Escolar Documentos
Profissional Documentos
Cultura Documentos
ThCTT3.4
I. I NTRODUCTION
Unmanned aerial vehicles (UAVs) are well suited to a
wide variety of applications. They have been the subject of
extensive research in recent years. In particular, their popularity in the recreational, research, and commercial sectors has
surged, especially in the eld of precision agriculture. Their
ease of use and maneuverability make them t to be used as
remote sensing platforms [1] and sampling devices [2] for
agricultural applications.
Some applications such as large scale crop spraying [3] do
not require extremely precise platforms, so moderate changes
in position are acceptable. On the other hand, other applications involving close interaction with the environment,
such as canopy sampling [4] or short-range shots in lm
production require a high degree of accuracy. Minimizing
changes in position under such circumstances becomes of
the foremost importance.
Wind is a major disturbance source for all ying platforms.
Being able to remain stable and minimize wind disturbances
would increase the range of possible applications for UAVs.
Reactive approaches to wind rejection have been the
subject of prior research. As part of their investigation, Orsag
et al. [5] developed a low level controller performing wind
rejection. However, their method did not use a separate
wind sensor and reacted to changes in attitude rather than
directly reacting to the wind. Passive approaches relying
{zhenrong.chen,k.stol}@auckland.ac.nz
2 B. A. MacDonald is with the Department of Electrical and Computer Engineering, University of Auckland, Auckland 1010, New Zealand
b.macdonald@auckland.ac.nz
978-1-4673-9333-1/16/$31.00 2016 IEEE
695
FRONT
1
4
Bx
w
v
By
Bz
(a)
x
y
z
Fig. 1.
(c)
Fig. 2. Top down view of the quadcopter and reference frames used.
(a) shows the wind frame W. (b) shows the inertial frame, O. , ,
and denote the counterclockwise rotation around the x-, y-, and z-axes
respectively. (c) shows the quadcopter. The body frame is denoted by B.
The forward axis of the quadcopter is coincident with B x.
system at the University of Auckland to transfer pose information to the ight controller through its GPS and compass
ports. This will allow for seamless transitions between indoor
and outdoor tests in the future.
III. M ODELING
This section describes the dynamics of the quadcopter
used. The assumptions made are listed and a complete
three-dimensional dynamic model is presented following
Luukkonens methodology [10]. Finally, a simplied two
dimensional model used for uniaxial wind rejection is presented.
A. Assumptions
The rigid body assumption is made for the frame and
motors. The model created assumes rotor friction and the
motor control dynamics are negligible. The mass is assumed
to be symmetrically distributed along the quadcopters roll,
pitch, and yaw axes.
C S C + S S
S S C C S
C C + S S
(1)
where Sa = sin a and Ca = cos a. The rotation matrix from
O
inertial to body frame is thus denoted by B
OR = B R .
Likewise, the transformation for angular velocities from
inertial to body frame is dened as
C C
O
S C
R
=
B
S
C S S S C
S S S + C C
C S
TABLE I
Q UADCOPTER PARAMETERS .
Description
Value
0.23 m
1, ,4
1.5 kg
Ixx
18 103 kg m2
Iyy
18 103 kg m2
Izz
36 103 kg m2
IM
6.9 106 kg m2
0.43 106 kg m2
8.2 106 kg m
(b)
Parameter
where
1
W = 0
0
= W
0
C
S
(2)
S
C S .
C C
C. Motor Modeling
A simple model accounting for the motors mass moment
of inertia IM , coefcient of viscous friction b, and lift
coefcient k is dened in this section. The total thrust B T
and total torque B applied on the body are obtained using
696
T=
4
i=1
4
BT
Ti , where B Ti = 0
0
i=1
0 ki2
(1)i (bi2 + IM i )
+ l cos i
sin i
(3)
WU
Faero
(4)
z
B Ti
D. Equations of Motion
Let the state vector x = x x , control input u = ,
and disturbance input ud = W Uu . The resulting nonlinear
system can be expressed as:
x
(11)
x = f (x, u, ud ) =
cf,x
2
g tan u m
(ud + x2 )
g=m 0
0
(1)i IM i
mg
= B mB V + B Rg + B T B FD (5)
mB V
O
IB = B IB V + + B B MD
(6)
0 9.81
x = Ax + Bu + Bd ud
y = Cx
(12)
(13)
E. Drag Modeling
The A, B, Bd , and C matrices are obtained by linearizing
the nonlinear system around
B
FD = sgn B Vapp cf B Vapp B Vapp
(7)
B
B
B
B
MD = sgn cm
(8)
x = 0, u = eq (ud )
where is the Hadamard product (element-wise multiplication) and sgn(a) is the signum function. cf and cm are
vectors of translational and rotational drag constants. The
apparent velocity vector B Vapp is a combination of the
translational velocity B V and the wind vector W U:
B
(14)
B O
W
O RW R U
Vapp = V
(9)
W
Uu W Uv W Uw
where W U =
and O
=
WR
diag (1, 1, 1). The wind frame of reference is shown in
Fig. 2a.
z = z
ud = z
where
0
=
1
Uu , = cf,x W Uu + x W Uu + x . (10)
Faero x,
0
, = 0
0
(16)
(17)
1 .
BT
Loiter
Controller
loiter
des
ff
Filtered pose
EKF
Transmitter
stick
positions
fff (ud )
ud
0
Uu
Sensor measurements
0
Fig. 5.
10
20
30
t (s)
Uv
40
Uw
50
60
u = loiter + eq
(18)
(21)
(19)
2
.
(20)
G = 0 2cf,x udgcos u
B. Loiter Results
The disturbance rejection performance of each of the
controllers was assessed in simulation. Each controller was
subjected to the wind prole shown in Fig. 5 over a period
of 60 s. The quadcopter was at rest at the start of each
simulation.
Fig. 6 shows the complete simulation results and demonstrates that all three controllers remained stable throughout
the simulation. The x-y trace of each controller over the
simulation period, in Fig. 6a, shows that the baseline trace
has a wider spread along the x-axis, which is expected since
it is the dominant wind direction. The DAC and nonlinear
y (m)
0.5
(a)
0
Dominant
wind
direction
0.5
1
0.5
0.5
x (m)
(b)
x (m)
1
0
1
(c)
y (m)
0.2
0
0.2
z (m)
0.1
(d)
0.1
4
(e)
( )
2
0
2
4
(f)
( )
20
10
0
(g)
( )
0.2
0
0.2
0
10
20
Baseline
30
t (s)
DAC
40
50
60
Nonlinear
699
y (m)
Industrial fan
0.2
Dominant
wind
direction
0.4
Quadcopter
0.2
Patterned mat
Fig. 7.
Test
ex,RMS (m)
ey,RMS (m)
ez,RMS (m)
|eRMS | (m)
BL
DAC
NL
0.33
0.18
0.10
0.11
0.11
0.11
0.030
0.030
0.030
0.35
0.21
0.15
TABLE III
RMS ERROR DURING EXPERIMENTAL LOITER TEST.
700
Test
ex,RMS (m)
ey,RMS (m)
ez,RMS (m)
|eRMS | (m)
BL
0.7
0.1
0.03
0.7
ACKNOWLEDGMENT
The authors would like to thank M. McCauley, A. Carrell,
S. Jiang, D. How, J. Kutia, and the Controls Group at the
University of Auckland for their assistance.
R EFERENCES
[1] D. Gatziolis, J. F. Lienard, A. Vogs, and N. S. Strigul, 3D Tree
Dimensionality Assessment Using Photogrammetry and Small Unmanned Aerial Vehicles, PLoS One, vol. 10, no. 9, Sep. 2015.
[2] J.-P. Ore, S. Elbaum, A. Burgin, and C. Detweiler, Autonomous Aerial
Water Sampling, Journal of Field Robotics, vol. 32, no. 8, pp. 1095
1113, 2015.
[3] B. Faial, G. Pessin, G. Filho, A. Carvalho, G. Furquim, and
J. Ueyama, Fine-Tuning of UAV Control Rules for Spraying Pesticides on Crop Fields, in Tools with Articial Intelligence (ICTAI),
2014 IEEE 26th International Conference on, Nov. 2014, pp. 527533.
[4] J. Kutia, K. Stol, and W. Xu, Canopy Sampling Using an Aerial
Manipulator: A Preliminary Study, in Unmanned Aircraft Systems
(ICUAS), 2015 International Conference on, Jun. 2015, pp. 477484.
[5] M. Orsag, T. Haus, I. Palunko, and S. Bogdan, State Estimation,
Robust Control and Obstacle Avoidance for Multicopter in Cluttered
Environments: EuRoC Experience and Results, in Unmanned Aircraft
Systems (ICUAS), 2015 International Conference on, Jun. 2015, pp.
455461.
[6] C. Mnsson and D. Stenberg, Model-Based Design Development and
Control of a Wind Resistant Multirotor UAV, Masters Thesis, Lund
University, Lund, Sweden, 2014.
[7] A. Cho, J. Kim, S. Lee, and C. Kee, Wind Estimation and Airspeed
Calibration using a UAV with a Single-Antenna GPS Receiver and
Pitot Tube, Aerospace and Electronic Systems, IEEE Transactions
on, vol. 47, no. 1, pp. 109117, Jan. 2011.
[8] B. Arain and F. Kendoul, Real-Time Wind Speed Estimation and
Compensation for Improved Flight, Aerospace and Electronic Systems, IEEE Transactions on, vol. 50, no. 2, pp. 15991606, Apr. 2014.
[9] DIYDrones. (2016, Feb.) Copter Multirotor UAV. [Online].
Available: http://copter.ardupilot.com/
[10] T. Luukkonen, Modelling and Control of Quadcopter, Independent
research project in applied mathematics, Aalto University, Espoo,
Finland, Aug. 2011.
[11] C. D. Johnson, Theory of Disturbance-Accommodating Controllers,
in Control and Dynamic Systems, C. T. Leondes, Ed. New York,
USA: Academic Press, 1976, vol. 12, ch. 7, pp. 387489.
VII. C ONCLUSIONS
A simplied wind-inclusive two-dimensional linear model
was created. Based on this mathematical model, two new
controllers were derived for uniaxial wind rejection: a DAC
using a linear control law to minimize wind disturbances and
a nonlinear feedforward controller computing the required
pitch angle to achieve static equilibrium. Both controllers
are modications of the existing APM 3.3 loiter controller.
The DAC was shown in a case study to decrease the RMS
position error of the UAV along the compensated axis by
66 % compared to the baseline. The nonlinear controller was
similarly shown to result in a decrease of 45 %. However,
the DAC was found to be more affected by changes in wind
speed due to its linear nature. The feasibility of rejecting a
5 m/s wind was proven by a sample experimental test with
the baseline controller.
The performance of the DAC and nonlinear controllers
when subjected to a large range of input wind speeds will
be assessed. Furthermore, the experimental UAV will be
upgraded to use precise pose measurements from a motion
capture system. This will enable more thorough physical
experiments. In order for the designed controllers to be
implemented, the quality of wind measurements taken by the
onboard pitot tube will need to be assessed. Subsequently, the
DAC and nonlinear controller will be implemented and tuned
on the system. Finally, their wind rejection performance will
be tested experimentally using an industrial fan.
701