Professional Documents
Culture Documents
ABSTRACT: This paper investigates the problem of controlling a team of quadrotors that cooperatively Review History:
transport a common payload. The main contribution of this study is to propose a cooperative control Received: 28 March 2018
algorithm based on a decentralized algorithm. This strategy is comprised of two main steps: the first Revised: 19 October 2018
one is calculating the basic control vectors for each quadrotor using Moore–Penrose theory aiming at Accepted: 22 October 2018
cooperative transport of an object and the second one is combining these vectors with individual control Available Online: 25 October 2018
vectors, which are obtained from a closed-loop non-linear robust controller. In this regard, a nonlinear
robust controller is designed based on Second Order Sliding Mode (SOSMC) approach using Extended
Kalman-Bucy Filter (EKBF) to estimate the unmeasured states which is capable of facing external Keywords:
disturbances. The distinctive features of this approach include robustness against model uncertainties quadrotor
along with high flexibility in designing the control parameters to have an optimal solution for the Second Order Sliding Mode Control
nonlinear dynamics of the system. Design of the controller is based on Lyapunov method which can Extended Kalman-Bucy Filter
provide the stability of the end-effecter during the tracking of the desired trajectory. Finally, simulation Cooperative Decentralized Control
results are given to illustrate the effectiveness of the proposed method for the cooperative quadrotors to External Disturbances
transport a common payload in various maneuvers.
147
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
approach, the performance of control system was improved stabilizes the object along three dimensional trajectories
by using a second-order sliding mode control (SOSMC) for based on a SOSMC incorporated with optimal Moore-
second-order uncertain plants. An adaptive SOSMC with a Penrose theory. We review and analyze experimental study
nonlinear sliding surface was introduced in [23] and a linear with teams of quadrotors cooperatively grasping, stabilizing,
switching surface was proposed for similar underactuated and transporting payloads in Section 5. It includes point-to-
systems in [24]. Wind as a disturbance source has also point path planning and trajectory tracking. Finally, Section
been considered in the flight process of the quadrotor to 6 concludes the paper with some future improvement
demonstrate the robustness of the control algorithm [25,26]. suggestions.
So, SOSMC is a suitable method for controlling uncertain 2- Mathematical Modeling and Coordinate Reference
(and possibly disturbed) nonlinear systems, specially UAVs, Frame
where the necessity of robust tracking schemes, as well as 2- 1- Dynamic Model of single quadrotor
fast convergent algorithms is crucial. Fig.1 shows the configuration of our quadrotor system that can
Hence, in this paper, a decentralized robust optimal controller fly in all directions and has no limit on maneuver. Two main
is designed for a group of UAVs in a formation which enables reference frames are considered: model frame, E = {Ex , E y , Ez },
manipulation of a common payload in three dimensions. Our and the body-fixed reference frame, B = {Bx , By , Bz } . In this
nomenclature, ξ = [ x, y, z ] and η [ϕ ,θ , Ψ ] are the position
T T
cooperative control algorithm is comprised of two parts. In the =
first part, basic control vectors, which guarantee the optimal and orientation vectors in reference E . Fi ( i = 1, 2,3, 4 ) is
performance of the coupled system, are determined using the thrust force produced by four rotors. The full quadrotor
Moore-Penrose theory. In the second part, individual control dynamical model with the x, y, z motions as outcome of a
vectors are designed for each quadrotor which can guarantee pitch, roll and yaw rotation is [27]:
robustness of the coupled system against uncertainties
¨ Iy − Iz jr l
= ϕ θψ Ix − Ix θΩ + I U 2 + ωϕ =( ϕ θ ψ + ϕ ψ )
and disturbances. Moreover, a combination of second
order sliding mode controller with an optimal cooperative x
algorithm is employed to control all the quadrotors such ¨ Iz − Ix jr l ¨
that a soft touching occurs between the object being carried = θ ϕψ − ϕΩ + U 3=
+ ωθ ( ϕ θ ψ − ϕ ψ)
Iy Ix y
and the cooperative quadrotors. A nonlinear filter equivalent ¨
Ix − Iy l
to well-known Extended Kalman-Bucy filter (EKBF) in a ψ ϕθ
= + U 4 + ωψ =− + ( ϕ θ ) +ω
continuous-time stochastic system has also been developed in Iz Iz
our work to improve the estimation accuracy and robustness (1)
−
implementation
and Ω + and+ ω x¨ = cosϕ sinθ cosψ + sinϕ sinψ l U + ω
=
of the system. The simple ϕ
design
θψ − θ
under noisy condition
ϕ ( ) 1 x
accurate estimation of state variables mi
make EKBF suitable for robotic ¨ −
manipulators and other ¨ l
industrial applications.= θ ϕψ − ϕΩ + + ωθ y ( cosϕ sinθ sinψ − sinϕ sinψ ) U1 + ω y
3=
m i
Finally, the proposed controller successfully rejects the ¨
− l
external disturbances (e.g.= ψ ϕθ
wind
effect)
for
+ + ωψ
cooperative z =− g + ( cosϕ cosθ ) U1 + ω z
quadrotors. The other advantages
of using approach are
this mi
from simple implementation and flexible parameter learning where
to design criteria customization. The simulation results show
the effectiveness and superior performance of the proposed
b b b b 2
control strategy for transporting a common payload in various
U1 lb 0 Ω1
−lb 0 2
maneuvers by cooperative quadrotors. = U 2 Ω2 (2)
The subsequent parts of this paper are as it follows. First of 0 −lb 0 lb 2
all, developing a model for a single quadrotor and modeling U
Ω
−k k −k k 32
3
U 4
of a team of UAV’s rigidly attached to a payload is discussed Ω4 ,
in Section 2. In Section 3, the Extended Kalman Bucy in which mi is considered as mass of one quadrotor; also k
Filter is described briefly. Section 4 proposes cooperative and b represent the thrust and drag coefficient, respectively.
control algorithms defined with respect to the payload that The input U1 is related to the total thrust; while the inputs
U 2 , U 3 , U 4 are related to the rotations of the quadrotor and
ω x , ω y ,ω z ,ωϕ ,ωθ ,ωΨ are external disturbances, jr denotes
the inertia of the z-axis and Ω = −Ω1 + Ω 2 − Ω3 + Ω 4 is the
overall residual propeller angular speed. The model in Eq. (1)
can be rewritten in a state-space form, X = f ( X,U ) , where
T
ÿ
=
X [ x1 ,…….., x12=
]T ϕ , ϕ , θ , θ ,ψ ,ψ , x, x , y, y , z , z
148
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
dx ( t ) f ( x ( t ) , t ) dt + G ( t ) dw ( t ) , ∀t ∈
=
(7)
=Zk h ( xk , t k ) + v k
where x ( t ) is the state vector and Z k is the measurement
vector. w ( t ) and v k denote the state and measurement noise,
a noise that apply to the system state variables in kalman filter
process. In filtering step, the estimation of error covariance
matrix is determined by:
P=
k ( ) (
Pk´ − Pk´ H T xˆ ´k , t k K xˆ ´k , t k H xˆ ´k , t k Pk´ ) ( ) (8)
Fig.2. The coordinate Cooperative quadrotor system [27] The estimated parameters by kalman filter have indicated by
coordinate frame B . The axes of the coupled system are notation prime. and Kalman filter gain is computed as:
described as body frame axes and each UAV has an individual
( ) ( ) ( )
−1
body frame Qi where=i {1, 2,…, n} and n is the number= of K xˆ ´k , t k H xˆ ´k , t k Pk´ H T xˆ ´k , t k + R k (9)
UAVs in the group. For extracting the dynamic model of
the coupled system we assumed the quadrotors are rigidly where R k represents the covariance matrix of the sensor
connected to the object and also, the center of mass of the noise. In prediction step, we have:
system is intended as the origin of the rigid-body coordinate
dxˆ ´ ( t )
system. Therefore, the motion formulation for each quadrotor
can be written as: dt
= f xˆ ´ ( t ) , t ( ) (10)
¨
mi z Qi =
−mi g + FBi , dP′ ( t )
( ) ( )
(3)
=F xˆ ´ ( t ) , t P′ ( t ) + P′ ( t ) F T xˆ ´ ( t ) , t + Q ( t ) (11)
where FBi is lift force generated by each quadrotor. By dt
applying Newton’s second law to the payload, we have: ( ) (
where F xˆ ´ ( t ) , t and H xˆ ´k , t k are the Jacobean matrixes )
of f ( x ( t ) , t ) and h ( x k , t k ) evaluated at xˆ ´ ( t ) respectively.
Moreover, Q ( t ) represents the power of measurement noise.
¨
mi Z QB = −mL g , (4)
After applying EKBF on system, the estimated is shown as
where mL is the mass of the object. Since, all the forces are
ϕˆ , ϕˆ , θˆ, θˆ,ψˆ ,ψˆ , xˆ , xˆ , yˆ , yˆ , zˆ, zˆ ,
T
149
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
4- Cooperative Control Strategy Therefore, four basic control vectors for n cooperative
In this section, the perfect state cooperative guidance quadrotors obtained according (16) and in the in the next T
controller of UAV presented in [28] will be developed to section the individual control vectors, FBdes , M xBdes , M yBdes , M zBdes ,
qn qn qn qn
achieve a decentralized control strategy with superior results will be established. Finally, by combination of two set of
regarding state estimation. Therefore, the problem considered mentioned vectors, the cooperative control vectors will be
in this part is separated in two steps: the first step focuses defined as it follows:
on designing the basic control vectors for n quadrotors T
U coop Qn = U Fqn , U Mzqn , U Mxqn , U Myqn FBdes des
, M xB des
, M yB des
, M zB (17)
using the Moore-Penrose theory based on estimated state qn qn qn qn
variables which are generated by EKBF, and the second step 4- 2- Deriving Individual Control Vectors of Cooperative
describes how to determine individual control vectors by Control Low
SOSMC. Finally, the cooperative control algorithm can be In this section, determining individual control vectors of
extracted by combining the individual control vectors and the cooperative control strategy based on second order sliding
basic control vectors. mode technique is supposed to asymptotically perform
4- 1- Deriving Basic Control Vectors of Cooperative Control position and attitude tracking of the quadrotors in order to
Law generate a robust solution. Now, to complete the cooperative
By using all force vectors of the coupled system in linear algorithm, using tracking errors of state variable(s), a
system (6), four equations for 4 quadrotors can be established switching sliding surface is considered for the whole
as follows: subsystem (rotational or transitional). The sliding manifolds
T
based on SOSMC are chosen as it follows [32, 33]:
FB , M xB , M yB , M zB = Xˆ u (12)
Sϕˆ α1ϕˆ (ϕ d − ϕˆ ) + α 2ϕˆ ϕd − ϕˆ
= ( )
where u ∈ includes 4 control input vectors for each n
4n
quadrotor: = ( )
Sθˆ α1θˆ θ d − θˆ + α 2θˆ θd − θˆ ( )
u FBq ,1 , M x , M y , M z ….., FBq ,n , M x , M yB , M zB
T
(13) Sψˆ α1ψˆ (ψ d −ψˆ ) + α 2ψˆ (ψ d −ψˆ )
=
q ,n (18)
S α ( x − xˆ ) + α x − xˆ ( )
Bq ,1 Bq ,1 Bq ,1 Bq ,n q ,n
= xˆ 1 xˆ d 2x d
Now, consider the following cost function to minimize the
control inputs [28]: S yˆ α1 yˆ
= ( y − yˆ ) + α ( y − yˆ )
d 2 yˆ d
=
J ∑(Λ Fi F + Λ Mxi M
2
q ,i
2
xq ,i + Λ Myi M 2
yq ,i + Λ Mzi M 2
zq ,i ) (14)
S zˆ α1zˆ
= ( z − zˆ ) + α ( z − zˆ )
d 2 zˆ d
i
in which, α1ϕˆ ,α 2ϕˆ ,…, α1zˆ , α 2 zˆ > 0 . The time derivatives of the
in which Λ Fi , Λ Mxi , Λ Myi , Λ Mzi are the weight of each of the sliding manifolds of rotational subsystem are obtained as it
control inputs. By considering the features of mathematical follows;
theory of Moore-Penrose inverse, point-wise minimization of
the function J will be accrued and an optimal control vector ( )
Sϕˆ = α1ϕˆ ϕd − ϕˆ + α 2ϕˆ ϕd − ϕˆ ( )
( ) ( )
can be written as [28 ]:
Sθˆ α1θˆ
= θd − θˆ + α 2θˆ θd − θˆ (19)
des T
Γ Xˆ T ( Xˆ Γ −2 Xˆ T ) −1 FBdes , M xB
u = * −2 des des
, M yB , M zB (15)
(
Sψˆ α1ψˆ ψ d −ψˆ + α 2ψˆ ψd −ψˆ
= ) ( )
where:
By generating S xˆ = −κ sgn ( S xˆ ) − Υ xˆ S xˆ , w =ϕ θ ψ
with
Γ = Γu22 , = −κ
x ( ) − Υx [1, 2,3] , the associated control inputs are
ith xˆ = ϕˆ ,θˆ,ψˆ , and ∈
Γ = dig Λ F 1 , Λ Mx1 , Λ My1 , Λ Mz1 ,..., Λ Fn , Λ Mxn , Λ Myn , Λ Mzn , defined as:
α1ϕˆ ˆ Iy − Iz jr ˆ
by assuming the weight of basic control vectors =U2
Ix
l
α 2ϕˆ
( )
ϕd − ϕˆ + ϕd − θψ ˆ
Ix Ix
+ θ Ω+
1
κ sgn ( Sϕˆ ) + Υ ϕˆ Sϕˆ − ωϕ
α 2ϕˆ 1
Λ Fi =
Λ F , Λ Mxi =
Λ Myi =
Λ Mzi , we have,
= ( )
I y α1θˆ ˆ Iz − Ix jr
θ d − θ + θ d − ϕψ + ϕˆ Ω +
1
( )
κ sgn Sθˆ + Υθˆ Sθˆ − ωθ (20)
( )
−1 U3 ˆ ˆ
Γ Xˆ T Xˆ Γ −2 Xˆ T
−2 U F , U Mz , U Mx , U My .
= l α 2θˆ Iy Ix α 2θˆ 2
qn qn qn qn
I z α1ψˆ
Also, according to this fact that all quadrotors have a common= U4
l
α
(ψ d −ψˆ ) +ψd − ϕθˆ ˆ Ix Iz− Iy + α12ψˆ [κ3 sgn ( Sψˆ ) + Υψˆ Sψˆ ] − ωψ
role in lifting the object off the ground and equal partnership, 2ψˆ
the body force F and yaw torque produced by each of them where κ1 ,κ 2 , κ 3 , Υ ϕˆ , Υ θˆ , Υψˆ > 0 . Now for transitional
is almost equal. Therefore, the basic control vectors which subsystem, the objective is to guarantee the state variables
include total force, yaw, pitch and roll moment generated for [ xˆ, yˆ , zˆ ] converge to the desired values [ xd , yd , zd ] ,
the coupled system structure are obtained as follows: respectively. The time derivatives of the sliding manifolds of
1 rotational subsystem are obtained,
n [1, 0, 0, 0, …..,1, 0, 0, 0]
ÿ
¨ ¨
1 0, 0, 0,1, ……, 0, 0, 0,1 S xˆ α1xˆ x d − xˆ + α 2 xˆ x d − xˆ
=
U Fqn n [ ]
U Mzqn Λ Mxy Λ Mxy
1
ψ ψ … ψ ψ (16) ÿ ¨ ¨
= Λ
S yˆ α1 yˆ y d − yˆ + α 2 yˆ y d − yˆ ,
y , C , S , 0, , y , C , S , 0
U Mxqn Mxy ∑ yi2 + n F
Λ
q1 q1 q1
ΛF
qn qn qn
= (21)
U Λ F
Myqn
Λ Mxy Λ Mxy
1
ÿ
¨ ¨
S zˆ α1zˆ zd − zˆ + α 2 zˆ z d − zˆ .
− xq1 , − Sψ q1 , Cψ q1 , 0, …, − xqn , − Sψ qn , Cψ qn , 0
Λ Mxy ΛF ΛF =
Λ ∑ xi + n
2
F
150
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
,
m α1zˆ α1θˆ ÿ
¨ ÿ ÿ
Iz − Ix
( ) 1 θd − Xˆ 4 qn + θ d − Xˆ 1qn Xˆ 5 qn
¨
=U1 zd − zˆ + z d + g + [κ sgn ( S zˆ ) + Υ zˆ S zˆ ] −ω z
I y α 2θˆ
(28)
µ3 α 2 zˆ α 2 zˆ 6 = Iy
, M ydes
( )
ÿ
l jr ˆ 1 −ω
where κ 4 ,κ 5 , κ 6 , Υ xˆ , Υ yˆ , Υ zˆ > 0 . In addition, the terms µ1 , µ2 , µ3
q ,n
+ X 1qn Ω +
κ t anh S + Υ S
Xˆ 3 qn Xˆ 3 qn θ
are: Ix α 2θˆ 2 Xˆ 3 qn
α1ψˆ ÿ
¨ ÿ ÿ
Ix − Iy
µ2 = ( cosϕˆ sinθˆsinψˆ − sinϕ sinψ ) , µ = ( ϕ θ 1 ψ+ ϕ ψ )µ = ϕ
I
θˆ
α
z 2ψˆ
ψ d − Xˆ 6 qn +ψ d − Xˆ 1qn Xˆ 3qn
Iz
) ( cosϕˆ sinθˆcosψˆ + sinϕˆ sinψˆ ) , µ = cosϕˆ cosθˆ
(23) M zdes =
ϕ ψ , µ1 = l 1
( )
q ,n
3
+ [κ 3 tanh S Xˆ + Υ Xˆ S Xˆ ] − ωψ
α 2ψˆ
To synthesize a stabilizing control law by second order
5 qn 5 qn 5 qn
sliding mode, the sufficient condition for stability of the In these equations Xˆ qn is the estimated state vector. Therefore:
system, i.e. V = ( S xˆ ) S xˆ . S xˆ < −. S xˆ , where ℒ is the positive
value ( > 0) , must be verified. The time derivative of V ( S x̂ ) =Ux
m α1xˆ
(
)
x − Xˆ 8 qn + xd +
µ1 α 2 xˆ d
1
κ tanh S Xˆ
α 2 xˆ 4 ( )+Υ Xˆ 7 qn
7 qn
S Xˆ − ω x
satisfies S xˆ . S x̂ < 0 . Thus, a Lyapunov function can be written
7 qn
1 (29)
as V = ∑ S xˆ T S xˆ , that one can describe as :
2 Uy =
m α1 yˆ
(
y − Xˆ 10 qn +
µ2 α 2 yˆ d )
yd +
1
α 2 yˆ 5 ( )
κ tanh S Xˆ 9 qn + Υ Xˆ 9 qn S Xˆ 9 qn − ω y
V ( S xˆ =
)
1 T
2
(
S xˆ S xˆ + S Tyˆ S yˆ + S zTˆ S zˆ + SϕTˆ Sϕˆ + SθTˆ Sθˆ + SψTˆ Sψˆ ) (24) and:
.
The chosen law for the attractive surface is the time derivative
of
= µ1 ( cosXˆ 1qn sinXˆ 3qn sinXˆ 5 qn + sinXˆ 1qn sinXˆ 5 qn )
( S ˆ ) = S ˆ . S ˆ = −κ S ˆ − Υ S 2ˆ ≤ 0
V x x x x x̂ x
=
(25) µ2 ( cosXˆ 1qn sinXˆ 3qn sinXˆ 5 qn − sinXˆ 1qn sinX
ˆ
5 qn ) (30)
Therefore, according to (25) the time derivative of final
Lyapunov function V ( S xˆ ) is as: µ3 = cosXˆ 1qn cosXˆ 3qn
−κ 4 S xˆ − Υ xˆ S xˆ2 − κ 5 S yˆ − Υ yˆ S y2ˆ − κ 6 S zˆ − Υ zˆ S zˆ2 4- 3- Final Cooperative Control Vectors
V ( S xˆ ) =
Now, according to Eq. (17) the cooperative control vectors
−κ1 Sϕˆ − Υ ϕˆ Sϕ2ˆ − κ 2 S ˆ − Υ ˆ S 2ˆ − κ 3 Sψˆ − Υψˆ Sψ2ˆ
θ θ θ
2
for each independently designed quadrotor can be described
¨
( l
)
ÿ
−κ 4 S xˆ − Υ xˆ α1xˆ xd − xˆ + α 2 xˆ x d − cos ϕˆ sin θˆ cosψˆ + sin ϕˆ sinψˆ U1 + ω x as follows:
m
U Fq1Fqdes
,1 + U Mxq 1 M xq ,1 + U Myq 1 M yq ,1 + U Mzq 1 M zq ,1
des des des
U coop Q1
2
¨
(
−κ 5 S yˆ − Υ yˆ α1 yˆ ( yd − yˆ ) + α 2 yˆ y d − cos ϕˆ sin θˆ sinψˆ − sin ϕˆ sinψˆ
l
)
U1 + ω y
U U Fq 2 Fqdes
,2 + U Mxq 2 M xq ,2 + U Myq 2 M yq ,2 + U Mzq 2 M zq ,2
des des des
m
2 coop Q2
= (31)
¨
( l
) U
coop Q3 U Fq 3 Fq ,3 + U Mxq 3 M xq ,3 + U Myq 3 M yq ,3 + U Mzq 3 M zq ,3
ÿ des des des des
−κ 6 S zˆ − Υ zˆ α1zˆ zd − zˆ + α 2 zˆ z d − − g + cos ϕˆ cos θˆ U1 + ω z (27)
m U coop Q
4
U Fq 4 Fqdes
¨
2
+ U Mxq 4 M xdes + U Myq 4 M ydes + U Mzq 4 M zdes
ÿ
ˆ ˆ Iy − Iz − jr θˆ Ω + l U + ω
−κ1 Sϕˆ − Υ ϕˆ α1ϕˆ ϕd − ϕˆ + α 2ϕˆ ϕ d − θψ
ÿ ÿ ÿ
,4 q ,4 q ,4 q ,4
ϕ
2
Ix Ix Ix
5- Simulation Results
2
ÿ
ÿ̈ ÿ Iz − Ix ÿ jr l In order to illustrate the efficiency of the designed controller,
−κ 2 Sθˆ − Υθˆ α1θˆ θd − θˆ + α 2θˆ θ d − ϕψ
ˆ ˆ − ϕˆ Ω + U 3 + ωθ
a set of numerical simulations are carried out. The parameters
Iy Ix y
¨ ÿ Ix − Iy l
2 for simulation of a sample cooperative quadrotors model
ÿ
−κ 3 Sψˆ − Υψˆ α1ψˆ ψ d −ψˆ + α 2ψˆ ψ d − ϕˆ θˆ + U 4 + ωψ are set as shown in Table 1 [30] and the factor of weight of
Iz Iz
control inputs is set as Λ = 1 . Mxy
Thus, under the control laws extracted from SOSMC approach ΛF 400
(U1 ,….,U 4 ,U x ,U y ) , all the system state trajectories can reach Table.1 Structural parameters [30]
at and stay on the associated sliding surfaces, respectively, Parameter Definition Value Unit
and guarantees asymptotic stability of closed loop system.
Remark 2. Note that signum function which appears in m Mass of quadrotor 1.80 kg
second order sliding mode controllers and filters can cause l Arm length 0.42
−5
m
3.35 ×10
2
chattering phenomena, or high-frequency oscillations of jr Rotor inertia
−3
kg m
2.16 ×10
2
control variables. This problem can be avoided by replacing Ix X inertia
−3
kg m
2.16 ×10
2
discontinuous signum function with an appropriate continuous Iy Y inertia kg m
−2
4.52 ×10
2
approximation such as sign ( si ) ≈ tanh ( si / ε i ) . Iz Z inertia
−6
kg m
2.98 ×10
2
Finally, individual control vectors in (20) and (22) that are b Trust factor
−2
N .s
2.25 ×10
2
generated using SOSMC method, represented by parameters d Drag factor N m.s
which are equivalent of ( Fqdes des des des
. Thus the g gravity 9.81 m / s2
,n , M x , M y , M z ) q ,n q ,n q ,n
individual control vectors for cooperative strategy are defined
151
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
Fig.3. Artificial wind gust generating setup Fig.3. The wind velocity component ω ( t ) is expressed with
respect to an inertial coordinate frame x, y, z . This disturbance
According to Item(4) Also in Eqs. (7)-(11) of the EKBF
has a time varying value which is introduced as follows:
estimator, Q ( t ) is considered as a diagonal matrix of
form Q = 10−3 * 7×7 ,10−4 * 5×5 , H ( xˆ ´k , tk ) = [ 12×12 ] and π ( t − 20 ) π ( t − 20 ) π ( t − 20 )
0.5sin + 0.2sin + 0.06sin
R k = 10−4 * 7×7 ,10−4 * 5×5 . Furthermore, state noise is set ω x ( t ) 20 10 5
π ( − ) π ( − ) π ( − ) m
as w ( t ) = wgn 10 ,1,10 , linear and measurement noise is
6 −4 ω ( t ) =
ù (t ) =
y
0.5sin
t 30
+ 0.2sin
t 30
+ 0.06sin
t 30
s (34)
assumed as v k = wgn 106 ,1,1, linear and the Jacobean matrix ω
z ( t ) 20 10 5
π ( t − 40 ) π ( t − 40 ) π ( t − 40 )
is set as: 0.5sin + 0.2sin + 0.06sin
20 10 5
0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 a1xˆ 6 0 a1xˆ 4 0 0 0 0 0 0 Remark 3. Genetic algorithm (GA) is a heuristic search
0 0 0 1 0 0 0 0 0 0 0 0 method used in artificial intelligence and computing. It is used
0 a3 xˆ 6 0 0 0 a3 xˆ 2 0 0 0 0 0 0 for finding optimized solutions to search problems based on
0 0 0 0 0 1 0 0 0 0 0 0
0 the theory of natural selection and evolutionary biology. After
F ( xˆ ´ ( t ) , t ) = 0 0 0 a2 xˆ 6 0 a1xˆ 4 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0 0 (32) an initial population is randomly generated, the algorithm
0 0 0 0 0 0 0 0 0 0 0 0
0 evolves the following three basics: reproduction, crossover,
0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 0 0 and mutation. The most important feature of this method is a
0 0 0 0 0 0 0 0 0 0 0 1 way to transform the system output to the cost function. This
0 0 0 0 0 0 0 0 0 0 0 0 cost function is as it follows:
where ∞
1
Xˆ T ( t ) ç Xˆ ( t ) + U T ( t ) ãU ( t ) dt
2 ∫0
=J (35)
I y − Iz Ix − I y Iz − Ix
=a1 = , a2 = , a3 (33)
Ix Iz Iy The GA operator values are determined after several
.
To investigate the effectiveness of proposed cooperative experiments. In this regard, the population size is set as 160,
controller, two cases are considered. The first case of grasping crossover is considered scatter and its value is 0.6, mutation
is a payload transporting by cooperative quadrotors from is set as 0.4 and type of probability is chosen as Gaussien.
different initial conditions of each UAVs for point-to-point Also η and γ are selected
= as η 10=3
[ 12×12 ] ,γ 102 [ 12×12 ] . The
path tracking. The second case considers grasping and then parameters of proposed controller, obtained by GA, have
manipulation by tracking a predefined trajectory. been presented in Table 2.
Hence, the wind field effect as a destructive factor is shown in
152
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
153
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
154
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
[3] J. Sanchez, R. Fierro, Sliding mode control for robot [6] S.C. LIU, D.L. Tan, G.J. Lui, Robust Leader-follower
formations, In: Proceedings of the 2003 IEEE International Formation Control of Mobile Robots Based on a Second
Symposium on Intelligent Control, Houston, USA, 2003. Order Kinematics Model, Acta Automatica Sinica, 33(9)
[4] O.A.A. Orqueda, R. Fierro, Robust vision-based (2007) 947–955.
nonlinear formation control, In: Proceedings of the 2006 [7] R. Babaei, A.F. Ehyaei, A Novel UAV Cooperative
American Control Conference, Minneapolis, USA, 2006. Approach for Industrial Inspec on and Monitoring,
[5] T.J. Tarn, A.K. Bejczy, X. Yun, Coordinated Control In: Proceeding of 31st International Power System
of Two Robot Arms, In: Proceedingog 1986 IEEE Conference (PSC), Tehran, Iran,2016.
International Conference Robotics and Automation , San [8] P. Cheng, J. Fink, S. Kim, V. Kumar, Cooperative
Francisco, USA, 1986. towing with multiple robots, Algorithmic Foundation of
155
R. Babaie, A. F. Ehyaei, AUT J. Model. Simul., 50(2) (2018) 147-156, DOI: 10.22060/miscj.2018.14252.5100
Robotics VIII. Springer Tracts in Advanced Robotics, 57 international conference on mechatronics , Istanbul,
( 2008) 101-116. Turkey , 2011
[9] N. Michael, J. Fink, V. Kumar, Cooperative manipulation [22] I. Eker, Second-order sliding mode control with
and transportation with aerial robots, Autonomous experimental application, ISA Transaction, 49(3) (2010)
Robots , 30(1) (2011) 73-86. 394–405.
[10] J. Fink, N. Michael, S. Kim, V. Kumar, Planning and [23] S. Mondal, C. Mahanta, A fast converging robust
control for cooperative manipulation and transportation controller using adaptive second order sliding mode, ISA
with aerial robots, Int. J. Rob. Res, 30(1) (2010) 324–334. Transaction, 51 (6) (2012) 713–721.
[11] M. Bernard, K. Kondak, I. Maza, A. Ollero, Autonomous [24] Z.Q. Guo, J.X. Xu, T.H. Lee, Design and implementation
transportation and deployment with aerial robots for of a new sliding mode controller on an underactuated
search and rescue missions, Journal of Field Robotics, wheeled inverted pendulum, Journal of the Franklin
28(6) (2011) 914–931. Institute , 351(4) (2014) 2261–2282.
[12] G. Wang, X. Nian, H. Wang, Cooperative control of [25] L. Besnard, Y.B. Shtessel, B. Landrum , Quadrotor
discrete-time linear multi-agent systems with fixed vehicle control via sliding mode controller driven by
information structure, J Control Theory , 10(3) (2012) sliding mode disturbance observer, Journal of the
403–409. Franklin Institute, 349 (2) (2012) 658–84.
[13] A.R. Patel, M.A. Patel, D.R. Vyas, Modeling and [26] C. Coza, C. Nicol, C.J.B. Macnab ,A. Ramirez-Serrano
analysis of quadrotor using sliding mode control, In: , Adaptive fuzzy control for a quadrotor helicopter robust
Proceeding of the 44th IEEE Southeastern Symp. on to wind buffeting, Journal of Intelligent and Fuzzy
System Theory, Jacksonville , 2012. Systems , 22(5)(2011) 267–83.
[14] L. Besnard, Y.B. Shtessel, B. Landrum, Quadrotor [27] R. Babaie , A.F. Ehyaei, Robust optimal motion planning
vehicle control via sliding mode controller driven approach to cooperative grasping and transporting using
by sliding mode disturbance observe , Journal of the multiple UAVs based on SDRE , Transaction of the
Franklin Institute, 349 (2)(2012) 658-684. Institute of Measurement and Control , 39(9)(2017)1391-
[15] B. Sumantri, N. Uchiyama, S. Sano, Y. Kawabata, 1408.
Robust tracking control of a quad-rotor helicopter [28] D. Mellinger, M. Shomin, N. Michael, V. Kumar ,
utilizing sliding mode control with a nonlinear sliding Cooperative Grasping and Transport Using Multiple
surface, J. System Design and Dynamics, 7(2)(2013) quadrotors, Distributed Autonomous Robotic Systems.
226-241. Springer Tracts in Advanced Robotics, 83( 2013) 545-
[16] B. Sumantri, N. Uchiyama, S. Sano, Least square 558.
based sliding mode control for a quad-rotor helicopter, [29] M. Rhudy, Y. Gu, M.R. Napolitano, An Analytical
In: Proceeding of IEEE/SICE Int. Symp. on System Approach for Comparing Linearization Methods in EKF
Integration, Kobe ,2013. and UKF, International Journal of Advanced Robotic
[17] J.J. Xiong , E.H. Zheng, Position and attitude tracking Systems, 10 (208) (2013).
control for a quadrotor UAV, ISA Transaction, 53(3) [30] M. Rhudy, Y. Gu , Understanding Nonlinear Kalman
(2014) 725-731. Filters Part I: Selection of EKF or UKF, Interactive
[18] R. Babaei, A. FEhyaei, Robust Control Design of Robotics Letters, Link: http://www2.statler.wvu.edu/~irl/
a quadrotor UA Based on Incremental Hierarchical page13.html, 2013.
Sliding Mode Approach, In: 25th Iranian Conference on [31] R. Babaie, A.F. Ehyaei, Robust Backstepping Control of
Electrical Engineering (lCEE), Tehran, Iran, 2017. a quadrotor UAV Using Extended Kalman Bucy Filter,
[19] L. Derafa, A. Benallegue, L. Fridman, Super twisting International Journal of Mechatronics, Electrical and
control algorithm for the attitude tracking of a four rotors Computer Technology, 5(16) (2015) 2276- 2291.
UAV, Journal of the Franklin Institute , 349 (2) (2012) [32] K. Runcharoo, V. Srichatrapimuk, Sliding Mode Control
685-699. of quadrotor, In: Proceeding of International Conference
[20] L.L.Vegan, B.C.Toledo, and A.G.Loukianov, Robust of Technological Advances in Electrical, Electronics and
block second order sliding mode control for a quadrotor, Computer Engineering (TAEECE), Konya, 2013.
Journal of the Franklin Institute , 349 (2) (2012) 719-739 [33] L.Wang, C. He, P. Zhu, Adaptive Sliding Mode Control
[21] M. Bouchoucha, S. Seghour, M. Tadjine, Classical for quadrotor Aerial Robot with Ι Type Configuration,
and second order sliding mode control solution to an International Journal of Automation and Control
attitude stabilization of a four rotors helicopter: from Engineering , 3(1) (2014) 20-26.
theory to experiment , In: Proceedings of the 2011 IEEE
R. Babaie, A. F. Ehyaei, Cooperative Control of Multiple quadrotors for Transporting a Common Payload, AUT J. Mod.
156