International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868

Foundation of Computer Science FCS, New York, USA

Volume 9 – No.5, August 2015 – www.ijais.org

Dynamic Modeling and Control of a HexaRotor using

Linear and Nonlinear Methods
Mostafa Moussid Adil Sayouti, PhD Hicham Medromi, PhD
Doctoral Student in Computer Professor Professor and Director
Engineering School Royal Naval of Morocco National Higher School of
National Higher School of (ERN) Electricity and Mechanics
Electricity and Mechanics

ABSTRACT S(  ) R3 : Skew symmetric matrix

A hexacopter aircraft is a class of helicopter, more specifically J r R3 : Inertia for each rotor
of multirotors. The hexacopter has several characteristics J xx/yy/zz R : The moments of inertia
(mechanically simple, vertical takeoff and landing, hovering l R : Distance to the center of gravity
capacities, agile) that give it several operational advantages over b R : Thrust constant [Ns2]
other types of aircraft. But its beneficts come at a cost: the
d R : The drag factor [Nms].
hexacopter has a highly nonlinear dynamics, coupled and
(kp, ki ,kd) R3 : Proportional, integral and derivative gain
underactuated which makes it impossible to operate without a
feedback controller action.
In this work we present a detailed mathematical model for a 1. INTRODUCTION
Vertical Takeoff and Landing (VTOL) type Unmanned Aerial This work will focus on the modeling and control of a hexarotor
Vehicle (UAV) known as the hexarotor. The nonlinear dynamic type UAV. The reason for choosing the hexarotor is in addition
model of the hexarotor is formulated using the Newton-Euler to its advantages (their increased load capacity and high
method, the formulated model is detailed including aerodynamic maneuverability.etc.), the research field is still facing some
effects and rotor dynamics that are omitted in many literature. challenges in the control field because the hexarotor is a highly
Three controls schemes, namely Proportional-Derivative- nonlinear, multivariable system and since it has six Degrees of
Integral (PID) controller, backstepping and sliding mode (SMC), Freedom but only four actuators, it is an under actuated system
have been applied to control the altitude, attitude, heading and [1].
position of the hexacopter in space. Simulation based Under-actuated systems are those having a less number of
experiments were conducted to evaluate and compare the control inputs compared to the system's degrees of freedom.
performance of three developed control techniques in terms of They are very difficult to control due to the nonlinear coupling
dynamic performance, stability and the effect of possible between the actuators and the degrees of freedom [2]. Although
disturbances. the most common flight control algorithms found in literature
This article focuses on modeling strategy and command of a are linear flight controllers, these controllers can only perform
kind hexarotor type unmanned aerial vehicle (UAV). These when the hexarotor is flying around hover, they suffer from huge
developments are part of the overall project initiated by the team performance degradation whenever the hexarotor leaves the
(EAS) of the Computer Laboratory, systems and renewable nominal conditions or performs aggressive maneuvers [3].
energy (LISER) of the National School of Electrical and The contributions of this work are: deriving an accurate and
Mechanical (ENSEM). detailed mathematical model of the hexarotor UAV, developing
linear and nonlinear control algorithms and applying those on
Keywords the derived mathematical model in computer based simulations.
Hexacopter; Vertical Takeo-and Landing; Nonlinear control; The work will be concluded with a comparison between the
Newton-Euler method; PID; backstepping; sliding-mode; developed control algorithms in terms of their dynamic
performance and their ability to stabilize the system under the
Nomenclature effect of possible disturbances.
Symbol Meaning
The paper remainder is organized as follows. In the next Section
RI : The earth inertial frame (RI -frame)
the mathematical formulation and the dynamic model of the
RB : The body-fixed frame (RB -frame)
hexacopter are described, while the applications of three
m R : Total mass of the hexa-rotor
different control techniques PID, Backstepping and Sliding-
g R : Gravity constant mode to hexarotor are presented in section III. In section IV, the
(Φ, θ, ψ) R3 : Euler angle of rotorcraft simulation results are given to highlight the proposed method,
ξ R3 : Position of the center of mass in the inertial while conclusion is drawn in the last section V.
frame
η R3 : Position of the orientation in the body frame 2. DYNAMIC MODEL OF THE
υ R3 : The translational velocity HEXAROTOR
ω R : The angular velocity The mathematical model of the hexacopter has to describe its
Kfa R3 : Propeller drag coefficient attitude according to the well-known geometry of this UAV.
Ωr R : Overall propeller speed (rad/s) More specifically, this aerial vehicle basically consists of six
J R3 : The diagonal inertia matrix propellers located orthogonally along the body frame. Figure 1

9
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

shows this configuration. There are three movements that Table 1. The main effects on the acting hexarotor
describe all possible combinations of attitude: Roll (rotation
around the X axis) is obtained when the balance of rotors 1, 2 Effects Fountainhead
and 3 (or 6, 5 and 4) is changed (speed increases or decreases).
By changing the angle, lateral acceleration is obtained; pitch Aerodynamics effects Rotating propellers
movement (rotation around the Y axis) is obtained when the
balance of the speed of the rotors 1 and 6 (or 3 and 4) is Inertiel counter torque Speed change of propellers
changed. The angle change results in a longitudinal acceleration;
Effect of gravity Position of the center of mass
yaw (rotation about the Z axis) is obtained by a simultaneous
change of speed of the motors (1, 3, 5) or (2, 4, 6). Gyroscopic effects Change in the direction of the drone
2.1 Hexacopter Kinematics L’effet de frottement Tous les mouvements du drone
This subsection describes the dynamical models of the Six
Rotor. The schematic structure of the hexacopter and the
rotational directions of the propellers are illustrated in Figure 1. In order to get equations of motion of entire system, the
In order to describe the hexacopter motion only two reference following assumptions have been made:
systems are necessary: earth inertial frame ( -frame) and body-  The hexacopter is a rigid body;
fixed frame (RB -frame).
 The hexacopter has a symmetrical structure;
 Tensions in all directions are proportional to the square
of the propeller speed.
To derive the dynamic model of the hexacopter (position and
attitude); the Newton-Euler formalism is used [4]. Therefore the
following equations are obtained:
RB
mI 33 0 33   V     mV   F 

RI  0 33 J       J 
    (1)

2.2.1 Forces
 Gravity force: Fg = [0 0 -mg]T.
Fig 1: The structure of hexarotor and its frames
6 6
Thrust force: Fp = RI [0 0  Fi]T = RI [0 0  b
B B

The orientation of the hexacopter is given by the three Euler i 1 i 1
angles, namely yaw angle ψ, pitch angle θ and roll angle ϕ that
 i ]T
2

together form the vector η = [ ϕ, θ, ψ]T. (ϕ and θ ]- , [; ψ

]-π ,π [ )  Rotor drag : Ft=kft  = I 33 [ kftx kfty kftz ]T x 
The position of the vehicle in the inertial frame is given by the
vector ξ = [x, y, z ]T. The vector of the drag forces, kft =diag(kftx, kfty, kftz).
So, the transformation related to the position and the angular 2
velocity from the body to the inertial frame is obtained  Air resistance : Ti= CtρAr2 = d i
respectively by:
Where A is a blade area, ρ the density of air, r the radius of the
blade and Ωi the angular velocity of a propeller.
 cos  cos cos sin  sin   sin cos  cos sin  cos   sin sin  
  2.2.2 Torques
 I   sin cos sin  sin sin   cos cos  cos  sin sin   cos sin  
B

  sin  sin  cos  cos  cos  

 
And
1 0  sin  
0 cos  sin  cos   where ω= Rr
Rr= 
0  sin  cos  cos  

2. 2 Applied forces and torques

The two main forces come from gravity and the thrust of the
rotors but to make the model more realistic rotor drag and air
friction is also included. The UAV rotorcraft system are quite Fig 2: Hexacopter Rotor Distances to Center of Gravity
complex. Their movements are governed by several effects • Actuator action :
either mechanical or aerodynamic. The main effects on the  Roll torque:Mx=
acting hexarotor have been listed in the following table:
2 2 1 2 2 2 2
bl(  2   5  ( 1   3   4   6 ))
2

10
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

12 
3
 Pitch torque: My= bl
2
( 1 
2
3 
2
4 
2
6 )   
2  u1   b b b b b b   2 
     2  (4)
    bl  bl bl bl  2 
 bl bl   3 
u 2  2
2 2 2 2 2 2

     bl 3
 Yaw torque: Mz= d ( 1   2  3   4  5   6 ) 2 2 2
 
bl 3 bl 3  bl 3   2 
u 3   0 0
 
4
The vector Mf is defined as: Mf = [Mx My Mz ]T    2 2 2 2
u 4   2 

• Torque aerodynamique resistance :  d d d d d d  5 
  2 
Ma =Kfa  = [Kfax  Kfay  Kfaz ] T
2 2 2 2
 6 
• gyroscopic effect from Propeller:
6
If the rotor velocities are needed to be calculated from the
Mgh =  Ωr  J r [0 0 (-1)i+1 i ]T control inputs, an inverse relationship between the control inputs
i 1 and the rotors' velocities is needed, which can be acquired by

Yaw counter torque : Mgh =[ 0 , 0 , Jr r ]T inverting the matrix in (4) to give,
With  r =  1   2   3   4  5   6 = (l u1 +2 u2 - u4)
2.2.3 Hexacopter mathematical model = (l u1 + u2 - u3 + u4)
The equations of motion, that governs the translational and the
rotational motion for the hexarotor with respect to the body = (l u1 - u2 - u3 - u4)
frame are (5)
= (l u1 -2 u2 + u4)
2.2.3.1 Translational dynamic
= (l u1 - u2 + u3 - u4)
m  = ΣF = Fp + Fg + Ft
= (l u1 + u2 + u3 + u4)
6
x = 1/m ( coscossin  sinsin ) (  Fi ) - kftx x /m
i 1

6 2.2.3.3 Total system model

y = 1/m ( cos sin sin  sincos )(  Fi )- kfty y
 /m (2) Finally, this derivation provides the 2nd order differential
i 1
equations for the aircraft’s position and orientation in space.
6 Applying relation (1) to (5) and rewriting the matrix equation in
z = 1/m ( cos cos )(  Fi )- kftz z /m - g
i 1
from of system, we obtain the following:
1
 = [   ( J yy - J zz ) - Kfax  - JrΩr  + u2].
2
2.2.3.2 Rotational dynamics
J  =-ω J ω + Mf - Ma- Mgh
J xx

1
 = [   ( J zz - J xx ) - Kfay  + JrΩr  + u3].
2

J xx  =   ( J yy - J zz )- Kfax  - JrΩr  +

2
J yy

1 1
[   ( J xx - J zz )- Kfaz  + u4]
2 2 2 2 2 2 2
+ bl(  2  5  ( 1  3   4   6 ))  = (6)
2 J zz
J yy  =   ( J zz - J xx )-Kfay  + JrΩr  +
2
(3) kftx 1
x = - x + uxu1
3 2 2 2 2 m m
+ bl ( 1  3   4   6 ) kfty 1
2 y = - y + uyu1
J zz  =   ( J xx - J yy )- Kfaz  +
2 m m
2 2 2 2 2 2
kftz cos  cos 
+ d ( 1   2   3   4   5   6 ) z = - z - g + u1
m m
The hexacopter’s total thrust force and torque control inputs u 1,
u2, u3, u4 are related to the six motor’s speed by the following with : ux = coscossin  sinsin .
equations: UT=[u1,u2,u3,u4] is the vector of (artificial) input uy = cos sin sin  sincos .
variables[5]:
The dynamic model presented in equation set (6) can be
rewritten in the state-space form x  = f(X,U). X R12 is the
vector of state variables given as follows:

11
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

[ϕ θ ψ x y z ] constituted of two subsystems, the angular rotations and the

X T =[ x linear translations, (Fig. 3)
1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11
x12]
x1= ϕ x7= x
x2= 1= x8= 7=
x3= θ x9= y
x4= 3= x10= 9=
x5= ψ x11= z
x6= 5= x12= 11=

2= = a1x4x6 + a2 + a3Ωrx4 + b1u2.

4= = a4x2x6 + a5 + a6Ωrx2 + b2u3.
6= = a7x2x4 + a8 + b3u4.
(7)
8= = a9x8 + uxu1.
10 = = a10x10 + uyu1. Fig 3: Connection of the two ideal subsystems of the overall
dynamical system
12 = = a11x12 + u1 – g. 2.2.4 Rotor Dynamics
The rotors are driven by DC-motors with the well known
To simplify, define, equations:
a9=- Kftx /m
a1=( J yy - J zz )/ J xx a2=- Kfax / J xx L = u – Ri - kcωm and J = τm – τd.
a10=- Kfty/m As we use a small motor with a very low inductance, the second
a4=( J zz - J xx )/ J yy a5=- Kfay/ J yy
order DC-motor dynamics may be approximated [5]:
a8=- Kfaz / J zz a11=- Kftz /m
a7=( J xx - J yy )/ J zz
J =- ωm - τd + u
a3=- Jr/Jxx a6=- Jr/Jyy By introducing the propeller and the gearbox models, the
b1= l/Jxx b2= l/Jyy b3= l/Jzz equation may be rewritten:
Rewriting the last equation (7) to have the angular accelerations m = - ωm - + u with =
in terms of the other variables (Rotational equation of motion), τ τ τ
The equation can be linearized around an operation point to
2 a1x4x6 + a2 + a3Ωrx4 + b1u2 the form m = -Aωm + B u + C.

4 = a4x2x6 + a5 + a6Ωrx2 + b2u3 In this next section, we present the application of two different
6 a7x2x4 + a8 + b3u4 control techniques Backstepping and Sliding-mode to hexarotor.

3. CONTROL OF HEXAROTOR
With the choice of the control input vector U, it is clear that the In this section, a control strategy is based on two loops (inner
rotational subsystem is fully-actuated, it is only dependant on the loop and outer loop). The inner loop contains four control laws:
rotational state variables x1 to x6 that correspond to ϕ, , θ, , roll command (ϕ), pitch command (θ), yaw control (ψ) and
ψ, respectively. controlling altitude Z. The outer loop includes two control laws
positions (x, y). The outer control loop generates a desired for
Rewriting the last equation (7) to have the accelerations in terms roll movement (θd) and pitch (ϕd) through the correction block.
of the other variables, we get translational equation of motion, This block corrects the rotation of roll and pitch depending on
the desired yaw (ψd). The figure below shows the control
strategy we will adopt Fig.4:
8 a9x8 + (cosx1cos x5 sin x3 + sinx1sin x5) u1
m

10 = a10x10 + (cosx1sin x3 sin x5 – sinx1cos x5) u1

m

12 a11x12 - g + (cosx1cos x3) u1

m

It is clear here that the translational subsystem is under_ actuated

as it dependant on both the translational state variables and the
rotational ones.
It is worthwhile to note in the latter system that the angles and
their time derivatives do not depend on translation components.
On the other hand, the translations depend on the angles. We can
ideally imagine the overall system described by (6) as Fig. 4. Synoptic scheme of the proposed control strategy

12
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

The hexarotor parameters used in the simulations are, ψd and d : Desired Yaw angle and rate of change
Table. 2 Parameters Used in the Simulations In order to design the PID controllers, nonlinear rotational
Parameter Description Value dynamics of hexarotor are linearized around zero, which are
Jxx MOI about body frame's x-axis 7.5e-3 kg.m2 given by,
Jyy MOI about body frame's y-axis 7.5e-3 kg.m2
Jzz MOI about body frame's z-axis 1.3e-2 kg.m2 ϕ(s) = 2(s); θ(s) = U3(s); ψ(s) = 4(s)
l Moment arm 0.23 m
Jr Rotor inertia 6e-5 kg.m2
m Quadrotor mass 0.650 kg
3.1.3 Position controller
Kf Aerodynamic force constant 3.13e-5 N s2 PID controller is defined for controlling the ex= xd – x and
KM Aerodynamic moment constant 7.5e-7 Nm s2 ey= yd – y positions. The control objective is to drive both
Rmot Motor circuit resistance 0.6 Ω values to zero (ex, ex)=(0,0). In this sense, the control laws are:
Kmot Motor torque constant 5.2 mNm/A
Kt Aerodynamic translation coeff diag(0.1,0.1,0.15) d= kp,x(xd - x) + kd,x( d- )+ ki,x ʃ(xd - x) dt
Kt Aerodynamic rotation coeff diag(0.1,0.1,0.15)
d = kp,y(yd - y) + kd,y( d- )+ ki,y ʃ(yd - y) dt
3.1 PID Controller for hexarotor
The classical PID linear controller has the advantage that
parameter gains are easy to adjust, is simple to design and has
good robustness. However some of the major challenges with
the hexarotor include the non-linearity associated with the
mathematical model and the imprecise nature of the model due
to unmodeled or inaccurate mathematical modeling of some of
the dynamics. Therefore applying PID controller to the hexarotor
limits its performance.
The purpose of the PID controller is to force the Euler angles to
follow desired trajectories. The objective in PID controller
design is to adjust the gains to arrive at an acceptable degree of
tracking performance in Euler angles.
After the mathematical model of the hexarotor along with its
open loop simulation is verified, a PID controller was developed.
The PID controller generates the desired control inputs for the
hexarotor. The block diagram for a PID controller is shown in
Figure 5.

Fig. 5 PID Controller Block Diagram

3.1.1 Altitude controller
U1 = kp,z (z –zd ) + kd,z ( – d )+ ki,z ʃ(z –zd ) dt.
Where
zd and d: Desired altitude and altitude rate of change.
Fig. 6. The PID Control Inputs and Simulation Response
3.1.2 Attitude controller
The control objective is to maintain the hexarotor in a constant 3.2 Backstepping Controller for Hexarotor
altitude (z). The PID controller for the θand ψ dynamics can Backstepping design refers to “step back” to the control input,
be given as and a major advantage of backstepping design is its flexibility to
avoid cancellation of useful nonlinearities and pursue the
U2 = kp,ϕ(ϕd –ϕ) + kd,ϕ( d – )+ ki,ϕ ʃ(ϕd –ϕ) dt (Roll angle) objectives of stabilization and tracking, rather than those of
U3 = kp,θ (θd –θ) + kd,θ ( – )+ ki,θ ʃ(θd – θ) dt (Pitch angle) linearization. Recursively constructed backstepping controller
d
employs the control Lyapunov function (CLF) to guarantee the
U4= kp,ψ(ψd –ψ) + kd,ψ ( d – )+ ki,ψ ʃ(ψd –ψ) dt (Yaw angle) global stability [10], [11].
Where In this section, a Backstepping controller is used to control the
2 attitude, heading and altitude of the hexarotor. The Backstepping
ϕd and d : Desired roll angle and rate of change s controller is based on the state space model derived in (7). Using
θd and : Desired Pitch angle and rate of change the backstepping approach, one can synthesize the control law
d
forcing the system to follow the desired trajectory. Refer to [6]

13
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

and [7] for more details. 3.2.2 Backstepping Control of the Linear
Considering the following system, Translations
The altitude control U1 and the Linear (ux,uy) Motion Control
xid - xi i [3,5,7,9,11] are obtained using the same approach described in 3.2.
ei =
x(i-1)d + k(i-1)e(i-1) - xi i [4,6,8,10,12] U1= [g-a11x12+ d + k11(-k11e11+e12)+ k12 e12+ e11]
with ki ˃ 0 i [2,…..,12] ux= ( ) [-a9x8+ d + k7(-k7e7+e8)+ k8 e8+ e7]
3.2.1 Backstepping Control of the Rotations uy= ( ) [-a10x10+ d + k9(-k9e9+e10)+ k10 e10+ e9].
Subsystem The sliding mode control inputs which were derived and
Using the backstepping approach, one can synthesize the control expressed in equation (8) were applied to the nonlinear model in
law forcing the system to follow the desired trajectory. Refer to (7) and responses are shown in fig.(7).
[7] and [8] for more details. For the first step we consider the
tracking-error ei = x1d - x1 and we use the Lyapunov theorem
by considering the Lyapunov function Vi positive definite and
it’s time derivative negative semi-definite:

i [3,5,7,9,11]
Vi=
Vi-1 + i [4,6,8,10,12]

For the first step we consider the tracking-error:

e1 = x1d – x1 = ϕd - x1
V1 = and 1 = e1 1= e1( d – x2)

The stabilization of e1 can be obtained by introducing a virtual

control input x2:
x2 = d+ k1 e1 1 = - k1 ≤ 0.

For the second step we consider the augmented Lyapunov

function:
e2 = d + k1 e1 – x2
V2 = +

And it’s time derivative is then: 2= e1 1+ e2 2

2= = e1(-k1e1+e2) + e2( d + k1 1- a1x4x6 – a2 – a3Ωrx4–b1

u 2)
The control input U2 is then extracted, satisfying:

2= -k1 – k2 ≤ 0.

U2 = [-a1x4x6 – a2 – a3Ωrx4+ d+ k1(-k1e1+e2)+ k2 e2+ e1]

Following exactly the same steps as the roll controller, the

control input U3 responsible of generating the pitch rotation and
U4 responsible of generating the yaw rotation are calculated to
be,
Fig. 7. The backstepping control inputs and simulation
Roll controller
response
U2 = [-a1x4x6 – a2 – a3Ωrx4+ d+ k1(-k1e1+e2)+ k 2 e 2+
e1 ] 3.3 Sliding-Mode Control of Hexarotor
Pitch controller Sliding mode control is a well-established nonlinear control
technique that displays certain degree of robustness against
U3 = [-a4x2x6 – a5 – a6Ωrx2+ d+ k3(-k3e3+e4)+ k4 e4+
uncertainties and disturbances. Its main advantage is that it does
e3 ] (8) not simplify the dynamics through linearization and has good
Yaw controller tracking. Although it is vulnerable to noise and it suffers from
chattering phenomenon, several approaches have been proposed
U4 = [-a7x2x4 – a8 + d+ k5(-k5e5+e6)+ k6 e6+ e5] to overcome these difficulties without giving concessions from
the robustness property of the scheme. The behavior is
composed of two phases, first the error dynamics is guided
toward a predefined subspace of the state space, which we call

14
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

the sliding manifold, and the behavior during this phase is called
the reaching phase. This phase is known for the high magnitude
control efforts that may undergo saturation sometimes. Second is
the sliding phase, during which the error vector obeys the
dynamical implications of the analytic description of the locus
itself, which is stable by design and the error terminates at the
origin inevitably. The control scheme takes its name from the
latter dynamic behavior and is called sliding mode control.
The hexarotor system was subdivided into the full-actuated and
under-actuated systems. The under-actuated system, to which
SMC was applied, was further subdivided into under-actuated
subsystems. Results showed good stability and robustness of the
system. Chattering effect of SMC was observed but minimized
Chattering
with a continuous approximation of a pre-determined “sign”
function
The basic sliding mode controller design procedure is performed
in two steps. Firstly, choice of sliding surface (S) is made
according to the tracking error, while the second step consist the
design of Lyapunov function which can satisfy the necessary Fig. 9 The sliding Control Simulation Response
sliding condition (S <0) [9][10]. The application of sliding
mode control to hexarotor dynamic is presented here by
obtaining the expression for control input. The sliding surface
are define,

Sϕ= e2 + λϕ e1 = 1d – x2 + λϕ( x1d – x1)

Sθ = e4 + λθ e3 = 3d – x4 + λθ( x3d – x3)
Sψ = e6 + λψ e5= 5d – x6 + λψ( x5d – x5)
Sx = e8 + λx e7= 7d – x8 + λx( x7d – x7)
Sy = e10 + λy e9= 9d – x10 + λy( x9d – x9)
Sz = e12 + λz e11= 11d – x12 + λz( x11d – x11)
Such that
ei = xid - xi
ei+1 = i i [1,….,11]
λi ˃ 0
Assuming here that V(Sϕ) = then, the necessary sliding Fig. 10. The sliding Control Inputs
condition is verified and lyapunov stability is guaranteed. The
chosen law for the attractive surface is the time derivative of
4. RESULTS AND DISCUTION
To be able to compare fairly between the three implemented
satisfying ( S < 0)
control techniques, the response graph of the system under the
effect of each the three controllers was plotted superimposed on
ϕ = - kϕsign(Sϕ) one another. Figure. 11 show the altitude response, the attitude
= 2 + λϕ 1 = - 2 + λϕ( 1d – x2 ) and heading responses respectively.
= - a1 x4 x6 – a2 x4 Ωr – b1 U2 + d + λϕ( d – x2 ).
U2 = [- a1 x4 x6 – a2 x4 Ωr + d+ λϕ( d – x2 ) – kϕsign(Sϕ)].
The same steeps are followed to extract U3, U4 and U1:
U2 = [- a1 x4 x6 – a2 x4 Ωr + d+ λϕ e2 – kϕsign(Sϕ)] (Roll)
(9)
U3 = [- a3x2 x6 – a4x2 Ωr + d+ λθ e4 – kθsign(Sθ)] (Pitch)

U4 = [- a5x2 x4– a4x2 Ωr + d+ λψ e6 – kψsign(Sψ)] (Yaw)

U1= [ g+ d+ λz e8– kzsign(Sz)] (Altitude)

ux= ( )[ d+ λx e10– kx sign(Sx)] (Linear x Motion)

uy= ( )[ d+ λy e12– ky sign(Sy)] ( Linear y Motion) Fig. 11 the altitude, the attitude and heading responses

The sliding mode control inputs which were derived and 4.1 Linear Operation
expressed in equation (9) were applied to the nonlinear model in In this paper three linear and nonlinear control schemes are used
(7) and responses are shown in fig. 9 and fig.10. to stabilize the attitude of hexarotor UAV. The methods
considered are PID, SMC, backstepping control. The employed
controllers developed to control the hexarotor model under

15
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

consideration gave comparable dynamic performances in terms Optimal LN A LN

of settling time and overshoot when they were deployed in near
hover stabilization of the hexarotor. Intelligent LN LN LN

The PID controller has been successfully applied to the Tracking ability A H H
hexarotor though with some limitations. The tuning of the PID Fast convergence A H LN
controller could pose some challenges as this must be conducted
around the equilibrium point, which is the hover point, to give Precision A H A
good performance. Simplicity H A LN
The SMC resulted in good stability and robustness of the Disturbance rejection LN H H
system; but an undesirable Chattering effect of SMC was
observed which was very notable in the attitude response unlike Noise (signal) H LN LN
the altitude. It minimized with a continuous approximation of a
Chattering LN H LN
predetermined “sign” function. The presence of the ‘‘sign’’ term
in the SMC's control law makes it a discontinuous controller. Legend: LN—low to none; A—average; H—high.
Shows that whenever the value of the surface s is positive, the
control law works to decrease the trajectory to reach the sliding 5. CONCLUSIONS AND FUTURE WORKS
surface (s = 0). Ideally it should continue sliding on the surface The goal of this work was to derive a mathematical model for
once hitting it, but due to the delay between the change of sign the hexarotor Unmanned Aerial Vehicle (UAV) and develop
and the change in the control action, the trajectory passes the nonlinear control algorithms to stabilize the states of the
surface to the other side. The main drawbacks of chattering are hexarotor, which include its altitude, attitude, heading and
that it causes the excitation of unmodeled system dynamics that position in space and to verify the performance of these
yields a possible instability of the system. In addition to that it is controllers with comparisons via computer simulations. The
associated with a high power consumption and possible actuator mathematical model of a hexarotor UAV was developed in
damage. These drawbacks make the SMC hard to be details including its aerodynamic effects and rotor dynamics
implemented on real systems. Backstepping control is a which we found lacking in many literatures; three control
recursive algorithm that breaks down the controller into steps techniques were then developed and synthesized; a linear
and progressively stabilizes each subsystem. Its advantage is that Proportional-Integral-Derivative (PID) controller, a nonlinear
the algorithm converges fast leading to less computational Sliding Mode Controller (SMC) and a nonlinear Backstepping
resources and it can handle disturbances well. The main controller. A complete simulation was then implemented on
limitation with the algorithm is its robustness is not good. To MATLAB/Simulink relying on the derived mathematical model
increase robustness (to external disturbances) of the general of the hexarotor. The simulation environment was used to
backstepping algorithm, an integrator is added and the algorithm evaluate the mentioned controllers and compare their dynamic
becomes Integrator backstepping control. The integral approach performances under different types of input conditions. The
was shown to eliminate the steady-state errors of the system, SMC and Backstepping controllers gave better performance
reduce response time and restrain overshoot of the control outside the linear hovering region due to their nonlinear nature.
parameters. Figure 11 show a quantitative comparison between The PID and Backstepping controllers gave better performance
the performance of the PID, SMC and Backstepping controllers than all the other controllers when the effect of wind was added
in terms of the settling time and overshoot of the system's to the system. The wind effect was modeled as extra forces and
response respectively. moments on the hexarotor body.

4.2 Nonlinear Operation As it can be seen from the experimental plots, the controller
When the controllers were used outside of the linear region introduced using the sliding-mode approach provides average
(away from hover), the PID controller failed to stabilize the results. This is partly due to switching nature of the controller
system due to the fact that PID comes out of a family of linear which introduces high frequency, low amplitude vibrations
controllers. On the other hand, the SMC and the Backstepping causing the sensor to drift. On the other hand, the backstepping
controller were able to stabilize the system with a good dynamic controller proves the ability to control the orientation angles in
performance. the presence of relatively high perturbations confirming by the
way some previous studies on under actuated systems. The SMC
In conclusion, PID controller is simple in structure; its controller has the problem of chattering.
performance may not be adequate under conditions of severe
disturbances. SMC is robust against disturbances yet it suffers An important part of this work was dedicated to finding a good
from the chattering phenomenon. Backstepping yield good control approach for hexarotors. Three techniques were explored
tracking results yet they need the availability of the system from theoretical development to final experiments.
nonlinearities, which is generally accepted as a drawback. In As evident from the review, no single algorithm presents the best
table.3 summarizes the comparison of the various algorithms as of the required features. It also been discussed that getting the
applied to hexarotors. The performance of a particular algorithm best performance usually requires hybrid control schemes that
depends on many factors that may not even be modeled. Hence, have the best combination of robustness, adaptability,
this table serves as guide in accordance with what is presented in optimality, simplicity, tracking ability, fast response and
this work. disturbance rejection among other factors.
Table.3 Comparison of control algorithms (PID, SMC, BS) The future work is to develop a fully autonomous vehicle with a
more powerful hybrid flight controller. For example, adding an
Characteristic PID Sliding mode Backstepping
integral action to the developed Backstepping controller
Robust A A LN combined with Frenet-Serret Theory will lead to the formulation
of an adaptive control algorithm robust to system uncertainties.
Adaptive LN H H Last but not least, implementing the developed control

16
 International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868
Foundation of Computer Science FCS, New York, USA
Volume 9 – No.5, August 2015 – www.ijais.org

techniques on real hexarotor hardware to give a more fair [5] Jun Shen, Qiang Wu, Xuwen Li, Yanhua Zhang, Research
comparison between their performances. The positive results of the RealTime Performance of Operating System, 5th
achieved through this development enhance the knowledge of International conference on Wireless Communications, pp.
this very unstable system and encourages us to continue towards 1-4, 2009.
full autonomy hexarotor.
[6] R. M. Murray and S. S. Sastry, A mathematical
6. REFERENCES introduction to robotic manip-ulation. CRC press, 1994.
[1] Hongning Hou, Jian Zhuang, Hu Xia, Guanwei Wang, and [7] A.W.A. Saif, M. Dhaifullah, M.A.Malki and M.E. Shafie,
Dehong Yu. A simple controller of minisize quad-rotor “Modiffied Backstepping Control of Quadrotor”,
vehicle. In Mechatronics and Automation (ICMA), 2010 International Multi-Conference on System, Signal and
International Conference on, pages 1701doi: Devices, 2012.
10.1109/ICMA.2010.5588802.
[8] A.A. Main, W. Daobo, “Modeling and Backstepping-based
[2] Jinhyun Kim, Min-Sung Kang, and Sangdeok Park. Nonlinear Control Strategy for a 6 DOF Quadrotor
Accurate modeling and robust hovering control for a Helicopter”, Chinese Journal of Aeronautics 21, pp. 261-
quadrotor vtol aircraft. Journal of Intelligent and Robotic 268, 2008.
Systems, 57(1-4):9-26, 2010.
[9] V.G. Adir, A.M.Stoica and J.F. Whidborne,“Sliding Mode
[3] Farid Kendoul. Survey of advances in guidance, navigation, Control of 4Y Octorotor”, U.P.B. Sci. Bull, Series D,
and control of unmanned rotorcraft systems. Journal of Vol.74, Iss. 4, pp. 37-51,2012.
Field Robotics, 29(2):315{378, 2012.
[10] K. Runcharoon and V. Srichatrapimuk, “Sliding Mode
[4] Simon János, Goran Martinovic, Navigation of Mobile Control of Quadrotor” International Conference of
Robots Using WSN‘s RSSI Parameter and Potential Field Technological Advances in Electrical, Electronics and
Method, Acta Polytechnica Hungarica, Journal of Applied Computer Engineering, pp. 552-556 May 9-11, 2013.
Sciences Vol.10, No.4, pp. 107-118, 2013.

17