See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317933218

Development and Experimental Verification of a Hybrid Vertical Take-Off and

Landing (VTOL) Unmanned Aerial Vehicle(UAV)

Conference Paper · June 2017

DOI: 10.1109/ICUAS.2017.7991420

CITATIONS READS

72 36,900

5 authors, including:

Haowei Gu Lyu Ximin

The Hong Kong University of Science and Technology The Hong Kong University of Science and Technology
12 PUBLICATIONS 332 CITATIONS 16 PUBLICATIONS 490 CITATIONS

SEE PROFILE SEE PROFILE

Fu Zhang
The University of Hong Kong
106 PUBLICATIONS 2,297 CITATIONS

SEE PROFILE

All content following this page was uploaded by Lyu Ximin on 01 November 2017.

The user has requested enhancement of the downloaded file.

 Development and Experimental Verification of a Hybrid Vertical
Take-Off and Landing (VTOL) Unmanned Aerial Vehicle(UAV)
Haowei Gu, Ximin Lyu, Zexiang Li, Shaojie Shen, and Fu Zhang

Abstract— In this paper, we present the design and ver-

ification of a hybrid vertical takeoff and landing (VTOL)
unmanned aerial vehicles (UAV) of the type named dual-
system or extra propulsion VTOL UAV. This paper features the
comprehensive system development of such VTOL UAVs from
all aspects, including the aircraft design and implementation,
onboard devices integration, ground station support, and long
distance communication. We proceed with aerodynamic anal-
ysis, mechanical design, and controller development. Finally,
we verify by experiment that this hybrid VTOL UAV has the
desired aerodynamic performance, flight stability, endurance
and range. In addition, with the designed flight controller, the
VTOL UAV can achieve full autonomous flight in a real outdoor
environment. It serves a good platform for future research, such
as vision-based precise landing, motion planing and quick 3-
D mapping, as well as service applications, such as medicine
delivery.

I. I NTRODUCTION
During the last decades, unmanned aerial vehicles (UAV)
have been widely used in both military and consumer areas
such as surveillance, tracking, monitoring and aerial pho-
tograpy. These UAVs are mainly classified into two types:
fixed-wing UAVs and rotary-wing UAVs and each type has
their own advantages and drawbacks. As for fixed-wing
UAVs, they have several features, such as high cruising speed
and altitude, high flight efficiency and large cabin, which
means they can fulfill many tasks that require the vehicle to
carry a heavy payload or achieve long endurance and range.
However, fixed-wing UAVs have their limits. A fixed-wing
airplane needs a runway or catapult to take off and land,
which imposes restrictions on its applications. Rotary-wing
UAVs, on the other hand, can take off and land vertically,
which makes these vehicles suitable for a wide variety of
environments. But rotary-wing vehicles do not have long
endurance or range. It is still a challenge to develop a small-
scale UAV, which is able to achieve long flight endurance
and range and at the same time takes off and land vertically.
Hybrid vertical takeoff and landing (VTOL) UAVs are a
suitable solution for this problem; they combine the concepts
of fixed-wing and rotary-wing UAVs in a single platform,
thus inheriting the advantages of both.
There are several types of hybrid VTOL UAVs, including
tail-sitter, tilt-rotors and dual-system UAVs.
*Research supported by Hong Kong ITF Foundation (ITS/334/15FP).
All authors are with the Department of Electronic and
Computer Engineering, Hong Kong University of Science and
Technology, Hong Kong, China. hguad@connect.ust.hk,
xlva@connect.ust.hk, eezxli@ust.hk,
eeshaojie@ust.hk, eefzhang@ust.hk
Fig. 1. The developed hybrid VTOL UAV (Coordinate systems are
defined in section III) Video available at: https://www.youtube.
com/watch?v=wRE4ICjHJ_0
A. Tail-sitter VTOL UAVs
A tail-sitter VTOL UAV takes off vertically on its tail, then
the whole aircraft including the fuselage tilts horizontally
for level flight. In [1] and [2], many aspects of a tail-
sitter have been studied, including design, manufacturing,
modeling, control and flight tests. However, since the same
motors are being used for both hovering and level flight, a
fixed-pitch propeller is usually not able to achieve optimal
efficiency in both stages while a varying-pitch propeller can
significantly complicate the mechanical design, maintenance
and manufacturing cost. In addition, the big wing area being
exposed to cross-wind during hovering introduces a con-
siderable disturbance that is likely to degrade the hovering
performance.
B. Tilt-rotor VTOL UAVs
Similar to a tail-sitter VTOL UAV, a tilt-rotor VTOL UAV
generates lift and forward thrust using the same propulsion
system. But instead of tilting the whole aircraft like a tail-
sitter VTOL UAV, the tilt-rotor VTOL UAV only tilts its
rotors. For vertical flight, such as takeoff, hovering and
landing, rotors of the propulsion system are angled such
that their plane of rotors is horizontal, lifting in the way
that a helicopter rotor does [3]. As the UAV gains speed,
the rotors are progressively tilted forward, with the plane
of the rotors eventually becoming vertical. In this mode,
the wing provides the lift, and the rotor provides thrust
as a propeller. In [4], [5] and [6], the structure design, presented. The designed aircraft is shown in Fig. 1. As
aerodynamic analysis and flight control were addressed. [7] validated by real flight tests, the developed UAV can achieve
shows another configuration of tilt-rotor, which also called the required flight range and fully autonomous operation.
tilt-wing VTOL UAV. The main drawback of a tilt-rotor The remainder of this paper is organized as follows. The
VTOL UAV is its complicated structure, which increases the aerodynamic and mechanical design, as well as onboard
aircraft weight and cost. devices integration, are detailed in section II. Position con-
trollers, attitude controllers and VTOL mixer are presented in
C. Dual-system VTOL UAVs section III. Section IV presents the details of the experiment
A dual-system VTOL UAV has two independent propul- setup and results discussion. And finally section V draws the
sion systems for hover and level flight. As seen in Fig. 2, the conclusion.
aircraft takes off vertically using the lift force generated by
the four lifting rotors. During this phase, the push motor II. S YSTEM D ESIGN
will be turned off and the whole aircraft behaves like a A. Airframe and Propulsion System Selection
quadrotor [8], its position as well as attitude are controlled As mentioned before, the designed aircraft is expected to
by the differential thrust of the four rotors. After reaching achieve a flight range of 30 km. For this purpose, a variety
the desired altitude, the aircraft starts transiting to level flight of devices need to be used, including video transmitter for
by turning on the push motor to gain forward speed. Once aircraft status monitoring, telemetry for long distance data
the cruise speed is reached, the four lifting rotors will be communication, radio control (RC) extender for emergency
turned off and the whole aircraft behaves like a conventional piloting, etc. Their estimated weights are summarized in
fixed-wing airplane. The control of the position and attitude table I, where the units are in grams.
is achieved by the push motor and control surfaces such
as the ailerons, elevators and rudders. Since the propulsion TABLE I
systems for hover and level flight are independent, they can TABLE OF DEVICE WEIGHT
be optimized for both flight modes separately. Other than
motors and propellers, all parts of a dual-system VTOL UAV Device Quantity Weight1
are assembled rigidly and no moving parts are involved. As Video transmitter 1 20
Telemetry 1 20
a result, they are easy to design, manufacture and maintain. Battery 1 900
Flight control board 1 30
Lifting Motor Lifting motors 4 70
ESCs for lifting motor 4 40
Pusher motor 1 80
Wing ESC for push motor 1 60
Others N.A. 200
Total weight N.A. 1750

The item ”Others” includes servo motors, wires, pitot-

tube, GPS, etc. The total weight estimated for all of the
Aileron Pusher Motor
devices is around 1.75 Kg . In our work, we intend to
adopt a commercially available fixed-wing airplane as the
starting point to develop a hybrid VTOL UAV due to their
proved flight stability and performance. Table II shows the
Vertical Tail
key configurations of a number of commercial-off-the-shell
fixed-wing airplanes.
Elevator Horizontal Tail
TABLE II
Fig. 2. Hybrid VTOL UAV explanation F IXED - WING A IRPLANE M ODELS

In this paper, we present the development of a dual-system Mini Sky- Ranger

UAV Talon
hybrid VTOL UAV from a holistic system design perspective. SkyHunter Hunter EX757-4
Wing Span (mm) 1238 1800 1718 1380
The primary goal of the designed aircraft is to deliver a Gross Weight (gram) 760 900 1050 750
lightweight parcel, such as first aid medicine, at a distance of MTOW (gram) 2500 3500 3000 1100
over 30 km, while maintaining a relative small size for better Payload Capacity
1740 2100 1950 350
portability. The designed unmanned aerial system has all the (gram)
key components for safe and reliable operation, including
autonomous navigation, robust flight controller, first-person- It can be seen that Mini Skyhunter has the smallest size
view (FPV) camera, long distance communication, ground while achieving the required payload capacity. A smaller size
station, etc. The detailed aerodynamic design, analysis, me-
chanical consideration and flight controller development are 1 The unit of weight is in gram
is preferable so to increase its maneuverability and ease of the thrust/gravity ratio is designed to be 1.3. The resulting
the flight test. The detailed parameters of such an airplane maximal thrust of each motor is therefore 0.81 kg, making
are summarized in table III, where mean chord, wing area DJIE310 a good choice.
and aspect ratio (AR) are measured on a real airplane. The In a similar way, thrust produced by the push motor is
cruise speed is recommended by the manufacturer. e is the designed to be 2 times greater than the drag force in order to
Oswald efficiency factor which is typically between 0.7 and gain sufficient maneuverability, such as during acceleration,
0.85 and can be computed from AR [9]. banking turns, and level flight. That is:
TABLE III 1
T = 2 · D = 2 × ρV 2 SCD
DETAILED PARAMETERS OF M INI S KY H UNTER 2
= 2 × 0.5 × 1.157 × 152 × 0.21 × 0.064 = 3.498N.
Mean Aerodynamic
(2)
Wing area Cruise speed AR e With this requirement, we investigate a few motors and
Chord (MAC)
17 cm 0.21m2 15 m/s 7.28 0.83 propellers designed for fixed-wing airplanes, as summarized
ρ MTOW Re CL CD in tableV. It can be seen that the Sunny Sky X2216 is the
1.157 kg/m3 2500 gram 2.3 × 106 0.896 0.064
right fit for to our needs.

TABLE V
To compute the Reynolds number Re, the commercial
PARAMETERS OF MOTORS AND PROPELLERS FOR FIXED - WING
software Profili is used. Supplying the airspeed; characteristic
AIRPLANE MODELS
linear dimension, which is defined as the mean aerodynamic
chord (MAC); and cruise altitude, which is 100 m for our
application to the software, yields the Reynolds number Thrust Efficiency Voltage
Re = 2.3 × 106 . Name KV
(N )
Propeller
(g/W ) (V )
Lift coefficient CL [9] is calculated for level flight where SunnySky X2826 880 4.9 12×6 0.068 14.8
the maximal takeoff weight M T OW = 24.5 N and cruise SunnySky X2216 880 3.92 10×4.7 0.072 14.8
SunnySky X2814 900 4.9 11×7 0.070 14.8
speed V = 15 m/s. As pointed out in [10], the total drag of
a typical airplane consists of three parts: parasitic drag D0 ,
wave drag Dw and induced drag Di . The wave drag occurs
when the airplane’s speed is near the speed of sound and can B. Mechanical Design
be safely ignored in our case where the flight speed is far Even though the main airframe is adopted from a
from the speed of sound. As shown in [11], the total drag, commercial-off-the-shell airplane, the integration of four
when parameterized by drag coefficient, can be expressed as lifting rotors for vertical takeoff and landing introduce new
problems that need to be carefully considered. A major
Swet CL2 challenge is the structural design and precise landing in
CD = · Cf e + (1)
S π · AR · e
where the typical value of Swet /S for the Mini SkyHunter
is around 4, as seen in [12], and the skin friction coefficient
Cf e is drawn from [13].
After determining the aerodynamic configuration of the
aircraft, we proceed to the propulsion system design. For
a hybrid VTOL UAV, there are two sets of propulsion
systems, namely the four lifting rotors for takeoff, landing
and hover flight, and the push motor for level flight. Table
IV summarizes the key parameters of a few motor-propeller
pairs specifically designed for multirotors.
TABLE IV
PARAMETERS OF MOTORS AND PROPELLERS FOR MULTIROTORS

Name KV Thrust (gram) Propeller Voltage (V)

DJI E310 960 800×4 9×4.5 11.1
DJI E800 350 2100×4 13×4.5 22.2
EMax MT2212 900 850×4 9×4.5 14.8
EMax MT2208 1500 690×4 6×45 14.8

It is clear that the thrust produced by the four lifting

rotors has to be greater than the force of gravity acting on
the vehicle. To keep a sufficient margin for maneuvering, Fig. 3. motor layout and inclination
 TABLE VI
the presence of cross wind. The four lifting motors are
S ELECTION OF DEVICES
located around the fuselage systematically along the X axis
and Y axis, as shown in Fig. 3. The vertical tails of the Communi-
main airframe, primarily designed for enhancing the direc- Name Figure Brand Frequency cation
tional stability of the aircraft during level flight, introduce
range
a considerable aerodynamic moment in the yawing during
hover flight in the presence of cross wind. This aerodynamic Depands
moment, together with other aerodynamic disturbances from My Fly on
AAT N.A.
the wing and fuselage, will affect the hover accuracy and Dream vedio
landing reliability. In order to mitigate this situation, a large transmitter
yawing moment that usually exceeds the reaction torque of
a typical motor-propeller pair is required. As such, the four
lifting motors of the vehicle are inclined along their diagonal Telemetry XTend 915MHz 32 Km
lines by a small angle δ. As shown in Fig. 3, the green circle
arrow indicates the direction of the motor reaction torque,
and the red arrow represents the inclination direction. With
Video
this design, the total moment in yawing is the sum of the LawMate 1.4GHz 35Km
torque produced by the motor thrust projected to the body transmitter
X − Y plane and the motor reaction torque. In addition, as
the motor inclination is along the motor diagonal lines, the
torque can be maximized with the same motor thrust and Radio
inclination angle. U Link 433M 40 Km
extender

C. Others
To constitute a comprehensive unmanned aerial system
that is able to perform a mission fully autonomously and
with the capability of communicating with a ground station
in real time to report its flight status, such as air data, position
and battery status, as well as for receiving commands from
the ground station in case of an emergency or mission re-
planning, several onboard devices need to be integrated.
After investigating the current commercial-off-the-shelf de-
vices, we have settled on the devices given in table VI.
As the designed aircraft is expected to achieve a flight
distance of 30 km, to support its long distance operation, an
automatic antenna tracker (AAT) is used extend the connec-
tion between the aircraft and the ground station. An AAT can
track the aircraft using its GPS signal, by way of an antenna
installed on the ground station that can always point to the
aircraft. By doing so, the data link can be greatly enhanced.
Telemetry is used to connect the vehicle and ground station
to download flight status and upload commands sent from
the ground station in real time. A lightweight first person
view (FPV) camera and the associated video transmitter are
used to monitor the aircraft’s surrounding environment in
real time. A radio extender is used to extend the operation
range of the radio controller. The flight controller used is
a Pixhawk autopilot open hardware and software platform
[14]. The Pixhawk autopilot has integrated the necessary
sensors for autonomous navigation, such as a magnetometer
for measuring the direction of south, an accelerometer for
measuring the direction of gravity, gyros for measuring
angular rate and a barometric altimeter for measuring the
altitude of the aircraft. With this open source platform, we
designed our own flight controller, as discussed in next
section, in detail. Other key components such as a pitot-
tube, GPS and data transmitter are also used although are
Camera ANAITV N.A. N.A.
Pixhawk 3DR N.A. N.A.
not shown in table VI.
Fig. 4 shows the entire unmanned aerial system that
has been implemented. The whole system consists of a
ground station, a pilot, and the unmanned aerial vehicle. The
ground station is a PC running Qgroundcontrol (QGC) [15]
software for logging and displaying the aircraft status, an
AAT for aircraft tracking and connection enhancement, and
an external monitor that displays the onboard FPV video as
well as the flight status of the aircraft such as the airspeed,
distance, altitude, speed and attitude. A pilot can maneuver
the aircraft directly via a radio controller if in manual testing
mode, or simply monitor the aircraft status and take over its
control in case of emergency if in fully autonomous flight
mode.
III. F LIGHT C ONTROL S YSTEM D ESIGN
The block diagram of our flight control system is shown in
Fig. 5. It can be seen that the whole flight system consists of
three level controllers: position controller, attitude controller,
and VTOL mixer.
Comprising the first level are three position controllers:
a transition controller, rotary-wing position controller, and
fixed-wing position controller, which are running simultane-
ously. A position controller receives the desired position from
 Vehicle with Autopilot
Enhanced Manual
Command
GPS Signal
Radio Extender
QGC Command Aircraft status FPV Video
Manual Command
Radio Controller Ground Station Monitor AAT
Fig. 4. Signal flow of whole unmanned aerial system
either human pilots or navigation way points and computes
the desired attitude and thrust, which are then fed to a
signal distribution module. The signal distribution module
maps the desired attitude from the transition controller as
well as, rotary-wing position controller to the same attitude
controller, called the VTOL attitude controller, and maps
the attitude from fixed-wing position controller to the cor-
responding fixed-wing attitude controller. The thrust of the
three position controllers are fed to a VTOL mixer for further
processing.
The second level controllers are two attitude controllers: a
fixed-wing attitude controller and VTOL attitude controller.
While the fixed-wing attitude controller directly computes
the desired moment from its attitude error, the VTOL atti-
tude controller will use either the transition or rotary-wing
position controller attitude set point to compute the desired
moment, depending on the current flight mode.
Depending on the current vehicle flight mode (i.e., rotary-
wing mode, fixed-wing mode or transition mode), the VTOL
mixer chooses different moment and thrust sources and maps
them to different actuators. For example, in rotary-wing
mode, the moment from the VTOL attitude controller and
the thrust from the rotary-wing position controller are used
to actuate the four lifting rotors. This results in an aircraft that
will behave like a typical quadrotor UAV. In transition flight
mode, the moment from the VTOL attitude controller and the
thrust from the transition controller are synthesized to actuate
the control surfaces such as ailerons, push motor, or the four
lifting rotors all together. More specifically, during forward
transition from rotary-wing mode to fixed-wing mode, the
push motor speeds up to a prescribed RPM to provide
forward thrust. Once the prescribed airspeed is reached,
the fixed-wing mode will be triggered. During backward
transition from fixed-wing mode to rotary-wing mode, the
push motor slows down gradually to a stop. At the same
time, the lifting rotors are turned on to control the vehicle’s
attitude and altitude. Similarly, once the vehicle’s airspeed
drops below some prescribed value, the rotary-wing mode
will be triggered. Finally, in fixed-wing mode, moment from
fixed-wing attitude controller and thrust from the fixed-wing
position controller are used to actuate the control surfaces
and push motor like a conventional fixed-wing airplane.
A. Position and Attitude Controller Implementation
In our scenario, the position controller including fixed-
wing position controller, rotary-wing position controller, and
transition controller. The fixed-wing position controller is
adopted from [16] and the rotary-wing position controller
from [17]. The transition controller will be presented in
section III-B. The fixed-wing attitude controller is a PID
controller adopted from [18]. The VTOL attitude controller
is a cascaded controller similar to [1]. These controllers are
all implemented on the existing framework of the Pixhawk
autopilot software stack [18].
B. Transition Controller
Here we focus on introducing the transition controller. The
key features of the transition controller in Fig. 5 are sending
fixed attitude command, the pitch and roll commands are
both set to 0, while the yaw is set to its current value at the
beginning of transition. The push motor is set to its maximum
power until the specified airspeed is reached.
The thrust command is determined by an altitude holding
controller, which is used to hold the aircraft altitude at the
value right before the transition. Recall that the rigid body
altitude dynamic can be written as
1 T
ḧz = r (Faero + FT ) + g. (3)
m 3
where r3 denotes the third column of the rotation matrix
R ∈ SO(3), which is the orientation of the body frame
relative to the inertial frame (seen in Fig. 1), hz is the
altitude in the inertial frame, Faero ∈ R3 and FT =
T

FT x FT y FT z ∈ R3 are the aerodynamic force and
motor thrust vector in body frame, respectively.
The reference altitude is described as hd , and the altitude
error is described as he = hd − hz . A simple altitude PID
controller is used to calculate motor thrust as follows:
Z
dhe
r3T FT + mg = kp he + ki he dt + kd . (4)
dt
The resulting thrust can be finally described as
kp he + ki he dt + kd dh
R
dt − mg
e
FT z = . (5)
r31
where r31 is the first entry of r3 ∈ R3 . It is worth noticing
that the aerodynamic force Faero , though neglected in the
controller design, can be compensated effectively by the
integral action. The computed thrust and attitude commands
are sent to both VTOL and fixed-wing attitude controller for
moment computation. A VTOL Mixer, as discussed below,
is finally used to distribute the required moment and thrust
to proper actuators (i.e. control surfaces and lifting motors)
depending on the current airspeed during transition.
 Thrust_1
Att_sp1 Vehicle Attitude Flight Mode
Transition
Controller
Att_sp1 VTOL Attitude Moment_1
Navigation Reference Thrust_2 Att_sp2 Controller
position Position Att_sp2
Way Points error Rotary-wing Signal
+_
Or Position Controller Distribution VTOL
Manual Inputs Fixed-wing Moment_2 Mixer PWM Aircraft
Thrust_3 Att_sp3 Attitude Controller Dynamics
Fixed-wing Att_sp3
Position Controller Thrust_1&Thrust_2&Thrust _3

Vehicle Position

Fig. 5. Architecture of the flight control system

where S is the total area including tails, c̄ is the mean aero-

dynamic chord (MAC), Cm0 is the pitch moment coefficient
xb when α = 0 and δi = 0. Cmα and Cmδ are, respectively, the
b pitch moment derivatives with respect to attack angle and
elevator deflection. As no rudder is used for this aircraft,
yb the moment produced by all control surfaces can be further
f1 f2
l1 written as
δ1 δ2 " 2MT x #   
 δ1
2
ρV Sl l1 CLδ 1 −1 0 δ2 
2MT y =
2 0 0 −1 (8)
ρV Sc̄Cmδ δ3
δ1 = δ2
where α1 and α2 are two different first order functions that
relate to airspeed. The moment and thrust produced by the
δi
δ3 four lifting rotors can be represented as:
 FT z    
Fig. 6. Vehicle configuration and effect of elevons cos(δ) 1 1 1 1 T1
MT x
 −1 1 −1 1  T2 
  
sin(δ)d+κ cos(δ)
=  T3  (9)

MT y
  
1 1 −1 −1

 (cos(δ)d−κ sin(δ)) sin(χ) 
C. VTOL Mixer MT z −1 1 1 −1 T4
(cos(δ)d−κ sin(δ)) cos(χ)
The available actuators of our vehicle are shown in Fig. 6. and
δ1 and δ2 are the angle of the left aileron and right aileron,
respectively. δ3 is the angle of the elevator. f1 and f2 Ti = ct $i2 (i = 1, 2, 3, 4) . (10)
are the lift forces produced by the left wing and right cq
κ = . (11)
wing, respectively. l1 is the distance between f1 , which acts ct
on the mean aerodynamic chord (MAC) and longitudinal where d is the length of each motor to the aircraft center of
axis. Because of symmetry, the distance between f2 and mass, as seen in Fig. 6 and χ is the azimuthal angle of the
longitudinal axis is l1 as well. According to [19], f1 and first rotor with respect to the lateral axis of the aircraft. δ is
f2 can be written as the inclination angle of the four lifting rotors, as presented
1 2 in the previous section. Ti is the rotor thrust and $i is the
f1 = ρV Sl (CL0 + CLα α + CLδ δ1 ) . rotation speed of each rotor (i = 1, 2, 3, 4). ct and cq are the
2 (6)
1 propeller thrust coefficient and propeller torque coefficient,
f2 = ρV 2 Sr (CL0 + CLα α + CLδ δ2 ) . respectively. κ is the ratio of torque to thrust coefficient.
2
It can be seen that the roll and pitch moment can be
where ρ is the air density, V is the airspeed, and Sl and produced by either control surfaces, as shown in Eq. (8)
Sr are the left and right wing area, respectively. Because of or by motor differential thrust, as shown in Eq. (9). In our
symmetry, Sl and Sr are the same. α is the attack angle. implementation, we assign a weighting factor α to weight
CL0 is the lift coefficient while α = 0, and δi = 0. CLα the contribution of the moment from the control surface and
and CLδ are, respectively, the lift coefficient derivatives with the motor differential thrust. That is, given the required roll
respect to attack angle and aileron deflection. Similarly, the moment L, the moment from the control surface and the
aerodynamic pitch moment can be written as motor differential thrust are, respectively, (1 − α)L and αL.
1 2 The way to allocate pitch moment is similar. The weighting
Maero = ρV S (Cm0 + Cmα α + Cmδ δ3 ) c̄. (7) factor α is in between zero and one, depending on the
2
current flight mode. In fixed-wing mode, α = 0, and the 2
required moments are solely generated by control surfaces,
1
the actuation signal of which are computed by inverting Eq.
(8). Accordingly, the required thrust is mapped to the push 0
motor. In rotary-wing mode, α = 1, the required moment as 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
well as thrust are solely generated by the four lifting rotors, 2000
the actuation signals of which are computed by inverting Eq. 1500
(9). In transition flight, the required moments are generated
by the control surface with portion 1 − α and by motor 1000
differential thrust with portion α. α = 1 − (0.6 VVT + 0.2) is 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
a function of airspeed V , indicating that the effectiveness of 40
the control surfaces increases as the airspeed V increases.
20
VT is a barrier speed that the aircraft has to exceed in
order to switch to full fixed-wing mode. The total thrust 0
produced by the four lifting rotors is computed from the 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
altitude holding controller presented previously, while the 32
push motor is set to a prescribed PWM to accelerate the
30
aircraft in a horizontal direction. In practice, the parameters
in Eq. (8) and Eq. (9) are not known, resulting a gain 28
mismatch when inverting these two equations. By properly 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
tunning the parameters of the flight controllers, these gain
mismatches can be effectively mitigated, as shown in real
Fig. 7. Flight status, throttle, airspeed and altitude during transition process
flight tests.

IV. E XPERIMENTS 50
Actual roll angle
With the designed dual-system VTOL UAV along with its Roll angle set point
flight control system, intensive experiments are conducted 0
to verify its aerodynamic performance, flight stability and
system level autonomy. The first experiment is designed
to verify the reliability of transition controller, the second -50
1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
experiment is to test the flight controller’s stability when the
aircraft is under active pilot control, the third experiment is 10
to test its range, endurance and the capacity of autonomous 0
flight, and the fourth experiment is to determine the optimal
flight speed. -10 Actual pitch angle
Pitch angle set point
A. Transition controller verification -20
1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
Since the transition is a necessary maneuver of the hy-
brid VTOL UAV to achieve both vertical takeoff and level 100
Actual yaw angle
flight, we designed an experiment to verify the reliability of 50 Yaw angle set point
the transition controller presented in previous sections. The
procedure of this experiment is as follows: the aircraft first 0
takes off in manual mode. After reaching the desired altitude,
the aircraft was switched to autonomous flight mode, where -50
1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58
the quadrotor position controller took place and controlled
the aircraft to a transition point. Once the transition point
is reached, the aircraft initiated a transition maneuver and Fig. 8. Roll, pitch and yaw angle during transition
subsequently flied as a fixed-wing airplane.
Fig. 7 and 8 show the transition process. The upper figure to fixed wing mode, and the push motor is controlled by
in Fig. 7 indicates the flight states where 0 means quadrotor a typical fixed-wing flight controller. Fig. 8 shows attitude
mode, 1 means transition mode and 2 means fixed wing flight response during the transition. It is worth noticing that the
mode. As mentioned in previous sections, after the transition roll and yaw angles are set to relative large values after the
maneuver is initiated, the push motor PWM increases linearly transition due to the bank turn maneuvering. This results
until the maximal value is reached, resulting an increased a slight altitude loss (i.e. 2m) and attitude error, which
airspeed. During the transition process, the altitude increased are caused by the low bandwidth of a typical fixed-wing
by 1m due to the increased aerodynamic lift. After the controller. These errors, in practice, are acceptable in our
airspeed reached the specified value, the flight state switched outdoor environment.
B. Manual Flight Test
As the whole vehicle is a new system whose aerodynamic
performance, robusteness and system reliability remain to
be verified, we designed a manual flight experiment for the
purpose of determining the minimal airspeed, verifying the 40
position controller and each chosen component of the whole 20

systems such as video transmitter, telemetry, and AAT. 0

160
This experiment is designed as follows: First, the hybrid 140
80
VTOL UAV takes off vertically in manual mode and lifts up 120
60
100
to about 50 m. Then, the test pilot triggers a switch command 80
40
20
to level flight mode. After the vehicle has fully transited to 60
40 0
the fixed-wing mode, the test pilot will control the aicraft 20 -20
such that it can approximately track a circle. Finally, the 0 -40
-20
hybrid VTOL UAV is switched to quadrotor mode and is -60

landed. Fig. 11. Trajectory of manual flight test

49

48 The data log for circling flight is shown in Fig. 9, Fig. 10

47 and Fig. 11. The upper figure in Fig. 9 shows the altitude
response. It can be seen that the actual altitude is well
46
regulated around the desired value, with about 2.6 m peak
45 Altitude masured by GPS to peak oscillation. This oscillation, in the same period of
Set point of altitude
44 the flight trajectory, is caused by cross wind and lead to a
0.8 1 1.2 1.4 1.6 1.8 2 2.2
similar oscillation in airspeed, as shown in the lower plot in
Fig. 9. Fig. 10 shows the roll, pitch and yaw angle of the
16.5
aircraft in manual mode where the attitude angle is directly
16 controlled by pilot. Due to the presence of cross wind, the
15.5 aircraft slowly drifts from its initial trajectory, the pilot thus
15
intentionally adjusted roll angles respectively at 1min, 1.2min
and 2min to maintain its trajectory within the proper area.
14.5 Airspeed
Fitting curve for airspeed
The resulting trajectory is shown in Fig. 11.
14
0.8 1 1.2 1.4 1.6 1.8 2 2.2
C. Fully Autonomous Flight and Flight Range Verification
As pointed out in section I, our primary goal is to
Fig. 9. Altitude and airspeed of manual flight test
develop a UAV which can take-off and land vertically and
achieve a 30 km range in a fully autonomous way. Thus,
this experiment is designed to verify whether our UAV can
50 achieve this goal.
Due to the limited flight test area, we cannot let the vehicle
0 fly straight forward for 30-kilometers. The compromised
solution is to follow a circle trajectory and calculate the
-50
0.8 1 1.2 1.4 1.6 1.8 2 2.2 distance by airspeed and endurance time. In this experiment,
we set the airspeed as 18 m/s (64.8 km/h). Then a minimal
10
flight endurance of 27.78 min is required to meet 30 km of
range. To verify this, a fully autonomous flight is performed
0
on our unmanned aerial system, including auto take-off,
-10 transition and level flight.
0.8 1 1.2 1.4 1.6 1.8 2 2.2
The log data is shown in Fig. 12, Fig. 14, Fig. 13 and Fig.
200 15. In Fig. 12, the upper plot shows the aircraft airspeed,
which oscillates in the same period of flight trajectory due
0 to the presence of cross wind. This results an oscillation
in current measurement (middle plot of Fig. 12), roll angle
-200 (upper plot of Fig. 13) and pitch angle (middle plot of
0.8 1 1.2 1.4 1.6 1.8 2 2.2
Fig. 13). Moreover, it is seen that our vehicle can fly
32 min, satisfying the required endurance (i.e. 27.78min).
Fig. 10. Roll, pitch and yaw angle of manual flight test Fig. 14 shows the positions. It is noticed that the altitude is
 22 200

20 100

18 0
16 -100
0 5 10 15 20 25 30 35
0 5 10 15 20 25 30 35
200
14
12
10 100
8
6
0
4 0 5 10 15 20 25 30 35
0 5 10 15 20 25 30 35
35
17
16 30

15 25
14
0 5 10 15 20 25 30 35
13
0 5 10 15 20 25 30 35

Fig. 14. Position of range test

Fig. 12. Airspeed, current and voltage range test

oscillating, which is, again, caused by the cross wind and

in the same period of flight trajectory. Fig. 15 shows the
resulting trajectory.

50

40

30

20 Fig. 15. Trajectory of range test

0 5 10 15 20 25 30 35

10
designed aircraft can be further improved by observing the
0 following:
• From Fig. 13 we can know that in the flight range test,
-10 the vehicle has a constant banking angle of about 33.7
0 5 10 15 20 25 30 35 deg. This means that only 100% · cos 33.7◦ = 83.2%
200 aerodynamic lift has been used to counterweigh the
aircraft’s force of gravity. The flight range for straight
level flight will be 1.2 times more.
0
• As we know, flight speed is cruicial in determining the
flight efficiency. As for our VTOL, 18 m/s may not be
-200 the optimal flight speed. Consequently, it is possible to
0 5 10 15 20 25 30 35 extend the flight range if the aircraft flys at an optimal
speed.
Fig. 13. Roll, pitch and yaw angle of range test To find the optimal flight speed, another experiment is
conducted. Similar to the second experiment, the aircraft is
set to loiter around a specified point at different test speeds.
D. Optimal Flight Speed Figure 16 shows the current of each case which oscillated
Although the autonomous flight test presented above has due to disturbance. The average current and voltage are
already realized our goal, the range and endurance of the summarized in table VII.
 12 full autonomous flight. The designed flight control system
10 16 m/s
8 consists of VTOL attitude controller, transition controller
6 and VTOL mixer based on our special aerodynamic char-
4
2 acteristics and mechanical design. Finally, experiments were
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
designed to determine the aerodynamic performance, verify
the controller’s stability and autonomy. As shown by these
12
10 17 m/s experiments, this prototype can meet the initial design goal.
8
6 In the future, we will optimize this prototype, in terms
4
2 of aerodynamics design and controller improvement. Actual
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 long distance experiments will be conducted to verify the
range of the communication devices. Based on this hybrid
12 VTOL UAV, several topics, such as precise landing, motion
10 18 m/s
8 planning, and 3-D mapping will be investigated.
6
4
2
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
R EFERENCES
12 [1] X. Lyu, H. Gu, Y. Wang, Z. Li, S. Shen, and F. Zhang, “Design
10 and implementation of a quadrotor tail-sitter vtol uav,” in 2017 IEEE
8
6 International Conference on Robotics and Automation (ICRA). IEEE,
4 19 m/s 2017(Accepted).
2 [2] F. Zhang, X. Lyu, Y. Wang, H. Gu, and Z. Li, “Modeling and flight
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
control simulation of a quadrotor tailsitter vtol uav,” in AIAA Modeling
and Simulation Technologies Conference, 2017, p. 1561.
[3] R. W. Prouty, Helicopter performance, stability, and control, 1995.
Fig. 16. This hybrid VTOL UAV is test in different airspeed to measure [4] R. T. Rysdyk and A. J. Calise, “Adaptive model inversion flight control
the current to determine cruising speed for tilt-rotor aircraft,” Journal of guidance, control, and dynamics,
vol. 22, no. 3, pp. 402–407, 1999.
[5] G. Flores, I. Lugo, and R. Lozano, “6-dof hovering controller design of
TABLE VII the quad tiltrotor aircraft: Simulations and experiments,” in Decision
and Control (CDC), 2014 IEEE 53rd Annual Conference on. IEEE,
A IRSPEED TEST RESULT 2014, pp. 6123–6128.
[6] D. Wu, H. Li, and S. Li, “Flight dynamics study of a small tilt rotor
Airspeed V (m/s) 16 17 18 19 uav with tail propeller,” in 16th AIAA Aviation Technology, Integration,
and Operations Conference, 2016, p. 3450.
Average Current I (A) 5.81 5.73 7.07 8.04
[7] E. Çetinsoy, S. Dikyar, C. Hançer, K. Oner, E. Sirimoglu, M. Unel,
V /I 2.75 2.97 2.54 2.36
and M. Aksit, “Design and construction of a novel quad tilt-wing uav,”
Mechatronics, vol. 22, no. 6, pp. 723–745, 2012.
[8] S. Bouabdallah, P. Murrieri, and R. Siegwart, “Design and control
of an indoor micro quadrotor,” in Robotics and Automation, 2004.
Proceedings. ICRA’04. 2004 IEEE International Conference on, vol. 5.
IEEE, 2004, pp. 4393–4398.
The expected range can be predicted by [9] D. P. Raymer, Aircraft design: a conceptual approach and Rds-
C V student, software for aircraft design, sizing, and performance set
·V =
R=t·V = · C. (12) (AIAA Education). AIAA (American Institute of Aeronautics & Ast,
I I 2006.
where R is the flight range, t is endurance time, C is battery [10] J. D.Anderson, Aircraft performance and design. McGraw-Hill, 1999.
[11] C. E. Jobe, “Prediction of aerodynamic drag.” DTIC Document, Tech.
capacity and I is the current drained from the battery. By Rep., 1984.
assuming that the battery capacity C is constant, the flight [12] (2008-11-24) 1th uas team rq-5a hunter2008 unmanned air system
range is proportional to the ratio of airspeed V to current 2008. [Online]. Available: http://slideplayer.com/slide/6286116/
[13] C. E. Jobe, “Prediction and verification of aerodynamic drag, part i:
I. As seen in table VII, the ratio V /I is maximal when prediction,” Progress in Astronautics and Aeronautics, edited by CE
the flight speed is set at 17 m/s, the corresponding flight Eugene, vol. 98, 1985.
range will be 2.97
2.54 = 1.17 higher than the case where V =
[14] L. Meier, P. Tanskanen, F. Fraundorfer, and M. Pollefeys, “Pixhawk:
A system for autonomous flight using onboard computer vision,” in
18m/s. Therefore, the flight range our aircraft can achieve Robotics and automation (ICRA), 2011 IEEE international conference
is expected to be 1.2 × 1.17 × 30 km = 42 km. on. IEEE, 2011, pp. 2992–2997.
[15] E. Zurich, “Qgroundcontrol: Ground control station for small air land
water autonomous unmanned systems,” 2013.
[16] S. Park, J. Deyst, and J. How, “A new nonlinear guidance logic
V. C ONCLUSION AND F UTURE W ORK for trajectory tracking,” in AIAA guidance, navigation, and control
In this study, a hybrid VTOL UAV was developed and conference and exhibit, 2004, p. 4900.
[17] D. Mellinger and V. Kumar, “Minimum snap trajectory generation and
tested. The design requirement was achieving a flight range control for quadrotors,” in Proc. of the IEEE Intl. Conf. on Robot. and
of 30 km. The work is based on a mini SkyHunter due to Autom., Shanghai, China, May 2011, pp. 2520–2525.
its size and payload capacity. On the basis of theoretical [18] (2017-02-11) Px4 firmware. [Online]. Available: https://github.com/
PX4/Firmware/tree/master/src/modules
calculation, selection of the propulsion system and mechan- [19] J. Roskam, Airplane flight dynamics and automatic flight controls.
ical design proceeded. A flight control system was designed DARcorporation, 1995.
to realize transition, improve flight stability and achieve

View publication stats