1

•A Novel Actuation Strategy for an Agile Bio-

inspired FWAV Performing a Morphing-coupled
Wingbeat Pattern
Ang Chen, Bifeng Song, Zhihe Wang, Dong Xue, Kang Liu

features regarding their forelimbs [Fig. 1(b)] [7], [8].

Abstract—Flying vertebrates exhibit sophisticated wingbeat Specifically, they all have an arm wing that comprises the
kinematics. Their specialized forelimbs allow for the wing humerus and forearm (radius and ulna) and a hand wing
morphing motion to couple with the flapping motion during their spanned by manus and other structures (feathers or membranes).
level flight. Previous flyable bionic platforms have successfully
applied bio-inspired wing morphing but cannot yet be propelled
To maintain rigidity and lightness, the wing articulation
by the morphing-coupled wingbeat pattern. Spurred by this, we constraints the forelimb motion primarily to those required
develop a bio-inspired flapping-wing aerial vehicle (FWAV) during flight [9]-[13]. Since bats can bend and stretch their
entitled RoboFalcon, which is equipped with a novel mechanism fingers, bird wings have fewer joints and less independent
to drive the bat-style morphing wings, performs a morphing- controllable degrees of freedom (DOFs) than bat wings. Due to
coupled wingbeat pattern, and overall manages an appealing flight. their wing skeletal system, birds can automatically couple
The novel mechanism of RoboFalcon allows coupling the
morphing and flapping during level flight and decoupling these
extension and flexion of their wing joints, making elbow and
when maneuvering is required, producing a bilateral asymmetric hand extend together and simplifying feathered wing morphing
downstroke affording high rolling agility. The bat-style morphing [13], [14]. In analogy, both these flapping flyers can tuck and
wing is designed with a tilted mounting angle around the radius at extend their wings on a unilateral side, which morphs the wing
the wrist joint to mimic the wrist supination and pronation effect shape and results in the span and area change [2], [15].
of flying vertebrates’ forelimbs. The agility of RoboFalcon is Span and area reduction of upstroke vary for different species
assessed through several rolling maneuver flight tests, and we
demonstrate its well-performing agility capability compared to
and flight velocities. Birds sweep their hand-wing backward
flying creatures and current flapping-wing platforms. Wind during upstroke while their arm-wing remains partially
tunnel tests indicate that the roll moment of the asymmetric extended, benefiting from independent overlapping feathers
downstroke is correlated with the flapping frequency, and the [15]-[18]. In contrast, microbats reduce their arm-wing span
wrist mounting angle can be used for tuning the angle of attack while keeping the hand-wing membrane relatively stretched,
and lift-thrust configuration of the equilibrium flight state. We and megabats bend fingers to retract the hand-wing [19], [23].
believe that this work yields a well-performing bionic platform
and provides a new actuation strategy for the morphing-coupled
It is commonly believed that birds and bats exhibit different
flapping flight. upstroke kinematics in level flight [18], but researchers have
also indicated that some bat species apply a hand-wing back-
Index Terms—Biomimetics, biologically-inspired robots, sweep which is similar to that of birds during upstroke when
mechanism design, flapping wing aerial vehicle. they fly faster [24]. Both birds and bats exhibit highly
specialized wrist osteology, making the hand-wing execute
supination during the upstroke and prevent hyperpronating
I. INTRODUCTION during downstroke by exploiting an interlocking mechanism

B IRDS and bats can change their wing planform, not only
during gliding but also in downstroke and upstroke phases
during the level flapping flight. Research suggests that birds
within their wrist osteology [25]-[27]. Flying vertebrates also
use wing morphing and wingbeat kinematics to enhance their
flight performance and agility. For instance, an extended wing
and bats reduce their wingspan and area during the upstroke by is advantageous for slow gliding and turning, while a swept
rapidly withdrawing the wings towards the body and increase wing is suitable for fast gliding [28]. Since bat wings are
these parameters during downstroke [Fig. 1(a)] [1], [2]. This proportionally heavier than bird wings [29], bats change wing
wingbeat kinematics presumably reduces drag and enhances inertia to execute falling and landing maneuvers, while birds
aerodynamic efficiency during level flight [3]-[6]. rely predominately on redirecting aerodynamic forces for low-
Despite the wing’s structure details are different, flying speed maneuvers [30]. To initiate a rolling maneuver, birds
vertebrates (including pterosaurs) share several standard generate bilateral asymmetries in the wing morphing and wing

This work was supported by National Natural Science Foundation of China, [email protected]; [email protected]; [email protected];
Grants No.11902103 and 11572255. (Corresponding author: Dong Xue.) [email protected]).
The authors are with the School of Aeronautics, Northwestern Polytechnical
University, Xi’an, Shaanxi 710072, China (e-mail: [email protected];
 2

trajectory to produce roll moment [30], [31]. This asymmetric Nevertheless, applying these wingbeat kinematics to a flyable
morphing coupled with wingbeat gives birds the high agility to platform imposes a significant engineering challenge for the
perform extreme maneuvers for foraging and to escape from mechanism and actuator design. Indeed, the wing extended and
predators. Both birds and bats can achieve their body tucked motion must be actuated at the same frequency as the
reorientation within a few wing-beats [29], [30], e.g., in the flapping motion, which is required for relatively high power
most extreme maneuver, a hummingbird can do a 180° yaw turn and high-speed actuators. However, bio-inspired flapping
even within four wingbeats [32]. wings need to be lightweight to reduce the moment of inertia
Inspired by the wing anatomical mechanism and versatile for high-speed flapping, which brings restrictions to equip
flight performance of birds and bats, many bionic robots and proper actuators for wing morphing coupled with flapping.
aircrafts have been developed to mimic the wing morphing Here, we develop a flyable bio-inspired flapping-wing aerial
motion of those airborne creatures. Some cases have vehicle (FWAV), the RoboFalcon, equipped with a specially
successfully improved flight agility and maneuverability by designed novel mechanism that achieves the wingbeat
using morphable articulated wings with membranous or kinematics of birds during level flight with a pair of bat-style
feathered surfaces [15], [33]-[38]. Unlike their natural membranous morphing wings [Fig. 1(c)]. Specifically, this
counterparts, most of those works are validated on conventional platform extends its wings during the downstroke phase to
platforms, which used propeller for thrust modulation or generate lift and thrust and tucks the wings in the upstroke
elevator/rudder for attitude control [15], [33]-[35]. While some phase. The proposed mechanism couples wing flapping motion
flapping-wing designs have also been used, these platforms with wing morphing motion and allows the servo to actuate the
actuate the wing morphing independently, which is not wing morphing motion during downstroke by briefly
associated with wing flapping [36]-[38]. decoupling morphing from flapping when maneuvers are
However, for flying vertebrates, the change of wingspan and required. The wrist joint of the morphing wing is designed with
area is coupled with wingbeat, because as already discussed, a pitch-up mounting angle to imitate the wrist supination of the
they use different wingspan and area for up- and down-strokes. bird and bat wing for tuning passive twist of the wing during
Current bio-inspired platform designs lack this ability for a the downstroke. Our design results in a well bionic-performing
wing-extended downstroke and a wing-tucked upstroke with FWAV platform with high rolling agility, as RoboFalcon
bio-inspired morphing wings, while investigations of this manages 90 degrees rolling maneuver within 2.5 wingbeat
wingbeat pattern for bionic robots or aircrafts are usually cycles. We validate RoboFalcon’s rolling agility by performing
limited to wind tunnel tests and simulations [39]-[42]. several flight experiments, where each experiment rolling

Fig. 1. The wingbeat pattern of avian and the forelimb anatomy of flying vertebrates inspired the design of a bionic FWAV. (a) Snapshot of a peregrine
falcon during level flight. The wing is tucked in upstroke and extended in the downstroke. (b) Wing anatomy of pterosaur, bat, and bird, exhibiting analogous
three-link structures. (c) The RoboFalcon FWAV in flight, as inspired by flying vertebrates, applying a novel actuation strategy for achieving the wingbeat
pattern of the bird in (a).
 3

maneuver is initiated with a one-shot bilateral asymmetric [Fig. 2(c)]. The flight controller placed in the body frame is an
downstroke flapping. We also analyze the flight characteristics autopilot (PixHawk mini) equipped with an integrated inertial
of four different wrist mounting angles and measure the roll measurement unit (IMU). Other accessories include an RC
moment of the bilateral asymmetric downstroke via wind tunnel receiver for control command input, a power management
tests. The results indicate that the wrist mounting angle could module for power input measuring, and a pitot tube for airspeed
affect the angle-of-attack and lift/thrust magnitude of the sensing [Fig. 2(d)]. With this configuration, it is possible to
equilibrium flight state, suggesting that birds’ and bats’ wrist record on a memory card the attitude angle and angular velocity
supination/pronation may play a similar role in level flight. in the 3-axis, the airspeed, and the power consumption during
Results also highlight that the roll moment generated by the the flight. RoboFalcon is powered by a Li-Po battery (3s
850mAh) and equipped with a brushless DC motor (SunnySky
asymmetric wingbeat pattern is correlated with flapping
X2212 KV980) to drive the flapping mechanism [Fig. 2(d)].
frequency, affording high roll control efficiency. Overall, we
believe that this work provides a new actuation strategy for
III. BIO-INSPIRED MORPHING WING
morphing-coupled wingbeat patterns taking into account lateral
control, and suggests a well-flying bionic platform for To simplify the structure and allow for a lightweight design,
investigating the flapping flight for flying vertebrates. we choose for the bio-inspired morphing wing design the bat-
style membranous wing structure over the bird-style feathered
II. PLATFORM DESIGN (Fig. 3). Despite naturally bird wings are proportionally lighter
than bat wings [15], [29], the artificial version of bat-style
To reproduce the actual flight effect of the flying creature and morphing wings affords a lighter design than bird-style wings
validate the flight capability of the mechanically synthesized because artificial feathers made from carbon or glass fiber are
wingbeat kinematics, the design of the flight platform is better heavier and less robust than real feathers [34].
referenced to medium- or large-sized birds which exploit this Our bat-style morphing wing’s inner section (arm wing)
flapping pattern during level flight. Our RoboFalcon FWAV comprises two main spars hinged to each other, representing the
platform is generally based on the morphological parameters of humerus and radius [Fig. 3(a)]. These spars are constrained by
the peregrine falcon (Falco peregrinus), which is designed with several linkage systems ensuring the extension and flexion of
a maximum wingspan of 1.2m and a total flight weight of 600g the arm that are determined by the linear distance between the
[Fig. 2(a)]. RoboFalcon is equipped with a conventional tail shoulder and the proximal end of lever 1 [Fig. 3(a), marked with
made from foam board that is strengthened utilizing carbon red two-way arrows]. Such design results in a one-DOF
fiber rods, with the latter being embedded in the leading edge sophisticated multi-link mechanism analogous to the wing
of the tail [Fig. 2(b)]. The tail involves only an elevator designs in [37], [38]. These spars are made by the same
providing flight pitch control. The body frame is made from technique as the body frame, and their junction forms a robust
2mm carbon fiber composite boards cut into specifically shaped carbon fiber revolute joint, the elbow joint [Fig. 3, (b) and (c)].
pieces utilizing CNC machinery to form a 3D puzzle structure The wing’s outer section (hand wing) is spanned by three

Fig. 2. RoboFalcon platform design and subsystem configuration. (a) Top view of RoboFalcon with the wings fully extended at level position. The wingspan
is designed concerning the maximum wingspan of the peregrine falcon [60], while the total weight is slightly lower than average for this species, which
enables a lower wing loading. (b) The tail made from foam board has a conventional elevator for pitch control, and a vertical stabilizer achieves the yaw
stability without a rudder. (c) The CNC-cut carbon fiber parts form the RoboFalcon’s body frame. (d) Avionics, sensors, and power transmission mechanism
mounted on the body frame.
 4

fingers, with the outermost being the strongest one forming the The anterior ends of these fingers are hinged to the distal end of
leading edge and the other two supporting the wing surface. The the radial spar. More specifically, the finger on the leading edge
fingers are designed to curve downward to help create a positive is connected to the radial spar with a revolute joint [wrist joint,
curvature airfoil and generate as much as possible lift [Fig. 3(b)]. Fig. 3(d)], while the other two are connected with ball-and-

Fig. 3. Bio-inspired morphing wing with the bat-style layout and other bionic details. (a) Top view of a fully extended wing (left) and tucked skeleton (right).
The morphing motion is driven by lever 1, denoted in red. (b) A rear view of a fully extended wing skeleton along with an arched finger is shown below. (c)
Details of the elbow joint hinge. (d) Enlarged top view of the wrist joint. Two of the fingers are fully actuated with pin-slot joints in morphing motion. (e)
The elbow joint is designed above the wing plane that helps to increase the camber of the membrane-covered arm wing, which is not easy to manage an
accurate aerodynamic profile. The 3D printed knuckle is tilted by a certain angle θ around the radial spar at the distal end, which allows the rotation axis of
the wrist hinge to be tilted relative to the wing plane and helps the out-of-plane motion of the hand wing. (f) The 3D-printed leading-edge support parts help
maintain the profile of the triangular area, in (a), spanned by the humeral and radial spar, which plays the role of propatagium.
 5

socket joints. These two fingers are entirely constrained by the radial spar. Hence, we can fine-tune the passive torsion of the
link rod [rod 3, Fig. 3(a)] that is posterior to the radial spar with hand wing by adjusting the mounting angle θ of the 3D printed
pin-slot joints [Fig. 3(d)], allowing the fingers to be actuated by knuckle of the wrist joint for optimal aerodynamic forces (as
the extension and flexion motion of the arm wing. thoroughly discussed later). Since the torsional stiffness of the
Given that bats and birds can both change their wing camber, radial spar remains the same, the larger θ, the weaker the
some researchers have indicated that these flying creatures passive torsion during the downstroke.
adjust their proximal wing camber by a flexible wing membrane
called the propatagium [Fig. 1(b)] combined with a change in IV. FLAPPING & MORPHING COUPLING MECHANISM
elbow joint posture [43]-[45]. Inspired by how bats and birds
A. Conical Rocker Mechanism
change their wing camber, we purposely place the elbow joint
at the highest point of the wing surface to help achieve a high To achieve the coupling linkage of wing flapping and wing
inner section camber [Fig. 3, (b) and (e)]. The wrist joint [Fig. morphing, we design a mechanism (Fig. 4) that allows flapping
3(d)] has only one degree of freedom, prohibiting wrist and morphing to be actuated at the same frequency. The phase
supination and pronation. Given that such supination and of the flapping motion can be 90 degrees ahead of the periodic
pronation exist in birds and bats, we circumvent this problem morphing motion, which enables the wing to extend during the
by an adequately adjusted wrist mounting angle, as mentioned downstroke and to tuck during the upstroke.
in the later text [46], [47]. This mechanism is named the Conical Rocker Mechanism
We construct the wrist joint with a pitching-up mounting (CRM). The mechanism uses a rocker arm [denoted by blue in
angle by attaching a 3D printed knuckle to the distal of the Fig. 4(a)] to convert the rotation motion of a gear [denoted by
radial spar [Fig. 3(e)]. The 3D-printed knuckle tilts at a certain orange in Fig. 4(a)] into the flapping motion of the humeral spar.
angle θ around the radial spar, and thus the rotation axis of the A shaft (shaft C) eccentrically hinged to the gear with a ball-
wrist joint is not perpendicular to the wing plane. This and-socket joint act as an oblique crank [denoted by green in
arrangement results in a pitching-up hand-wing that can Fig. 4(a)]. The shaft C is confined in a shaft hole of the rocker
partially offset the passive wrist pronation (hyperpronating) due arm and is only allowed to rotate and slide with respect to the
to the aerodynamic load on the wing during the fully extended rocker arm [Fig. 4(b)]. The rocker arm and the humeral spar are
downstroke. Since the wing skeleton is retracted during the connected by a shorter shaft, while another longer shaft,
upstroke, the pitching-up wrist joint causes the hand-wing to oriented at 90° to the shorter one, is restrained by the body
bend downward even more, achieving the same supination and frame. The cross shaft [denoted by red in Fig. 4, (a) and (c)]
flexion movement that birds and bats do in upstrokes [25], [26]. formed by these two shafts restrains the motion of the rocker
The elastic membranes have been successfully applied to bat-
and the humeral spar. As the gear rotates, shaft C drives the
inspired wings [37], [38], however, during a constantly
rocker arm to sway in a conical trajectory [Fig. 4(d)], while the
morphing motion, the elastic damping of these materials
imposes extra energy consumption. Instead, bats actively center of the cross shaft is located right at the apex of the cone.
change their membrane stiffness to keep the wing surface flat As actuated by the rocker, the cross shaft oscillates around the
during the morphing process [48]. Considering the difficulty of longer shaft causing the humeral spar to rotate up and down,
fabricating the flexible membrane materials that can actively affording wing flapping.
and rapidly change the stiffness over a large scale, we cover the This mechanism has many advantages in adjusting flapping-
wing skeleton with the unstretchable polyester fabric as wing wing parameters such as amplitude, the tilt angle of the stroke
membrane to avoid burdening the morphing motion actuator plane, and dihedral angle. For instance, the flapping amplitude
[Fig. 3(a)]. This arrangement may cause unrestrained collapse is defined by the apex angle of the cone, which is described by
and wrinkling of the wing surfaces during skeleton retracting R
and ensure a solid and stable surface when the wing is fully Φ  2 arctan (1)
H
extended. This collapse and wrinkling do not highly impact the where Φ is the flapping amplitude angle, R the radius of the
flight (it may cause some extra drag, but RoboFalcon is still crank (distance between the rotation axis of the gear and the
flyable), as discussed later. Two lightweight 3D printed parts ball-and-socket joint of the shaft C), and H the height from the
are mounted in front of the humeral and the radial spar to
rotation plane of the crank to the center point of the cross shaft.
support the front wing membrane, representing the propatagium,
Since shaft C is allowed to slide in the rocker arm, we can
to form a relatively rigid leading edge [Fig. 3(a), 3(f)]. The 3D-
printed parts are designed to preserve the aerodynamic shape of change H by moving the center point of the cross shaft laterally,
the leading edge and avoid interference when the skeleton is thus changing the flapping amplitude. Similarly, because shaft
retracting. Additionally, a 3D printed rib [Fig. 3(f), rightmost] C is hinged to the gear by a ball-and-socket joint, the cross shaft
hinged to the elbow joint helps support the wing membrane in is allowed to move upward or downward to force the rocker arm
the gap between the two 3D printed parts [Fig. 3(a)]. to sway in an oblique conical trajectory, thus changing the
As a result, the raised elbow and twisted wrist help create a dihedral angle. Moreover, tilting the longer shaft of the cross
highly cambered membranous morphing wing with a rigid but shaft around the lateral or the vertical axis causes a change in
deformable leading edge. The wing can be passively twisted the tilt angle of the stroke plane or the wing to sweep back or
under aerodynamic loads to generate lift and thrust during the forth. By adjusting the position of the cross shaft relative to the
downstroke. The two ends of the link rod 3 [Fig. 3(a)] are body frame with appropriate actuation methods, we can adjust
equipped with ball-and-socket joints, which allows the passive these parameters dynamically without interfering with flapping.
torsion of the outer section skeleton to depend primarily on the
 6

These advantages of CRM could help develop the insect- or morphing and flapping are coupled, wing morphing cannot be
hummingbird-inspired FWAV. However, it should be noted used independently for attitude and maneuvering control, as
that in this article, we focus more on CRM’s feature of birds and bats do.
achieving the coupling motion of flapping and morphing.
B. Morphing decoupler
To ensure the wing is fully extended during the downstroke
and the wing morphing motion can be controlled independently
while maneuvers are required, we introduce a morphing
decoupler to meet these requirements [Fig. 5]. The decoupler
comprises three sliders: motor input slider (MIS), servo input
slider (SIS), and output slider (OS) [Fig. 5, (a), (b), (c)]. All
these sliders are constrained to allow only linear sliding along
the longer shaft of the cross shaft. The MIS linked with the
rocker arm is powered by the BLDC motor and moves
reciprocally. The SIS is actuated (or locked) by a servo mounted
at the posterior end of the longer shaft through a link rod [Fig.
5(a)]. MIS and SIS are processed by a five-axis CNC machine,
and both have sliding arms located outside the longer shaft with
specially shaped slots on them [Fig. 5(c)]. The sliding arm of
MIS holds the sliding arm of SIS within it, sharing a pin located
in their slots [Fig. 5(b)]. The pin is linked to the OS, and the OS
is hinged to lever 1 in Fig. 3(a). This design constrains the OS
sliding by MIS and SIS, which means that the wing morphing
motion can be actuated by both the BLDC motor and servo.
Specifically, the relationship between the motion of these three
sliders is divided into two cases. In the first one, when the MIS
slides in a relatively anterior position, the pin is simultaneously
in the linear slot of MIS and the curved slot of SIS. Hence, the
pin can slide with respect to the MIS while being limited by the
SIS [shown on the upper side in Fig. 5(d)].
Regarding the second case, when the MIS slides posteriorly,
the pin is in the curved slot of MIS and the linear slot of SIS,
and thus, the pin moves together with the MIS and can slide
with respect to the SIS [shown downside in Fig. 5(d)]. In this
way, the OS, which outputs power to the wing, will be driven
by the motor or locked by the servo in due time with the motion
of the pin. If the gear turns in a direction as in Fig. 5(a)
(indicated by yellow arrow), the wing would be fully extended
during the downstroke because the wing is decoupled from the
MIS that is moving anteriorly and locked in the extended state
by the SIS. When the MIS moves at the rear with the gear
Fig. 4. Wingbeat actuation mechanism for flapping and morphing coupling. rotation, which occurs during the upstroke, the wing is again
(a) The Conical Rocker Mechanism (CRM) is mounted in the body frame,
with its components marked in different colors. (b) The mechanical coupled to the MIS and performs the morphing motion actuated
constraints of the rock arm and shaft C. (c) The cross shaft is the critical part by the motor. Equipped with the morphing decoupler, the CRM
of CRM. The longer shaft acts as the flapping axis of the wing, while the can drive the wing to create the wingtip and wrist trajectory as
shorter as the rotation axis of the shoulder when performing wing morphing.
(d) The action effect of CRM and the conical trajectory of the rocker arm. shown in Fig. 5(e), which is significantly similar to the bird
(e) The flapping motion is phase lead the morphing actuation motion by 90°. trajectory of previous studies [49].
To derive the conical swing of the rocker arm into a linear Furthermore, since the SIS locks the wing in the extended
reciprocating motion, a link rod [denoted by purple in Fig. 4(a)] state during the downstroke, the morphing motion of the wing
is hinged between the rocker arm and a slider [denoted by can be controlled by the rear-mounted servo [Fig. 5(f)].
magenta in Fig. 4(a)] on the longer shaft. With the slider’s Therefore, the left and right wings can be actuated
reciprocating motion, we can easily actuate the wing skeleton independently to cause a bilateral asymmetric change in the
to extend and tuck periodically by hinging lever 1 [in Fig. 3(a)] wingspan and area during the downstroke phase. In this way,
to the slider. Thus, the wing morphing motion will be coupled the servos on both sides can ignore the time sequences of wings
with the wing flapping motion with 90 degrees phase lag [Fig. flapping and move their respective SISs to the appropriate
4(e)]. However, this arrangement will cause the wing to extend positions to ensure that each wing is or is not fully extended
fully only when it flaps down to the level position. Since during each downstroke. This design greatly eases the task of
 7

Fig. 5. Wing-morphing decoupling mechanism for roll control. (a) The morphing decoupler is mounted on the longer shaft of the cross shaft, with each part
denoted in different colors. (b) Enlarged view of the morphing decoupler. For the sake of clarity, the trimetric views of these three sliders (MIS, OS, SIS) are
shown above to exhibit their 3D geometry. (c) The top view of MIS and SIS aligned on the longer shaft. Both sliders have hockey-stick-shaped slots on their
sliding arms to guide the pin’s motion in (b). (d) With the morphing decoupler, the wing skeleton is fully extended during downstroke by decoupling from the
MIS (shown above) and tucked during upstroke by coupling with MIS again (shown below). (e) The wingtips and wrists trajectory of the wingbeat kinematics
driven by CRM with the application of morphing decoupler. (f) The servo actuates the wing to tuck in during downstroke because of MIS decoupling (see
supplementary video for more details).
 8

the wing morphing actuator and reduces the difficulty of lateral downstroke at an 8m/s level flight (with tethered string
attitude control for the FWAV with a complex wingbeat pattern. protection, see Section VII). Then the vehicle preserves the
The servos do not need to run at the same frequency as the wing typical symmetric flapping during the open-loop rolling
flapping, affording lightweight and low-powered servos. For dynamic response [Fig. 6, (c) and (d)]. Each rolling maneuver
example, to perform a left rolling maneuver, we can input a utilizes the same servo value to ensure that each asymmetric
command to make the left servo pull the left SIS backward to downstroke has the same area difference between the left- and
limit the maximum wingspan on that side and then allow the right-wing. An aircraft’s agility is commonly defined as the
wing to flap for a few cycles. During this process, the wings on rapidity of change in speed and direction, typically measured
the two sides flap downward asymmetrically to generate a through the agility metric that includes angular acceleration and
the time it takes to achieve a desired rotation angle change [50],
differential lift and then tuck and flap upward symmetrically
[51]. In some studies, the time normalized by the wingbeat-
driven by the CRM (see supplementary video). This feature
cycle period is often used to quantify the flapping-wing flight
allows the RoboFalcon to be controlled manually without an
maneuvers [52]. To compare the agility differences between
autopilot. various bio-inspired flapping-wing aerial vehicles and flying
creatures, we use the agility metric measured as the time
V. ROLLING MANEUVER & FLIGHT TEST normalized by wingbeat-cycle that the flapping wing flyers
Although the morphing decoupler allows the servo to actuate require for a 90-degree attitude change. In Fig. 6(e), we
at any moment during the flapping flight, we need a definite and compare the rolling agility of RoboFalcon against other
controllable bilateral asymmetric downstroke to realize how the flapping wing vehicles and flying creatures.
morphing-coupled wing-flapping pattern utilizes the Our results indicate that the asymmetric downstroke based
asymmetric downstroke to perform a roll maneuver. For this on the morphing-coupled flapping notably improves the rolling
purpose, we equip the RoboFalcon with a simple Hall-sensor- agility of the flapping wing flight [Fig. 6(d)]. During the
based closed-loop control system to detect the flapping asymmetric downstroke, the vehicle starts to roll left, and the
sequences of the wings [Fig. 6, (a) and (b)]. The Hall switch, roll angle decreases after maneuver initiation, indicating a lack
mounted at the body frame [Fig. 6(a)], is triggered by a magnet of lateral stability for the RoboFalcon. This lack of lateral
embedded in the gear when the motor drives the wings flapping stability is partially the reason for its high rolling agility, with
down to the level position. As the gear rotates, the rolling and the roll rate being higher in each downstroke than in the
gliding controller [Fig. 6(b)], based on STM32 F103 (see upstroke, which is probably due to the rightward yaw angle of
Section VII), determines the current flapping cycle period by the vehicle during the rolling maneuver. The adverse roll-yaw
measuring the time interval between two recent sequential coupling due to asymmetric wing morphing was also found in
triggers of the Hall switch. Once the roll command from the a previous study on a propeller-powered platform in a gliding
autopilot is received, the controller can manage the action time state [34]. Nevertheless, for the asymmetric morphing coupled
of the servo with the known wingbeat cycle period to ensure with flapping, the mechanism of adverse yaw generation could
that the unilateral wingspan is limited only during the next be more complex, which is not within the scope of this paper.
single downstroke phase (see Section VII). Thus, we obtain a The RoboFalcon achieves a 90-degree rolling maneuver within
single and stable bilateral asymmetric downstroke among a 2.5 wingbeats, which is an appealing performance compared to
sequence of standard symmetric flapping. Furthermore, we previous bio-inspired flapping-wing robots [32], [37], [52], and
program the rolling and gliding controller to stop the motor only at the same level compared to some flying vertebrates [29] [Fig.
when the Hall switch is triggered. By adding a glide-lock to stop 6(e)].
the gear reversal [Fig. 6(A)], the wing will be locked at the level In our untethered outdoor flight test, RoboFalcon flies quite
position, ensuring for RoboFalcon the gliding ability and the well, managing a climb, banking turn, and rapid rolling
capability to use the bilateral asymmetric wing morphing maneuver [Fig. 6(f) and supplementary video]. This agile
during gliding to achieve the banking turn of [33], [34]. vehicle applies a simple PID controller of the autopilot for pitch
To demonstrate RoboFalcon’s roll control and high agility, and roll control and can be remotely operated by a human pilot.
we perform several rolling maneuver experiments to measure The power consumption for an 8m/s level flight is about 50W,
pitch, yaw, and roll angles (airspeed, power consumption, linear enabling a flight endurance of 11 min.
acceleration, and angular velocity were also recorded). Each
maneuver is initiated with a single cycle bilateral asymmetric
 9

Fig. 6. The rolling maneuver for agility assessment and the flight test. (a) The glide-lock mechanism. The Hall switch (denoted in blue) is for wing
position checking. The pawl (denoted in red) stops the gear from reversing when the wing flaps down to the level position. (b) The block diagram of the
rolling and gliding control system. The autopilot controls the elevator servo directly, while the control of the motor on the wing servos is via the rolling
& gliding controller. (c) The reference frame of the Euler angle. (d) The time evolution of the Euler angles during the rolling maneuver. The maneuver
is initiated with a one-shot bilateral asymmetric downstroke (denoted in yellow). Symmetric wing-extended downstrokes are indicated in dark gray and
wing-tucked upstrokes in light gray. (e) Comparison of the wingbeat number for performing a 90-degree orientation change for different flapping flyers.
The data are acquired from [29], [32], [61]-[67] for the animals and [32], [37], [52] for the robots. (f) RoboFalcon performing a banking turn (shown
left), a climb (shown upper left), and a rapid rolling maneuver (shown lower left) during flight tests.
 10

flapping frequencies at an airspeed of 8m/s and angles of attack

ranging from 2 to 12 degrees placed below stall.
VI. WIND TUNNEL TEST Significant cyclical aerodynamic effects characterize the
To assess the aerodynamic effect of the wrist mounting angle flapping wing, and thus analyzing the steady flight state is not
trivial. From the cycle-averaged perspective, the flapping wing
(wrist supination) and the characteristics of the bilateral
asymmetric wingbeat kinematics, we measure the aerodynamic aerodynamics can be analyzed regarding the method of the
forces with a 6-axis load cell under various wing configurations fixed wing [53] (see Section VII). The cycle-averaged lift and
cycle-averaged net thrust of various flapping frequencies versus
in a low-speed wind tunnel (see Section VII). Firstly, we test the
effects of four different wrist mounting angles θ (10°, 15°, 20°, angle of attack are illustrated in Fig. 7(a).
and 25°) on the flight characteristics of a steady level flight state. During the trimed steady level flight, the average
aerodynamic forces acting on the vehicle are at equilibrium, i.e.,
The tests involve measuring lift and net thrust with various
the cycle-averaged net thrust should be zero, and the cycle-

Fig. 7. The effects of wrist mounting angle and the roll moment of the asymmetric downstroke. (a) The cycle-averaged lift and net thrust of four different
wrist mounting angles. The discrete data are fitted to obtain a continuous model (see Supplementary Materials). The yellow line indicates the thrust-drag
equilibrium state among various configurations of flapping frequencies and angles of attack. (b) The cycle-averaged lift of the thrust-drag equilibrium state
of different wrist mounting angles versus angles of attack. The black dashed line indicates the lift-weight equilibrium state. The wrist mounting angle affects
the angle of attack of the equilibrium flight state. (c) The roll moment of different flapping frequencies in one wingbeat cycle, with the shaded region
indicating the upstroke. The roll moment during upstroke basically remains the same, while the peak value of the roll moment in the downstroke phase
becomes larger as the frequency increases. (d) The cycle-averaged roll moment of different angles of attack correlates linearly with the flapping frequency.
 11

averaged lift would be equal to the weight of the RoboFalcon VII. MATERIALS AND METHODS
(600g). The equilibrium state is expressed as:
A. Bio-inspired morphing wing design and fabrication

 Tnet  0

 (2) The wing skeleton mechanism is a one-DOF sophisticated

L  W
 multi-link mechanism consisting of two sets of four-bar linkage
For a given airspeed value, the cycle-averaged lift and net thrust mechanisms. We determine the lengths of the humeral spar,
are the functions of the angle of attack and the flapping radial spar, and hand wing, using the vector closed-loop
frequency. Since the data of various flight states measured from equation to solve the lengths of the link rods and levers by
the wind tunnel are discrete, it is hard to find the lift-weight and giving the angles of the shoulder, elbow, and wrist at the fully
thrust-drag equilibrium state directly. To overcome that, we fit expanded and completely tucked states of the wing skeleton
the cycle-averaged lift and cycle-averaged net thrust data with (see Supplementary Text for details). The lengths of the spars
quadratic polynomial (see Section VII). In Fig. 7(a), the yellow and hand wings are determined concerning the statistical bird
line indicates the thrust-drag equilibrium state. For the net (bat) wing data [58], [59].
thrust graph, this is the level straight line past the zero point, We mainly construct the wing skeleton from carbon fiber
while for the lift graph, this curve indicates the lift of the thrust- composite boards. The link rods, levers, and fingers are directly
drag equilibrium state versus the angle of attack. cut from the 3k carbon fiber board (thicknesses of 2 and 1.5 mm)
We compare the lift curves of the thrust-drag equilibrium with a CNC milling machine (JINGDIAO Carver S400). The
state of the four wrist mounting angles in Fig. 7(b) with the lift- humeral and radial spars are made from the same carbon fiber
weight equilibrium state indicated by the black dashed line. The board parts that are cut into a specific shape to form a 3D puzzle
intersection of the black dashed line and the lift curves indicates structure to constitute the hinge knuckles at two ends of the
the lift-weight and thrust-drag equilibrium states of the steady spars. The leading-edge support parts, wrist hinge knuckles, and
level flight. other accessories, e.g., ball joint and servo mounting base, and
The results indicate that a smaller θ allows the vehicle to fly are 3D printed (Raise3D Pro2 Plus) utilizing the Polylactic
at a larger trimmed angle of attack. This could be because a (PLA) and Polycarbonate (PC) materials. We employ the
smaller θ involves a stronger downstroke twist of the hand-wing, KINGMAX KM0940 digital servo (stall torque of 4.3kg·cm)
and thus the vehicle requires a larger trimmed angle of attack to for the wing morphing actuating, while for the wing membrane,
compensate for the pitch-down (wrist hyperpronating) of the the ripstop polyester fabric (210T), which is unstretchable and
wing to keep the wing’s local angle of attack in the right range. airtight. The wing membrane is only glued to the fingers, main
Also, the mounting angle θ affects the available lift (the region spars, and leading-edge support parts of the skeleton to ensure
of lift above 600g in Fig. 7(a) and thrust (the region of thrust no stretching during the skeleton folding process.
above the equilibrium line in Fig. 7(a) at this airspeed. For our
trials [Fig. 7(a)], the available lift increases with θ, while the
B. Rolling & gliding controller
available thrust maximizes at θ=20°. This suggests that birds
and bats could use wrist supination-pronation movements to The rolling and gliding controller is constructed to generate
adjust the angle of attack of the steady level flight state and a one-shot roll command for the rolling maneuver to test and
modulate the lift and thrust magnitude. control the wing stop position for the gliding state. The
Then, we measure the roll moment of a 25° wrist mounting controller is equipped with an STM32F103C8T6
angle wing configuration at an airspeed of 8m/s under various microprocessor running at 72MHz that contains a 6-channel
flapping frequencies and angles of attack (0° to 12°). The PWM input to receive the operation command from the
average lift increases with the flapping frequency for all wing autopilot and a 3-channel PWM output to control the motor
configurations, resembling previous studies on bendable or speed and the two servos for wing morphing actuation [Fig.
foldable flapping wing aerodynamics [41], [54], [55]. Since the 6(B)]. An OH3144 unipolar Hall-effect switch is connected to
roll moment of the RoboFalcon is achieved by the bilateral lift the microprocessor via an external interrupt.
differential generated by the bilateral asymmetric downstroke, When wings flap down to the level position, the magnet
the correlation between lift and flapping frequency is also embedded in the main gear triggers the Hall switch to send an
shown in the roll moment. Fig. 7(c) presents the roll moment of interrupt signal to the rolling and gliding controller. During
the morphing-coupled flapping that applies the bilateral flapping, the controller measures the time interval of two
asymmetric downstroke in a wingbeat cycle (α=12°). The adjacent interrupt signals and calculates the flapping frequency.
cycle-averaged rolling moments of different angles of attack When the rolling and gliding controller receives a predefined
versus flapping frequency are illustrated in Fig. 7(d). In general, one-shot roll command, it takes over the roll control authority
the roll moment remains at zero during the upstroke and only from the autopilot and waits for the Hall switch signal. Once the
magnifies during the asymmetric downstroke. For the higher controller receives the signal, it delays half a wingbeat cycle
flapping frequency, the larger the peak value of the roll moment and then outputs a left-wing morphing command with a
(negative values correspond to left roll). This means that the duration of three-quarters of a wingbeat cycle to cover the next
aerodynamic forces during downstroke dominate the creation downstroke so that we can get a single bilateral asymmetric
of the roll moment, which is similar to the results of the study downstroke for the rolling maneuver test.
on birds [30]. When the throttle is lower than a certain threshold, the
controller reduces the motor speed and stops the motor after
receiving the external interrupt signal. This affords to stop the
wings right at the level position, and the glide-lock limits the
 12

gear from reversing to handle the aerodynamic load during combining an angle of attack and a flapping frequency is
gliding [Fig. 6(a)]. measured for more than five wingbeat cycles. Each state’s lift
and net thrust data is cycle-averaged, and the maximum RMSE
for different cycles is lower than 36.96g and 8.35g, respectively.
C. Rolling maneuver test set up We fit the data utilizing a quadratic polynomial to obtain a
We perform five rolling maneuver tests for recording the continuous model and determine the equilibrium point of each
open-loop dynamic response of a single bilateral asymmetric wing configuration in Fig. 7(a) and 7(b). The fitted results are
downstroke. The tests are performed within a hangar to avoid presented in Supplementary Materials.
wind interference, and the vehicle is tethered with a thin nylon For the roll moment measurement, the angle of attack ranges
string (diameter 0.55mm) for protecting itself from hitting from 0 to 12 degrees with a 4-degree increment, and the
ground and walls. The nylon string is slack during the flight and flapping frequency is roughly limited within the range of
is light enough, presenting negligible additional weight and 1.4~2.6 Hz to protect the load cell. The curve of roll moment
drag. The RoboFalcon is launched at the gliding state by versus normalized time in Fig. 7(c) is filtered with the 5th order
throwing it. Immediately afterward, the motor starts and the Butterworth low-pass filter (cut-off frequency of 12Hz), and the
vehicle goes into a closed-loop controlled stabilized flapping maximum RMSE of the cycle-averaged roll moment in Fig. 7(d)
flight. Then, the operator sends a predefined roll command to is lower than 0.007N·m.
the autopilot, enabling the rolling and gliding controller to take
over the roll control and perform the open-loop roll maneuver
initiated with the bilateral asymmetric downstroke. VIII. DISCUSSION
The autopilot, Pixmini, is used for logging linear acceleration,
angular velocity, attitude angle, airspeed, power consumption, Inspired by the wingbeat kinematics of birds and bats during
and the control inputs. The airspeed sensor (sensirion sdp31) level flight, we develop a flapping wing aerial vehicle (FWAV),
and the control inputs are sampled at 100Hz and 50Hz, entitled RoboFalcon, equipped with the novel Conical Rocker
respectively, while the remaining data at 200Hz. We calculate Mechanism (CRM) to achieve this morphing-coupled flapping
the mean and standard error of the attitude angle data for the pattern. The CRM provides the coupling motions required for
five rolling maneuver flights, illustrated in Fig. 6(d). All data wing flapping and morphing and makes this platform achieve
timings are normalized by the wingbeat cycle and aligned to the the wingbeat kinematics where the wings are fully extended
wingbeat that produced the asymmetric downstroke. The initial during the downstroke and tucked during the upstroke. The
yaw angle is aligned to zero. untethered outdoor flight tests demonstrate the appealing flight
ability achieved by our RoboFalcon and highlight the
effectiveness of the proposed wingbeat kinematics.
D. Wind tunnel measurements Additionally, our design introduces a mechanical decoupler
We use a low-speed low turbulence intensity wind tunnel to allow for the independent control of wing morphing. The
housed at Northwestern Polytechnical University. The airspeed actuation source of the wing morphing alternates during
is provided by the wind tunnel precisely, and a tilting platform flapping between the BLDC motor and servo, enabling an
accurately controls the angle of attack. We use a PWM signal adjustable flapping pattern for each side and managing a rolling
generator to control the throttle value for the flapping frequency maneuver. To validate the agility improvement of our roll
adjustment, and the accurate frequency value is determined by control strategy, we perform several rolling maneuver
recording the time interval of two adjacent signals from the experiments employing the rolling and gliding controller. The
onboard Hall-effect switch. We use a 6-axis load cell (ME results highlight that the bilateral asymmetric downstroke
K6D40 force/torque sensor) for the aerodynamic forces enables RoboFalcon to perform a 90-degree roll maneuver
measurements, with the sample rate set to 1000Hz. The within only 2.5 wingbeat cycles, which is considered a highly
RoboFalcon is mounted on top of the load cell, while the latter appealing performance among nature flyers and bio-inspired
is fixed with the tilting platform. With this installation, the FWAVs. The wind tunnel data indicate that the roll moment
gravity of RoboFalcon needs to be offset at every tested angle generated by the bilateral asymmetric downstroke increases
of attack. Since the lift and thrust are acquired in the coordinate with the flapping frequency, similar to the increase in the lift
system of the load cell, it is necessary to convert the acquired with frequency, which has been validated in previous studies of
forces into the wind axis coordinate system for level steady bendable or foldable flapping wings [41], [54], [55]. However,
flight analysis. The coordinate system conversion is as follows it is important to note that the increase in roll moment for a bird-
 L   sin α cos α  Fx   sin α  or bat-inspired FWAV has not been reported yet. The frequency-
       G 
  (3) regulated roll moment affords RoboFalcon a high roll control
Tnet  cos α  sin α Fz  cos α efficiency. During RoboFalcon’s flight, the downstroke
where the L is the lift, Tnet the net thrust, α the angle of attack, dominates in creating the roll moment during one wingbeat
G the gravity of RoboFalcon, Fx the force measured in the x- cycle, consistent with the previous study on birds [30].
axis of the load cell, and Fz the force measured in the z-axis of Furthermore, our designed bio-inspired morphing wing
the load cell. applies a wrist mounting angle that affords to fine-tune the
For each wing configuration measurement, the angle of passive torsion of the hand wing during the downstroke. The
attack ranges from 0 to 12 degrees with a 2-degree increment, wind tunnel data indicate that the angle of attack of the
while the throttle value increases with a 5% increment to cover equilibrium state decrease with the increase of the wrist
the flapping frequency approximately from 2 to 4 Hz. Each state mounting angle. This result suggests that the wrist supination-
 13

pronation movements of birds and bats could be used to trim that can mimic most of the wingbeat kinematics of birds and
the angle of attack in a steady level flight. Adjustment of the bats with complex degrees of freedom.
wrist mounting angle provides an engineering solution to
optimize the lift and thrust combination of the morphing-
coupled flapping wing. APPENDIX
This work results in a novel actuation strategy for a flapping
wing vehicle to couple the wing-morphing motion with the MULTI-LINK MECHANISM DESIGN OF THE MORPHING WING
wingbeat motion. It introduces a new rolling attitude control of SKELETON
the bird- and bat-inspired FWAV, simplifying roll control with
To design the morphing wing skeleton, the lengths of the
complex flapping kinematics and reducing the burden on the
humerus (lh), radius (lr), and manus (lm) need to be given. It is
wing morphing actuator. Additionally, our setup allows
also necessary to determine the angle of the shoulder joint (θs),
exploiting less powerful and lighter-weighted actuators (like
elbow joint (θe), wrist joint (θw), and the displacement of the OS
typical servo motor) and ultimately meeting the weight
(output slider, xA) when the skeleton is fully extended and
constraints during flight.
tucked (subscripted by e and t). Given these parameters, as
The wingbeat kinematics of RoboFalcon is very similar to
shown in Fig. 8 and Table I, the vector closed-loop equation can
that of birds, enabling a high level of bionic performance.
be used to solve for the length of the remaining linkages.
Therefore, our platform is suitable for studying birds’ and bats’
Referring to Fig. 8(c), the vector closed-loop equations for
aerodynamic and kinetic mechanisms during steady-level flight
the one slider-rocker mechanism and the two four-bar linkage
and maneuvers with bilateral asymmetric wingbeat patterns.
mechanisms in the wing skeleton structure can be expressed as
The bat-style wing of RoboFalcon can accommodate the flight   
requirements well despite the membrane wrinkles when the OA  OC  CA (4)
wing is tucked. Our developed wing adopts several bionic    
features, i.e., 3D printed rigid leading-edge support and the CB  BE  ED  DC  0 (5)
raised elbow joint to mimic the function of the propatagium    
[43]-[45], which provides a design reference for the bio- EF  FH  HG  GE  0 . (6)
inspired morphing wings based on vertebrate-wing anatomy to The orthogonal decomposition of the above equation along
increase the camber of their inner section (arm wing). The final with the xy direction yields
platform is agile and can fly quite well. Based on wind tunnel 

 x A  c cos θ1  (b  a ) cos θ2
 (7)
data, the RoboFalcon has considerable available lift beyond the 
 0  c sin θ1  (b  a ) sin θ2

equilibrium point and thus has significant load capacity to carry

 b cos θ2  e cos θ3  i cos θ4  d cos θ1  0
larger capacity batteries and onboard equipment for more 
 (8)
extended missions. The RoboFalcon is also more stealthy due 
 b sin θ2  e sin θ3  i sin θ4  d sin θ1  0

to the similarity of its wingbeat pattern to that of a bird. 

 f cos θ3  j cos θ5  h cos θ6  ( g  i ) cos θ4  0
It should be noted that, in this study, the wing based on the  . (9)
unstretchable membrane may generate sufficient lift for flight, 
 f sin θ3  j sin θ5  h sin θ6  ( g  i ) sin θ4  0

but eventually, the efficiency is reduced due to the additional The variables in Fig. 8(c) are related to the given parameters
drag caused by the wrinkles. Since even in nature, birds fly in Fig. 8, (a) and (b), as follows
more efficiently than bats [15], future work shall investigate a θs  θ1

bird-style feathered wing for this platform, to allow an 


θe  θ1  θ4
enhanced aerodynamic profile when the wings are tucked [34]. 
θ  θ  θ
This work mainly focuses on providing an engineering solution  w 4 6 . (10)


to achieve the morphing-coupled flapping motion and l h  cd
validating the agility benefits from the bilateral asymmetric 


downstroke without delving into the energy efficiency lr  g


advantage of this wingbeat pattern. Therefore future works shall Substituting (10) into (7), (8), (9), we obtain
also involve more detailed and precise experimental  x A  (lh  d ) cos θs  (b  a ) cos θ2


measurements to analyze the efficiency advantages of this  (11)
 0  (lh  d ) sin θs  (b  a ) sin θ2


wingbeat pattern after equipping some sort of single-degree-of-
freedom feathered morphing wings. Also, this platform can 
 b cos θ2  e cos θ3  i cos(θe  θs )  d cos θs  0
only achieve the wingbeat pattern of birds and bats in level 
 (12)
b sin θ2  e sin θ3  i sin(θe  θs )  d sin θs  0


flight, but not the completely different pattern in the take-off
and perching states used by these flying vertebrates.
 f cos θ3  j cos θ5  h cos(θw  θe  θs )  (lr  i )cos(θe  θs )  0
Considering that the CRM within RoboFalcon has the potential  (13)
 f sin θ3  j sin θ5  h sin(θw  θe  θs )  (lr  i )sin(θe  θs )  0
ability to adjust the flapping amplitude, stroke plane, dihedral
angle, and sweep angle, RoboFalcon can be expected to couple By using the trigonometric relationship and substituting
these parameters with flapping by deploying appropriate among (11), (12), (13) to eliminate the variable θ2, θ3, θ5, we
actuation strategies to achieve the wingbeat patterns of birds or obtain
bats during take-off and perching, such as stroke plane tilting
and amplitude modulation [29], [56], [57]. The latter shall drive ( x A  (lh  d ) cos θs ) 2 +(lh  d ) 2 sin 2 θs =
(b  a) 2 (14)
research towards developing a flyable bio-inspired aerial robot
 14

x A  (lh  d ) cos θs 2 TABLE I

(d cos θs  i cos(θe  θs )  b )  LINKAGE LENGTH AND STATE PARAMETERS OF THE MORPHING WING
ba
(15) SKELETON
(lh  d )sin θs 2
(d sin θs  i sin(θe  θs )  b )  e2 Given parameter Derived linkage length (mm)
ba
lh 110 a 15.3646
lr 180 c 33.5769
f x  (lh  d )cos θs linkage length
( (d cos θs  i cos(θe  θs )  b A ) lw 370 d 76.4231
e ba (mm)
b 20 e 73.6915
h cos(θw  θe  θs )  (lr  i )cos(θe  θs )) 2 
.(16) f 30 g 180
f (lh  d )sin θs
( (d sin θs  i sin(θe  θs )  b )
e ba θse 51° h 20.1844
h sin(θw  θe  θs )  (lr  i )sin(θe  θs )) 2  j 2 θee 110° i 16.9713
extended state
θwe 147° j 202.8699

Since b and f are already given in Table I, the above three xAe 45mm
equations can be used to construct two sets of equations for the θwt 20°
extended and tucked state, respectively, via the parameters θet 41°
given in Table I to derive the remaining six unknown variables. tucked state
The remaining parameters of each linkage length of the wing θwt 35°
skeleton are also shown in Table I. xAt 65mm
The given partial linkage lengths and the extended and tucked state
parameters are shown at the left as the design constraints. Other linkage
lengths derived from kinematic relations are shown at right.

FITTING RESULTS OF CYCLE-AVERAGED LIFT AND NET THRUST

Fig. 8. Kinematic diagram of the morphing wing mechanism. The lengths

of the humerus, radius, and manus are referenced to the forelimb structure
of flying vertebrates while meeting the wing span constraints, which are
the given design parameters. (a) and (b) show the fully extended and
Fig. 10. Fitting surfaces of the cycle-averaged lift versus angle of attack and
tucked states of the wing skeleton, respectively, and the angles of the flapping frequency for different wrist mounting angles.
shoulder, elbow, and wrist joints for these two states are given in the
diagrams as the design constraints for determining the remaining linkage
length. (c) Reference diagram for mechanism kinematic analysis.
 15

TABLE II
FITTING RESULTS OF CYCLE-AVERAGED LIFT AND NET THRUST
Cycle-averaged lift Cycle-averaged net thrust
Model L=z0+aα+bF+cα2+dF2+fαF T=z0+aα+bF+cα2+dF2+fαF
Wrist mounting angle θ (°) 10 15 20 25 10 15 20 25
z0 9.940 155.857 191.549 202.331 -127.799 -63.082 -80.520 -67.745
a 22.879 12.322 14.278 13.544 1.402 -0.231 2.939 2.594
Parameters

b 97.421 29.811 11.376 10.631 18.077 -17.318 -4.745 -11.970

c -0.095 0.092 -0.029 -0.038 -0.222 -0.213 -0.417 -0.444
d -0.664 12.722 17.959 18.423 7.276 12.767 10.466 10.952
f 0.576 2.769 2.656 3.454 -0.453 -0.6618 -0.939 -1.184
R value 0.99969 0.99948 0.99936 0.99864 0.99952 0.99950 0.99825 0.99851
RSME (g) 3.958 3.015 4.808 6.631 1.270 1.016 2.127 1.960
Number of point 70 42 59 50 70 42 59 50
For the model expressions in the table, L is the cycle-averaged lift, T the cycle-averaged net thrust, α the angle of attack, and F the flapping frequency.

[6] D. K. Riskin, A. Bergou, K. S. Breuer, S. M. Swartz, “Upstroke wing

flexion and the inertial cost of bat flight,” Proc. R. Soc. B., vol. 279, no.
1740, pp. 2945-2950, Apr. 2012.
[7] J. M. V. Rayner, “The evolution of vertebrate flight,” Biol. J. Linn. Soc.,
vol. 34, no. 3, pp. 269-287, Jul. 1988.
[8] E. Bell, B. Andres, A. Goswami, “Integration and dissociation of limb
elements in flying vertebrates: a comparison of pterosaurs, birds and bats,”
J. Evol. Biol., vol. 24, no. 12, pp. 2586-2599, Sep. 2011.
[9] A. A. Biewener, K. P. Dial, “In vivo strain in the humerus of pigeons
(Columba livia) during flight,” J. Morphol., vol. 225, no. 1, pp. 61-75, Jul.
1995.
[10] S. Swartz, M. Bennett, D. Carrier, “Wing bone stresses in free flying bats
and the evolution of skeletal design for flight,” Nature, vol. 359, no. 6397,
pp. 726-729, Oct. 1992.
[11] U. Norberg, Vertebrate flight. Berlin, Germany: Springer-Verlag, 1990.
[12] S. Swartz, K. Middleton, “Biomechanics of the bat limb skeleton: scaling,
material properties and mechanics,” Cells Tissues Organs, vol. 187, no.1,
pp. 59-84, Dec. 2008.
[13] R. J. Vazquez, “The automating skeletal and muscular mechanisms of the
avian wing (Aves),” Zoomorphology, vol. 114, no. 1, pp. 59-71, Mar.
1994.
[14] A. K. Stowers, L. Y. Matloff, D. Lentink, “How pigeons couple three-
dimensional elbow and wrist motion to morph their wings,” J. R. Soc.
Interface, vol. 14, no. 133, Aug. 2017, Art. no. 20170224.
[15] D. D. Chin, L. Y. Matloff, A. K. Stowers, E. R. Tucci, D. Lentink,
“Inspiration for wing design: How forelimb specialization enables active
flight in modern vertebrates,” J. R. Soc. Interface., vol. 14, no. 131, Jun.
2017, Art. no. 20170240.
[16] B. W. Tobalske, T. L. Hedrick, A. A. Biewener, “Wing kinematics of
avian flight across speeds,” J. Avian Biol., vol. 34, no. 2, pp. 177-184, Jul.
2003.
[17] C. J. Pennycuick, “Power requirements for horizontal flight in the pigeon
Fig. 10. Fitting surfaces of the cycle-averaged net thrust versus angle of attack
Columba livia,” J. Exp. Biol., vol. 49, no.3, pp. 527-555, Dec. 1968.
and flapping frequency for different wrist mounting angles.
[18] A. Hedenström, L. C. Johansson, G. R. Spedding, “Bird or bat: comparing
airframe design and flight performance,” Bioinspir. Biomim., vol. 4, no.
1, Mar. 2009, Art. no. 015001.
[19] M. Wolf, L. C. Johansson, R. V. Busse, Y. Winter, A. Hedenström,
“Kinematics of flight and the relationship to the vortex wake of a Pallas’
REFERENCES long tongued bat (Glossophaga soricina),” J. Exp. Biol., vol. 213, no. 12,
pp. 2142-2153, Jun. 2010.
[20] R. V. Busse, A. Hedenström, Y. Winter, L. C. Johansson, “Kinematics
[1] S. Poore, A. Sánchez-Haiman, G. Goslow, “Wing upstroke and the and wing shape across flight speed in the bat, Leptonycteris yerbabuenae,”
evolution of flapping flight,” Nature, vol. 387, no. 6635, pp. 799-802, Jun. Biol. Open, vol. 1, no. 12, pp. 1226-1238, Dec. 2012.
1997. [21] A. Hedenström, L. C. Johansson, “Bat flight: aerodynamics, kinematics
[2] C. J. Pennycuick, “The Membrane Wings of Bats and Pterosaurs,” in and flight morphology,” J. Exp. Biol., vol. 218, no. 5, pp. 653-663, Mar.
Modelling the Flying Bird. Elsevier, 2008, pp. 135-160. 2015.
[3] C. J. Pennycuick, “Span-Ratio Analysis Used to Estimate Effective Lift: [22] U. M. Norberg, “Aerodynamics, kinematics, and energetics of horizontal
Drag Ratio in the Double-Crested Cormorant Phalacrocorax Auritus from flapping flight in the long-eared bat Plecotus auritus,” J. Exp. Biol., vol.
Field Observations,” J. Exp. Biol., vol. 142, no. 1, pp. 1-15, Mar. 1989. 65, no. 1, pp. 179-212, Aug. 1976.
[4] M. Rosén, G.R. Spedding, A. Hedenström, “Wake structure and wingbeat [23] H. D. Aldridge, “Kinematics and aerodynamics of the greater horseshoe
kinematics of a house-martin Delichon urbica,” J. R. Soc. Interface, vol. bat, Rhinolophus ferrumequinum, in horizontal flight at various flight
4, no. 15, pp. 659-668, Jan. 2007. speeds,” J. Exp. Biol., vol. 126, no. 1, pp. 479-497, Nov. 1986.
[5] K. J. Park, M. Rosén, A. Hedenström, “Flight kinematics of the barn [24] T. Y. Hubel, N. I. Hristov, S. M. Swartz, K. S. Breuer, “Changes in
swallow (Hirundo rustica) over a wide range of speeds in a wind tunnel,” kinematics and aerodynamics over a range of speeds in Tadarida
J. Exp. Biol., vol. 204, no. 15, pp. 2741-2750, Aug. 2001. brasiliensis, the Brazilian free-tailed bat,” J. R. Soc. Interface, vol. 9, no.
71, pp. 1120-1130, Jan. 2012.
 16

[25] R. J. Vazquez, “Functional osteology of the avian wrist and the evolution complexity of bat wing kinematics,” J. Theor. Biol., vol. 254, no. 3, pp.
of flapping flight,” J. Morphol. vol. 211, no. 3, pp. 259-268, Mar. 1992. 604-615, Oct. 2008.
[26] U.M. Norberg, Functional osteology and myology of the wing of the dog- [48] J. A. Cheney, N. Konow, K. M. Middleton, K. S. Breuer, T. J. Roberts,
faced bat Rousettus aegyptiacus (É. Geoffroy) (Mammalia, Chiroptera). E. L. Giblin, S. M. Swartz, “Membrane muscle function in the compliant
Z. Morph. Tiere, vol. 73, no. 1, pp. 1-44, 1972. wings of bats,” Bioinspir. Biomim., vol. 9, no. 2, May. 2014, Art. no.
[27] J. H. Ostrom, S. O. Poore, G. E. Goslow Jr, “Humeral rotation and wrist 025007.
supination: important functional complex for the evolution of powered [49] B. Tobalske, K. Dial, “Flight kinematics of black-billed magpies and
flight in birds,” Smithson. Contrib. Paleobiol, vol. 1999, no. 89, pp. 301- pigeons over a wide range of speeds,” J. Exp. Biol., vol. 199, no.2, pp.
309, 1996. 263-280, Feb. 1996.
[28] D. Lentink, U. K. Müller, E. J. Stamhuis, R. de Kat, W. van Gestel, [50] R. Bitten, “Qualitative and quantitative comparison of government and
L. L. Veldhuis, P. Henningsson, A. Hedenström, J. J. Videler, industry agility metrics,” J. Aircr., vol. 27, no. 3, pp. 276-282, 1990.
J. L. van Leeuwen, “How swifts control their glide performance [51] R. Dudley, “Mechanisms and implications of animal flight
with morphing wings,” Nature, vol, 446, no. 7139, pp.1082-1085, Apr. maneuverability,” Integr. Comp. Biol., vol. 42, no. 1, pp. 135-140, Feb.
2007. 2002.
[29] A. J. Bergou, S. M. Swartz, H. Vejdani, D. K. Riskin, L. Reimnitz, G. [52] M. Karásek, F. T. Muijres, C. D. Wagter, B. D. Remes, G. C. de Croon,
Taubin, K. S. Breuer. “Falling with style: bats perform complex aerial “A tailless aerial robotic flapper reveals that flies use torque coupling in
rotations by adjusting wing inertia,” PLoS Biol. vol. 13, no. 11, 2015, Art. rapid banked turns,” Science, vol. 361, no. 6407, pp. 1089-1094, 2018.
no. e1002297. [53] P. Nian, B. Song, J. Xuan, W. Yang and Y. Dong, “A Wind Tunnel
[30] T. L. Hedrick, J. R. Usherwood, A. A. Biewener, “Low speed Experimental Study on the Flexible Flapping Wing With an Attached
maneuvering flight of the rose-breasted cockatoo (Eolophus roseicapillus). Airfoil to the Root,” IEEE Access, vol. 7, pp. 47891-47903, 2019.
II. Inertial and aerodynamic reorientation,” J. Exp. Biol., vol. 210, no. 11, [54] P. Nian, B. Song, J. Xuan, W. Zhou, D. Xue, “Study on flexible flapping
pp. 1912-1924, Jun. 2007. wings with three dimensional asymmetric passive deformation in a
[31] I. G. Ros, M. A. Badger, A. N. Pierson, L. C. Bassman, A. A. Biewener, flapping cycle,” Aerosp. Sci. Techno., vol. 104, Sep. 2020, Art. no.
“Pigeons produce aerodynamic torques through changes in wing 105944.
trajectory during low speed aerial turns,” J. Exp. Biol. vol. 218, no. 3, pp. [55] A. A. Wissa, Y. Tummala, J. E. Hubbard Jr, M. I. Frecker, “Passively
480-490, Feb. 2015. morphing ornithopter wings constructed using a novel compliant spine:
[32] F. Fei, Z. Tu, J. Zhang, X. Deng, “Learning Extreme Hummingbird design and testing,” Smart Mater. Struct., vol. 21, no. 9, Aug. 2012, Art.
Maneuvers on Flapping Wing Robots,” in 2019 IEEE International no. 094028.
Conference on Robotics and Automation (ICRA), IEEE, 2019, pp. 109- [56] A. M. Berg, A. A. Biewener, “Wing and body kinematics of takeoff and
115. landing flight in the pigeon (Columba livia),” J. Exp. Biol., vol. 213, no.
[33] M. D. Luca, S. Mintchev, G. Heitz, F. Noca, D. Floreano, “Bioinspired 10, pp. 1651-1658, May. 2010.
morphing wings for extended flight envelope and roll control of small [57] K. D. Earls, “Kinematics and mechanics of ground take-off in the starling
drones,” Interface focus, vol. 7, no. 1, Feb. 2017, Art. no. 20160092. Sturnis vulgaris and the quail Coturnix coturnix,” J. Exp. Biol., vol. 203,
[34] E. Chang, L. Y. Matloff, A. K. Stowers, D. Lentink, “Soft biohybrid no. 4, pp. 725-739, Feb. 2000.
morphing wings with feathers underactuated by wrist and finger motion,” [58] A. Azuma, “Flight by Gliding,” in The biokinetics of flying and swimming,
Sci. Robot., vol. 5, no. 38, Jan. 2020, Art. no. eaay1246. American Institute of Aeronautics and Astronautics, 2006, pp. 47-50.
[35] E. Ajanic, M. Feroskhan, S. Mintchev, F. Noca, D. Floreano, [59] U. M. Lindhe-Norberg, A. P. Brooke, W. J. Trewhella, “Soaring and non-
“Bioinspired wing and tail morphing extends drone flight capabilities,” soaring bats of the family pteropodidae (flying foxes, Pteropus spp.): wing
Sci. Robot., vol. 5, no. 47, Oct. 2020, Art. no. eabc2897. morphology and flight performance,” J. Exp. Biol., vol. 203, no. 3, pp.
[36] A. Ramezani, X. Shi, S. Chung, S. “Hutchinson, Bat Bot (B2), a 651-664. Feb. 2000.
biologically inspired flying machine,” in 2016 IEEE International [60] C. M. White, N. J. Clum, T. J. Cade, and W. G. Hunt, “Peregrine Falcon
Conference on Robotics and Automation (ICRA), IEEE, 2016, pp. 3219- (Falco peregrinus),” in The Birds of North America, Cornell Lab of
3226. Ornithology, Ithaca, NY, USA, 2002.
[37] A. Ramezani, S.-J. Chung, S. Hutchinson, “A biomimetic robotic [61] J. Iriarte-Díaz, S. M. Swartz, “Kinematics of slow turn maneuvering in
platform to study flight specializations of bats,” Sci. Robot., vol. 2, no. 3, the fruit bat Cynopterus brachyotis,” J. Exp. Biol., vol. 211, no. 21, pp.
Feb. 2017, Art. no. eaal2505. 3478-3489, Nov. 2008.
[38] Festo AG & Co. KG, “BionicFlyingFox: Ultra-lightweight flying object [62] P. Windes, D. K. Tafti, R. Müller, “Kinematic and aerodynamic analysis
with intelligent kinematics,” 2018, [Online]. Available: of a bat performing a turning-ascending maneuver,” Bioinspir. Biomim.,
https://www.festo.com/net/SupportPortal/Files/492827/Festo_BionicFlyi vol. 16, no. 1, 2020, Art. no. 016019.
ngFox_en.pdf. [63] I. G. Ros, M. A. Badger, A. N. Pierson, L. C. Bassman, A. A. Biewener,
[39] J. W. Bahlman, S. M. Swartz, K. S. Breuer, “Design and characterization “Pigeons produce aerodynamic torques through changes in wing
of a multi-articulated robotic bat wing,” Bioinspir. Biomim., vol. 8, no. 1, trajectory during low speed aerial turns,” J. Exp. Biol., vol. 218, no. 3, pp.
Feb. 2013, Art. no. 016009. 480-490, Feb. 2015.
[40] J. Colorado, A. Barrientos, C. Rossi, K. S. Breuer, “Biomechanics of [64] J. Y. Su, S. C. Ting, J. T. Yang, “How a Small Bird Executes a Sharp
smart wings in a bat robot: morphing wings using SMA actuators,” Turning Maneuver: A Mechanical Perspective,” Exp. Mech., vol. 52, no.7
Bioinspir. Biomim., vol. 7, no. 3. Feb. 2012, Art. no. 036006. pp. 693-703, 2012.
[41] P. T. Chen, S. P. Joshi, S. Swartz, K. Breuer, G. W. Reich, “Bat-inspired [65] T. L. Hedrick, A. A. Biewener, “Low speed maneuvering flight of the
flapping flight,” In 22nd AIAA/ASME/AHS Adaptive Structures rose-breasted cockatoo (Eolophus roseicapillus). I. Kinematic and
Conference, 2014, pp. 1120. neuromuscular control of turning,” J. Exp. Biol., vol. 210, no. 11, pp.
[42] B. Parslew, W. J. Crowther, “Simulating avian wingbeat kinematics,” J. 1897–1911. Jun. 2007.
Biomech., vol. 43, no. 16, pp. 3191-3198, Dec. 2010. [66] S. N. Fry, R. Sayaman, M. H. Dickinson, “The aerodynamics of free-flight
[43] H. Wagner, M. Weger, M. Klaas, W. Schröder, “Features of owl wings maneuvers in Drosophila,” Science, vol. 300, no. 5618, pp. 495-498, Apr.
that promote silent flight,” Interface focus, vol. 7, no. 1, Feb. 2017, Art. 2003.
no. 20160078. [67] F. T. Muijres, M. J. Elzinga, J. M. Melis, M. H. Dickinson, “Flies evade
[44] P. G. Puranik, G. Gopalakrishna, N. Chari, “Flight parameters of the looming targets by executing rapid visually directed banked turns,”
short-nosed fruit bat Cynopterus marginatus,” Proc. Indian Acad. Sci., vol. Science, vol. 344, no. 6180, pp. 172-177, Apr. 2014.
83, no. 4, pp. 160-165, Apr. 1976.
[45] R. E. Brown, J. J. Baumel, R. D. Klemm, “Anatomy of the propatagium:
the great horned owl (Bubo virginianus),” J. Morphol., vol. 219, no. 2, pp.
205-224, Feb. 1994.
[46] N. S. Proctor, P. J. Lynch, “Manual of ornithology: avian structure &
function,” Yale University Press, 1993.
[47] D. K. Riskin, D. J. Willis, J. Iriarte-Díaz, T. L. Hedrick, M. Kostandov, J.
Chen, D. H. Laidlaw, K. S. Breuer, S. M. Swartz, “Quantifying the