Ocean Engineering 219 (2021) 108324

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Design, fabrication, and characterization of a multimodal hybrid aerial

underwater vehicle
Di Lu a,b,c , Chengke Xiong a,b,c , Hexiong Zhou a,b,c , Chenxin Lyu a,b,c , Rui Hu a,b,c , Caoyang Yu a,c ,
Zheng Zeng a,c ,∗, Lian Lian a,c ,∗
a State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China
b School of Naval Architecture Ocean & Civil Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China
c School of Oceanography, Shanghai Jiao Tong University, Shanghai, 200030, China

ARTICLE INFO ABSTRACT

Keywords: Hybrid aerial underwater vehicle is a novelty able to fly and swim, which is extensively suitable for missions
Unmanned aerial underwater vehicle like water sampling, observing semi-submerged structure, and underwater exploration, etc. Motivated by
Unmanned aerial–aquatic vehicle the efficiency of fixed-wing unmanned aerial vehicles (UAVs) and underwater gliders in their own specific
Flying submarine
environments and the maneuverability of rotary-wing UAVs, this paper presents an improved design of a
Unmanned aerial vehicle
multimodal HAUV capable of level and vertical flight, hovering, and underwater glide. Specially, to balance
Underwater glider
flight payload and weight under water, an innovative configuration is proposed, which includes a newly
developed lightweight pneumatic buoyancy system and excludes the linear actuator commonly used for pitch
control of gliders. Moreover, motor arms can fold for better hydrodynamics. All these tradeoffs demand an
adequate match between the fuselage aerodynamics and propulsion property, and special management of
buoyancy and weight to guarantee predicted performances. Therefore, key design principles are proposed and
elaborated. Based on the principles, a prototype, named Nezha III, was fabricated and tested. The prototype’s
performance characterization in the regions of underwater, flight, and water/air transition are presented. The
results demonstrate the desired performance of the prototype under different modes, and notably prove its
outstanding capability of diving to 50 m depth.

1. Introduction Taher et al., 2016). A multi-robot system was established to carry out
multi-domain monitoring of coral reefs (Shkurti et al., 2012). A new
Conventionally, missions scenarios in the ocean that supported project that builds a collaborative team of heterogeneous robots for
by surface vessels incur huge expenditure in terms of the cost of obtaining a multi-domain awareness on a floating structure like ship
time, labor and other resources. Besides, oceanographic phenomenon is and iceberg is recently reported (Ross et al., 2019). The heterogeneous
highly variable and sometime unpredictable. The in-situ observations
multi-robot teams benefit the ocean monitoring by providing rich data
would fail if the vessels cannot reach the spot in time. The advent
collected from air, surface, and underwater environments. But it is not
of unmanned vehicles has benefited the underwater technique and
changed the way the ocean is observed. They provides a view of easy and inexpensive to integrating different vehicles into one system
the interior ocean with higher spatial and temporal resolution (Davis harmoniously, and thus it is a great challenge to organize and control
et al., 2002). Based on the great development of unmanned vehicles, such system well.
there are increasing interests in coordinating multiple unmanned ve- Lately, the field of hybrid aerial underwater vehicle (HAUVs) that
hicles for carrying out complex missions efficiently under advanced enable operations in both air and water is newly emerging. Thanks
control techniques (Peng et al., 2020). Recently, researchers are even to the HAUV’s capabilities of locomotion in different medium, scien-
attempting the joint use of different unmanned vehicles, including tists can minimize the complexity of multi-domain monitoring mis-
unmanned aerial vehicles (UAVs), unmanned surface vehicles (USVs) as sions since data can be collected from different environment by only
well as unmanned underwater vehicles (UUVs), for ocean monitoring. one platform. Other applications HAUVs can be extensively used for
Researchers achieved progress in building and operating a distributed
includes but not limited to intelligence gathering, surveillance and
heterogeneous autonomous sensor network that combined UAVs, USVs,
reconnaissance, inspection, search and rescue.
and UUVs in coastal environments (Valdivia y Alvarado et al., 2011;

∗ Corresponding authors at: State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China.
E-mail addresses: zheng.zeng@sjtu.edu.cn (Z. Zeng), llian@sjtu.edu.cn (L. Lian).

https://doi.org/10.1016/j.oceaneng.2020.108324
Received 4 March 2020; Received in revised form 12 July 2020; Accepted 28 October 2020
Available online 4 November 2020
0029-8018/© 2020 Elsevier Ltd. All rights reserved.
D. Lu et al. Ocean Engineering 219 (2021) 108324

Multi-domain locomotion imposes a considerable challenge to the

vehicle design. The existing approaches to HAUVs popularly consider
endowing two main subsets of UAVs, i.e., fixed-wing and rotary-wing
UAVs, with underwater functionality. As their aerial counterpart, fixed-
wing HAUVs are outstanding for flight endurance, range and cruise
speed. An early instance of fixed HAUVs was Test Sub which was ini-
tially launched from a mothership aircraft and glided to the surface and
dived (Young, 2014). However, it could not take off from the surface.
Lately, a single aerial rotor fixed-wing HAUV, EagleRay, was tested and
characterized through flight, underwater cruise, and repeated water–air
transition tests. It had been experimentally proved to be fully functional
in distinct medium (Weisler et al., 2017). Later, a similar fixed-wing
HAUV except for an aft water rotor being added for underwater ma-
Fig. 1. 3D rendering picture of the proposed HAUV integrating the features of
neuver was fabricated and it completed a full cycle of multi-domain fixed-wing UAVs, rotary-wing UAVs, and UGs.
mission (Stewart et al., 2018). To optimize the water transition profile
with better hydrodynamics for the fixed-wing HAUVs, new aquatic
micro aerial vehicle (AquaMAV) designed with morphing wings was
fabricated and its performance was also experimentally tested and char-
developed. But the take-off propulsion used a one-time-only use water
acterized. This design addresses the shortages of the current systems,
jet thruster, which denies repeated water–air transition (Siddall et al.,
and improve the HAUV’s capability of conducting underwater missions
2018, 2017; Siddall and Kovač, 2017).
at increased depth over longer periods as well as higher levels of
In the last five years, the effort in rotary-wing HAUVs has presented
autonomy. This upgrade presents three innovative achievements:
distinctive prospect in well-controlled multi-domain observations. The
first known rotary-wing HAUVs was proposed conceptually and mod- 1. Use a lightweight super high pressure pneumatic buoyance con-
eled in 2014 (Drews et al., 2014). Recently, the same team presented trol system as a substitute for conventional buoyancy system of
their first but still in-production prototype with four upper aerial rotors UGs. This system is testified to enable dives to the depth of 50
and four lower underwater rotors (da Rosa et al., 2018). To cross m for our prototype, which to our best knowledge makes the
different medium seamlessly and repeatedly, an HAUV, Naviator2, was vehicle be the first one of its kind that can make such deep dive.
evolved from a multirotor. It had four dual-aerial-propellers propul- 2. Make a trade-off design that abandoned the translational mech-
sions to drive the vehicle underwater and in the air (Maia et al., anism (but keep the rotational one for tuning glide) of the inner
2015, 2017). Moreover, a design that combined a ballast system with movable mass commonly used in UGs to further balance the
a quadrotor-based flying system resulted in a different HAUV, Loon flight payload and the underwater weight. Due to the inability
Copter, whose buoyancy and attitude in the water were controlled of longitudinally regulating the mass distribution for the vehicle,
by pumping in or out the ambient fluid (Alzu’bi et al., 2018). These the buoyancy–pitch coupling of the vehicle is studied in depth
rotary-wing HAUVs have been experimentally proved with superiority and took full advantage of as an innovative strategy for pitch
in terms of smooth, controllable, and repeated cross-domain transitions. control of planar glide.
Like their aerial counterparts, rotary-wing HAUVs are suitable for 3. Design folding rotor arms for better hydrodynamics underwater,
high-resolution observations where high maneuverability and hovering and theoretically and practically address the consequent over-
ability are required. However, all these advantages are at the cost of the turn caused by rise of the center of gravity (CG) during the
fixed-wing HAUV’s endurance, range, and speed. unfolding process when the vehicle prepares to take off from the
Note that the basic structures of most present designs for HAUV surface.
are closely related to that of typical UAVs. Although these vehicles are
capable of well flight and water/air transition, their abilities to move Rest of the paper is organized as follows. In Section 2 the concept
underwater are still limited in terms of working depth and underwater of the proposed multimodal HAUV and the nominal maneuvers are
endurance, which weakens their utility in the ocean. defined, while key design principles of the vehicle are elaborated in
Among the existing UUV technology, underwater gliders (UGs) Section 3. Then, Section 4 details the fabrication of the prototype
have been proved to be the most efficient platform and already been in terms of five subsystems. The performance characterization of the
widely used for long-range, deep, and long-term missions (Edwards prototype in different operation regimes are presented in Section 5.
et al., 2017). Inspired by the features of UGs and given the respective Finally, Section 6 comes to the conclusion.
downsides of existing HAUVs, the concept of novel multimodal HAUV
that coordinates flight endurance, speed, maneuverability as well as un- 2. Operation overview
derwater efficient locomotion and endurance for a single platform was
firstly proposed in our previous work (Lu et al., 2018). This multimodal The concept of the proposed HAUV temps to obtain a higher level
HAUV basically integrated the functionalities of UGs and tail-sitter of integration of fixed-wing and rotary-wing UAVs as well as UGs,
UAVs (a combined form of fixed-wing and rotary-wing UAVs), indicat- enabling larger flight payload and deeper and longer operation under
ing the combined ability of underwater glide as well as vertical and water. A conceptual design is presented in Fig. 1.
level flight. Later, a proof-of-concept prototype, Nezha, was designed In a full cycle of operation (Fig. 2), the vehicle is expected to
and experimentally tested (Lu et al., 2019), proving the feasibility of the perform level and vertical flight, hovering, surface drifting, underwater
multimodal HAUV concept. In Nezha’s design, a pneumatic buoyancy glide as well as water–air transition for different mission requirements.
system was employed for underwater gliding. Nevertheless, Nezha was To achieve that, the propulsion system and other actuators of UAVs
embarrassed by its limited underwater operation. First, it could dived and UGs are adapted accordingly for the vehicle’s different operation
no more than 5 m. Second, precise control on gliding performance modes.
was denied. Third, it had no turning ability underwater. Fourth, its
hydrodynamics was undesirable due to the rotor arms. Fifth, it had a 2.1. Level flight
small payload capacity.
In this paper, a brand-new design of a multimodal HAUV is pre- Being capable of level flight, the vehicle can travel a far distance
sented. Then, a proof-of-concept prototype, named Nezha III, was rapidly and efficiently in oceanic mission scenarios.

2
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 4. The vehicle drifting at the water’s surface with its arms folded can always hold
an upright position.

After finishing the transition, the vehicle flies vertically in the air and
the thrust mainly acts against the gravity, ensuring a new equilibrium
Fig. 2. An illustration of multiple operations for the proposed vehicle in a cross-domain state of vertical flight. In a vertical flight, the vehicle not only has
mission scenario. Solid arrow lines indicate small range movements while dotted arrow
lines indicate long range travels.
sufficient thrust that allows for heavier takeoff weight but also benefits
from good maneuverability.
Like the rotary-wing UAVs, the vertical flight maneuver highly
consumes energy. However, some special scenarios do require a precise
control and excellent maneuverability for an HAUV. For example, a
detailed monitoring in a confined environments or a close-up view of
feature of interest like offshore oil platform can be only satisfied by
a slow flight. Other applications such as ocean photography and water
sampling (Ore et al., 2015) need the platform to execute a zero-velocity
loitering in air. Therefore, integrating the abilities of vertical flight and
hovering makes the vehicle versatile for extended applications.
Besides the considerations above, vertical flying maneuver is bene-
ficial for the takeoff and landing on the water’s surface, which will be
discussed in later subsections.

Fig. 3. (a) Switch from vertical flight to level flight along with increasing forward
2.3. Surface drifting
speed. (b) Switch from level flight to vertical flight along with decreasing forward
speed. Red and green arrow lines respectively indicate increasing and decreasing thrust The vehicle can naturally drift at the sea surface in an upright posi-
force of corresponding rotors. (For interpretation of the references to colour in this tion. For an HAUV, it is useful to make itself function as a free-drifting
figure legend, the reader is referred to the web version of this article.)
surface buoy that has been utilized for many scientific activities (Her-
bers and Janssen, 2016; Gonzalez-Haro et al., 2016). In addition,
floating on surface guarantees a good wireless communication for con-
Inspired by the mechanism of tail-sitter UAVs, the vehicle can tinuous data transfer, and may support a solar charger for the vehicle
switch to level flight without extra tilting mechanism, thus saving in the future.
weight and simplifying the vehicle design (Zhang et al., 2017). Fig. 3(a) Surface drifting requires no extra mechanism on the present system
demonstrates the switching maneuver from vertical to level flight. The to realize. Fig. 4 demonstrates the vehicle stabilized at the water’s
rotors (motor/propeller combinations) in different sides of the wings surface with its arms folded. The pneumatic variable buoyancy system
spin in different rates to produce a moment to pitch the vehicle forward has a bladder, namely the actuator responsible for stable flotation and
along with the increasing speed. Consequently, the induced lift of the gliding locomotion underwater, mounted outside on the top. To float
fixed wings increases enough to counteract the gravity while the force at the surface, the bladder is inflated to such a volume that the vehicle
against the drag is dominated by the thrust. After that, the thrust, the becomes positively buoyant. Meanwhile, the rotor arms fold down to
lift, the gravity, and the drag will finally reach a new equilibrium, lower the CG under the center of buoyancy (CB) so that the vehicle
which settles the vehicle into a stable level flight. could stay at a stable upright floating position.
The air drag is normally much smaller than the vehicle’s gravity,
therefore the power of rotors in a level flight is lower compared with 2.4. Underwater glide
that in a vertical flight. In our previous work (Lu et al., 2019), flight
simulations were carried out to qualitatively evaluate the decrease of One of the novelties of the proposed vehicle is merging the func-
power consumption happened when the vehicle changed to level flight. tionalities of UGs and UAVs. This is the first attempt to adapt the
This fact of the proposed vehicle will be verified and characterized mechanism of UGs in a flyable platform, and thus a try to extend the
quantitatively in Section 5. underwater endurance for HAUVs.
The only driving force pushing a UG forward is the horizontal
2.2. Vertical flight and hovering component of lift induced by the gliding motion due to the inequality
between gravity and buoyancy. Without any propulsion, gliding motion
The proposed vehicle also features the capabilities of vertical flight is quite energy efficient. For the proposed vehicle, a glide strategy is
and hovering as rotary-wing UAVs. thus implemented solely via the pneumatic buoyancy system because
To change from level to vertical flight, the thrust of rotors on the linear movable mass is abandoned for weight reduction. For turning
different sides of the wings differs to create a pitch moment that brings direction underwater, an inner movable mass that can rotate 360◦
the vehicle back to vertical as shown in Fig. 3(b). As the angle of around the central axis is added to the vehicle. Rotating the movable
attack (AOA) of the wings increases and the horizontal component of mass will roll the vehicle to turn. When the movable mass is located
the thrust decreases, the vehicle gradually slows down due to the drag. at the most bottom side, the CG moves far off the central axis to the

3
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 5. Diagram of buoyancy and gravity of the vehicle in downward (left) and upward
(right) glides.

bottom so that the vehicle acquires enough ‘‘bottom-heaviness’’ for

stable gliding (Leonard and Graver, 2001). Also note that six arms fold
and all rotors keep idle for better hydrodynamics in a glide.
Fig. 5 demonstrates different managements of buoyancy with re-
spects to the gravity in downward and upward glides. Regulating
buoyancy and the CB position relative to the CG can control the gliding
performance. In a downward glide, the vehicle becomes negative buoy-
ant and descends as a result of the deflation of the bladder. Meanwhile,
the CB gradually shifts behind the CG as the bladder shrinks, which
results in a trim by the head. In an upward glide, the bladder is inflated
with the compressed air stored inside an inner cylinder in advance.
The expanding bladder increases buoyancy and changes the CB position
to the front of the CG. Then the vehicle is trimmed by the stern and
ascends. In both descent and ascent cases, the lift of the wings enables a
horizontal forward displacement of the vehicle. Eventually, the vehicle
moves forward by repeating alternating descent and ascent, i.e., a
sawtooth trajectory.
Unlike UGs, a significant characteristic of the proposed vehicle is
the exclusion of the longitudinal regulation of movable mass for the
purpose of maximum reduction of flight payload without losing the
underwater gliding ability. Due to this tradeoff design, the pitch motion
of the vehicle is coupled with the change of buoyancy, making the Fig. 6. Workflow diagram of full operations of the water-to-air transition and the
vehicle highly underactuated and thus harder to control when moving reverse transition.
underwater. In the next section, the design principle helping address
this problem is elaborated.
Fig. 6 shows the procedures of water–air transitions. From underwa-
2.5. Water–air transition ter gliding to flying, the vehicle firstly inflates the bladder to surface.
Then the inner movable mass rotates to a specific position where the CG
Smooth cross-domain transition is the key to bridging underwater lies below the CB along with the central axis, resulting in the body in an
locomotion and flying. Vertical takeoff and landing (VTOL) is chosen upright position. As seen in the figure, the CG will rise over the CB as
as the water–air transition strategy for the following reasons. First, the arms rise up. Without proper buoyancy management, the couple of
VTOL permits heavier takeoff weight. For small or medium HAUVs, gravitational and buoyant forces will produce an overturning moment
their payload capacities are very limited. However, heavier waterproof instead of a restoring moment. Thus, the bladder needs to be further
structures and devices are almost inevitable in designing an HAUV expanded to guarantee metacentric stability before the arms unfold.
for deeper operation. Applying VTOL is thus most likely to make This design concern is fully discussed in the following section.
successful takeoff for a heavy vehicle. Second, wings are designed To achieve a successful landing, same procedure but in reverse order
to be floodable to avoid the great extra buoyancy due to their large is taken. The arms are folded right after the vehicle lands on the surface
volume. VTOL maneuver helps reduce the negative effect of the water safely and all rotors are shut down. Then, the movable mass is shifted
entrained in the floodable wings when exiting water and relieve the to the position where the most bottom heaviness and gliding stability
impact when entering water. As the model of passively draining water are met. After that, the vehicle is ready for underwater glide.
indicates (Stewart et al., 2019), the faster the egress speed is, the
greater the hydrodynamic force imposed on the vehicle is. In another 3. Design considerations
concern, ingress speed and attitude must be controlled. Otherwise, the
vehicle may suffer from severe water impact or even crash in the air- The variable used in this paper are defined in Table 1. As in Fig. 7,
to-water ingress (Young, 2014). That situation might get harsher for two reference frames, i.e., the inertial frame and the body fixed frame,
larger and heavier platform. So, implementing VTOL maneuver is good are used. The inertial frame is defined as conventional north-east-
for the vehicle slowly crossing the water’s surface under well control. down frame while the body fixed frame is defined in front-right-down
Third, VTOL is naturally a part of the maneuver of vertical flight. Thus orientation with the origin coincident with the centroid of the fixed
no extra design is needed. fuselage (excludes bladder).

4
D. Lu et al. Ocean Engineering 219 (2021) 108324

Table 1
Nomenclature.
Variable Description
𝛼 Angle of attack, tan 𝛼 = 𝑢∕𝑤
𝛽 Rotation angle of rotatable mass around the central axis
𝐵 Total buoyancy of the vehicle
( )
𝐷 Drag force, 𝐷 = 𝐾𝐷0 + 𝐾𝐷 𝛼 2 𝑉 2
𝛥 Mass of fluid displaced by the body except for the bladder
𝜂 Heeling angle of the vehicle floating at the surface
𝑔 Gravitational acceleration
𝐺 Total vehicle gravity
𝐺𝑍 Length of righting lever
𝐺𝑀 Metacentric height
𝐺𝑝 Gravity of rotatable mass
Fig. 7. Frame assignment and definitions of the elements of the weight and buoyancy. 𝐺𝑠 Gravity of stationary body mass
𝐽2 Component of total inertia of vehicle/fluid system in 𝐞2 body axis
𝐾𝑠 Constant coefficients related to lift, drag, and viscous moment
( )
𝐿 Lift force, 𝐾𝐿0 + 𝐾𝐿 𝛼 𝑉 2
Normally, the lateral positions of the mass of fluid displaced by the 𝑚𝑖 Element of sum of stationary body and added mass in 𝐞𝑖 body axis
bladder 𝑚𝑏 and the nonuniformly distributed mass 𝑚𝑤 , i.e., 𝑦𝑏 and 𝑦𝑤 , (𝑖 = 1, 3)
𝑚𝑏 Mass of fluid displaced by the bladder
are designed to be zero for the symmetry of the rolling behavior and
𝑚𝐵 Mass of total displaced fluid, 𝑚𝐵 = 𝑚𝑏 + 𝛥
the passive stability of yaw motion under water. 𝑚̄ Rotatable mass
In flight, the propulsion system (six motor/propeller combinations) 𝑚ℎ Uniformly distributed hull mass
supports different flying maneuvers. As for underwater locomotion 𝑚𝑤 Point mass for nonuniform mass distribution
𝑚𝑠 Stationary body mass, 𝑚𝑠 = 𝑚ℎ + 𝑚𝑤
and the surface floatation, the vehicle’s behavior is controlled via the
𝑚𝐺 Total vehicle mass, 𝑚𝐺 = 𝑚𝑠 + 𝑚̄
pneumatic buoyancy system and the rotatable mass. ( )
𝑀𝐷𝐿 Viscous moment, 𝐾𝑀0 + 𝐾𝑀 𝛼 𝑉 2
𝑃𝑝𝑖 Component of linear momentum of 𝑚̄ in 𝐞𝑖 body axis, (𝑖 = 1, 3)
𝑃𝑤𝑖 Component of linear momentum of 𝑚𝑤 in 𝐞𝑖 body axis, (𝑖 = 1, 3)
3.1. Concern about planar glide 𝑞 Angular velocity of the vehicle around 𝐞2 body axis
𝑅𝑝 Offset distance of 𝑚̄ from the central axis
𝐫𝑏 Position of 𝑚𝑏 in body coordinates, 𝐫𝑏 = (𝑥𝑏 , 𝑦𝑏 , 𝑧𝑏 )
As aforementioned, the presented vehicle is more underactuated
𝐫𝐵 Position of 𝑚𝐵 in body coordinates, 𝐫𝐵 = 𝑚𝑏 𝐫𝑏 ∕𝑚𝐵
than UGs when moving underwater because it has only one actuator 𝐫𝑤 Position of 𝑚𝑤 in body coordinates, 𝐫𝑤 = (𝑥𝑤 , 𝑦𝑤 , 𝑧𝑤 )
(the bladder). Once the bladder’s volume changes, the pitch attitude 𝐫𝑝 Position of 𝑚̄ in body coordinates,
( )
and the buoyancy will vary synchronously, forbidding separate controls 𝐫𝑝 = (𝑥𝑝 , 𝑦𝑝 , 𝑧𝑝 ) = 𝑥𝑝 , 𝑅𝑝 cos 𝛽, 𝑅𝑝 sin 𝛽
𝐫𝑠 Position of 𝑚𝑠 in body coordinates, 𝐫𝑠 = (𝑥𝑠 , 𝑦𝑠 , 𝑧𝑠 ) = 𝑚𝑤 𝐫𝑤 ∕𝑚𝑠
of pitch attitude and buoyancy. To obtain the intended performance of
𝐫𝐺 Position of 𝑚𝐺 in body coordinates,
the glide in vertical plane, proper weight and buoyancy managements ( )
𝐫𝐺 = (𝑥𝐺 , 𝑦𝐺 , 𝑧𝐺 ) = 𝑚𝑠 𝐫𝑠 + 𝑚𝐫 ̄ 𝑝 ∕𝑚𝐺
are crucial for the design and achieved by exploiting the nature of 𝜃 Pitch angle
pitch–buoyancy coupling. 𝑇 Control thrust of propulsion
𝜏 Control torque of propulsion
The linearized model of proposed system in a steady planar glide is 𝜏𝑅 Restoring moment
obtained below as similar to that of the UG system (Graver, 2005), 𝑢 Component of linear velocity of the vehicle in 𝐞1 body axis
1 [ ( ) ( )] 𝑤 Component of linear velocity of the vehicle in 𝐞3 body axis
√
𝛿 𝑞̇ = 𝑚 𝑔 𝑥𝐺 sin 𝜃0 − 𝑧𝐺 cos 𝜃0 − 𝑚𝑏0 𝑔 𝑥𝑏 sin 𝜃0 − 𝑧𝑏 cos 𝜃0 𝛿𝜃 𝑉 Speed in vertical plane, 𝑉 = 𝑢2 + 𝑤2
𝐽2 𝐺
𝜉 Glide path angle, 𝜉 = 𝜃 − 𝛼
1 ( )
− ̄ 0 𝑥𝑝 + 𝑚𝑤
𝑚𝑢 ̄ 0 𝑧𝑝 + 𝑚𝑤 𝑢0 𝑥𝑤 + 𝑚𝑤 𝑤0 𝑧𝑤 𝛿𝑞
𝐽2
[ ( )( ) ]
1 ( )
+ 𝑚3 − 𝑚1 𝑤0 + 𝐾𝑀0 + 𝐾𝑀 𝛼0 2𝑢0 − 𝐾𝑀 𝑤0 𝛿𝑢
𝐽2
1 (
[ ) ( )( ) ] (1) 1 [( ) ] 1 ( )
+ 𝑚3 − 𝑚1 𝑢0 + 𝐾𝑀0 + 𝐾𝑀 𝛼0 2𝑤0 + 𝐾𝑀 𝑢0 𝛿𝑤 𝛿 𝑤̇ = 𝛥 + 𝑚𝑏0 − 𝑚𝐺 𝑔 sin 𝜃0 𝛿𝜃 + 𝑚̄ + 𝑚𝑤 + 𝑚1 𝑢0 𝛿𝑞
𝐽2 𝑚3 𝑚3
{[( )( ) ]
1
1 ( ) 𝑧𝑝 − 𝐾𝐿0 + 𝐾𝐿 𝛼0 2𝑢0 − 𝐾𝐿 𝑤0 cos 𝛼0
+ 𝑔 𝑥𝑏 cos 𝜃0 + 𝑧𝑏 sin 𝜃0 𝛿𝑚𝑏 − 𝛿 𝑃̇ 𝑝1 𝑚3
𝐽2 𝐽2 ( )
𝑥𝑝 𝑧 𝑥 + 𝐾𝐿0 + 𝐾𝐿 𝛼0 sin 𝛼0 ⋅ 𝑤0
+ 𝛿 𝑃̇ − 𝑤 𝛿 𝑃̇ + 𝑤 𝛿 𝑃̇ [( )( ) ]
𝐽2 𝑝3 𝐽2 𝑤1 𝐽2 𝑤3 + 𝐾𝐷0 + 𝐾𝐷 𝛼02 2𝑢0 − 2𝐾𝐷 𝛼0 𝑤0 sin 𝛼0
1 [( ) ] 1 ( ) ( ) }
𝛿 𝑢̇ = 𝛥 + 𝑚𝑏0 − 𝑚𝐺 𝑔 cos 𝜃0 𝛿𝜃 + −𝑚̄ − 𝑚𝑤 − 𝑚3 𝑤0 𝛿𝑞 − 𝐾𝐷0 + 𝐾𝐷 𝛼02 cos 𝛼0 ⋅ 𝑤0 𝛿𝑢
𝑚1 𝑚1
{[( )( ) ] {[( )( ) ] (3)
1 1
+ 𝐾𝐿0 + 𝐾𝐿 𝛼0 2𝑢0 − 𝐾𝐿 𝑤0 sin 𝛼0 − 𝐾𝐿0 + 𝐾𝐿 𝛼0 2𝑤0 + 𝐾𝐿 𝑢0 cos 𝛼0
𝑚 𝑚3
( 1 ) ( )
− 𝐾𝐿0 + 𝐾𝐿 𝛼0 cos 𝛼0 ⋅ 𝑤0 − 𝐾𝐿0 + 𝐾𝐿 𝛼0 sin 𝛼0 ⋅ 𝑢0
[( )( ) ] [( )( ) ]
− 𝐾𝐷0 + 𝐾𝐷 𝛼02 2𝑢0 − 2𝐾𝐷 𝛼0 𝑤0 cos 𝛼0 + 𝐾𝐷0 + 𝐾𝐷 𝛼02 2𝑤0 + 2𝐾𝐷 𝛼0 𝑢0 sin 𝛼0
( ) } ( ) }
− 𝐾𝐷0 + 𝐾𝐷 𝛼02 sin 𝛼0 ⋅ 𝑤0 𝛿𝑢 + 𝐾𝐷0 + 𝐾𝐷 𝛼02 cos 𝛼0 ⋅ 𝑢0 𝛿𝑤
{[( )( ) ] (2) 1 1 ̇ 1 ̇
1 − 𝑔 cos 𝜃0 𝛿𝑚𝑏 − 𝛿𝑃 − 𝛿𝑃
+ 𝐾𝐿0 + 𝐾𝐿 𝛼0 2𝑤0 + 𝐾𝐿 𝑢0 sin 𝛼0 𝑚3 𝑚3 𝑝3 𝑚3 𝑤3
𝑚
( 1 )
+ 𝐾𝐿0 + 𝐾𝐿 𝛼0 cos 𝛼0 ⋅ 𝑢0 where 𝐾𝑠 are the constant coefficients relative to drag 𝐷, lift 𝐿, and
[( )( ) ]
− 𝐾𝐷0 + 𝐾𝐷 𝛼02 2𝑤0 + 2𝐾𝐷 𝛼0 𝑢0 cos 𝛼0 viscous moment 𝑀𝐷𝐿 due to wing and body shown in Fig. 8, and the
( ) } variables denoted with subscript ‘‘0’’ are the corresponding value at
+ 𝐾𝐷0 + 𝐾𝐷 𝛼02 sin 𝛼0 ⋅ 𝑢0 𝛿𝑤
planar gliding equilibria while the variables with a former ‘‘𝛿’’ represent
1 1 ̇ 1 ̇
+ 𝑔 sin 𝜃0 𝛿𝑚𝑏 − 𝛿𝑃 − 𝛿𝑃 the small perturbation from their respective equilibrium value.
𝑚1 𝑚1 𝑝1 𝑚1 𝑤1

5
D. Lu et al. Ocean Engineering 219 (2021) 108324

Therefore, it has,
√ √
𝑉𝑡 = ||𝑢𝑡 || + ||𝑤𝑡 || > ||𝑢0 || + ||𝑤0 || = 𝑉0
2 2 2 2

which says an increase of velocity. Contrarily, a negative 𝛿𝑢 and a

positive 𝛿𝑤 happen if 𝛿𝑚𝑏 is negative. Then, one can infer a decrease
in velocity (𝑉𝑡 < 𝑉0 ) with Theorem 1.
In a steady downward glide, 𝜃0 < 0. The vehicle has positive velocity
components in 𝐞1 and 𝐞3 direction due to a positive AOA, i.e., 𝑢0 > 0 and
𝑤0 > 0. Increasing the bladder’s buoyancy (𝛿𝑚𝑏 > 0) produces negative
𝛿𝑢 and 𝛿𝑤. Recalling Theorem 1, it is clear that 𝑉𝑡 < 𝑉0 . However, if 𝛿𝑚𝑏
is negative, there are positive variations in both velocity components
(𝛿𝑢 > 0 and 𝛿𝑤 > 0). Following the studies in the upward glide, it can
Fig. 8. Lift, drag, and moment on the vehicle gliding underwater.
be readily proved that 𝑉𝑡 > 𝑉0 .
In conclusion, the bladder mounted at the forward end of the fuse-
lage affects the gliding performance in such a way: inflation (deflation)
Substituting 𝛿𝜃 = 𝛿𝑞 = 𝛿𝑢 = 𝛿𝑤 = 0 and 𝛿𝑃𝑤𝑖 = 𝛿𝑃𝑝𝑖 = 0 (𝑖 = 1, 3) of the bladder will continuously trim the vehicle by the stern (head) and
into Eqs. (1), (2), and (3), the influence of the bladder on the vehicle’s increase upward (downward) gliding trend. Therefore, design parame-
dynamic can be independently observed as follows: ters of the gravity and buoyancy that needed for the desired gliding
1 ( ) performance are discussed based partially on a extreme downward
𝛿 𝑞̇ = 𝑔 𝑥𝑏 cos 𝜃0 + 𝑧𝑏 sin 𝜃0 𝛿𝑚𝑏 (4)
𝐽2 glide equilibria and a extreme upward glide equilibria.
1 The model of a vehicle in a steady planar glide holds the following
𝛿 𝑢̇ = 𝑔 sin 𝜃0 𝛿𝑚𝑏 (5)
𝑚1 relationships (Graver, 2005),
( √ )
1 ( )
𝛿 𝑤̇ = − 𝑔 cos 𝜃0 𝛿𝑚𝑏 (6) 1 𝐾𝐿 𝐾
𝑚3 𝛼= tan 𝜉 −1 + 1 − 4 𝐷 cot 𝜉 𝐾𝐷0 cot 𝜉 + 𝐾𝐿0 (7)
2 𝐾𝐷 𝐾2
( ) 𝐿
where 𝜃0 ∈ − 𝜋2 , 𝜋2 . with
Because the bladder is mounted at the head of the vehicle and √
⎛ ⎛ 𝐾 ⎛𝐾 ( 𝐾 )2 𝐾 ⎞⎞⎞
symmetric around 𝐞1 axis, it has 𝑥𝑏 > 0 and 𝑧𝑏 = 0. Accordingly, a small 𝜋 ⎜ 𝐿0 − 𝐿0 𝐷0
𝜉 ∈ ⎜− , tan−1 ⎜2 𝐷 + ⎟⎟⎟
increase of 𝑚𝑏 , i.e., 𝛿𝑚𝑏 > 0, results in an increase of pitch in vertical ⎜ 2 ⎜ 𝐾𝐿 ⎜ 𝐾𝐿 𝐾𝐿 𝐾𝐷 ⎟⎟⎟
⎝ ⎝ ⎝ ⎠⎠⎠
plane due to 𝛿 𝑞̇ > 0. Conversely, reducing buoyancy of the bladder, √
⎛ ⎛ 𝐾 ⎛𝐾 ( 𝐾𝐿0 )2
𝐾𝐷0 ⎞⎞ 𝜋 ⎞
i.e., 𝛿𝑚𝑏 < 0, leads to decrease in pitch angle due to 𝛿 𝑞̇ < 0. 𝐿
The following theorem will be used to analyze the change of glide ∪ ⎜ tan−1 ⎜2 𝐷 ⎜ 0 + + ⎟⎟ , ⎟ ,
⎜ ⎜ 𝐾𝐿 ⎜ 𝐾𝐿 𝐾𝐿 𝐾𝐷 ⎟⎟ 2 ⎟
velocity caused by the variation of 𝑚𝑏 . ⎝ ⎝ ⎝ ⎠⎠ ⎠
and
Theorem 1. 𝑦 = (𝛿𝑥 + 𝑥)2 − 𝑥2 is a real-valued continuous function ( ( ) ( ))
1
𝑚𝐺 − 𝑚𝐵 = − sin 𝜉 𝐾𝐷0 + 𝐾𝐷 𝛼 2 + cos 𝜉 𝐾𝐿0 + 𝐾𝐿 𝛼 𝑉 2 , (8)
for 𝛿𝑥 with 𝛿𝑥 ∈ 𝑅 and a real constant 𝑥. If 𝑥 > 0, 𝑦 > 0 when 𝑔
𝛿𝑥 ∈ (−∞, −2𝑥) ∪ (0, ∞) and 𝑦 < 0 when 𝛿𝑥 ∈ (−2𝑥, 0). If 𝑥 < 0, 𝑦 > 0 ( ) ( )
𝑚𝐺 𝑔 𝑥𝐺 cos 𝜃 + 𝑧𝐺 sin 𝜃 − 𝑚𝐵 𝑔 𝑥𝐵 cos 𝜃 + 𝑧𝐵 sin 𝜃
when 𝛿𝑥 ∈ (−∞, 0) ∪ (−2𝑥, ∞) and 𝑦 < 0 when 𝛿𝑥 ∈ (0, −2𝑥). ( ) ( ) 2 (9)
= 𝑚3 − 𝑚1 𝑢𝑤 + 𝐾𝑀0 + 𝐾𝑀 𝛼 𝑉 .
Proof. The given function is equal to 𝑦 = 𝛿𝑥 (2𝑥 + 𝛿𝑥), which has two with 𝜃 = 𝜉 + 𝛼, 𝑢 = 𝑉 cos 𝛼, and 𝑤 = 𝑉 sin 𝛼.
different roots for a given non-zero constant 𝑥, i.e., 𝛿𝑥 = 0 or 𝛿𝑥 = −2𝑥. First of all, consider the case that the bladder shrinks completely
It is easy to tell the sign of 𝑦 for different 𝛿𝑥. (𝑚𝑏 = 0), which implies the maximum descending rate and nose-down
Denote 𝑉𝑡 the updated velocity of vehicle after a non-zero 𝛿𝑚𝑏 is pitch angle. Given a desired glide speed 𝑉𝑑_𝑑𝑜𝑤𝑛 and a glide path angle
applied in an arbitrarily small time interval 𝛿𝑡 > 0. It has, 𝜉𝑑_𝑑𝑜𝑤𝑛 in this condition, the total mass of the vehicle and the buoyancy
√ of the fuselage should be designed as below deriving from Eqs. (7) and
𝑉𝑡 = 𝑢2𝑡 + 𝑤2𝑡 (8),
( ( )
with the updated components of velocity, 𝑢𝑡 and 𝑤𝑡 , given as, 1 2
𝑚𝐺 − 𝛥 = − sin 𝜉𝑑_𝑑𝑜𝑤𝑛 𝐾𝐷0 + 𝐾𝐷 𝛼𝑑_𝑑𝑜𝑤𝑛
𝑔
𝑢𝑡 = 𝑢0 + 𝛿𝑢 ( )) (10)
2
+ cos 𝜉𝑑_𝑑𝑜𝑤𝑛 𝐾𝐿0 + 𝐾𝐿 𝛼𝑑_𝑑𝑜𝑤𝑛 𝑉𝑑_𝑑𝑜𝑤𝑛 .
𝑤𝑡 = 𝑤0 + 𝛿𝑤
𝛿𝑡 Note that
𝛿𝑢 = 𝛿 𝑢d𝑡
̇
∫0 𝑚𝐵 𝐫𝐵 = 𝑚𝑏 𝐫𝑏 , (11)
𝛿𝑡
𝛿𝑤 = ̇
𝛿 𝑤d𝑡 then the CG position can be selected by Eq. (9) as
∫0
1 (( )
Note that 𝛿𝑢 and 𝛿𝑤 are the consequence of changing 𝑚𝑏 with small 𝑥𝐺 cos 𝜃𝑑_𝑑𝑜𝑤𝑛 + 𝑧𝐺 sin 𝜃𝑑_𝑑𝑜𝑤𝑛 = 𝑚3 − 𝑚1 𝑢𝑑_𝑑𝑜𝑤𝑛 𝑤𝑑_𝑑𝑜𝑤𝑛
𝑚𝐺 𝑔
magnitude. ( ) ) (12)
2
In a steady upward glide, 𝜃0 > 0. Note that vehicle has a negative + 𝐾𝑀0 + 𝐾𝑀 𝛼𝑑_𝑑𝑜𝑤𝑛 𝑉𝑑_𝑑𝑜𝑤𝑛 .
AOA (Leonard and Graver, 2001), which gives 𝑢0 > 0, 𝑤0 < 0. A
Then, take the bladder with its maximum volume for gliding into
positive 𝛿𝑚𝑏 yields a positive 𝛿 𝑢̇ and a negative 𝛿 𝑤.
̇ Therefore, 𝛿𝑢 and
consideration, and denote 𝑚𝑏_𝑚𝑎𝑥 the mass of fluid displaced by the
𝛿𝑤 share the same sign as 𝑢0 and 𝑤0 respectively, and then the below
bladder. The vehicle must glide upward with maximum speed and ex-
relationships hold,
treme nose-up pitch attitude under such condition. However, it should
|𝑢𝑡 | = |𝑢0 + 𝛿𝑢| = |𝑢0 | + |𝛿𝑢| > |𝑢0 |
| | | | | | | | be clarified that the maximum volume of the bladder corresponding to
|𝑤𝑡 | = |𝑤0 + 𝛿𝑤| = |𝑤0 | + |𝛿𝑤| > |𝑤0 | 𝑚𝑏_𝑚𝑎𝑥 is not the same as the extreme volume that the bladder should
| | | | | | | |

6
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 11. Vehicle stability diagram showing the CG, the CB, and the metacenter (M).
The vehicle stays upright (left) and heels over to one side (right) with an angle, 𝜂.
Fig. 9. Sketch of cross-section of the rotatable mass rotating around the central axis
viewing from the aft. Orange sector represents the rotatable mass. 𝐺𝑠 and 𝐺𝑝 are the
gravity of 𝑚𝑠 and 𝑚̄ respectively. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.) in (−90◦ , 90◦ ). As in Fig. 9(a), the mass is fixed in the most bottom
place, i.e., 𝛽 = 0◦ , when the vehicle glides steadily in vertical plane. In
this situation, the vehicle has enough restoring moment against heeling
caused by lateral disturbance. In terms of rolling to the left or right
for the vehicle, the mass moves to the position where 𝛽 ∈ (0◦ , 90◦ )
or 𝛽 ∈ (−90◦ , 0◦ ) as shown in Fig. 9(b). To adjust itself to upright
orientation after surfacing, the vehicle will shift the rotatable mass to
the most upper side, i.e., 𝛽 = ±180◦ , as depicted in Fig. 9(c). By doing
so, the CG reconverges to the longitudinal axis and locates below the CB
as shown in Fig. 10. A successful transition between gliding and vertical
floating makes demands on the mass distribution of the vehicle and the
displaced fluid as discussed below.
For the vehicle vertically floating at the surface, it naturally has

𝑚𝐵 = 𝑚𝑏 + 𝛥 = 𝑚𝐺 .

On one hand, the fact that the CG lies below the CB yields that
Fig. 10. Buoyant and gravitational force conditions of the vehicle floating at the 𝑥 𝑚
surface vertically. 𝑥𝐵 = 𝑏 𝑏 > 𝑥𝐺 , (15)
𝑚𝐵
which is equal to
𝑥 𝑚
never expand over to prevent burst. This will be further discussed in 𝑥𝑏 > 𝐺 𝐺 . (16)
Section 3.3. 𝑚𝐺 − 𝛥
Denote 𝑉𝑑_𝑢𝑝 and 𝜉𝑑_𝑢𝑝 the desired glide speed and glide path angle On the other hand, the requirement that 𝑧𝐺 comes to zero when 𝛽 =
respectively in this extreme upward gliding condition. The required ±180◦ yields
𝑚𝑏_𝑚𝑎𝑥 can then be calculated with Eq. (8) as |
𝑚𝑠 𝑧𝑠 + 𝑚̄ 𝑧𝑝 |
( ) 1( ( ) |𝛽=±180◦ (17)
2 = 0.
𝑚𝑏_𝑚𝑎𝑥 = 𝑚𝐺 − 𝛥 − − sin 𝜉𝑑_𝑢𝑝 𝐾𝐷0 + 𝐾𝐷 𝛼𝑑_𝑢𝑝 𝑚𝑠 + 𝑚̄
𝑔
( )) (13)
2 Substituting 𝑧𝑝 = 𝑅𝑝 cos 𝛽 into Eq. (17) gives
+ cos 𝜉𝑑_𝑢𝑝 𝐾𝐿0 + 𝐾𝐿 𝛼𝑑_𝑢𝑝 𝑉𝑑_𝑢𝑝 .
𝑚𝑤
Also, the appropriate installation position of the bladder is obtained 𝑅𝑝 = 𝑧 , (18)
𝑚̄ 𝑤
from Eq. (9) as where 𝑅𝑝 is the offset distance of the rotatable mass from the 𝐞1 axis.
1 ( ( )
𝑥𝑏 cos 𝜃𝑑_𝑢𝑝 + 𝑧𝑏 sin 𝜃𝑑_𝑢𝑝 = 𝑚𝐺 𝑔 𝑥𝐺 cos 𝜃𝑑_𝑢𝑝 + 𝑧𝐺 sin 𝜃𝑑_𝑢𝑝
𝑚𝑏_𝑚𝑎𝑥 𝑔 3.3. Concern about metacentric stability
(( ) ( ) ))
2
− 𝑚3 − 𝑚1 𝑢𝑑_𝑢𝑝 𝑤𝑑_𝑢𝑝 + 𝐾𝑀0 + 𝐾𝑀 𝛼𝑑_𝑢𝑝_𝑚𝑎𝑥 𝑉𝑑_𝑢𝑝 .
When floating at the surface, the vehicle is exposed to harsh en-
(14) vironment conditions like ocean wind, waves, surface currents, etc.
Before takeoff or right after landing, the rotor arms are unfolded. The
Recall that 𝑧𝑏 = 0, 𝑥𝑏 can be determined via Eq. (14). unfolded arms, however, will rise the CG significantly, which adversely
affects the floating stability. An effective way to retain the restoring
3.2. Concern about mode transition ability is making sure that the new rising CG is still below the CB by
properly managing the weight of foldable components. Another way
Vertical floating at the surface is a necessary position for the vehicle referring to metacentric stability can still prevent the vehicle from
right after landing or just before takeoff. Also it is a crucial period in overturn even if the CG rises over the CB. As the rising distance of the
the transition between swimming and flying. As for typical UGs, they rotors in the unfolding process is too much to arranged the CG still
cannot do a vertical floating due to their design with the limitation below the CB, the later scheme was embraced in our final design.
of mass distribution. Therefore, a rotating mechanism that rotates the Fig. 11 illustrates the principle of metacentric stability when the
movable mass 360◦ freely is newly developed for the proposed vehicle CG is above the CB. In the figure, point M represents the metacenter at
to realize the motion switch between a normal gliding and a vertical which the vertical line through the CB intersects the longitudinal body
floating. axis. The torque couple of the downward gravitational force and the
Fig. 9 shows the rotatable mass positions in different operation upward buoyant force will act to bring the tilting body back to vertical
stages. For the gliding operation, the mass position 𝛽 should be limited as long as point M lies above the CG.

7
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 12. Force condition of a steady level flight of the vehicle.

The torque couple, also known as restoring torque, is

𝜏𝑅 = 𝐵 ⋅ 𝐺𝑍, (19) Fig. 13. Exterior photograph of Nezha III demonstrating the bladder, the propulsion
system for flight, the folding mechanisms, the main housing, the fixed wings, and the
where an estimation of righting lever 𝐺𝑍, the distance between CG supporting frame.
and point Z, can be obtained with the metacentric height (𝐺𝑀), the
distance between the CG and point M, as
( )
𝐺𝑍 = 𝐺𝑀 ⋅ sin 𝜂. (20) 𝐷 𝛼𝑑 , 𝑉𝑑
𝑇 = , (23)
The relationship only holds at small angles of heel, 𝜂. cos 𝛼
Therefore, good metacentric stability benefits from large 𝐺𝑀. Note ( )
𝜏 = −𝑀𝐷𝐿 𝛼𝑑 , 𝑉𝑑 + 𝑚𝐺 𝑔𝑥𝑔 cos 𝛼𝑑 . (24)
that point M is only treated as a fixed point at small heeling angles;
otherwise, M is no longer fixed and have to be recalculated at different Eqs. (22), (23), (24) imply that the efficiency for a level flight of the
submerged body. Ways of increasing metacentric height include height- proposed HAUV is determined by the lift-drag ratio and drag property
ening point M or lowering the CG. The former scheme requires the matching with the vehicle’s weight.
submerged portion of the bladder can shift to the tilting side as much Since the wings and the vehicle’s shape dominate the aerodynamics,
as possible for a given 𝜂. And the more this shift distance is for the Eq. (22) provides an essential relationship between wing design, overall
same 𝜂, the greater the restoring torque 𝜏𝑅 will be, which thus enhances configuration, and weight management. Eqs. (23) and (24) show the
the metacentric stability. However, there is a critical heeling angle, 𝜂𝑐 , needed control input to maintain a desired level flight, which assists
where the metacentric stability vanishes. Any inclination greater than in selection of motor/propeller combination or a rough evaluation for
𝜂𝑐 will lead to a torque created by the gravity and buoyancy in the flight efficiency.
same direction of the heel and force the vehicle to capsize rather than Eqs. (10), (12), (13), (14), (16), (18), (22), (23), (24), and the
righting it back. guideline in Section 3.3 are the critical principles for a proper design
of the proposed vehicle.
3.4. Concern about level flight in longitudinal plane
4. Prototype vehicle
Transition to level flight is a process where the increasing lift of
wings gradually takes over the decreasing thrust to counteract the As a realization of the improved concept of the multimodal HAUV
gravity. However, in a steady level flight, the thrust overcomes not compared with its first prototype, Nezha III was designed and fab-
only the drag force but also partially the gravity. Therefore, the match ricated. Nezha III is not intended to be an optimized design, but
between the vehicle’s aerodynamics and the propulsion property de- rather a proof-of-concept demonstration that can achieve the desired
cides the performance of a steady level flight. Note that control surfaces functionality.
widely used in fixed-wing UAVs were not adopted in our design since
the propulsion system can generate control moments in all directions, 4.1. Vehicle configuration
which also contribute to payload reduction of the vehicle.
Given a desired AOA, 𝛼𝑑 , and a desired forward speed, 𝑉𝑑 , the equi- Nezha III is completed with great use of off-the-shelf components as
librium of level flight can be represented in inertial frame coordinates much as possible for being cost efficient and easy maintenance. Figs. 13
as below according to Fig. 12: and 14 show its exterior and interior details.
( ) Nezha III mainly comprises a hexacopter propulsion system, a pair
𝑇 cos 𝛼𝑑 − 𝐷 𝛼𝑑 , 𝑉𝑑 = 0
( ) of fixed wings, folding mechanisms, an outside bladder, a supporting
𝐺 − 𝑇 sin 𝛼𝑑 − 𝐿 𝛼𝑑 , 𝑉𝑑 = 0 (21) frame, and a tubular main hull protecting the pneumatic buoyancy
( )
𝜏 + 𝑀𝐷𝐿 𝛼𝑑 , 𝑉𝑑 − 𝑚𝐺 𝑔𝑥𝑔 cos 𝛼𝑑 = 0, system, rotatable mass, avionics and batteries stored inside. Nezha III
has a gross weight of 18 kg, wingspan of 1650 mm, diagonal wheelbase
where 𝑇 and 𝜏 are the control thrust and control pitching moment of
of 965 mm, and primary body length of 577 mm excluding the antenna.
six rotors.
The sealed chamber of the tubular fuselage is formed with a 160-
Therefore,
( ) mm-outer-diameter, 5-mm-wall-thickness acrylic cylinder, two acrylic
𝐺 𝐿 𝛼𝑑 , 𝑉𝑑 bulkheads, and two aluminum sealing flanges. Sealing flange has dou-
( ) = ( ) + tan 𝛼, (22)
𝐷 𝛼𝑑 , 𝑉𝑑 𝐷 𝛼𝑑 , 𝑉𝑑 ble O-ring seals on its side and one O-ring seal on its end surface to

8
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 15. Buoyancy control system diagram. HP and LP denote high pressure and low
pressure respectively.

The barometer is not waterproof so that it is placed inside the main

hull. In this case, the vehicle’s hull needs to not only be fully sealed
under water but also be breathable in air for the barometer functioning
normally. Therefore, a vent piston driven by a micro gear motor with
screw shaft is designed and fabricated in-house. With a 6-V direct
current (DC) power supply, the piston can fully open or seal in 10 s
according to the control command.
As the design principles in Section 3 suggest, the external rubber
bladder is placed at the top and is circularly symmetric about the
Fig. 14. Interior photograph of Nezha III showing the pneumatic buoyancy system,
the rotation mechanism, avionics, and power supply.
longitudinal body axis. According to Section 3.3, the bladder should has
a large enough radius for good metacentric stability. In our prototype,
an annular bladder is used instead of a spherical one. Given a certain
bladder’s volume, an annular shape can spread that volume over an
make the enclosure fully waterproof. The inside dry compartment pro-
area with a larger radius while a round shape occupies that volume in
vides a protected zone for the apparatus including pneumatic buoyancy
the middle.
system, rotatable mass distribution system, batteries, and electronic
Inevitably, deviation of the mass and buoyancy properties of the
instrument. There are 7 cable penetrators and 2 connectors mounted
fabricated vehicle from the designed ones is introduced into the system
at the upper bulkhead. They are designed for the wires and pneumatic
due to the machining error, estimation errors, etc. So, the prototype has
tubes pass-through. Besides, a pressure sensor for depth measurement
been trimmed by adding ballast (e.g. stainless steel).
and a venting device are installed on the bulkhead.
A pair of off-the-shelf wings for UAVs is implemented to provide
4.2. High pressure pneumatic buoyancy engine
lift for flight and underwater glide. The wings are made of balsa wood
with the unilateral wingspan of 730 mm and the mean chord length of
The buoyancy engine consists of the air supply (a carbon fiber cylin-
291 mm. Because of the great amount of displacement and the low den-
der), the pressure regulator (a pressure reducing valve), two actuators
sity of the wings, several holes with a diameter of 10 mm were drilled (two solenoid valves for inflation and deflation respectively), the rubber
through the leading edge and trailing edge of the wings to permit free bladder, and the safety devices (a check valve preventing backflow of
flooding and draining. There is a rectangular carbon fiber tube in each ambient water and a safety valve preventing bladder failure). Fig. 15
wing for enhanced strength. The wings are symmetrically placed on shows the systematic diagram of the buoyancy system.
both sides of the fuselage with two 3D-printed resin connectors. Three The cylinder can store 0.22 L compressed air under the pressure of
stainless steel adjustable hose clamps pass through the connectors and 30 MPa at most. Our prototype is designed for shallow operation, and
fasten them on the main hull together with the wings. The wings can every fraction of weight counts for the flight performance. A tradeoff
be shifted along the hull in vertical direction to trim the CB and the CG selection of equipment is achieved with the lightweight valves and
together. tubes that can operate under the pressure up to 2 MPa. Therefore, the
The hexacopter configuration is a trade-off design between carrying air pressure is lowered from 30 MPa to 2 MPa by the pressure reducing
capacity and weight of propulsion system. Six 360-mm carbon fiber valve before the air flows to the rest components.
tubes and two customized 2.5-mm carbon fiber plates compose the The solenoid valves are separately controlled via different relays.
supporting frame structure of the propulsion system. In addition, a 3D- Unlike proportional solenoid valves, the solenoid valves utilized in the
printed nylon circular connector is used to attach the frame to the main system are either on or off, which means the flow speed cannot be
hull by a hose clamp. Position of the frame is also adjustable in vertical continuously adjusted online. To control the buoyancy effectively, the
direction but is intentionally set close to the top for the emergence of time windows of opening valves are the only control effort in practice.
the propellers from the water. A proportional–differential (PD) controller is developed to adjust the
When the vehicle stands on the ground, its high center of gravity duration of valve opening. Since the valves open and close alternately,
adversely affects the stability. Situation might get worse when the the output of the controller is actually the working frequency of the
vehicle needs to tilt to resist cross wind during landing. Therefore, four valves.
extended customized triangular carbon fiber plates fixed around the In our early design, a pressure sensor was employed to obtain an
main hull serve as four orthogonal landing support frame. accurate estimation of the bladder volume. The sensor was not able

9
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 16. Sketch of cross section of rotatable mass viewing from aft side. Gray rectangles
are the two rotatable batteries; red arrow represents the gravitational force vector of
rotatable mass; black curved arrow illustrates the measured direction of rotation angle
of the mass. (For interpretation of the references to colour in this figure legend, the Fig. 17. 3D rendered picture of the folding mechanism except its actuator (a 6-V DC
reader is referred to the web version of this article.) gear motor).

to place inside the bladder due to its size. Instead, it was mounted 360◦ . The reading jump may lead to an overshooting rotation over 360◦
outside the bladder measuring the air pressure through a branch of if the revolution continues in its previous direction, causing excessive
inflation air tube. However, the test results showed an unacceptable distortion of wire and damaging the wire.
deviation between the measurements and the actual value of pressure The batteries are cuboid as shown in Fig. 16 so that they are
in the bladder during inflation or deflation. This is because the gas arranged into a V shape to get an adequate eccentric distance of the
compressibility and the friction loss (the loss of pressure) cannot be mass, i.e., 𝑅𝑝 = 2.2cm. Then, a steady planar glide can be performed
ignored in a fast pipe flow, resulting in an obvious deviation between when the vehicle is most bottom-heavy at 𝛽 = 90◦ , and a vertical
the pressure of the measuring point and the pressure inside the bladder. attitude can be achieved when the CG is vertically in line with the CB
Thus, the method that used the information of the bladder’s pressure along the longitudinal body axis at 𝛽 = 270◦ .
to control a glide was abandoned.
In the present form, the control scheme is completed without the 4.4. Folding mechanism for foldable arms
need of pressure data. The discussion in Section 3.1 hints a one-to-one
correspondence between buoyancy, gliding attitude, and glide speed. Nezha III equips with 6 folding mechanisms in all, everyone of
Since the buoyancy and the glide speed are difficult to estimate, the which is responsible for an arm. Fig. 17 shows the 3D rendered struc-
attitude data can be easily and accurately measured by the attitude ture of the folding mechanism in two different extremity. The actuator
sensor and fed back to the PD controller to realize the close-loop control of the mechanism is a 6-V gear motor sealed by double O-rings in
of glide motion. The test results demonstrate the feasibility of the a cuboid aluminum alloys shell. Epoxy resin seal is used for wire
proposed control scheme. penetration in the bottom of the shell. A brass thread rod is used in
Note that the number of gliding cycles varies with the operating combination with a stainless steel barrel nut to translate turning motion
depth, and additional exhaust inevitably happens when the controller of the motor into linear motion of the nut. Then the nut leads a lever
operates. Besides, operating depth has an influence on the working rate arm to rotate 90◦ around a hinge. By detecting the locked rotor current
of buoyancy engine. This affects the consumption of the compressed of motor, it is easy and confident to know the arms are fully folded or
air during the buoyancy adjustment. Advanced control scheme that unfolded, and thus shut down the motor immediately. The mechanism
can adaptively manage the change of inflation and deflation dynamics is self-locking without power when the barrel nut sits at both ends of
due to the variation of depth is potentially helpful with that problem. the thread rod.
However, it is beyond the scope of this work. Therefore, the evaluation Fig. 18 demonstrates the folding and unfolding process of the rotor
and the optimization of the gliding endurance is not included in this arms. Six arms are designed to move synchronously to prevent that the
work. CG offsets so far from the longitudinal axis that the couple of gravity
and buoyancy turns the vehicle upside down when it floats at the
4.3. Rotating mechanism for movable mass water’s surface. As shown in the figure, it takes respectively about 11
s and 13 s for the mechanism to fully lower and open the arms.
As shown in Fig. 14, the rotation mechanism is comprised of the
rotational actuators (a micro gear motor and a worm drive), the offset 4.5. Propulsion
mass (two LiPo batteries weighing 1.724 kg), and the 360-degree hall
angle sensor. The selected worm gear with a gear ratio of 40:1 not only Six rotors propel the vehicle in flight. A rotor is a combination of
is self-locking without power but amplifies the output torque of the gear a waterproof 400-KV BLDC motor with weight of 280 g and a foldable
motor, making it possible to apply a micro gear motor with weight of 10 carbon fiber propeller with a 46.7-cm (18.4-in) diameter, a 16.8-cm
g, an output torque up to 2 kg.cm, and nominal voltage of 9 V. To know (6.6-in) pitch. Due to the gravity, the blades of the propeller can be
the rotational mass position, the angle sensor measures the rotation of a folded freely during the arms folding to avoid hard interference with
shaft that revolves in sync with the mass through a 1:1 nylon spur gear the wings. Again, the blades can fully extend when the motors spin
pair. The rotating mechanism moves at a constant revolution speed of fast. Two 4-in-1 55-A electronic speed controllers (ESCs) are used to
7◦ ∕s until the it reaches the prescribed position, but it is limited to only control motor speed. The first three motors occupy the three channel
one full rotation to avoid the twisting damage of the wires. of an ESC while the rests occupy the other ESC. Thus, every ESC has
Fig. 16 displays a cross section view of the rotatable mass illustrat- one redundant channel for backup.
ing the measurement of the rotation angle 𝛽. Note that the zero angle in The motors are exposed in water. To waterproof it, the motor coils
practical situation represents the leftmost position of the rotatable mass are coated with waterproof paint, and original stainless steel bearings
instead of the bottommost as depicted in Fig. 9. This adjustment avoids are replaced by ceramic ones. The system weight benefits from the
the measurements near critical values (0◦ and 360◦ ). This is necessary utilization of 4-in-1 ESCs at the price of drastic heating. For a good heat
because a jump from 0◦ to 360◦ or the opposite will likely happen in dissipation condition, ESCs are mounted outside the hull. Meanwhile,
the presence of disturbance when the sensor reading gets close to 0◦ or they are sealed by epoxy resin with good thermal conductivity. Also,

10
D. Lu et al. Ocean Engineering 219 (2021) 108324

4.6. Avionics and power systems

The avionics of Nezha III are classified into 3 main subsystems

(power system, flight system, and underwater control system) in terms
of different functions. Most of avionics are shielded inside the acrylic
main hull. Apart from those instruments submerged in ambient water
as previously stated, the global positioning system (GPS) and radio
telemetry antennas are waterproofed using epoxy resin seal and are
mounted on the external top as well with their cables passing through
waterproof cable penetrators to the inside. Fig. 20 is the electrical
system diagram for Nezha III.
A desire controller that can enable both level and vertical flight
is still in developing. In the preliminary stage of the development for
Nezha III, the success of the water/air transition is the most crucial
and essential concern. Therefore, for the current system, the flight con-
troller is established with a commercial multirotor autopilot, which can
ensure stable vertical flight and VTOL maneuver with easy integration
and good compatibility for the present design. However, it should be
pointed out that the autopilot limits flight attitude up to 45◦ , which
precludes practical level flight in the field tests. On the other hand, a
micro control unit (MCU) that responsible for underwater operation has
been developed and completed with full basic functions, i.e., regulating
the pitch and buoyancy states, rotating the movable mass, folding
or unfolding the arms, collecting data and communicating with the
operator, etc. With these two controllers, Nezha III can react to the
ground control station to move in both air and water.
The flight remote control link is established via an onboard small
2.4 GHz SBus receiver. The autopilot is powered by a power man-
agement unit (PMU). A global navigation satellite system (GNSS) and
compass compact module collects the vehicle’s position and orientation
information and feeds them back to the autopilot. Once the autopilot
receives the pilot’s command from the remote controller (RC), it will
Fig. 18. Demonstration sequence of folding (left) and unfolding (right) tests in act on two ESCs accordingly to control rotations of each motor. Note
laboratory.
that there is a one-way data link where a pulse width modulation
(PWM) signal is sent from the autopilot to the MCU. The PWM signal
is manually tuned within 1000 ∼ 2000 via a channel of RC to reboot the
system for emergency. The MCU will reset itself if the signal is greater
than 1700. This is practically useful when the underwater operation
program crashes unexpectedly.
Real-time remote control is denied under water. Therefore, Nezha
III is preprogrammed to execute certain actions including planar glide,
turning glide, switch between glide and vertical float, folding or un-
folding arms. There is a bidirectional 433 MHz wireless link between
PC ground station and the vehicle. So, control signals from the ground
station can be sent to the vehicle while data onboard can be transmitted
back to the station as well. Once the command is received, the MCU
will react correspondingly. To adjust buoyancy, the MCU directly turns
on or off the relays to control the solenoid valves. To activate the vent
piston or move the rotatable mass, the MCU controls a dual-h-bridge
DC motor driver with rated current of 1.5 A for each channel to drive
the actuating motors. The same three motor drivers are instructed by
the MCU and share the same control signal for the synchronization of
six folding arms. Note that there is a relay in the power supply line
Fig. 19. Statics thrust test results in terms of angular velocity, thrust, power, and of the autopilot. This relay is directly controlled by the MCU. When
efficiency versus throttle. Note that efficiency = ouput power/input power.
receiving an reset command for autopilot from the PC ground station,
the MCU will trigger a transient power outage for the autopilot via the
relay, following by powering the autopilot again to reboot it.
the ESC’s aluminum cooling shield is deliberately left to be directly Sensors for different measurements are crucial for both flight and
exposed in air when waterproofing the ESC. underwater maneuvers. The autopilot kit has a self-contained set of
Fig. 19 shows the results of static thrust tests for a single mo- sensors for a stable flight. However, sharing the flight data stored in the
autopilot with the MCU is still one of our undergoing work. Therefore,
tor/propeller combination in air. In the tests, the rotor can generate
an extra set of sensors is employed for motion control underwater.
maximum thrust of 5.1 kg at 6920-rpm rotation speed at full throt-
Depth is measured by an ultra-small pressure sensor with a verified
tle. The vehicle weight occupies 60 percent of the maximum 30.6-kg resolution and accuracy of 1 cm. Attitude and heading information
takeoff weight, leaving the propulsion with enough thrust margin for underwater is provided by an attitude and heading reference system
disturbance rejection in flight. (AHRS). Note that the AHRS orientation is adjusted in such a way

11
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 20. Schematic of the electrical system. The abbreviations in the diagram refer to: FM—folding mechanism; GNSS—global navigation satellite system; ESC—electronic speed
controller; ADC-analog-to-digital converter; AHRS—attitude and heading reference system.

that its x, y, and z axes respectively point to the right, the head, and
the bottom. The adjustment avoids the singularity of 90◦ pitch angle
simply and effectively when the vehicle is upright. Accessible position
information of the vehicle at the surface or in flight is enabled by
another GPS module. As to the fulfillment of controlling the vent piston
and the folding arms, there are four channels of current sensor data
signaling the MCU to judge whether the actuators reach their extremity
positions. Moreover, the angle sensor is needed for the rotatable mass
as discussed in Section 4.3. The measurements of rotation angle of
the rotatable mass and operating current of the folding mechanisms
are collected via a 16-Bit analog-to-digital (ADC) module that can be
accessed via I2C communication. The MCU is able to access to those
collected data as well as depth measurement with its I2C interface via
an I2C multiplexer.
Four 6000-mAh 6S LiPo (22.2 V) batteries and six different buck
converters constitute the electric power distribution system. As stated
in Section 4.3, two batteries are placed on the rotating mechanism to
serve as offset mass. The other two batteries are fixed in a specified Fig. 21. First order filtered estimation of vertical velocity. The upper figure displays
location to meet the overall arrangement designed in Section 3. The the calculated velocity and filtered velocity as a function of measured depth. The lower
batteries directly power the PMU, the two ESCs, and the two solenoids figure demonstrates the deviation between the measured depth and estimated depth.
with nominal voltage of 24 V. To isolate the control system from high-
power actuators, two independent buck converters stepping down the
high voltage of the batteries to 9 V are used respectively to power the 5.1. Underwater glide
MCU and the actuators for vent piston and rotatable mass. Meanwhile,
three buck converters with peak output current of 3 A feed 6 V voltage The underwater gliding performance was field tested in a lake with
to the motor drivers of the folding mechanisms. Regarding the else average depth of 30 m and maximum depth over 100 m.
devices operating under low power including sensors, a 5-V buck
The vehicle states were measured with sampling time of 0.18 s.
converter powers them all.
The AHRS system has poor accuracy for velocities along three axis. For
5. Vehicle performance characterization the assessment of the motion speed, the vertical speed of the vehicle
was obtained through dividing the difference between two consecutive
Nezha III was tested to verify the feasibility of the proposed vehicle, depth measurements by the sample interval. However, this calculated
and its multi-domain locomotion performance was characterized in velocity was subject to the sensor uncertainty. The random error up
three operating regimes, i.e., underwater glide, flight, and transition to 0.01 m introduced by the depth measurement (mentioned in Sec-
between water and air. tion 4.6) was magnified by 5.5 times in the velocity calculation, causing

12
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 23. (a) Photograph of upward glide in a 45◦ pitch position. (b) Photograph of
downward glide in a −20◦ pitch position.

Fig. 22. A cycle of free glide to the target depth of 40 m. in the figure, the second stage lasted about 2.2 min while the fourth
stage took about 1 min. The average rising velocity is about twice faster
than the average diving velocity. The diving velocity reached 0.64 m/s
an unpleasant high frequency measurement noise with amplitude of at about −71◦ pitch angle while the rising velocity accelerated to 0.94
0.056 m/s. Therefore, a first-order filter below was implemented for m/s at near 90◦ pitch angle. A significant feature that the increasing
a better estimation of the vertical speed, downward velocity gradually came to equilibrium is captured from 140
s to 155.3 s. This equilibrium is reached after the bladder cannot be
𝑣̂ 𝑘+1 = 0.5𝑣𝑘 + 0.5𝑣̂ 𝑘 squeezed any more, i.e., 𝑚𝑏 = 0 and 𝑚̇ 𝑏 = 0. Therefore, as for the dive of
where 𝑣𝑘 is the 𝑘th calculated vertical velocity, 𝑣̂ 𝑘 is the 𝑘th filtered the prototype with the bladder completely compressed, the maximum
velocity. Fig. 21 demonstrates the filter performance through two cycles velocity is approximately 0.64 m/s along with the max pitch of −71◦ .
of dive. The upper subfigure shows that the filtered velocity is more At the beginning of the third stage, a dissonance between the
precise than the calculated one. The deviation between the measured variations of pitch angle and downward velocity are observed. The
depth and the estimated depth reconstructed with filtered velocity is pitch angle increased rapidly over 0◦ in 21.1 s and over 40◦ in 56.9
demonstrated in the lower subfigure, where it is obvious that the higher s after the inflation started at 155.3 s. This was an aggressive nose-up
the speed is, the bigger the error of the estimation is. maneuver. However, the vehicle still descended in the meantime, and
Note that the velocity information is mainly used in the assessment finally touched the lake bottom at 51-m depth before it started rising
of the gliding motion rather than completing the gliding control loop at 270 s. Consequently, a large AOA was encountered by the vehicle,
in this work. The vehicle has the feature of pitch–buoyancy coupling. deteriorating the hydrodynamic performance. The problem is mainly
Adjusting the buoyancy will accordingly change the attitude and affect caused by the great downward inertia of the body at the maximum
the gliding velocity. Since the AHRS system can measure the attitude descending rate and the slow buoyancy regulation in deep water. As for
accurately, the gliding motion controller is thus designed to actuate the latter fact, it is found out that the prototype has a large trim by the
the buoyancy engine to track the desired pitch attitude instead of the stern with about 40◦ at neutral buoyancy. Besides, the great ambient
buoyancy or velocity. pressure will slow down the inflation. Therefore, the vehicle was still
negative buoyant but decelerated during pitch angle changing from
5.1.1. Free planar glide −71◦ to 40◦ . This behavior is repeatedly observed in the transition from
The couple relation of the linear and pitch motions of the prototype a downward to an upward glide of the following gliding experiments.
was firstly studied through the experiment of a free gliding to 40 m The braking problem is common for UG due to the lack of direct control
depth (Fig. 22). The free glide was defined as a four-stage sequence of force from the propeller. A practical solution is to adjust the vehicle’s
operation. The first stage (from 0 s to 21.4 s) started with the bladder buoyancy in advance of reaching the target depth. But this control
deflation at the surface and continued until the vehicle dived over 1 m strategy is not included in this work since our preliminary goal is to
below the surface. The second stage (from 21.4 s to 155.3 s) was a free identify the system’s basic functionalities.
downward glide with no inflation or deflation control from 1 m to 40 It should also be pointed out that the sharp deceleration from 155.3
m. In this stage, the bladder was passively compressed by the ambient s to 200 s in the figure is not the true velocity response. Because the
pressure, and, therefore, its displaced volume varied with depth. The pressure sensor was mounted at the head, the rapid pitch motion due
third stage (from 155.3 s to 315 s) lasted from the moment the vehicle to the inflation diminished the difference between two sequent depth
descended over 40 m until it rose back to 40 m again. During this time, measurements even if the vehicle was still descending, making the
the inflation valve kept opened to inflate the bladder without extra calculated velocity experience a sharp decrease. But the velocity of CG
control. The last stage (from 315 s to 377 s) consisted of a free upward was still faster than the calculated one.
glide from 40 m to surface with no inflation or deflation control. During In the fourth stage, the vehicle had a free and accelerating rise
the period of the glide, the rotatable mass was fixed at 𝛽 = 90◦ to reject without any control. During the period, the decreasing ambient pres-
rolling disturbance, and, thus, maintain 𝜙 = 0◦ . The rolling stability in sure made the bladder continuously expand without active inflation.
the upward and downward glides is proved to be relatively good in the Thus, the buoyancy unceasingly increased and the CB kept shifting
presence of current. forward, which resulted in an upward acceleration and an increasing
In the free downward glide, the bladder kept shrinking as the pitch motion.
depth increased, resulting in an increasing nose-down attitude and glide
speed. In the free upward glide, the nose-up attitude and glide speed 5.1.2. Controlled planar glide
also showed increasing tendency due to the continuous expansion of Controlled planar glide was further studied by applying a well-
the bladder as a result of the decreasing ambient pressure. As shown tuned PD controller for pitch motion. Figs. 23 and 24 respectively show

13
D. Lu et al. Ocean Engineering 219 (2021) 108324

Table 3
Roll attitudes achieved by different rotatable mass positions 𝛽 in steady downward
glides with −20◦ pitch angle.
𝛽 (◦ ) 10 30 50 70 90 110 130 150 170
Roll angle (◦ ) −20 −13 −5.5 −1.9 −0.3 3 5 12 18

Table 4
Roll attitudes achieved by different rotatable mass positions 𝛽 in steady upward glides
with 45◦ pitch angle.
𝛽 (◦ ) 10 30 50 70 90 110 130 150 170
Roll angle (◦ ) −18.4 −10 −5.5 −1.5 0 3 6 9 12.6

Fig. 24. Three cycles of planar glide with a well-tuned PD algorithm for pitch control.
The target pitch angles for downward and upward glides are −20◦ and 45◦ respectively.

Table 2
Vertical velocities and pitch angles of five steady glide paths.
𝜃𝑑 (◦ ) 45 −10 −20 −30 −45
Vertical velocity (m/s) −0.21 0.07 0.11 0.18 0.31

the photograph and the results of a representative controlled planar

gliding with a −20◦ pitch angle in descent and 45◦ pitch angle in
ascent. During the tests, the rotatable mass was fixed at the position
of 𝛽 = 90◦ to prevent the vehicle from rolling. The vehicle started the Fig. 25. Turning glides with 𝛽 = 15◦ and a resultant roll to the left in a shallow
operation.
controlled inflation for the bladder to float up right after it dived over
8 m, and activated the controlled deflation to descend once it rose
to 2 m. Disturbance on pitch dynamics was normally experienced by
the vehicle when it surfaced and partially emerged from water. The a symmetrical rolling behavior as a function of the rotatable mass
results prove the effectiveness of the utilized PD algorithm for pitch position from 10◦ to 170◦ . But the symmetry of rolling behavior is
attitude tracking. Note that the PD controller is tuned well only for deteriorated when gliding upward. Obviously in the upward glides, the
operations in shallow water. As the operation depth gets deeper, the vehicle rolled to the left by 18.4◦ when the rotatable mass was in the
inflation and deflation characteristics of the buoyancy engine change leftmost position (𝛽 = 10◦ ) while it rolled to the right by only 12.6◦
due to the increased ambient pressure. Therefore, the parameters tuned when the rotatable mass was in the rightmost position (𝛽 = 170◦ ).
well for the shallow zone will worse the performance of PD controller This is mainly attributed to the non-uniform expansion of the bladder.
working in deeper depth. A control scheme that can tackle the varying Due to the manufacturing issue, the inflated off-the-shelf bladder has
properties of the actuators of the system is a challenge and still under its right and up sides bigger than the left and the bottom. Therefore,
developing. These tests primarily studied the motion characteristics of the CB will offset toward the right and up from the center when the
the prototype including the relationship between the vertical velocity vehicle floats up by inflating the bladder, which makes a roll to the
and the pitch angle in a steady glide. These features are independent left greater than a roll to the right.
of the operation depth or the performance of actuators. Based on the rolling characteristic, turning maneuver was then
A group of controlled planar gliding tests were conducted with tested and demonstrated by fixing rotatable mass in a position where
different desired pitch angles 𝜃𝑑 from −45◦ to 45◦ . Each test composed 𝛽 = 15◦ in a controlled glide with −20◦ downward pitch angle and
of at least five cycles of glide with the same desired pitch angle to 45◦ upward pitch angle. The results are recorded in Fig. 25. In the
obtain an average result. Table 2 shows the statistics of the vertical downward glide, the vehicle had an average roll of 16.8◦ to the left
velocity and the pitch angle in steady glides. Recall that the afore- and thus yawed to the left with the rate of 2.8◦ /s. In the upward glide,
mentioned glide paths with pitch angle from 0◦ to 40◦ indicate the the vehicle maintains an average roll of 13.6◦ to the left but performs a
unstable transition stage from descent to ascent with large AOA and yaw motion in a reverse direction (to the right) with the rate of 0.5◦ /s.
poor hydrodynamics. Therefore, there is no analysis for these glide The opposite change of yawing direction after the transition between
paths. a downward path and an upward path is normal for UGs, which is
primarily determined by the position of wings (Graver, 2005). Never-
5.1.3. Turning glide theless, the turning rate of the descent and the ascent are quantitatively
Since the prototype with no control surface, roll–yaw coupling is different. A leading cause is the top–bottom asymmetry of the used
the only approach to turning glide. airfoil, which is originally designed for aircraft. This airfoil can provide
To study the rolling behavior of the prototype first, tests were good hydrodynamic force at a positive AOA (in a downward glide),
carried out by fixing the rotatable mass in a variety of positions 𝛽 in but has poor hydrodynamics at a negative AOA (in an upward glide).
controlled glides with −20◦ downward pitch angle and 45◦ upward Therefore, the steering ability got worse in the upward glide with a
pitch angle. The roll response to different rotatable mass positions steeper pitch angle of 45◦ . Besides that, the offset of the CB position
are listed in Tables 3 and 4. It is found out that the vehicle has caused by the asymmetrical inflation of the bladder made the heel angle

14
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 26. The prototype floated at the surface in different poses after unfolding the
arms. (a) the average outer and inner diameter of the bladder are 290 mm and 160 mm
respectively; (b) the average outer and inner diameter of the bladder are 420 mm and
250 mm respectively.

to the left in descent greater than that in ascent. Therefore, the vehicle
had a slower turning behavior when floating up.
Fig. 27. Lift, drag, and moment on the vehicle in level flight.
5.2. Steady float at the surface

In practice, the prototype vertically rests on the water’s surface with

the arms either folded or unfolded after the rotatable mass shifts to the
position of 𝛽 = 270◦ . Folding the arms will lower the CG position below
the CB, which always produces a restoring moment to right the tilting
body. On the contrary, the CG is higher than the CB when the arms
are unfolded. As discussed in Section 3.3, the bladder should maintain
its maximum volume to provide the vehicle with sufficient metacentric
stability to float steadily at the surface.
Fig. 26 demonstrates two cases with different bladder inflation sizes,
where the bladder had an average 290-mm outer diameter and 160- Fig. 28. Lift-to-drag ratio variation as a function of AOA.
mm inner diameter for the first case and had its maximum dimension
with an average 420-mm outer diameter and 250-mm inner diameter
for the second case. Note that the prototype in both cases were positive
buoyant. In the first case (Fig. 26(a)), the vehicle with its arms unfolded
was initially placed in an upright position at the surface but rapidly
overturned. In the second case (Fig. 26(b)), the vehicle was successfully
stabilized at the surface thanks to the enough metacentric stability
provided by the bladder. To evaluate the metacentric stability, the
vehicle was forced to heel toward different directions to find the critical
heel angle at which the vehicle can no longer right itself. Through
a serials of tests, a 28◦ minimum angle of vanishing stability was
identified. This fact evidences a certain degree of seakeeping quality Fig. 29. The velocities of steady level flight with different AOA.
for the vehicle.

5.3. Level flight

by the vehicle at a zero AOA. Though the flight slows down as the AOA
increases, the speed hardly changes when the AOA exceeds 10◦ .
Because of the experiment site constraints and the attitude adjust- Based on the aerodynamic data and the above velocity information,
ment limitation of the utilized flight controller, level flight has yet to the requisite control forces of steady level flight can be determined
be practically performed and tested for the prototype. To character- through Eqs. (22), (23), and (24), which are shown in Fig. 30. The
ize the level flight performance, computational fluid dynamics (CFD) greatest control efforts, i.e., 142.38 N (14.53 kg) maximum thrust and
technique was applied for the present study. −7.09 N⋅m maximum pitch torque of the propulsion, are needed for
Consider the vehicle in steady horizontal flight with different AOAs a completely horizontal flight (𝛼 = 0◦ ). The most efficient level flight
and velocities. The aerodynamics of the prototype are shown in Figs. 27 happens at 6◦ AOA and 31.5 m/s with 44.95 N (4.59 kg) minimum
and 28 as a function of AOA and velocity. As known from Fig. 28, thrust and −1.47 N⋅m pitch torque. Given the vehicle’s weight of 18 kg
lift-to-drag (L/D) ratio is not determined by the cruise velocity but the and the propulsion characteristics in Fig. 19, it can be inferred that
AOA. The peak L/D ratio 3.92 happens at 7◦ AOA. the propulsion power will reduce up to 79.9 percent when the vehicle
Recall that Eqs. (22), (23), and (24) always hold in a steady level transitions from vertical to level flight.
flight. By using Eq. (22) and extrapolating from the data sets in Figs. 27
and 28, the forward velocities of steady level flight with different AOA 5.4. Vertical flight
are calculated (Fig. 29). The results show that the vehicle will fly faster
in a level flight with a more horizontal attitude which indicates a To test the stability and the maneuverability of the prototype in ver-
smaller AOA. And the fastest speed of about 66.4 m/s can be reached tical flight, trials including hovering and flying a rectangle path were

15
D. Lu et al. Ocean Engineering 219 (2021) 108324

5.5. Water/air transition

The transition consists of two stages. The first one is the transition
between underwater glide and vertical float at the surface. The second
one is the water-exit or water-entry process.
Fig. 33 demonstrate the transition maneuver of the prototype be-
tween a glide and a vertical float. Before takeoff, it requires the
maximum bladder volume and the rotatable mass position of 𝛽 = 270◦
for the vehicle prior to raising the arms. These maneuvers set the vehi-
cle in an upright position and prevent overturn. If the vehicle capsizes,
refolding the arms can bring the body back to vertical orientation again.
Fig. 30. Control thrust and moment required to maintain level flight at different AOA.
If the vehicle tempts to glide after landing, the arms will be folded at
first. After that, the rotatable mass shifts to the position of 𝛽 = 90◦ to
incline the body forward and obtain enough bottom-heaviness. Then
the vehicle can start gliding by deflating the bladder.
Figs. 34 and 35 respectively show the time-stamped photograph
sequence of water entry and exit of the prototype. In the tests, the
surface condition was not quite good. The vehicle encountered random
wave and wind disturbance. Despite the unfavorable surface condition,
the takeoff and landing maneuvers were repeatedly successful. The
prototype was able to reject some disturbance and go through the
water’s surface smoothly. The results strongly prove the benefits of
a well metacentric stability design and a VTOL water/air transition
profile to the proposed HAUV.
In addition, good maneuverability of the vertical flight allows for
any transition velocity within the speed range of the vehicle. However,
Fig. 31. Photograph of the prototype hovering above the lake surface subject to slight the exit and entry velocities should be optimized with the concerns of
wind.
energy saving and structural integrity in the presence of water impact.
In the air-to-water ingress, the fast speed indicates a low power of
the propulsion. To the extremity, powering off all the rotors gives the
vehicle a free fall into water. But the faster the entering speed is, the
heavier the impact force exerted on the fuselage is. This increases the
risk of structural fracture. Moreover, lowering the propulsion power
may weaken the ability of disturbance rejection. Therefore, the optimal
entering velocity trajectory should be selected with a careful and
comprehensive consideration.
In the water-to-air egress, it takes time for the floodable wings to
passively drain out the entrained water. The drain time is dominated
by the motion of the wings, the interior structure and the open of the
wings (Stewart et al., 2019). If the vehicle leaves water fast, the adverse
effect of the hydrodynamic damping and the entrained water becomes
significant, making the vehicle weigh heavier temporarily. The power
consumption of the propulsion will consequently increase. Though a
slow water-exit process helps to mitigate the problem, it prolongs the
duration of the exit and thus expends more energy. Therefore, the
takeoff velocity should be optimized by considering the aspects of the
Fig. 32. The rectangle GPS trajectory (black line) of the prototype in vertical flight water resistance, the draining dynamics of the entrained water, and the
under the pilot’s control. (For interpretation of the references to colour in this figure energy consumption.
legend, the reader is referred to the web version of this article.)

6. Conclusion

carried out under manual control. Note that the prototype automati-
An improved concept of the multimodal HAUV has been presented
cally stabilize its attitude and position under the control of autopilot,
in this paper. It is designed to be versatile for performing level and
but it can only fly a trajectory according to the pilot’s command sent
vertical flight, hovering, surface drifting, and underwater glide, which
via RC.
aims at extending application scenarios and operation endurance as
Fig. 31 shows the prototype maintaining a steady hover in the air.
As other researches have studied, the wings impose an adverse effect many as possible. Compared to the current systems, four following
on the tail-sitter UAV subject to crosswind. In our experiment, the improvements have been made. First, a high pressure pneumatic buoy-
prototype indeed suffered from slight wind pushing the vehicle away. ancy system is designed for longer and deeper operation underwater.
But the vehicle succeeded in stabilizing itself. Second, a rotating mechanism is added for turning glide and switching
Fig. 32 displays the actual rectangle trajectory that the prototype operation modes in the water. Third, hexacopter configuration is im-
flied under the pilot’s control. Note that the breeze was encountered plemented for greater payload capacity. Fourth, folding arm structure
frequently by the vehicle while it flied above the lake surface. The is fulfilled for better hydrodynamics of glide.
practical operation proves the flight stability and maneuverability of To ensure the viability of the proposed concept, the paper presents
the vehicle, and evidences the capability of carrying out tasks in design principles related to the planar gliding performance, the realiza-
confined space. tion of mode switch between glide and vertical float, the metacentric

16
D. Lu et al. Ocean Engineering 219 (2021) 108324

Fig. 33. Practical demonstration of transition between gliding and vertical floating modes. The upper arrow line shows the transition steps from gliding to vertical floating while
the bottom arrow line shows the reverse transition steps.

Fig. 34. Photograph sequence of water entry process in the presence of slight wave and wind.

Fig. 35. Photograph sequence of water exit process in the presence of slight wave and wind.

stability at the surface, and the level flight performance. These princi- is also crucial. Moreover, the optimization of the velocity profile of the
ples are key to the proper management of the buoyancy and the weight water/air transition plays a significant role in real applications.
of the vehicle.
With the basis of the above design considerations, a proof-of-
concept prototype, named Nezha III, has been fabricated and tested. CRediT authorship contribution statement
Due to the site-specific constraints and the attitude limitation of the
commercial flight controller, level flight have yet to be practically
Di Lu: Conceptualization, Methodology, Formal analysis, Investiga-
performed. Numerical method was implemented instead to evaluate
tion, Writing - original draft, Visualization. Chengke Xiong: Formal
level flight performance. The simulation results confirm a better ef-
analysis, Investigation, Data curation, Visualization. Hexiong Zhou:
ficiency of the level flight than the vertical flight. In the vertical
flight, the good maneuverability and stability of the prototype were Investigation, Formal analysis. Chenxin Lyu: Investigation, Data cu-
demonstrated even though there existed disturbance caused by the ration. Rui Hu: Investigation, Data curation. Caoyang Yu: Method-
winds. The motion performance underwater was assessed by a series ology, Software. Zheng Zeng: Conceptualization, Supervision, Project
of gliding tests. The free glide experiment confirmed the buoyancy administration. Lian Lian: Supervision, Funding acquisition.
engine could operate at 50 m depth, and the results also confirm
the predicted pitch–buoyancy coupling of the proposed vehicle. In
addition, pitch is much more sensitive to the buoyancy change than Declaration of competing interest
the velocity. Then, controlled planar glides were achieved by applying
a PD controller for pitch attitude tracking. In the controlled glide The authors declare that they have no known competing finan-
experiments, the quantitative relationship between the pitch angle and cial interests or personal relationships that could have appeared to
the steady vertical velocity of glides have been obtained. After that, influence the work reported in this paper.
rolling behavior and the consequent roll–yaw coupling performance
were evaluated through turning tests. Finally, the ability of repeated
water/air transition was successfully proved even in the presence of Acknowledgment
wave and wind disturbances.
There are still lots of works to be done to improve the present
design. More elaborate analysis of the aero/hydrodynamics of the This work was supported in part by the National Natural Science
vehicle (especially for the airfoil) are needed to optimize the design. Foundation of China under Grant 41706108, in part by the Shanghai
A comprehensive controller that can effectively deal with the variation Sailing Program, China under Grant 17YF1409600, and in part by
of the buoyancy engine’s performance versus depth is underway. Devel- the open project of Pilot National Laboratory for Marine Science and
opment of a flight controller compatible with level and vertical flight Technology (Qingdao), China under Grant QNLM-2016ORP0104.

17
D. Lu et al. Ocean Engineering 219 (2021) 108324

References Zhang, F., Lyu, X., Wang, Y., Gu, H., Li, Z., 2017. Modeling and flight control simulation
of a quadrotor tail-sitter vtol uav. In: 2017 AIAA Modeling and Simulation
Alzu’bi, H., Mansour, I., Rawashdeh, O., 2018. Loon copter: Implementation of a Technologies Conference, pp. 1–13.
hybrid unmanned aquatic–aerial quadcopter with active buoyancy control. J. Field
Robotics 35 (5), 764–778.
da Rosa, R.T.S., Evald, P.J.D.O., Drews, P.L.J., Neto, A.A., Horn, A.C., Azzolin, R.Z.,
Botelho, S.S.C., 2018. A Comparative study on sigma-point Kalman filters for Di Lu received the B.Eng. degree in Naval Architecture &
trajectory estimation of hybrid aerial-aquatic vehicles. In: IEEE/RSJ International Ocean Engineering from Huazhong University of Science
Conference on Intelligent Robots and Systems, IROS, pp. 7460–7465. and Technology, Wuhan, China, in 2015. He is currently
Davis, R.E., Eriksen, C.C., Jones, C.P., et al., 2002. Autonomous buoyancy-driven pursuing the Ph.D. degree in Naval Architecture, Ocean
underwater gliders. In: Technology and Applications of Autonomous Underwater & Civil Engineering at Shanghai Jiao Tong University,
Vehicles. pp. 37–58. Shanghai, China. His research interest includes underwater
Drews, P.L.J., Neto, A.A., Campos, M.F.M., 2014. Hybrid unmanned aerial underwa- vehicles, hybrid aerial underwater vehicles, nonlinear time
ter vehicle: Modeling and simulation. In: IEEE/RSJ International Conference on varying systems, adaptive control, and backstepping control.
Intelligent Robots and Systems, IROS, pp. 4637–4642.
Edwards, D., Arnold, N., Heinzen, S., Strem, C., Young, T., 2017. Flying emplacement
of an underwater glider. In: OCEANS 2017 - Anchorage, pp. 1–6.
Gonzalez-Haro, C., Autret, E., Isern-Fontanet, J., Tandeo, P., Garello, R., 2016. Ocean
surface current reconstruction: On the transfer function between infrared SST and
along-track altimeter observations. In: Oceans 2016 MTS/IEEE Monterey, pp. 1–5. Chengke Xiong received her Bachelor’s degree in Naval
Graver, J.G., 2005. Underwater Gliders: Dynamics, Control and Design (Thesis). Architecture & Ocean Engineering from Huazhong Univer-
Princeton university Princeton, NJ. sity of Science and Technology. She is currently a Ph.D.
Herbers, T.H.C., Janssen, T.T., 2016. Lagrangian Surface wave motion and Stokes drift student in School of Naval Architecture, Ocean & Civil
fluctuations. J. Phys. Oceanogr. 46 (4), 1009–1021. Engineering at Shanghai Jiao Tong University. Her Ph.D.
Leonard, N.E., Graver, J.G., 2001. Model-based feedback control of autonomous works are focused on optimization algorithms, adaptive
underwater gliders. IEEE J. Ocean. Eng. 26 (4), 633–645. ocean sampling and path planning of unmanned marine
Lu, D., Xiong, C.K., Lyu, B.Z., Zeng, Z., Lian, L., 2018. Multi-mode hybrid aerial vehicles. She is also an IEEE student member.
underwater vehicle with extended endurance. In: 2018 Oceans - MTS/IEEE Kobe
Techno-Oceans, OTO, pp. 1–7.
Lu, D., Xiong, C.K., Zeng, Z., Lian, L., 2019. A multimodal aerial underwater vehicle
with extended endurance and capabilities. In: 2019 International Conference on
Robotics and Automation, ICRA, pp. 4674–4680.
Maia, M.M., Mercado, D.A., Diez, F.J., 2017. Design and implementation of mul- Hexiong Zhou received the Bachelor’s degree in electrical
tirotor aerial-underwater vehicles with experimental results. In: 2017 IEEE/RSJ engineering and automation from Northwestern Polytechni-
International Conference on Intelligent Robots and Systems, IROS, pp. 961–966. cal University, China, in 2015, and received the Master’s
Maia, M.M., Soni, P., Diez-Garias1, F.J., 2015. Demonstration of an Aerial and degree in control science and engineering at Harbin Engi-
submersible vehicle capable of flight and underwater navigation with seamless neering University, China, in 2018. He is currently working
air-water transition. arXiv preprint arXiv:1507.01932. toward the Ph.D. degree at State Key Laboratory of Ocean
Ore, J.-P., Elbaum, S., Burgin, A., Detweiler, C., 2015. Autonomous aerial water Engineering, Shanghai Jiao Tong University. His research in-
sampling. J. Field Robotics 32 (8), 1095–1113. terests include path planning, optimal guidance, navigation
Peng, Z., Wang, J., Wang, D., Han, Q., 2020. An overview of recent advances in and tracking control of autonomous underwater vehicles
coordinated control of multiple autonomous surface vehicles. IEEE Trans. Ind. Inf. (AUVs).
1.
Ross, J., Lindsay, J., Gregson, E., Moore, A., Patel, J., Seto, M., 2019. Collaboration
of multi-domain marine robots towards above and below-water characterization of
floating targets. In: 2019 IEEE International Symposium on Robotic and Sensors Chenxin Lyu received the B.Eng. degree in ocean technol-
Environments. ROSE, IEEE, pp. 1–7. ogy from Dalian University of Technology, Dalian, China,
Shkurti, F., Xu, A., Meghjani, M., Gamboa Higuera, J.C., Girdhar, Y., Giguére, P., in 2019. He is currently working toward the M.S. degree at
Dey, B.B., Li, J., Kalmbach, A., Prahacs, C., Turgeon, K., Rekleitis, I., Dudek, G., the School of Oceanography, Shanghai Jiao Tong University,
2012. Multi-Domain monitoring of marine environments using a heterogeneous Shanghai, China. His research interests include underwater
robot team. In: 2012 IEEE/RSJ International Conference on Intelligent Robots and vehicles, hybrid aerial underwater vehicles, and adaptive
Systems, IROS, pp. 1747–1753. control.
Siddall, R., Kennedy, G., Kovač, M., 2018. High-power propulsion strategies for aquatic
take-off in robotics. In: Bicchi, A., Burgard, W. (Eds.), Robotics Research: Vol. 1.
Springer International Publishing, Cham, pp. 5–20.
Siddall, R., Kovač, M., 2017. Fast aquatic escape with a jet thruster. IEEE/ASME Trans.
Mechatronics 22 (1), 217–226.
Siddall, R., Ortega Ancel, A., Kovač, M., 2017. Wind and water tunnel testing of a Rui Hu received the Bachelor’s degree in automation
morphing aquatic micro air vehicle. Interface Focus 7 (1), 20160085. science and engineering from South China university of
Stewart, W., Weisler, W., Anderson, M., Bryant, M., Peters, K., 2019. Dynamic modeling Technology. Guangzhou, China, in 2019. He is currently
of passively draining structures for Aerial–Aquatic unmanned vehicles. IEEE J. working toward master degree at the school of oceanog-
Ocean. Eng. 1–11. raphy, Shanghai Jiao Tong University, Shanghai, China.
Stewart, W., Weisler, W., MacLeod, M., Powers, T., Defreitas, A., Gritter, R., Ander- He focuses on control system of hybrid aerial underwater
son, M., Peters, K., Gopalarathnam, A., Bryant, M., 2018. Design and demonstration vehicles now.
of a seabird-inspired fixed-wing hybrid UAV-UUV system. Bioinspiration Biomim.
13 (5), 056013.
Taher, T., Viswanathan, V., Varghese, T., Jiang, H., Patrikalakis, N., Cloitre, A., 2016.
Multi-domain autonomous mobile network for sensing. In: OCEANS 2016 MTS/IEEE
Monterey. IEEE, pp. 1–6.
Valdivia y Alvarado, P., Taher, T., Kurniawati, H., Weymouth, G., Khan, R.R.,
Leighton, J., Papadopoulos, G., Barbastathis, G., Patrikalakis, N., 2011. A coastal Caoyang Yu received the B.E. and Ph.D. degrees in marine
distributed autonomous sensor network. In: OCEANS’11 MTS/IEEE KONA. IEEE, engineering from the School of Naval Architecture and
pp. 1–8. Ocean Engineering, Huazhong University of Science and
Weisler, W., Stewart, W., Anderson, M.B., Peters, K.J., Gopalarathnam, A., Bryant, M., Technology, Wuhan, China, in 2013 and 2018, respectively.
2017. Testing and characterization of a fixed wing cross-domain unmanned vehicle From September 2017 to February 2018, he was a Visiting
operating in aerial and underwater environments. IEEE J. Ocean. Eng. 43 (4), Research Student with the Faculty of Engineering and the
969–982. Environment, University of Southampton, Southampton, UK.
Young, T., 2014. Design and testing of an unmanned aerial to underwater vehicle. In: He is currently with the School of Oceanography, Shang-
14th AIAA Aviation Technology, Integration, and Operations Conference. In: AIAA hai Jiao Tong University. His research interests include
AVIATION Forum, American Institute of Aeronautics and Astronautics. guidance and control for marine robotic vehicles.

18
D. Lu et al. Ocean Engineering 219 (2021) 108324

Zheng Zeng received the B.Eng. degree in electrical and Lian Lian received the B.S. and M.S. degrees in naval
electronic engineering from Hunan University, Changsha, architecture and ocean engineering and the Ph.D. de-
China, in 2010, and the Ph.D. degree from Flinders Uni- gree in technology management from Shanghai Jiao Tong
versity, Adelaide, SA, Australia, in 2015. He is currently University, Shanghai, China, in 1982, 1985, and 2012,
an Associate Research Professor at Shanghai Jiao Tong respectively. She has been a Full Professor at Shanghai Jiao
University, Shanghai, China. Principal areas of expertise lie Tong University since 1998, and served as a member of
in autonomous marine vehicles, including optimal guidance, Expert Group of National Hi-Tech Program (863 Program)
navigation, and control systems. He is the Project Leader from 1999 to 2011 and a member of Expert Group of
for multiple projects funded by Shanghai Sailing Program, the National Science Foundation of China from 2008 to
Qingdao National Laboratory for Marine Science and Tech- 2011. She is currently the Vice Dean of the Institute
nology and National Natural Science Foundation of China. of Oceanography, Shanghai Jiao Tong University, and a
Prof. Zeng is a reviewer for multiple international scien- member of the IEEE/OES Administrative Committee, as well
tific journals, a committee member of the IEEE-OES/MTS as the IEEE/OES Shanghai Chapter Chair. Her research
OCEANS’16 Shanghai conference, and he currently serves mainly focuses on underwater vehicles and marine obser-
as the Secretary of the IEEE-OES Shanghai Section. vation. She has been in charge of eight projects as PI/Chief
Designer with total funding over 90 million RMB, which
include national and municipal research projects funded by
the National Hi-Tech 863 Program, National Key Project,
National Natural Science Foundation of China, National
R&D Program of Marine Technology, as well as the R&D
Program of Science and Technology Committee of Shanghai
Municipal Government.

19