You are on page 1of 8

2017 IEEE International Conference on Robotics and Automation (ICRA)

Singapore, May 29 - June 3, 2017

Design and Dynamic Analysis of a Transformable HOvering Rotorcraft


(THOR)
Jun En Low, Luke Thura Soe Win, Danial Sufiyan Bin Shaiful, Chee How Tan, Gim Song Soh and
Shaohui Foong, IEEE Member

Abstract— This paper describes the Transformable HOvering


Rotorcraft (THOR), a prototype Unmanned Aerial Vehicle
(UAV) that explores a novel approach in combining the range
and speed of a horizontal flying platform with the hovering and
maneuverability of a rotor-wing. This is achieved by integrating
a tailless flying wing configuration with a single-axis rotor, or
monocopter. By maintaining full utilization of all aerodynamic
surfaces and propulsion sources in both flight modes, this
method represents the most structurally efficient approach to
achieving a cruising mode and a hovering mode on the same
frame. Using a dual servo and motor configuration, we propose
an under-actuated system that is able to achieve controllability
in 4 degrees of freedom while in its horizontal cruising mode
and in 5 degrees of freedom while in its hovering mode. In both
indoor and outdoor experiments, the UAV is able to transition
between either flight modes seamlessly and repeatedly without
the need for any additional mechanisms and actuators.
Fig. 1. Clockwise from top left: QuadRanger, Firefly6, Project Wing and
I. INTRODUCTION a monocopter developed at SUTD based on works by Fregene et al.
With the maturity of unmanned aerial vehicle (UAV) tech-
nology, designs can be divided into four distinct types; fixed-
wings, rotor-wings, flapping wings and aerostats. Fixed-wing to switch between and operate in either flight mode, as can
configurations are able to provide a quick response at long- be seen in Google’s Project Wing.
range, rotor-wings are useful in areas that are congested, However wing area on most current hybrids are still used
limited or undeveloped, flapping wings achieve remarkable exclusively for fixed-wing flight. We thus propose the fusion
levels of flight efficiency in small scale applications (micro- of a fixed-wing configuration with the monocopter rotor-
UAVs) [1] and aerostats are able to remain airborne al- wing, a nature-inspired vehicle that spins its entire body
most indefinitely. In between these are hybrid UAVs, their to achieve highly efficient flight, to increase the structural
fundamental motivation being to make a complementary efficiency of the hybrid by keeping the wing area fully
combination of two or more configurations. In this paper, utilized in both flight modes. Notable works on the concept
we look at designs that combine the fixed-wing’s range with can be found in [2], [3], [4] and [5]. Historically, the
the rotor-wing’s versatility; giving users quick, long-ranged monocopter has lost out to other forms of rotor-wings due to
capabilities in even the most obstacle rich environments. its fully rotating frame and demanding control requirements.
Probably the most straightforward approach is to fuse the However, with the shift in interest from manned to unmanned
two design types onto the same frame, an example being flight and the miniaturization of reliable microelectronic and
the QuadRanger by PX4. These hybrids use an independent actuator systems, this platform has encountered a resurgence.
system for each flight mode’s control and propulsion. On An example of a hybrid that uses it is Dzyne’s ROTORwing
a greater level of structural efficiency, some designs use an [6] where it is used for the Vertical Take-Off and Landing
additional actuator to reorient some of the propulsion or con- (VTOL) portion of its flight. This concept however, has a
trol systems during a transition step. This reduces the amount tail configuration which is only active during forward flight
of redundant components in either flight mode, an example and a de-spin motor that is used only in transition.
being the BirdEyeView FireFly6. Even more extreme, some In this paper we introduce the Transformable HOvering
hybrids use the propulsion and control systems themselves Rotorcraft (THOR) and two of its flight modes, Hovering
Mode (H-MOD) and Cruising Mode (C-MOD), with the
*Research supported by the SUTD-MIT International Design Centre objective of developing a structurally efficient hybrid UAV.
(IDC) and by the Temasek Laboratories Defence Innovation Research
Programme (DIRP) IGDSP1502041. In Section II, we establish the operational principles and the
The authors are with the Engineering Product Development Pillar, necessary design parameters of such a setup, culminating in
Singapore University of Technology & Design (SUTD), 8 Somapah a dynamic model described in Section III. In Section IV we
Road, 487372, Singapore. Email: {junen low, thura soe,
danial sufiyan, cheehow tan, sohgimsong, conduct a series of experiments to assess our approach before
foongshaohui@sutd.edu.sg} we conclude the paper with Section V.

978-1-5090-4633-1/17/$31.00 ©2017 IEEE 6389


II. OPERATIONAL PRINCIPLES AND DESIGN chord point coincident to the pitch axis, thus requiring the
ANALYSIS dimensions of the wing to be constrained to ensure that the
In designing a fixed-wing rotor-wing hybrid, we face sum A.C. of each wing does remain so.
two obstacles; the positioning of the two flight modes’
equilibrium points and the effects of rotational dynamics on
large wing areas. To solve these, we attempt to hold true to
structural efficiency by avoiding additional systems and to
instead draw solutions from the concept’s inherent strengths.
A. Flight Mode Arrangements
Like any aircraft in steady flight, both the THOR’s H-
MOD and C-MOD needs to be in equilibrium with respect
to its center of mass. Conventionally, this is solved by
strategically arranging the masses on the craft to balance
out the aerodynamic forces and torques. There is an added
Fig. 2. Important Dimensions for Wing Equilibrium
complexity to this problem on the THOR as it has two
flight modes on the same frame and their respective sources
of aerodynamic forces are different; one rotary and one To do this, we first define the parameters that are affected.
translational. There are two mass based solutions to this, the For each wing, c stands for the chord length and r stands
first is to position the flight modes such that the equilibrium for the span-wise distance. Subscript 0 represents the root
point on the body frame for both states are near equal, the chord position, subscript m for motor, subscript p for pitch
second is to implement a component that passively shifts its axis, subscript n for the nth blade element along the wing
body frame position depending on the mode of operation. and subscript R for wing tip. For each blade element, there
Considering that the fundamental equations of motion are are two aerodynamic forces acting on the A.C., a lift force
deeply rooted in the first solution and accounting for the dLn and a drag force dDn defined by:
added complexity of a significant non-rigid mass in both the 1
flight modes and the transition step, it is desirable to focus dLn = ρcn Cl U 2 drn (1)
2
on the first solution for now.
It is useful to start the equilibrium analysis from the 1
simpler flight mode, the C-MOD. In steady flight, a span- dDn = ρcn Cd U 2 drn (2)
2
wise symmetrical design can be used to maintain C-MOD
roll and yaw equilibrium (this design will also prove to where ρ is the density of air, Cl is the lift coefficient, Cd is
be useful later for the H-MOD). For pitch equilibrium, we the drag coefficient, U is the freestream velocity and dr is
attempt to reduce all the torques operating about the C- the infinitesimal width of the element. Using a linear taper,
MOD’s pitch axis and to do this, we break the craft down cn is defined to be the chord length expressed as a function
into two parts, the wings and the body. of the span-wise distance:
On each wing, the forces and torques contributing to a  
cR − c 0
pitching moment can be resolved to act on three distinct cn = rn + c 0 (3)
points, the propeller center (P.C.), the aerodynamic center rR
(A.C.) and the center of mass (M.C.). If all these points Since the C-MOD’s flight objective is to cruise long
were arranged to be collinear to one another, the pitch torque distances, ρ can be taken to be constant. Similarly, the
can be taken to be negligible. Let us set this line to be the freestream velocity can be assumed to be constant across
pitch axis and define it to stretch from wing tip to wing the wing as the craft is not expected to conduct highly agile
tip. To align the P.C. to the pitch axis, we chose a leading maneuvers. Lastly, with no wing twist in C-MOD, the lift and
edge mounted motor with its rotation axis collinear to the drag coefficient, which are functions of the angle between
wing chord. This takes advantage of the positive angle of the freestream velocity vector and blade chord (known as
attack of the wing during operation to keep the propellers the angle of attack or α), can also be taken as a constant
cleared from the ground (especially for take-off and landing). across the wing. This leaves cn to be the only span-wise
Consequently, with the front end loaded with the motor, a affected component. Thus for wing aerodynamic forces and
leading edge taper is useful as not only does it improve the masses to give zero torque about the pitch axis:
lift distribution of each wing [7] but it also balances the
 LR
mass of the motor such that the M.C. of each wing is on the
pitch axis. Finally to align the A.C., a symmetrical airfoil s · dL = 0 (4)
L0
is used. This positions the A.C of each blade element a
quarter chord away from the leading edge with negligible  
rR c(r)
aerodynamic torques [7]. However with the existence of
(cm − cp )mm + [(c − cp ) ρh] dc dr = 0 (5)
the taper, not every blade element will have its quarter r0 0

6390
Fig. 3. From left to right, H-MOD, Transition Step, C-MOD. Note how H-MOD yaw corresponds to C-MOD roll with respect to the body frame.

where s is the torque arm between the blade elements’ achieving a pitch and roll equilibrium. Also due to the con-
quarter chord points and the pitch axis, c(r) is the function tinuous rotation, two new factors need to be accounted for,
described in (3) and mm is the total mass of the motor centripetal forces and rotor-wing aerodynamics. The former
system. Note that dL can be swapped with dD in (4) to the is resolved by placing the H-MOD’s yaw axis orthogonal and
same effect. We have also approximated the blade element centered on the C-MOD’s pitch axis. This takes advantage of
thickness to be a constant, h. This is because most of the the span-wise symmetry mentioned earlier to ensure that the
prototypes that were tested used flat plate airfoils, a type inertia tensor is as close to diagonal as possible. For the latter,
of symmetrical airfoil that is easy to build and maintain. the aerodynamic forces become more complex and would
NACA airfoil-ed prototypes had their wing masses balanced require the implementation of additional tools, like the Blade
by adding small weights at the edges to compensate for the Element Momentum Theory (BEMT) [9], to determine the
aforementioned assumption. Through (4) and (5) cP and cm subsequent aerodynamic torques. Here, we take advantage of
are constrained to be: wing symmetry to simplify the problem. By placing identical
wings at a 180◦ offset to one another, we are able to cancel
c2R + cR c0 + c20 the aerodynamic torques experienced in H-MOD hover.
cp = (6)
2 (cR + c0 ) Hence, with a specific wing design and orthogonally ar-
ranged flight modes, we need only to implement a transition
ρh rR  2  mechanism that rotates the wings 180◦ from each other.
cm = c + 3cp cR − cR c0 − cp c0 − c2R (7)
6 mm 0 Consequently, we propose a structurally efficient transition
Hence given a specific wing material, total wing area, mechanism where flap control rotates the entire wing. To
wingspan and motor, (6) and (7) can be used to define the effect a transition, the craft will rotate each wing 90◦ in
position of the pitch axis and the motor relative to the rest opposite directions. This is illustrated in Fig. 3.
of the wing such that its three influential points, the P.C, the B. Reference Frame Definitions
M.C. and the A.C., are all collinear.
With the flight configurations fully defined and before
This leaves us with the body component for C-MOD
proceeding to our analysis of rotational dynamics on large
equilibrium. By ensuring its size is small relative to the wings
wing areas, we feel it is now apt to introduce a set of
and by using a span-wise symmetrical, perforated structure,
coordinate systems. First, the Earth frame XE YE ZE , where
we approximate the aerodynamic contribution of the body to
gravity(g) is defined to be g = [ 0 0 −1 ]||g||. Next,
be the drag force:
the pilot’s reference frame XP YP ZP , which is set up such
that ZE and ZP are equivalent and XE is aligned to the
DF = −kquad U 2 − klin U (8)
pilot’s front. Finally, the body reference frame xyz, which
where kquad and klin are the drag coefficients as seen in [8]. is pinned to the craft’s center of mass such that the y-axis
The force acts at the M.C. of the craft and parallel to the is equivalent to the pitch axis described in Section II.A
freestream velocity, U . With the pitch axis set via the wing and the z-axis points in the direction of C-MOD forward.
design, the components on the body are arranged accordingly The corresponding body frame rotations are thus θ, φ, and
to ensure that the body component’s own M.C. is also at the ψ. In describing velocities in the body reference, we use
midpoint of the pitch axis. v = [ẋ ẏ ż] and ω = [θ̇ φ̇ ψ̇]. These conventions are
With the C-MOD equilibrium resolved, we can now deter- illustrated in Fig. 3. Note that while the terms roll, pitch and
mine a position for the H-MOD. Due to its continuous rota- yaw corresponds to the C-MOD x, y and z-axis, H-MOD
tion along its yaw axis, steady H-MOD flight is dependent on roll and pitch corresponds to the XP and YP -axis.

6391
C. Effects of Propulsion and Body Rotational Configurations introduce a relationship between ωmotor and motor forces
Using the large wing areas associated with fixed-wing and torques. We use the approximation as seen in [11]:
flight for rotor-wing maneuvers produces a pair of unusual
2
design problems. Firstly, gyroscopic precession while in H- F = kF ωmotor (11)
MOD produces a significant wing twist while in operation.
Secondly, the craft is also subject to pronounced cases of 2
M = kM ωmotor + Imotor ω̇motor (12)
asymmetric blade loading.
In the case of gyroscopic precession, the rotating mass where kF is the motor lift constant and kM is the motor drag
of the motor system produces a torque on the rotating body constant. With both motors rotating in the same direction, the
frame as described by the equation: torque term, M , would cause C-MOD to roll. However, this
is preferable over an unbalanced H-MOD wing twist as a
τgyro = Imotor ωmotor θ̇ (9) simple flap correction can be used to resolve the former.
where Imotor is the inertia tensor of the motor, ωmotor . Asymmetrical blade loading [12] is a manifestation of non-
Given the large wing area on the THOR, this produces a uniform freestream velocities relative to the motor system’s
pronounced twist along the wing while the craft operates in rotating propeller. The result is a shift in the P.C. of each mo-
its H-MOD, thus changing the effective α across the wing. tor depending on the inflow distribution. Both THOR flight
This in turn affects the lift and drag forces experienced along modes experience some form of asymmetrical blade loading.
each wing. To estimate the change in α, we use a linear In H-MOD, the θ̇ term will cause the inflow distribution to
approximation based on (9): be biased towards the outer side of the motor system. In
C-MOD, the phenomena manifests itself when the motors
α = ktwist τgyro y (10) are operated at a high angle of attack, a common situation
given that flap control rotates the entire wing along with
where ktwist is the twist coefficient and y is the spanwise the motor. Depending on the motor rotation, this will either
distance from body frame origin. It can be seen that to reduce shift the distribution away from or towards the body origin.
changes in α, it would be useful to keep θ̇ constant during H- With a Type A Rotation Configuration, we use the following
MOD operation, much like in helicopter control [10]. It can equation to approximate the shift in the body frame:
also be seen that, depending on the directions of rotation of
ωmotor and θ̇, the effects on α can be exacerbated or reduced, ⎡ ⎤
0
as illustrated in Table I.
δP ABL = −δSBL = ⎣ kf lap α + krot θ̇ ⎦ (13)
TABLE I 0
ROTATION C ONFIGURATIONS
where δP ABL is for the port and δSABL is for the starboard
Config. Craft Port Starboard Precession motor. kf lap and krot are defined to be the loading coeffi-
Rotation Motor Motor Effect cients. Similar to the motor torque term, the flight controller
(θ̇) (ωmotor ) (ωmotor ) (α) will also need to correct for this phenomenon, this time via
A Positive Positive Positive Positive
B Positive Positive Negative Unbalanced
a motor thrust correction.
C Positive Negative Positive Unbalanced
D Positive Negative Negative Negative
E Negative Positive Positive Negative
F Negative Positive Negative Unbalanced
G Negative Negative Positive Unbalanced
H Negative Negative Negative Positive

The direction of rotation is taken relative to their respective


directions of thrust. Rotational configurations that are ’Un-
balanced’ are undesirable as one wing will have a positive
twist while the other will have a negative twist, thus requiring
the flight controller to do additional compensations during H-
MOD flight. For the remaining configurations (A, D, E and
H), while the magnitudes of ωmotor might defer between
the motors, the twist effect is still in the same direction for
both wings, thereby reducing the relative differences in blade
twist. Within these, Types A and H are the most appealing
as a positive increase in angle of attack on both wings will
give greater lift.
However, this design choice for H-MOD has an unde- Fig. 4. P-Factor Effects on a CCW rotor
sirable impact on the C-MOD. It is useful at this point to

6392
III. DYNAMIC MODEL
For the dynamic model of this craft, we choose the body
frame as the workspace for two reasons. Firstly, the on-board
Inertial Measurement Unit (IMU) outputs data relative to the
body frame. Secondly, the aerodynamic and motor forces
and torques are related to the velocities acting relative to the
craft’s wings, which requires only one rotation about the y-
axis to be translated to the body frame. For motor forces,
the angle of interest is γ which is defined to be the angle
between the wing chord and the body frame (also known as
the flap input). For aerodynamic forces, we define the angle
between the freestream velocity (U ) and the body frame to
be β. The lift force and drag force acting on each wing is
defined to act perpendicular and parallel to the freestream
velocity vector respectively. Combined, these angles form Fig. 7. H-MOD Free Body Diagram
the angle of attack (α) as illustrated in Fig. 5.
It is important to note that the magnitude and location of
the lift and drag forces on the wing are calculated differently
depending on the flight mode. C-MOD uses the equations (1)
and (2) while H-MOD uses definitions described in [3] with
blade twist (10) implementation from [13]. For simplicity
in notation, we combine these forces for each mode into
the vector A for each wing and define their point of action
Fig. 5. The wing relative angles α, β and γ
relative to body origin using the vector sP W for the port
For the effects of gravity on the body frame, we use the wing and sSW for starboard. Likewise, we define the position
equation: of the motors as the vectors sP M and sSM . Hence, we have
the equations of motion:
W = mRg g (14) 1
v̇ = [FP + FS + AP + AS + W + DF ] (15)
m
where m is the mass of the craft and Rg is the rotation matrix
linking the body frame to the world frame.
With these, we can establish the Free-Body Diagrams as ω̇ = I −1 [(sP M + δP ABL )FP + (sSM + δSABL )FS
seen in Fig. 6 and Fig. 7. As mentioned in Section II.A, each +MP + MS + sP W (AP ) + sSW (AS )] (16)
wing in H-MOD experiences centripetal forces that cancel
IV. EXPERIMENTAL RESULTS
each other out, hence their omission in Fig. 7. However, it is
also acknowledged that these forces produce a phenomenon Given the atypical nature of this hybrid, there is con-
known as blade flap [9] about the z-axis, the implementation siderable work to be done to improve the dynamic model
of which has been deferred to future work. Lastly, we assume and develop an effective controller for flight. In this paper,
the H-MOD has negligible DF as the airflow velocity due we present two experiments towards this direction; the first
to θ̇ this close to the x-axis is small. explores effects of gyroscopic precession as described in
Section II.C and the second explores steady flight. The
prototype used for these is shown in Fig. 8.

Fig. 6. C-MOD Free-Body Diagram.


Fig. 8. THOR vIII Prototype in H-MOD Configuration

6393
In this prototype, a direct drive mechanism is used for flap
control. As this puts a lot of stress on each actuator, a rotary
bearing is added to reduce bending load. Also added onto the
system to aid testing is an SD card reader to log flight data.
A close up of the electronics can be seen in Fig. 9 while its
specifications can be found in Table II.

Fig. 10. Force/Torque ground setup to observe effects of gyroscopic


precession in H-MOD

Fig. 9. Close Up of THOR vIII Electronics

TABLE II
THOR V III S PECIFICATIONS L IST

Flight Controller Teensy 3.2


IMU SENtral MPU9250 + BMP280
ESC Afro 20A
Motor Emax RS2205-2600KV Fig. 11. Force/Torque sensor data for Rotation Configurations A, B and
D
Servo Futaba BLS173SV
Power 1300mAh 25C 3S LiPo
Airfoil Flat Plate
Wing Material Plywood and Monokote to the wings thereby affecting the amount of lift generated.
Wingspan 105cm It should be noted that while the difference is noticeable,
Root Chord Position 10cm
Airfoil Thickness (h) 3mm the effect on the angle of attack must be small as the Type
Overall Length 16cm A configuration did not twist the wing into a stalling angle
Height 9cm of attack. Interestingly, the Type B configuration, which is
Weight (m) 532g supposed to give unbalanced angles, is only marginally less
Rotation Config. A
effective at generating lift than the Type A. This could be due
to an inefficiency of the second wing operating in the wake
of the first or it could be because the negative twist causes
A. Gyroscopic Precession Experiment the motor to dedicate more of its thrust into the generation
To analyze the wing twist effect due to gyroscopic pre- of θ̇ thereby increasing lift via the U term. Also of note is
cession, we mounted the craft onto a ATI 6-axis Capacitive that the marginal difference in lift does not give Type B any
Force/Torque (F/T) Sensor. An MTO-050 rotary bearing is form of merit as the instability of the configuration predicted
used to link the craft to the sensor such that the craft is free in Section II.C could not be observed with the craft fixed in
to rotate in H-MOD hover, as can be seen in Fig. 10. The the x-y plane.
rotation configurations as described in Table I were tested
with a constant flap input of 10◦ and throttle input of 50%. B. Flight Experiments
These correspond to the optimal angle of attack for flat plate As a first step to developing an effective flight controller
airfoils and the expected throttle input to achieve hover. for the THOR, we aim to achieve steady flight in both
The results for a Positive (Type A), Unbalanced (Type B) modes using the same frame. To do this, we implemented
and Negative (Type D) effect on α are illustrated in Fig. a Proportional-Integral controller for C-MOD and a Propor-
11 where the raw data was sampled at 60Hz. To remove tional controller for H-MOD. A magnetometer was fixed onto
noise, a moving mean (MM) of 15 data points was used. The the pilot to determine his reference frame, which was then
results corroborate the theory that precession can give twist transmitted to the craft along with a throttle, roll, pitch and

6394
yaw input via a remote controller. The feedback control loop around 6 revolutions per second while in hover. We believe
of the craft is illustrated in Fig. 12. With this setup, three the 12 Hz component to correspond to the blade flapping
tests were conducted; an outdoor C-MOD test, an indoor phenomena mentioned in Section III.
H-MOD test and an indoor transition test.

RC Orientation Desired Yaw,


Desired Throttle Motor
RC Throttle
Controller
RC Pitch  
   
THOR
RC Roll RC Filter 
    
RC Yaw Flap
Desired Pitch, Controller
RC Mode
Desired Roll,

  
Heading
Hardware
Fusion


 

Fig. 12. THOR Feedback Control Diagram

Given that the primary use of the flight mode is to cruise


long distances, our first test had the craft flown along a Fig. 14. Roll Response in H-MOD hover
linear course after which pitch control was input by the pilot
to dive and climb the craft. The results of this experiment
is illustrated in Fig. 13 where the C-MOD can be seen to
actively maintain a pitch, roll and yaw equilibrium given no
aggressive inputs.

Fig. 15. Raw Roll Data and Single-Sided Amplitude Spectrum computed
via discrete Fourier transform

For the transition tests, we wanted to verify if the actuators


were strong enough and fast enough to execute a transition in
mid-flight. To do this, we brought the craft up to an altitude
of about 2m in either mode and attempted a transition,
Fig. 13. Pitch Response in C-MOD forward flight producing some interesting results. In the transition between
C-MOD to H-MOD, it was found that by putting the C-MOD
The second test was set up to assess the craft’s ability into a climb, the craft is able to reliably maintain altitude
to maintain and control hover flight. Much like the single while switching its wings into its H-MOD configuration. We
wing monocopters in [3], [4] and [5], the THOR’s dual wing suspect that, given strong enough motors, the craft operates
configuration was able to passively maintain a near stationary akin to a bi-copter during the transition step after-which the
hover without additional pilot input. In terms of control, rotating wings take over as the primary source of lift. In tran-
we present the roll data only since pitch and roll control sition from H-MOD to C-MOD, the initiation of an upwards
outputs are identical relative to the body frame. Sampled facing C-MOD was once again useful in maintaining craft
at 100Hz, a moving mean of 12 data points was used to altitude during transition. In fact, the motors were strong
smoothen the data. When compared to the roll input and enough to almost instantaneously break the craft out of H-
magnetometer orientation data, the craft was found to be MOD rotation and push the craft into translational flight.
responsive to the pilot’s input. These are illustrated in Fig. Interestingly, the craft was found to consistently transit from
14. To understand the cyclic response of the raw roll data in H-MOD to C-MOD with the z-axis pointing upwards. As
Fig. 14, we computed its Fast Fourier Transform, illustrated is illustrated in Fig. 16, the craft always transits to C-MOD
in Fig. 15. The 6 Hz frequency corresponds to the craft’s with green side up. We suspect this to be because the positive
angular velocity which the on-board IMU corroborated to be angle of attack of the wings in either mode requires one

6395
wing to rotate less than the other wing during a transition. [5] K. Fregene and C. L. Bolden, “Dynamics and control of a biomimetic
The motor on this wing therefore pushes the craft into the single-wing nano air vehicle,” in Proc. IEEE ACC ’10, Baltimore, MD,
June 30 - July 2 2010, pp. 51–56.
forward direction while the other wing is still in the process [6] R. A. G. Mark Allan Page, Matthew Robert McCue, “Long endurance
of executing its own transition, thereby tipping the craft the vertical takeoff and landing aircraft,” U.S. Patent 8 991 751, Mar. 31,
right side up. 2015.
[7] J. John D. Anderson, Fundamentals in Aerodynamics. Singapore:
McGraw Hill, 2011, ch. 4.7, 5.3.
[8] M. Bangura and R. Mahony, “Nonlinear dynamic modeling for high
performance control of a quadrotor,” in Proc. AARA ACRA ’12,
Wellington, New Zealand, Dec.3 - 5 2012, pp. 115–124.
[9] J. G. Leishman, Principles of Helicopter Aerodynamics. New York,
NY: Cambridge University Press, 2008, ch. 3.
[10] R. W. Prouty, Helicopter Performance, Stability, and Control. Mal-
abar, FL: Krieger Publishing Company, 2005, ch. 9.
[11] T. Luukkonen, “Modelling and control of quadcopter,” Aalto Univer-
sity, School of Science, Espoo, Finland, Independent research project
Fig. 16. Transition from H-MOD to C-MOD in applied mathematics, Aug. 2011.
[12] F. A. Administration, Flight Training Handbook: Ac 61-21A. Okla-
homa City, OK: Aviation Book Co, 1980, ch. 17.
[13] Y. H. P. Jun En Low and S. Foong, “Analysis of wing twist effects
V. CONCLUSIONS on hover flight dynamics of a single rotor aerial craft,” in Proc. IEEE
AIM ’16, Banff, Canada, 12 - 15 2016, pp. 323–328.
In this work we present a hybrid UAV testbed aimed at
achieving maximum structural efficiency by using the same
aerodynamic surfaces and propulsion sources for fixed-wing
and rotor-wing flight. We establish a design that allows these
two systems to coexist and complement each other on the
same frame without the implementation of any additional or
single mode only systems. Our approach leverages on physi-
cal design parameters to simplify the equilibrium problem of
both flight modes while also accounting for the craft’s unique
flight conditions. We verify these by developing a prototype
with a feedback controller. Experimental data shows that the
craft is able to maintain its equilibrium in both modes.
Given that the craft borrows its concepts from two dif-
ferent fields of aerodynamics, we believe the system gives
strong justification for developing new tools for analysis and
control that would have uses beyond this particular craft. Our
current research work is focused on three fronts; the first is
further verification and improvements to the model of the
system’s dynamics, the second is to implement additional
design parameters that improve the craft’s response to pilot
input and lastly, to implement a more elaborate controller for
both flight modes and the transition step.

ACKNOWLEDGMENT
This work was funded jointly by the SUTD-MIT Interna-
tional Design Centre and the Temasek Laboratories@SUTD,
Singapore. The authors would like to thank Kris Winer for
providing the IMU for this project.

R EFERENCES
[1] T. J. Mueller, Ed., Fixed and Flapping Wing Aerodynamics for Micro
Air Vehicle Applications, ser. Progress in Astronautics and Aeronau-
tics. Reston, VA: American Institute of Aeronautics and Astronautics,
Inc, 2001, vol. 195.
[2] A. Kellas, “The guided samara: Design and development of a control-
lable single-bladed autorotating vehicle,” M. Sc. thesis, Massachusetts
Institute of Technology, Boston, MA, Aug. 2007.
[3] J. Houghton and W. Hoburg, “Fly-by-wire control of a monocopter,”
Massachusetts Institute of Technology, Boston, MA, Experimental
Projects II Tech. Rep., May 2008.
[4] D. J. P. Evan R Ulrich and J. S. Humbert, “From falling to flying:
the path to powered flight of a robotic samara nano air vehicle,”
Bioinspiration & Biomimetics, vol. 5, no. 4, p. 045009, Nov. 2010.

6396

You might also like