49th IEEE Conference on Decision and Control

December 15-17, 2010

Hilton Atlanta Hotel, Atlanta, GA, USA

Fault Recovery of an Under-Actuated Quadrotor Aerial Vehicle

M. Ranjbaran and K. Khorasani

Abstract Miniature Unmanned Aerial Vehicles (UAVs) with

ability to vertically take off and land (as in quadrotors)
exhibit advantages and features in maneuverability that have
recently gained strong interest in the research community.
Reliability of control systems require robustness and fault
tolerance capabilities in presence of anomalies and unexpected
failures in actuators, sensors or subsystems. Development of
an autonomous fault diagnosis and recovery system that can
cope with these faults has attracted a lot of interest in the
past several years. Particularly, for small aerial vehicles due to
hardware redundancy limitations design of a reliable control
system plays an important role in ensuring acceptable and
efcient performance.
In this paper, an autonomous fault recovery scheme is
proposed in response to actuator faults in an under-actuated
quadrotor aerial vehicle. A self-recovery mechanism, which
extends the capabilities of the quadrotor system to operate
under the presence of faults is proposed. The solution developed
takes into account the management of the control authority by
incorporating the post-fault model of the actuator.

I. INTRODUCTION
Unmanned Aerial Vehicles, or UAVs are becoming widely
used as valuable tools in todays society. As their application
both in the military and in the industrial sector increases,
potential miniature UAVs have steadily gained interest in the
research community. UAVs have several basic advantages
over manned systems including increased manoeuvrability,
low cost, reduced radar signatures and less risk to crews.
Vertical take off and landing type UAVs exhibit further
advantages in manoeuvrability features. Such vehicles are
to require little human intervention from the take-off to the
landing [1]. Quadrotors have become an exciting new area of
unmanned aerial vehicle research in the past few years. It is
an aircraft that is lifted and propelled by four rotors in a cross
conguration and its basic motions are generated by varying
the speeds of all the four rotors. The quadrotor rotorcraft is
not a new conguration. It already existed in 1920s [2]. The
uniqueness of this type of UAV is in its vertical landing/takeoff capability, hovering ability, great maneuverability and
being simple to manufacture.
The quadrotor is a 6 Degree of Freedom (DOF) device
with only four actuators, which make it an under-actuated
vehicle with unstable dynamics and highly coupled states.
This research is supported in part by a Strategic Projects Grant and
a Discovery Grant from the Natural Sciences and Engineering Research
Council of Canada (NSERC).
M. Ranjbaran was with the Department of Electrical and Computer
Engineering, Concordia University, Montreal, Quebec, H3G 1M8 Canada.
She is now with McGill University.
K. Khorasani is with the Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, H3G 1M8 Canada
(Email:kash@ece.concordia.ca).

978-1-4244-7746-3/10/$26.00 2010 Crown

In order to develop a reliable control to guarantee a stable

autonomous ight, development of simple and robust control
laws stabilizing the quadrotor has become an important area
of investigation.
Enhanced reliability and safety of complex and autonomous systems due to occurrence of actuator faults are
expected to be achieved by incorporating Fault Diagnosis,
Isolation and Recovery (FDIR) mechanisms in the design
of the control system. The FDIR module is in charge of
detecting, identifying, isolating and generating a recovery
procedure to allow acceptable performance of the system
when it is subject to a fault. The main objective of this
module is to enhance the reliability, performance and survivability of the system.
In general, methods for implementing fault detection and
isolation are classied into two categories, namely processhistory based methods and process-model based methods.
The rst approach depends on the knowledge collected
from past experiences and availability of a large amount of
historical data, while the latter relies on interactions between
various dynamical system components and variables and
a priori knowledge about the process.
The goal of the fault recovery mechanism is to select an
optimal possible conguration of the non-faulty actuators,
sensors, and components in the system where a fault has
occurred and diagnosed, to maintain the quality of the system
performance despite the presence of faults.
Over the past decade various approaches for fault recovery
have been proposed in the literature. One of the existing
active approaches for the fault recovery problem is through
adaptive control methods. The following methods fall under
this fault recovery category:
Model Reference Adaptive Control (MRAC) or Model
Following method [3], [4], and
Adaptive feedback linearization [5], [6].
A number of researchers have also developed various
control methods to stabilize a quadrotor. The work done in
[7] and [8] have used optimal Linear Quadratic Regulator
(LQR) for the controller design. Lyapunov theory is also used
as another design technique [9] and [10]. According to this
method, it is possible to ensure, under certain conditions, the
asymptotic stability of the aerial vehicle. Backstepping and
sliding mode control have also been used in [11], [12] and
[13]. In these works the convergence of the quadrotor internal
states is guaranteed, however, the computations required are
relatively excessive.
A feedback linearization method was rst used in [14]
to make the quadrotor track a reference trajectory. They
developed a nonlinear state space dynamic model and used

4385

an exact global feedback linearization and non-interacting

control law for controlling the translational motion and yaw
angle outputs.
The method developed in [14] was also used in [15].
In their work a PD controller was designed to control
the y-axis and the yaw angle, and a feedback linearization
controller was implemented to control the x-axis and the
z-axis states (translational motions). In [16], a feedback
linearization scheme with a high-order sliding mode observer
was developed for a quadrotor and in simulations it was
shown to be quite robust against wind disturbances and noise.
In [11], feedback linearization and adaptive sliding mode
control schemes for a quadrotor are also developed and their
capabilities are compared. Given that the quadrotor is an
under-actuated system, the possible set of available solutions
for control and fault recovery is rather limited. As shown
subsequently, an adaptive feedback linearization strategy is
employed for the purpose of fault recovery that will yield an
acceptable behavior and performance in presence of certain
types of faults in the vehicle actuators.
II. T HE Q UADROTOR M ODEL
The quadrotor simply consists of four dc motors on which
propellers are mounted in a cross conguration. Each propeller is connected to the motor through the reduction gears.
All the propellers axes of rotation are xed and parallel. The
front and the rear propellers rotate counter-clockwise, while
the left and the right ones turn clockwise. This conguration
of opposite pair directions removes the need for a tail rotor
(needed instead in the standard helicopter structure).
In Figure 1 the schematic of a simplied quadrotor structure is shown where i (rads1 ) refers to the propellers
rotation speed. While at hovering, all the four propellers
rotate at the same speed which implies that i = H for
i = 1, 2, 3, 4 to counterbalance the acceleration due to gravity.
Although the quadrotor has 6 DOF, it is just equipped with
four propellers, hence it is not possible to reach a desired
set point for all the DOF and the system is under-actuated.
However, by selecting four controllable variables properly,
it is possible to design a controller that will ensure the
vehicle could reach a desired height and attitude. Below, we
separately present the dynamical models corresponding to
the vehicle and the actuators.
1) Dynamic Model: Reference [17] provides a mathematical model of the quadrotor that is derived from the NewtonEuler formulation, that is

x = (cos sin cos + sin sin ) Um1

y = (cos sin sin sin cos ) Um1

z = g + (cos cos ) U1
m
(1)
= ( Iyy Izz ) Jt p T + U2

Ixx
Ixx
Ixx

Jt p
xx

IzzII
T + UIyy3
) + Iyy
= (

yy

= ( Ixx Iyy ) + U4
Izz

T

Izz

where x y z
represents the position of the quadrotor
in the inertial frame, the attitude state variables in the body

Fig. 1. Simplied quadrotor model representation at hovering and the

coordinate systems (the body and the earth reference frames).


T
frame
represents the roll, the pitch and the
yaw angles, respectively, m (Kg) is the overall mass, g refers
to the acceleration due to gravity, Ixx , Iyy and Izz (N m s2 ) are
the inertia moments in the body xed frame and Jt p (N m s2 )
is the total rotational moment of inertia around the propeller
axis. Furthermore, U1 denotes the normalized total lift force,
and U2 , U3 and U4 correspond to the control inputs of the
roll, the pitch and the yaw moments, respectively, and T
refers to the overall residual propeller angular speed. These
input moments are dened according to the equations
U = LUT T

(2a)

(2b)
T = 1 + 2 3 + 4
T
is the movement vector
where U = U1 U2 U3 U4
T

is the thrust vector. The
and T = T1 T2 T3 T4
constant matrix LUT is dened according to

1
1
1
1
0 l
0
l

LUT =
(3)
l
0
l
0
db db db db


where l(m) is the distance between the center of the quadrotor and the center of a propeller, b and d denote the thrust and
the drag coefcients, respectively, and Ti is the thrust force
generated by each rotor and is proportional to the square of
each propellerss speed, that is we have
Ti = b2i

(4)

2) Actuator Model: The rotors are driven by DC motors.

The model consists of three elements in series for the stator,
that is, the motor resistance Rmot , the motor inductance L, and
the back-EMF voltage which is given by e = ke m . Applying
the Kirchhoffs current law to the DC motor circuit results
in
di
u = Rmot i + L + ke m
(5)
dt
where u (V ) is the input voltage to the motor, ke (V s rad 1 )
is the back-EMF constant, and m (rad s1 ) is the motor
angular speed. The dynamics of the motor is described by
the following representation

4386

Jtm m = Mt Ml

(6)

where Jtm (N m s2 ) is the total motor moment of inertia,

m (rad s2 ) is the motor angular acceleration, Mt (N m) is
the motor torque, and Ml (N m) is the load torque. Since the
motor is small and assuming a low inductance, the rotor
dynamics can be approximated by:
u ke m
Ml
Rmot
k2
ke
= e m Ml +
u
Rmot
Rmot

Jtm m = kt

(7)

The actual motor system is composed of the motor itself, the

gear box and the propeller. Considering the propeller and the
gearbox, the load torque experienced by the motor is given
by the equation
d
Ml = 3 m2
(8)
r
where d(Nms2 ) denotes the aerodynamic drag factor and r
and refer to the gearbox reduction ratio and efciency
factor, respectively. The total inertia seen by the motor can
also be described as
Jtm = Jm +

Jp
r2

(9)

where Jm (N m s2 ) is the rotor moment of inertia about the

motor axis and J p (N m s2 ) is the rotor moment of inertia
about the propeller axis. Equation (7) can be rewritten
according to (8) and (9) as follows

(Jm +

Jp
ke2
d
ke

m 3 m2 +
u
m
r2
Rmot
r
Rmot

(10)

Equation (10) is formulated with respect to the motor axis.

It is possible to reformulate this equation from the propeller
axis as shown below
2
i = ke r2 i d2i + ke ru
(J p + r Jm )
Rmot
Rmot
2

(11)

From equation (11), the total rotational inertia around the

propeller axis Jt p (N m s2 ) can be dened as Jt p = (J p +
ke2
1
2
r2 Jm ). By setting Rmot
Jt p r = t , equation (11) can be
rewritten as:
i = 1 i d2i + 1 u

t
ke rt

Ti = 2bi i
2b
2bi
= 2i 2db3i +
ui
t
ke rt

2
2d
2 b
= Ti Ti Ti +
Ti ui
t
ke rt
b

Since the model (13) is nonlinear, a suitable approach

would be to linearize this dynamics around an operating
point, T0 . The rst order Taylor series approximation yields
the linearized model as given by
Ti = At Ti + Bt ui +Ct

(14)

In the above equation, the parameters At , Bt and Ct are the

linearized coefcients and are dened as follows:

3d

At = 2

T0

b
b
Bt = ke rt T0
(15)

C = d T 2
t

b 0

The set point corresponding to the linearizing condition

could be calculated from the fact that at hovering the total
thrust should be equal to the gravitational force effective on
the quadrotor. In other words, we have
4

Ti = mg

(16)

i=1

Using the linearized dynamic equation for the thrust Ti

developed in (14), it is possible to nd the dynamic equations
from the input voltage to the propellers to the movement
moments. For this purpose it is useful to write the dynamic
equation
in (14) in a matrix form for the thrust vector T =

T
T1 T2 T3 T4
, as follows
T = AT T + BT u +CT

(17)

where AT = At I(44) and BT = Bt I(44) are constant matrices


T
and CT = Ct 1 1 1 1 . The notation I(44) refers

T
to a 4 4 identity matrix and u = u1 u2 u3 u4
is
dened as the vector of the input voltages to the propellers.
By left-multiplying the transfer matrix given by equation
(3) between U and T , that is LUT , with (17) the following
equation is obtained
LUT T = (LUT AT )T + (LUT BT )u + (LUT CT )

(12)

The dynamic equation of the propeller angular speed

i (rad s1 ) is dened according to equation (12). On the
other hand, the input moments to the quadrotor dened in
equation (2a) are related to the propellers speed. Hence, it
is possible to derive input moment dynamic equations and
complete the quadrotor model by considering the effects of
the motor on the dynamics of the entire system.
From equations (4) and (12) the input voltage to the
propeller i, ui , and to the thrust Ti dynamic equation could
be obtained as follows:

(13)

(18)

It should be noted that:


LUT AT = LUT (At I(44) ) = At (I(44) )LTU = AT LUT
LUT BT = BT LUT
From equation (2a), it is possible to rewrite equation (18) as
U = AT U + (LUT BT )u + (LUT CT )

(19)

Since the quadrotor motion can be assumed close to the

hovering condition, small angular changes occur (especially
for the roll and the pitch angles). Since the rates of change
in and are small, the terms due to the gyroscopic

4387

effects appearing in the dynamic equations of and in

(1) are also negligible. These assumptions are also veried
through simulation results. Moreover, since the structure of
the quadrotor is symmetric, the body moments of inertia Ixx
and Iyy are equal. This fact will also simplify the dynamic
equation of in (1). If the altitude z reaches a desired setpoint given by zd , then z 0.
As stated earlier, in the hovering condition the total thrust
should be equal to the gravitational force effective on the
quadrotor, in other words:
4

U1 = Ti = mg

(20)

i=1

Therefore, if z 0 and and are sufciently close to

zero, then U1 mg.
By assuming U1 mg and 0, it is possible to
simplify the dynamic equations of the x and y states that
are dened in (1). Considering all the above assumptions,
the dynamic equations of the quadrotor system including
the dynamic equations for the movement vector is specied
according to the following model, that is

z = g + (cos cos ) Um1

y = g sin

x = g sin
= UIxx2
(21)

U3

= Iyy

= UIzz4

U = AT U + (LUT BT )u + (LUT CT )
The control algorithm that is to be designed should provide
the appropriate signals to the actuators. Since there are
only four propellers, no more than four variables can be
controlled in the loop. It is possible to dene the position
of thequadrotor in
T space completely by the linear position
E = x y z
and the yaw angle (heading angle) .
These four variables are indeed selected for control purposes
in this work. It can be seen that the system is partitioned into
four semi-decoupled subsystems as the outputs z, y, x and
can be controlled by U1 , U2 , U3 and U4 , respectively.
III. L OSS OF E FFECTIVENESS (LOE) FAULT M ODELING
In case of a loss of effectiveness (LOE) fault in an actuator,
the output speed of the quadrotor becomes different from the
commanded output desired by the controller, that is
i = ki ci

0 < < ki < 1

(22)

where i refers to the actual output from the ith actuator and
ci is the commanded output by the controller. Therefore,
the resulting thrust force from this actuator varies according
to the following equation
Ti = b2i
= b(ki ci )2

(23)

The dynamics of Ti dened in equation (14) would also

change due to the LOE fault, that is

Ti = 2bki2 ci ci
= ki2 At + ki2 Bt ui + ki2Ct

(24)

or in other words,
Ti = Ati Ti + Bti ui +Ct

i = 1, 2, 3, 4

(25)

where Ati = ki2 At and Bti = ki2 Bt . It should be noted that we

have assumed that the only coefcients subject to change
due to a fault are the Ati and Bti and the coefcient Ct would
stay unaffected. The term Ct is proportional to the drag and
proportional to the inverse square of the thrust factor, which
makes it a relatively small constant value.
Equation (25) can now be represented in a matrix form,
that is
T = AT 0 T + BT 0 u +CT
(26)
where

At1
0
AT 0 =
0
0

Bt1
0
BT 0 =
0
0

Ct
0
CT =
0
0

0
At2
0
0
0
Bt2
0
0
0
Ct
0
0

0
0
0
At4

0
0
0
0
0
Bt3
0
Bt4

0
0
0
0
Ct 0
0 Ct
0
0
At3
0

(27a)

(27b)

(27c)

Now if the thrust dynamics for all the actuators are not
identical, the dynamic equations of the movement vector U
change since AT = At I and BT = Bt I. Therefore, it is necessary to derive the dynamic equations of the movement vector
while the actuators do not have the same characteristics, in
other words when Ati = At j and Bti = Bt j for i, j = 1, . . . , 4,
i = j.
The relationship between the movement vector U and T
was dened in equation (2a). By left-multiplying equation
(26) with LUT , we would obtain
LUT T = (LUT AT 0 )T + (LUT BT 0 )u + (LUT CT )

(28)

From equation (2a), equation (28) can be rewritten as follows

1
U = (LUT AT 0 LUT
)U + (LUT BT 0 )u + (LUT CT )

(29)

If the actuators have the same parameters, in other words

1
Ati = At and Bti = Bt for i = 1, . . . , 4, then (LUT AT 0 LUT
)=
AT = At I44 as expected for the healthy system.
IV. A DAPTIVE F EEDBACK L INEARIZATION R ECOVERY
C ONTROL S TRATEGY
At the end of Section III, the dynamic equations of the
movement vector were derived in equation (29). In this
equation the contribution of each actuator to the resulting
movement vector is specied. As discussed earlier, in case
of a LOE fault the parameters of the actuators also change.
A parameter estimation algorithm is now presented in this
section to provide an estimate of the faulty actuator severity

4388

and to develop a nonlinear adaptive controller to guarantee

stability and recovery of the closed-loop system.
The recovery controller is designed by considering the
dynamic equation of the movement vector that is obtained
in equation (29). The following equations are derived for
designing the feedback linearization controller for the (z,U1 ),
(y,U2 ), (x,U3 ) and ( ,U4 ) subsystems, such that the input
appears in the output derivatives equation, namely
z(3) = sin cos
y(5) = g

U1
U1
U 1
cos sin
+ cos cos
m
m
m

U 2
U2
cos + g sin + 2g sin + g 3 cos
Ixx
Ixx

U 3
U3
x(5) = g cos g sin 2g sin g 3 cos
Iyy
Iyy
(3) =

U4
Izz

F2 =

(31)

(32)

(33)

1
4 (At1 + 3At )

1
2 (At1 + At )

b
4d (At1 + At )

At

l
4 (At1 + At )
d
4b (At1 + At )

1
2 (At1 + At )
d
2lb (At1 At )

lb
4d (At1 At )
1
4 (At1 + 3At )

Bt1

+
lBt1

db Bt1

Bt

Bt

Bt

lBt

Bt

lBt

d
b Bt

db Bt

d
b Bt

u1

u
2

u3

u4

F1 =

U1
m

U1
m

cos cos
m

u1

u2

u
3
u4

= F3

cos cos
m
g cos
Ixx l

cos cos
m

cos cos
m

g cos
Ixx l

0
g cos
Iyy l
1
d
Izz ( b )

0
1 d
Izz ( b )

(37b)

(37c)

1 d
Izz ( b )

Bt1

Bt

0
0

Bt

1
0


0


0

Bt

w1

w2
F1 A t1 F2

w3

w4
(38)

where A t1 and Bt1 are the estimates of the unknown parameters

 At1 and Bt1 andT the new control input vector
is to be selected so that the
W = w1 w2 w3 w4
stability and control (LQR) of the feedback linearized system
is achieved.
By applying the control law as dened in (38) to the
system (36), the closed-loop dynamics can be written as

z(3)

(5)
y

x(5)

(3)

U1

U
2

U3

U4

0
0
Bt1

0
B
0
t
= F1 + At1 F2 + F3

0
0 Bt

0
0
0

Bt1
0
0
0

B
0
0
t

F 1
3
0

0
B
0
t

0
0
0 Bt

Bt

w1

w2
F1 A t1 F2

w3

w4
(39)

Let us dene the parameter estimation error as the difference between the actual value of the unknown parameter
and its estimate, i.e. At1 = At1 A t1 and Bt1 = Bt1 Bt1 .
Therefore, equation (39) can be rewritten as

+ (LUT CT )

(35)

The above dynamic equation is now substituted in equations

(30) to (33) which can be rewritten in a compact matrix
form. Note that we are seeking a form to separate the terms
that are related to the unknown variables At1 and Bt1 . This
is achieved in the following equation as shown below
(3)

z
Bt1 0 0 0
u1
y(5)
0 Bt 0 0 u2

x(5) = F1 + At1 F2 + F3 0
0 Bt 0 u3
0
0 0 Bt
u4
(3)
(36)
where

cos cos
m

The input signal u is dened as

where At1 = k2 At and Bt1 = k2 Bt .

The dynamics of the movement vector as dened in (29)
can be written as
U 1

U
2

U 3

U 4

g cos
l
1
lb
Iyy ( 4 U1 + 2 U3 + 4d U4 )
1
d
d
1
Izz At ( 4b U1 + 2lb U3 + 4 U4 )

F3 = g cos

Iyy l
1
d
Izz ( b )

(30)

It should be noted that the relative degree of the system is

equal to the order of the system and no internal dynamics
exists in designing the feedback linearization controller.
Without loss of generality, let us assume that the LOE
fault has occurred in the rst actuator and the other three
actuators are healthy, that is

T1 = At1 T1 + Bt1 u1 +Ct
(34)
Ti = At Ti + Bt ui +Ct
for i = 2, 3, 4

cos cos 1
b
( 4 U1 12 U3 4d
U4 )
m

b
sin cos cos sin +
At ( 34 U1 + 12 U3 + 4d
U4 ) + 4Ct
U2
g
3g Ixx sin + g 3 cos Ixx cos At U2
U

l
1
lb
3g Iyy3 sin g 3 cos + g cos
Iyy At ( 4 U1 + 2 U3 4d U4 )
1
d
d
3
Izz At ( 4b U1 2lb U3 + 4 U4 )

z(3)
y(5)
x(5)
(3)

= F1 + At1 F2

Bt1
Bt1

+ Bt1 F3

(37a)

4389

w1

2
0
0 0 0
F 1
F1 A t1 F2

3
w3

0
0 0 0
w4
0
0 0 0

w1
w1
w
w
2
2
+ F3 F31
F1 A t1 F2 =
+ At1 F2
w3
w3

+ F3

w4
1
Bt1

0
0
0

0
0
0

0
0
0

w1

w
0

1
F 2
3 w
0
3
w4
0

w4

F1 A t1 F2

(40)

Let

1
Bt1

0
F4 = F3
0
0

0
0
0

0
0
0

0 1
F3

0
0

w1
w2
w3
w4




At1 = A t1 = F2T PX + X T PF2


Bt1 = Bt1 = F4T PX + X T PF4

F1 A t1 F2

(41)

Hence, the linearized dynamics of system (40) can be rewritten as

(3)

z
w1
y(5) w2

(42)
x(5) = w3 + At1 F2 + Bt1 F4
w4
(3)
The control inputs w1 , w2 , w3 and w4 are dened according
to the following equations
(3)

w1 = zd k1z (ez ) k2z (ez ) k3z (ez )

(43a)

(5)
(4)
(3)
w2 = yd k1y (ey ) k2y (ey ) k3y (ey ) k4y (e)
y k5y (ey )

(43b)
(5)
(4)
(3)
w3 = xd k1x (ex ) k2x (ex ) k3x (e)
x k4x (e)
x k5x (ex )
(43c)
(3)
w4 = d k1 (e ) k2 (e ) k3 (e )
(43d)

(48)

where X is the state vector dened in equation (45), and P is

a (16 16) matrix that is obtained by solving the following
Lyapunov equation, that is
AT P + PA = I1616

(49)

where the A matrix is dened according to system (46)

and the notation I1616 denotes a 16 16 identity matrix.
It should be noted that since A is a Hurwitz matrix, P is a
positive denite matrix [18].
The following theorem provides a sufcient condition for
stability of the resulting closed-loop recovered system.
Theorem 1: The state trajectories of the closed-loop system
(46) that is subjected to the LOE fault in the rst actuator
and which has employed the update law for the parameter
estimation errors At1 and Bt1 given by equation (48) are
globally stable in the sense of Lyapunov.
Proof: To carry out the stability analysis, let us choose the
following radially unbounded Lyapunov function candidate

In the above equations zd , yd , xd and d denote the

desired output variables and ez = z zd , ey = y
yd and ex = x xd and e = d are dened
as
The gain vectors Kz =
 the trackingT error signals.

T
, Ky = k1y k2y k3y k4y k5y , Kx =
k1z k2z k3z

1
1
V = (X T PX + A2t1 + B2t1 )
(50)
2
2
The derivative of this function along the state trajectories of
the closed-loop system (46) yields

and K = k1 k2 k3 are
k1x k2x k3x k4x k5x
obtained from the LQR design procedure.
From the equations (42) and (43a) to (43d), it is possible
to write the dynamic equations of the error signals ez , ey , ex
and e as follows

Now, by using (48) the above expression simplies to

T

V = X T X

(44)

T

(3) (3)
(4) (4)
X = ey ex ey ex ez ey ex e ez ey ex e ez ey ex e
(45)

Therefore, one could specify the state-space representation

of the system according to

F4 =




0
0

...
...

0
0

= X T AT PX + X T PAX

F2
F4

T

(46)

116

(47a)

116

(47b)

T

(52)

It can be concluded from (49) that

It is possible to represent the above equation in the statespace form. The selected state variables for this purpose is
as follows

F2 =

+ At1 (F2T PX X T PF2 ) + Bt1 (F4T PX X T PF4 )

= X T (AT P + PA)X

= A F2 + B F4
t1
t1

X = A1616 X + At1 F2 + Bt1 F4

(51)

V = (AX + At1 F2 + Bt1 F4 )T PX + X T P(AX + At1 F2 + Bt1 F4 )

(3)

ez + k1z ez + k2z ez + k3z ez

(5)
(3)
ey + k1y e(4)
y + k2y ey + k3y ey + k4y ey + k5y ey

e(5) + k e(4) + k e(3) + k e + k e + k e

1x x
2x x
3x x
4x x
x
5x x
(3)
e + k1 e + k2 e + k3 e

where

V = X T PX + X T PX + At1 At1 + Bt1 Bt1

T

All the eigenvalues of the A matrix are negative by proper

selection of the ki parameters. In other words, A is a Hurwitz
matrix. Now, let us dene the update law for the parameter
estimation errors At1 and Bt1 according to the following
rules:

(53)

which guarantees the negative semi-deniteness of the function V . This implies that the origin is a globally stable
equilibrium point of system (46) given that the Lyapunov
function is radially unbounded.
Remark: Provided that the external desired reference trajectories are persistently exciting, one can then ensure that
A t1 At1 and Bt1 Bt1 , implying that the fault recovery
system can independently be used to isolate and identify the
severity of the injected fault.
In this section, we have studied the case of a LOE fault
recovery by assuming that the fault has occurred in the
rst actuator. The same method could be applied to design
three other controllers that accommodate the LOE fault in
the other three actuators. In other words, the fault recovery
module contains four different controllers and based upon
the information that one can receive from the fault detection
and isolation unit the proper controller can then be selected
and invoked. In each of these controllers, two parameters,
namely, Ati and Bti that are related to the faulty ith actuator

4390

are assumed to be unknown and the parameters related to the

healthy actuators are assumed to be all known. It is clear that
similar approach can be employed to design a controller for
the healthy system where all the parameters are considered
to be known.
(a)

x (m)

10

5
Recovered

10

20

30

Delayed recovery
Healthy

40

50
time (sec)

60

Faulty

70

80

90

100

90

100

(b)
6

4
y (m)

Healthy
Faulty
Recovered
Delayed recovery

10

20

30

40

50
time (sec)

60

70

80

(c)
6

z (m)

4
Healthy

Delayed recovery

Recovered

of a delayed fault detection and isolation by initiating the

recovery procedure 8 seconds after the fault occurrence.
Figures 2 and 3 show the linear position and the Euler
angles of the system, respectively, in response to the commanded trajectory for the healthy system as well as the
faulty and the recovered scenarios by using our proposed
fault recovery algorithm with and without the delayed fault
detection and isolation information. Figure 4 depicts the
estimated parameters of the faulty actuator corresponding
to this case. Table I shows the means and the standard
deviations of the steady state tracking error signals for x,
y, z and under four different scenarios where the system
is operating, namely, (I) healthy condition, (II) with a faulty
actuator and no fault recovery solution invoked, (III) with a
fault recovery mechanism invoked, and (IV) with a delayed
fault recovery mechanism invoked. It can be seen from
Figures 2 and 3 and Table I that our proposed fault recovery
mechanism has a considerable effect on the performance of
the system in presence of a 25% LOE fault by reducing the
steady state error between the desired and the actual outputs.
It should be noted that as intuitively expected, the sooner
the fault recovery is initiated, the improved performance and
smaller error signals are obtained.

Faulty

(a)
0.5
0

10

20

30

40

50
time (sec)

60

70

80

90

100

Fig. 2. Linear position in response to the commanded trajectory corresponding to a 25% LOE fault in the rst actuator subject to as well as
without the fault recovery mechanism invoked: (a) x-axis, (b) y-axis, and
(c) z-axis measured in meters.

roll (rad)

Healthy
Faulty
Recovered

Delayed recovery

0.5
0

10

20

30

40

50
time (sec)

60

70

80

90

100

60

70

80

90

100

(b)

V. S IMULATION R ESULTS

0.5
0

10

20

30

40

50
time (sec)
(c)

1.5
Delayed recovery

yaw (rad)

In this section, the behavior of the quadrotor system that

is embedded with the fault recovery mechanism is studied
through simulation results. The quadrotor model that is
used for simulation is the OS4 that is developed in Ecole
Polythechnique Federal De Lausanne [17]. The mathematical
model that is used for control design is partially nonlinear
(the actuator dynamics is linearized for control design),
however, we have applied the controller to the fully nonlinear
model of the quadrotor and have considered the full actuator
dynamics. In simulations additive white Gaussian noise is
also added to the input and output channels to simulate a
more realistic environment. The noise power is selected so
that the signal to noise ratio is approximately 15 db.
The commanded trajectory starts at the position (x, y, z) =
(0, 0, 0) while the roll, the pitch and the yaw angles are
initially set to zero. The commanded trajectory is to y
from the initial point to the nal point (5, 5, 5)(m) in 10
seconds and hovering at the nal point. The fault trajectory
considered here assumes a 25% partial LOE failure in the
rst actuator at t = 20 sec. We are assuming that a fault detection and isolation mechanism exists to alarm the occurrence
of the fault in the faulty actuator without much delay and
the fault recovery mechanism is initiated after the detection
and isolation of the fault. The performance response of our
proposed fault recovery scheme is also evaluated in case

pitch (rad)

0.5

1
Faulty
Recovered

0.5

Healthy

0
0

10

20

30

40

50
time (sec)

60

70

80

90

100

Fig. 3. Euler angles in response to the commanded trajectory corresponding

to a 25% LOE fault in the rst actuator subject to as well as without the
fault recovery mechanism invoked: (a) Roll angle (rad), (b) Pitch angle (rad)
and (c) Yaw angle (rad).

Next, the performance of the quadrotor system is evaluated

in case of a more severe LOE fault. Specically, we consider
a 50% LOE fault in the rst actuator in a mission similar
to the previous case. Table II shows the means and the
standard deviations of the steady state tracking error signals
for x, y, z and under four different scenarios where the
system is operating, namely, (I) 50% LOE in the rst actuator
and without fault recovery solution invoked, (II) with an
immediate fault recovery mechanism invoked, (III) with a 5

4391

TABLE I
M EAN

(S TDV ) OF THE TRACKING ERROR

SIGNALS FOR 25% LOE. L EGEND : (I) HEALTHY CONDITION , (II) WITH
AND STANDARD DEVIATION

A FAULTY ACTUATOR AND NO FAULT RECOVERY INVOKED ,

FAULT RECOVERY INVOKED , AND

(IV)

(III) WITH A

WITH A DELAYED FAULT

RECOVERY INVOKED .

Error

(I)

(II)

(III)

(IV)

ex Mean (m)
ex Stdv (m)
ey Mean (m)
ey Stdv (m)
ez Mean (m)
ez Stdv (m)
e Mean (rad)
e Stdv (rad)

-0.0537
0.2809
-0.0486
0.1632
-0.0542
0.1884
1.3461e-004
8.7942e-004

4.2883
2.3790
-0.0484
0.1634
-1.0619
0.4445
0.8659
0.4471

0.4226
0.4586
-0.0484
0.1644
-0.1645
0.1684
0.0947
0.0566

2.8756
1.8593
-0.0483
0.1637
-0.7317
0.3586
0.5818
0.3505

TABLE II
M EAN

(S TDV ) OF THE TRACKING ERROR

50% LOE. L EGEND : (I) FAULT IN THE FIRST ACTUATOR

AND STANDARD DEVIATION

SIGNALS FOR

AND WITHOUT FAULT RECOVERY INVOKED ,

FAULT RECOVERY INVOKED ,

(II) WITH AN IMMEDIATE

(III) WITH A 5 SECOND DELAY IN

INITIATING THE FAULT RECOVERY, AND

(IV)

WITH A

10

SECOND DELAY

IN INITIATING THE FAULT RECOVERY.

Error

(I)

(II)

(III)

(IV)

ex Mean (m)
ex Stdv (m)
ey Mean (m)
ey Stdv (m)
ez Mean (m)
ez Stdv (m)
e Mean (rad)
e Stdv (rad)

11.8699
5.2980
-0.0484
0.1637
-4.3580
1.1027
2.9836
1.0285

0.3915
0.4485
-0.0484
0.1631
-0.1567
0.1702
0.0885
0.0548

4.5412
3.5770
-0.0484
0.1639
-0.8790
0.6828
0.8147
0.3263

6.5079
4.1577
-0.0482
0.1637
-1.5712
0.8786
1.5082
0.8225

second delay in initiating the fault recovery, and (IV) with a

10 second delay in initiating the fault recovey. As expected,
the fault recovery mechanism performance is considerably
better when there is smaller delay in the detection and
isolation of the fault.
(a)

(b)

15.58

2.5

Bt2

At1

15.6
15.62

15.64
15.66

20

40
60
time (sec)

80

100

1.5

20

40
60
time (sec)

80

100

Fig. 4.
Estimated actuator parameters in response to the commanded
trajectory corresponding to a 25% LOE fault in the rst actuator with the
fault recovery mechanism: (a) At1 and (b) Bt1 .

VI. C ONCLUSION
In this paper, a fault recovery mechanism is proposed
for reconguring the control system from LOE faults in
the quadrotors actuators. An adaptive feedback linearization
technique is employed for the controller design and global
stability of the system with the fault recovery mechanism
is shown. This is accomplished by introducing a parameter
estimation algorithm and by deriving proper update laws

for the faulty actuator parameters subject to changes in

the quadrotor actuators due to the presence of the LOE
fault. Simulation results are also presented to evaluate the
performance of the proposed fault recovery mechanism in
presence of an LOE fault in one of the actuators. It is
observed that by employing the fault recovery algorithm the
steady state tracking errors of the system outputs reduce
considerably when compared to the responses obtained from
the faulty system that has not employed our proposed fault
recovery strategy.
R EFERENCES
[1] L. Derafa, T. Madani, and A. Benallegue, Dynamic modelling and
experimental identication of four rotors helicopter parameters, in
IEEE International Conference on Industrial Technology, pp. 1834
1839, 2006.
[2] P. Castillo, R. Lozano, and A. Dzul, Modelling and Control of MiniFlying Machines. Springer, 2005.
[3] M. Bodson, A recongurable nonlinear autopilot, in Proceedings of
the AIAA Guidance, Navigation, and Control Conference, 2002.
[4] M. Bodson, Multivariable adaptive algorithms for recongurable
ight control, IEEE Transactions on Control Systems Technology,
1997.
[5] R. Patton, Fault-tolerant control systems: The 1997 situation, in IFAC
Symposium on Fault Detection Supervision and Safety for Technical
Processes, 1997.
[6] A. T. Vemuri and M. M. Polycarpou, Neural network based robust
fault diagnosis in robotic systems, IEEE Transactions on Neural
Networks, vol. 8, pp. 14101420, 1997.
[7] J. How, B. Bethke, A. Frank, D. Dale, and J. Vian, Real-time
indoor autonomous vehicle test environment, IEEE Control Systems
Magazine, 2008.
[8] I. Cowling, J. Whidborne, and A. Cooke, Optimal trajectory planning
and lqr control for a quadrotor uav, in Proceedings of the International
Conference Control, 2006.
[9] R. L. P. Castillo and A. Dzul, Stabilization of a mini rotorcraft
having four rotors, in Proceedings of 2004 IEEE/RSJ International
Conference on Intelligent Robots and Systems, pp. 2693 2698, 2004.
[10] A. Palomino, S. Salazar-Cruz, and R. Lozano, Trajectory tracking for
a four rotor mini-aircraft, in Proceedings of the 44th IEEE Conference
on Decision and control, and the European Control Conference,
p. 2505 2510, 2005.
[11] D. Lee, H. J. Kim, and S. Sastry, Feedback linearization vs. adaptive
sliding mode control for a quadrotor helicopter, International Journal
of Control, Automation, and Systems, pp. 419428, 2009.
[12] H. Bouadi, M. Bouchoucha, and M. Tadjine, Sliding mode control
based on backstepping approach for an uav type-quadrotor, International Journal of. Applied Mathematics and Computer Sciences, pp. 12
17, 2007.
[13] R. Xu and U. Ozguner, Sliding mode control of a quadrotor helicopter, in Proceedings of the 45th IEEE Conference on Decision and
Control, 2006.
[14] V. Mistler, A. Benallegue, and N. MSirdi, Exact linearization and
noninteracting control of a 4 rotorshelicopter via dynamic feedback,
in Proceedings of 10th IEEE International Workshop on Robot and
Human Interactive Communication, 2001.
[15] E. Altug, J. P. Ostrowski, and R. Mahony, Control of a quadrotor
helocopter using visual feedback, in Proceedings of the IEEE International Conference on Robotics and Automation, pp. 72 77, 2002.
[16] A. Mokhtari and A. Benallegue, Dynamic feedback controller of
euler angles and wind parameters estimation for a quadrotor unmanned
aerial vehicle, in Proceedings of IEEE International Conference on
Robotics and Automation, 2004.
[17] S. Bouabdallah, Design and Control of Quadrotors With Application
to Autonomous Flying. PhD thesis, Ecole Polythechnique Federale De
Lausanne, 2007.
[18] J. J. Slotine and W. Li, Applied Nonlinear Control. Prentice Hall,
1991.

4392