 Time variation of the UAV yaw angle for
different values of   and  r between 0.4 and 1.4
 The choice of the values for the positive constants ACKNOWLEDGMENT
  ,   ,   and  p ,  q ,  r influence the responses of the This work was supported by CNCSIS-UEFISCSU, project
system. While the constants are positive, the system is stable; PN II-RU “High-precision strap-down inertial navigators,
the bigger the values of constants are, the better the system based on the connection and adaptive integration of the nano
properties are (the overshoot and the transient regime period and micro inertial sensors in low cost networks, with a high
decrease). In Fig. 7, Fig. 8 and Fig. 9, I present the time degree of redundancy”, code102/2010.
variations of the three attitude angles for different values of the
constants   ,   and   (values between 0.4 and 1.4). The REFERENCES
best results are obtained for the maximum value (in this case [1] H. Chao, Y. Cao, Y. Chen, “Autopilots for small fixed-wing unmanned
air vehicle: A survey”, IEEE International Conference on Mechatronics
1.4). For example, if one considers the case of the roll angle and Automation, 2007, pp. 3144-3149.
stabilization, for   0.4,  p  0.4 , the overshoot is 2 deg, [2] Y.C. Paw, Synthesis and Validation of Flight Control for UAV.
Dissertation submitted to the Faculty of the Graduate School of the
while, for   1.4,  p  1.4 , it decreases ten times to the University of Minnesota, December, 2009.
value 0.2 deg . The same conclusion may be obtained regarding [3] N.M. Jodeh, P.A. Blue, A.A. Waldron, “Development of small
unmanned aerial vehicle research platform: Modeling and simulating
the transient regime period – the increase of the constants   with flight test validation”, AIAA Modeling and Simulation
Technologies Conference and Exhibit, no. 2006-6261, 2006.
and  p from the values 0.4 to the values 1.4 leads to the [4] N.K. Peyada, A.K. Ghosh, ”Aircraft parameter estimation using a new
filtering technique based upon a neural network and Gauss-Newton
decrease of the transient regime period from 13 seconds to 5 method”. Aeronautical Journal, April 2009, 113, (1142), pp. 243-252.
seconds (61.53% decrease); same remarks for the pitch and [5] Y.C. Paw, ”Attitude control of Mini-UAV using backstepping control”.
yaw angles. Nonlinear Systems Project, ME8282, October 2007.
[6] D. Blaaw, Flight Control System for a Variable Stability Blended-wing-
Body Unmanned Aerial Vehicle, Master of Science Thesis, University
VII. CONCLUSIONS of Stellenbosch, South Africa, March 2009.
[7] T.L. Grigorie, Strap-down inertial navigation systems. Optimization
The focus of the backstepping controller synthesis process
studies. Sitech Publisher, Craiova, 2007, 326 pp.
was the choice of the error variables and of the Lyapunov [8] T.L. Grigorie, L. Dinca, J. Corcau, “Aircrafts’ Altitude Measurement
functions. The deflections of the control surfaces have been Using Pressure Information: Barometric Altitude and Density Altitude”,
obtained – equations (43) and (44). The obtained attitude WSEAS Transactions on Circuits and Systems, 7, 2010, pp. 503-512.
[9] T.L. Grigorie, D.G. Sandu, “The influences of the gyro sensors’ errors
controller may be implemented on UAVs; the technique that on the attitude calculus”. IEEE Signal Processing Society - 49th
has been used for its design is a general one and may be used International Symposium ELMAR-2007 focused on Mobile Multimedia,
to all kind of flying objects, offering very good operation and Zadar, Croatia, 12-14 September, 2007, pp. 85-92.
[10] D. Jung, P. Tsiotras, “Modeling and hardware-in-the-loop simulation for
characteristics of the output signals. a small unmanned aerial vehicle”, AIAA Guidance, Navigation and
The design of the three controllers and the backstepping Control Conference and Exhibit, no. 2007-2768, 2007.
technique has directly used the nonlinear equations and the [11] I. Kaminer, O. Yakimenko, V. Dobrokhodov, K. Jones, “Rapid flight
test proto-typing system and the fleet of UAVs and MAVs at the naval
Lyapunov analysis. The method has an important advantage postgraduate school”, Proceedings of the 3rd AIAA Unmanned
with respect to the classical techniques (for example the design Unlimited Technical Conference, 2004.
of PID controllers): the nonlinear system (describing the [12] J. Renfrow, S. Liebler, J. Denham, “F-14 flight control law design,
dynamics of UAV) must not be linearised in a trim point. This verification and validation using computer aided engineering tools”,
Proceedings of the Third IEEE Conference on Control Applications,
linearization must be done if we want to design PID vol. 3, pp. 359-364, 1994.
controllers. The linearization of the UAV dynamics limits the [13] M. Tischler, Advances in aircraft flight control. Taylor and Francis Ltd,
applicability of PID controllers. 1996.
[14] L. Dinca, J. Corcau, “EDS and EFIS systems on board of modern
Using the backstepping method, the author obtained the transport aircraft”, 10th International Conference on Applied and
expressions of the elevator, rudder and aileron deflections that Theoretical Electricity, Craiova, Romania, 2010.
control the UAV, at every moment, and stabilize it to the [15] M.V. Cook, Flight Dynamics Principles. Elsevier Butterworth-
Heinemann, 1997.
desired values of the attitude angles and angular rates. [16] J.H. Blakelock, Automatic Control of Aircraft and Missiles. Wiley-
The paper author made complex simulations in Matlab/ Interscience, 1991.
Simulink by using data from the UAV flight tests. The obtained [17] B. Etkin, L.D. Reid, Dynamics of Flight Stability and Control, 3rd ed.
John Wiley & Sons, 1996.
results are very good (UAV attitude angles and the UAV
[18] I.K. Peddle, Acceleration Based Maneuver Flight Control System for
angular rates tend to their imposed values in a proper transient Unmanned Aerial Vehicles, University of Stellenbosch, 2007.
regime without big overshoots). [19] S.C. Kriel, A Comparison of Control Systems for the Flight Transition
While the constants   ,   ,   and  p ,  q ,  r have positive of VTOL Unmanned Aerial Vehicles, University of Stellenbosch, 2008.
[20] M. Chiaramonti, G. Mengali, “Control laws for a formation of
values, the system is stable; the bigger the values of constants autonomous flight vehicles”. Aeronautical Journal, April 2009, 113,
are, the better the system properties are (the overshoot and the (1147), pp. 609-616.
transient regime period decrease).

View publication stats