Fault-Tolerant Lyapunov-Gain-Scheduled PID Control of a Quadrotor UAV

Abderrahmen Bouguerra1*, Djamel Saigaa 2, Kamel Kara3, Samir Zeghlache1

1
Department of Electrical Engineering, M’sila University, Zip postal 166 Ichbilia 28000, M’sila, Algeria
2
Departments of Electronics, M’sila University, Zip postal 166 Ichbilia 28000, M’sila, Algeria
3
Departments of Electronics, Blida University, Zip postal 270 Blida, Algeria
* Corresponding author’s Email:[email protected]

Abstract: The work has done in this paper concern the passive fault tolerant control. Based on Gain-Adaptive
Proportional-Integral-Derivative (PID) using the approach from the theory of Lyapunov and their application to the
model vertical flying drone Quadrotor type, the PID controller with fixed parameters may fail to provide acceptable
control performance. To improve the PID control effect, new designs of the Lyapunov gain Scheduled PID controller
(LGSPID) were presented in this paper. The proposed techniques were applied to the Quadrotor, where adaptive PID
controllers were proposed for fault-tolerant control system in the presence of actuator faults. The parameters of PID
controller were adjusted by an adaptation algorithm gradient type, used to tune in real-time the controller gain, the
proposed adaptive PID controller was compared with the conventional PID. The obtained results confirm the
effectiveness of the proposed method.
Keywords: Adaptation algorithm gradient type; Gain-adaptive PID; LGSPID; Quadrotor model; PID control.

each a rotor to change the thrust and torque pro-

1. Introduction duced by each. Quadrotors are classified as rotor-
craft, as opposed to fixed wing aircraft, because
Fault-Tolerant Control (FTC) is a relatively new their lift is derived from four rotors [5].
idea that makes possible to develop a control feed- PID controllers are the most familiar controller
back that allows keeping the required system perf- in the society of automation and control, due to
ormance in the case of faults[1]. The control their simple structure and wide variety of usage.
strategy can perceive fault tolerant when there is an These kinds of controllers are classified into two
adapta- tion mechanism that changes the control main categories in terms of parameters selection
law in the case of faults. strategies. In the first group, controller gains are
Another solution is to use hardware redundancy fixed during operation while in the second group,
in sensors and/or actuators. In general, FTC gains change based on the operating conditions.
systems are classified into two distinct classes [2]: In the first group, gains are tuned by the
passive and active. In passive FTC [3, 4], designer and remain invariable during the
controllers are designed to be robust against a set of operation.
presumed faults, therefore there is no need for fault One of the most familiar methods for choosing
detection. In contrast to passive ones, active FTC control gains in this group is Ziegler-Nichols me-
schemes react to system components faults actively thod [6].
by re- configuring control actions and by doing so In most applications, due to structural changes
the system stability and acceptable performance is the controlled system may lose its effectiveness,
maintained. therefore the PID gains need to be continuously
A Quadrotor is an aircraft that is lifted and retuned during the system life span. To reduce the
propelled by four rotors. The Control of Quadrotor effort of retuning the gains and also in order to
can be achieved by varying the relative speed of

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 1

increase system's performance, in the second group
of controllers, the gains are adapted online.
A number of methods have been proposed in
documents for PID gain scheduling [7]. A stable
gain-scheduling PID controller is developed based
on grid point concept for nonlinear systems.
Different gain scheduling methods have been
studied and compared [8, 9] and a new PID scheme
is proposed in which the controller gains are
scheduled by a fuzzy inference scheme. Many Figure1 The Quadrotor in an inertial frame
methods and research work in this domain in
[10-14]. And an intelligent control scheme uses a  Free stream air velocity is zero.
fuzzy switching mechanism, grey prediction and  The motors dynamics is relatively fast and
genetic algorithm (GA) in [15]. The interested can be neglected.
readers can find a brief review of different fuzzy  The flexibility of the blade is relatively
PID structures in [16]. small and can be neglected.
In this work, adaptive Lyapunov Gain Sche-  Drag is supposed to be linear, thus obeying
duled PID control approaches for a Quadrotor Stokes’s law.
system was proposed. The dynamic model [23] is:
A PID controller is used to approach a law
unknown ideal control online. Contrary to the work  u
1 sin 
of [17-19], wherein the adaptation law is selected to  x  
ensure the decrease of a Lyapunov function  m
candidate on the output error, the adaptive law, in  y  u1 cos  . sin 
this work was selected to minimize the gradient
method a quadratic criterion of error at the input of
 m
 u
the system, the error between the unknown ideal
control and output of the PID controller [17-19].  z  1 cos  . cos   g
The remainder of this paper is organized as  m (1)
follows. The model of the Quadrotor is described in 
Section II. The Lyapunov Gain Scheduled PID    
(LGSPID) strategy is designed in Section III.    
Section IV presents the simulation results to de- 
monstrate the effectiveness of the FTC Controller.    
Concluding remarks are provided in Section VI.
The generalized moments are:
2. Model of the auadrotor UAV
 
In this section, the general dynamic model of a    
Quadrotor UAV was studied.   (2)
A body-fixed frame B(O ', x, y, z) and the earth-  
fixed frame E(O, X , Y , Z ) were assumed to be at
the center of gravity of the Quadrotor UAV, where   (3)
the z-axis was pointing upwards, as seen in Figure
1. where    x, y, z   R 3    ,  ,   S3 and J  TT I T .
The orientation of Quadrotor UAV that referred
to as roll, pitch and yaw was given by a vector   s 0 1
( , , ) which was measured with respect to the T  c s c 0  (4)
c c 0 
earth coordinate frame E.  s
Based on some basic assumptions as given be-
low:
 Design is symmetrical.
 I xx 0 0   I xx 0 0 
I   0 I yy 0    0 I xx 0  (5)
 Quadrotor body is rigid.   
 Propellers are rigid.  0 0 I zz   0 0 2 I xx 

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 2

 The coriolis and centripetal vector denoted by 3. Ftc Strategy
C  ,   defined as below and computed as given by
In this section, adaptive PID controllers were to
(9). best approximate the ideal command unknown (14)
[24].

 J  
1 
C  ,     J 
T
(6)
 2    (  f ( x )  v   s   tanh( s /  ))
uu  (14)
g ( x)
 1  c c 2 2
c s c  s 
  with α> 0, β> 0 and ε a small positive constant.
J  I xx c s c 2  c
2

 0
 (7)
  s 0 1  Where：

d
  ) e(t ),  0
n 1
  s 2 c   s 2 c  s s c   c 2 c  c 
2 2
s (t )  ( (15)
  dt
J  I  s s c   c 2 c  s 2
 xx
0
 (8)
  c 0 0  where: e(t )  yd (t )  y (t ) .
Choose the Lyapunov function:
C 1,1
 C1, 2  C1, 3  0
 1
C  I xx ( c s 2   s c s   c )
2
V s2 (16)
2 ,1
2

 I xx s c s
C 2, 2
The derivative of (10) along paths (9) is bounded

C 2,3
  I xx c (9) by:

  I xx ( c s 2   c c 2 )
2
C V   s
2
3,1 (17)
C   I xx ( c c 2   s 2 )
 3, 2 The derivative of the filtered error can be
C 3, 3
0 written as [24]

where m denotes the mass of the rotorcraft and s  v  f ( x)  g ( x)u (18)

I XX  I YY  ml , I ZZ  2ml .
2 2

with: v  yd ( n)  kn1e( n1)  ...  k1e , k j  Cnj11 n  j

Where (see Figure 1) [20, 21]:
where λ is positive constant.
u1  F1  F2  F3  F4 (10) The three PID controller gains kp, ki and kd were
considered here as adjustable parameters. To do this,
u4  d  F1  F2  F3  F4  (11) an adaptive mechanism would be developed to
minimize a quadratic criterion of the error between
u3   F2  F4  l (12) the ideal unknown command u  and the command
u pid provided, resulting from the PID controller.
u2   F3  F1  l (13) The ideal control law (14) was then approximated
by a PID controller of the form.
Table 1 The parameters of the quadrotor rotorcraft [22]
t
de(t )
Definition Parameter Value 
u  u pid  k p e(t )  k i e( )d  kd
0
dt
(19)
Lever length l 0.232m
Mass of Quadrotor m 0.52kg where: u   (e)l .
T

Drag coefficient d 7.5e-7N ms2

T
Thrust coefficient b 3.13e-5N s2  t
de(t ) 
 (e)  e(t )  e( )d  (20)
Rotor inertia Jr 6e-5N kgm2  0
dt 
Gravitational
g 9.81 m/s2  l is the vector of parameters adjusted in the control,
acceleration
which is defined by: l  k p ki kd 

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 3

 From equation (18) X position υ=1.27, λ=2.1
α=20, ε=0.001, β=2,
 Ki
s   s   tanh( s /  )  g ( x)(u  u ) (21) υ=1.4, λ=2
α=8, ε=0.001, β=1.2,
with: u   T (e)l  is optimal command and θ* Kd
υ=5, λ=3
is the optimal parameters. α=12, ε=0.001, β=6,
Kp
From equation (21) υ=3.7, λ=2
LGS-PID α=13, ε=0.001, β=3.7,
Ki
g ( x)eu  s   s   tanh(s /  ) (22) Y position υ=5.4, λ=2.5
α=9, ε=0.001, β=2,
Kd
From (22), the law of adaptation parameters is υ=5.2, λ=2
given by α=7, ε=0.001, β=4,
Kp
l    (e) s   s   tanh( s /  ) 
(23) LGS-PID
υ=7, λ=1.7
α=8, ε=0.001, β=3,
Ki
Ψ angle υ=5, λ=2
α=11.4, ε=0.001, β=2,
4. Simulation Result Kd
υ=2, λ=1
The proposed LGS-PID control scheme (fig.2)
presented in this paper was tested on a model of the 2.5 2.5
full Quadrotor helicopter model in presence of the
2 2
actuators faults.
X position (m)

Y position (m)
It is assumed that a loss of control effectiveness 1.5 1.5

of 40% by echelon the faults were taking place in 1 1

the command u1, u2, u3 and u4 at time instant t=35s PID control PID control
and ends on time t=55s. 0.5 Desired 0.5 Desired
LGS-PID control LGS-PID control
The synthesis parameters are selected as Table 2. 0 0
0 20 40 60 80 0 20 40 60 80
Time (s) Time (s)

3.5 3.5

3 3

2.5 2.5
yaw ange (rad)
Z position (m)

2 2

1.5 1.5

1 1 PID control
PID control
Desired 0.5 Desired
0.5
LGS-PID control LGS-PID control
0 0
0 20 40 60 80 0 20 40 60 80
Time (s) Time (s)

0.1
0.1
PID control PID control
Figure 2 Synoptic scheme of the proposed control 0.05
LGS-PID control 0.08 LGS-PID control
pitch angle (rad)

roll angle (rad)

strategy 0.06

0 0.04
Table 2 Synthesis parameters of the proposed
controller -0.05
0.02

0
Definition Parameter Value
-0.1 -0.02
PID (φ) Kp =4.5, Ki =1.5, Kd 0 20 40 60 80 0 20 40 60 80
Kp, Ki, Kd
angle =1.5 Time (s) Time (s)

PID (θ) Kp =3.5, Ki =0.5, Kd Figure 3 Comparison between PID control and
Kp, Ki, Kd LGS-PID control
angle =2
α=20, ε=0.001, β=10,
Kp
υ=3, λ=5
LGS-PID α=10, ε=0.001, β=20,
Ki
Z position υ=4, λ=20
α=18, ε=0.001, β=2,
Kd
υ=0.5, λ=1
LGS-PID Kp α=22, ε=0.001, β=12,

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 4

 12 LGS-PID control 6 LGS-PID control control, the LGS-PID gains were time-varying to
PID control
10 PID control adapt to uncertainties, disturbances as can been
4
clearly in figure 6. It can be seen in Figure 3, 4 and
U1 (N.m)

U2 (N.m)
8

6 2 5 that after the fault occurs, Kp decreased to avoid

4 system pass due to increase in tracking error. The
0
2 derivative gain Kd remained fixed with a high value
0 -2 to make a fast response to sudden changes in
0 20 40 60 80 0 20 40 60 80
Time (s) Time (s) tracking error. When the system stopped de-
scending (losing altitude) Kd decreased to let the
6 PID control 6 PID control system recover faster and go back to its desired
LGS-PID control 5 LGS-PID control
position. After the fault, integrator gain Ki also
4 4
increased to help the recovery process.
U3 (N.m)

U4 (N.m)

3
2 2
1 5. Conclusion
0
0
-1 In this paper, we presented the Fault-Tolerant
-2
0 20 40 60 80 0 20 40 60 80 Lyapunov-Gain-Scheduled PID Control of the full
Time (s) Time (s)
Quadrotor helicopter in the presence of the fault.
Figure 4 Commands u1, u2, u3 and u4 of PID control Firstly, we started by the development of the dy-
and LGS-PID control namic model of the Quadrotor taking into account
the different physics phenomena, after we are in-
100
60
15 terested in proposing the FTC controller based on
Kd gain of X position

Lyapunov method. Simulation results also validated

Kp gain of X position

Ki gain of X position

80
40
10
60
20 that the presented FTC had a satisfactory tracking
40 0 5
performance and was robust to the external dis-
20 -20 turbances.
0 -40 0
0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
Time (s) Time (s) Time (s)
References
60 15

[1] M. Witczak, V. Puig, and S. Montes de Oca, “A

Kd gain of Y position
Kp gain of Y position

Ki gain of Y position

150
40
10 fault-tolerant control scheme for non-linear
100 20
discrete- time systems: Application to the
50 0
5 twin-rotor system”, In: Conf. Rec. 2010 IEEE Int.
-20
Conf. Control and Fault-Tolerant Systems (SysTol),
0
0 20 40 60 80 0 20 40 60 80
0
0 20 40 60 80 pp.861–866, 2010.
Time (s) Time (s) Time (s)

100 150 16
[2] Ł.Dziekan, M.vWitczak and J. Korbicz, “Active
fault-tolerant control design for Takagi-Sugeno
Kd gain of Z position
Kp gain of Z position

Ki gain of Z position

80 14
100 fuzzy systems”, Bulletin of the polish academy of
60
50
12
sciences technical sciences, Vol.59, No.1, 2011.
40 10

0
[3] J. Chen, R. Patton and Z. Chen, “An LMI approach
20 8
to fault tolerant control of uncertain systems”, In:
0
0 20 40 60 80
-50
0 20 40 60 80
6
0 20 40 60 80 Conf. Rec. 1998 IEEE Int. Conf. Decision and
Time (s) Time (s) Time (s) Control, pp.175–180, 1998.
60 3.5
50
3 [4] Z. Qu, C. Ihlefeld, J. Yufang and A. Saengdeejing,
“Robust fault-tolerant self-recovering control of
Kp gain of roll angle

Kd gain of roll angle

Ki gain of roll angle

40 40 2.5
30 2
non-linear uncertain systems”, Automatica. Vol.39,
20 1.5
20
1
pp.1763–1771, 2003.
10
0 0.5
0
0
[5] J. G. Ziegler and N. B. Nichols, “Optimum settings
-10
0 20 40 60 80
-20
0 20 40 60 80
-0.5
0 20 40 60 80
for automatic controllers”, ASME Trans. Vol.64, pp.
Time (s) Time (s) Time (s) 759–768, 1942.
Figure 5 Gains Kp, Ki and Kd of LGS-PID control [6] T. C. T. Ng, F. H. F. Leung and P. K. S. Tam, “A
simple gain scheduled PID controller with stability
The time evolutions of the LGS-PID gains are consideration based on a grid-point concept”, In:
illustrated in Figure 5. Unlike those of the PID Conf. Rec. 1997 IEEE Int. Conf. Industrial Elec-

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 5

 tronics, pp.1090–1094, 1997. [19] S. Bouabdallah and R. Siegwart, “Backstepping
and Sliding-mode Techniques Applied to an Indoor
[7] F. Karray, W. Gueaieb and S. Al-Sharhan, “The Micro Quadrotor”, In: Conf. Rec. 2005 IEEE Int.
hierarchical expert tuning of PID controllers using Conf. Robotics and Automation, Barcelona, Spain,
tools of soft computing” IEEE Trans. Systems Man, pp.2247-2252, 2005.
and Cybernetics-Part B: Cybernetics, Vol.32, No.1,
pp.77–90, 2002. [20] R. Xu and U. Ozguner, “Sliding mode control of a
Quadrotor helicopter”, In: Conf. Rec. 2006 IEEE
[8] Z. Zhao, M. Tomizuka and S. Isaka, “Fuzzy gain Int. Conf. Decision & Control, San Diego, USA, pp.
scheduling of PID controllers”, IEEE Trans. on 4957-4962, 2006.
Systems Man and Cybernetics, Vol.23, No.5, pp.
1392–1398, 1993. [21] M. Önder Efe, “Robust low altitude behavior
control of a Quadrotor rotorcraft through sliding
[9] K. Yu and J. Hsu, “Fuzzy gain scheduling PID modes”, In: Conf. Rec. 2007 Mediterranean.
control design based on particle swarm Control and Automation, Athens, Greece, pp.23-35,
optimization method” In: Second International 2007.
Conference on Innovative Computing, Information
and Control. Kumamoto, 2007. [22] P. Castillo, R. Lozano and A. E. Dzul, “Modeling
and Control of Mini-Flying Machines”, Springer-
[10] Zulfatman and M. F. Rahmat, “Application of self- Verlag London Limited 2005.
tuning fuzzy PID controller on industrial hydraulic
actuator using system identification approach”, Int. [23] A. Boubakir, S. Labiod and T. M. Guerra, “Com-
J. on Smart Sensing and Intelligent Systems, Vol.2, mande PID Adaptative des Systèmes non Linéaires
pp.246–261, 2009. Affines en la Commande”, CGE’06, EMP, ALGER,
April, 2009.
[11] Y. Guo and T. Yang, “A new type of computational
verb gain-scheduling PID controller”, In: Inter-
national Conference on Counterfeiting Security
and Identification in Communication, Chengdu, pp.
235–238, 2010.
[12] L. Yao and C. Lin, “Design of gain scheduled fuzzy
PID controller”, World Academy of Science, Engi-
neering and Technology, Vol.1, pp.152–156, 2005.
[13] I. M. M. El Emary, W. Emar and M. J. Aqel, “The
adaptive fuzzy designed PID controller using
wavelet network”, Journal of Computer Science
and Information System, Vol.6, No.2, pp.141–163,
2009.
[14] J. Jih-Gau and T. Kai-Ti, “Design and realization of
a hybrid intelligent controller for a twin rotor mimo
system”, Journal of Marine Science and
Technology, Vol.21, No.3, pp.333–341, 2013.
[15] B. Hu, G. K. I. Mann and R. G. Gosine, “A sys-
tematic study of fuzzy PID controllers-function
based evaluation approach”, IEEE Trans. on Fuzzy
Systems, Vol.9, No.5, pp.699–712, 2001.
[16] J. T. Spooner and K. M. Passino, “Stable adaptive
control using fuzzy systems and neural networks”,
IEEE Trans. on Fuzzy Syst., Vol.4, No.3, pp.
339-359, 1996.
[17] W. D. Chang, R. C. Hwang, and J. G. Hsieh, “A
self-tuning PID control for a class of nonlinear
systems based on the Lyapunov approach”, Journal
of Process Contr., Vol.12, No.2, pp.233-242, 2002.
[18] M. Feki, “An adaptive feedback control of lineariz-
able chaotic systems,” Chaos, Solitons and Frac-
tals, Vol.15, No.5, pp.883-890, 2003.

International Journal of Intelligent Engineering and Systems, Vol.8, No.2, 2015 6