Accepted Manuscript

Design and performance evaluation of an amphibious spherical robot

Maoxun Li, Shuxiang Guo, Hideyuki Hirata, Hidenori Ishihara

PII: S0921-8890(14)00249-8
DOI: http://dx.doi.org/10.1016/j.robot.2014.11.007
Reference: ROBOT 2397

To appear in: Robotics and Autonomous Systems

Received date: 30 August 2014

Revised date: 29 October 2014
Accepted date: 7 November 2014

Please cite this article as: M. Li, S. Guo, H. Hirata, H. Ishihara, Design and performance
evaluation of an amphibious spherical robot, Robotics and Autonomous Systems (2014),
http://dx.doi.org/10.1016/j.robot.2014.11.007

This is a PDF file of an unedited manuscript that has been accepted for publication. As a
service to our customers we are providing this early version of the manuscript. The manuscript
will undergo copyediting, typesetting, and review of the resulting proof before it is published in
its final form. Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
 Design and Performance Evaluation of an
Amphibious Spherical Robot
Maoxun Li1,*, Shuxiang Guo2,3, Hideyuki Hirata2, Hidenori Ishihara2
1
Graduate school of Engineering, Kagawa University, 2217-20, Hayashichou, Kagawa, Japan
2
Faculty of Engineering, Kagawa University, 2217-20, Hayashichou, Kagawa, Japan
3
Beijing Institute of Technology, China
* Corresponding author E-mail: [email protected]

Abstract – This paper presents an amphibious spherical monitoring, sample collection, and recovering objects on land
robot that consists of a sealed upper hemispheroid, two quarter and underwater.
spherical shells, and a plastic circular plate. It has a plastic shelf As the typical amphibious robots, snake-inspired robots
for carrying the micro-robots, and four actuating units for propel themselves through undulating their bodies on land and
movement. Each unit is composed of a water-jet propeller and
two servomotors, each of which can rotate 90º in the horizontal
underwater. ACM-R5 is one of the snake-like robots with
and vertical directions. The robot is capable of motion on land, as good property of water-proof and high obdurability [1]. Unlike
well as underwater. The robot is capable of three walking gaits; snake robots, a number of amphibious robots employ different
therefore, we describe experiments on various terrains to propulsion methods to cope with different environments. The
evaluate the walking motion performance, including stability and actuator of an amphibious robot named Whegs is a
velocity. Additionally, plenty of underwater experiments are combination of propellers and legs, which allows the robot to
conducted to evaluate the underwater performance, containing move on rough terrain and underwater [2]. AmphiRobot-II is
horizontal and vertical motions, and to verify the fixture and an amphibious biomimetic fish-like robot with a wheel–
deployment mechanism for the micro-robot. propeller–fin mechanism and a specialized swivel mechanism.
[3]. More specifically, the wheel–propeller–fin mechanism
Index Terms – Amphibious Spherical Robot, Quadruped
Walking, Walking Gait, Water-jet Propeller.
functions as a driving wheel for crawling on land, as a
common screw propeller or a pectoral fin in water. With the
Nomenclature specialized swivel mechanism, AmphiRobot-II is able to
freely switch between fish-like (lateral) and dolphin-like
, , torque on the servomotor (dorsoventral) swimming motions. A salamander-like
, moment arm of the servomotor amphibious robot named Salamander Robot can mimic the
ability of motion transformation from terrestrial to aquatic
, normal forces exerted on the surfaces
locomotion by utilizing body undulation and limb walking,
, friction and coefficient of friction and vice versa [4]. An amphibious mobile robot with a
rotational velocity of motor shaft spherical rotary paddle mechanism is reported in Ref. [5]. The
, incoming flow and central flow velocities spherical rotary paddle mechanism functions as a rotary
a linear combination of and paddle in water, as a wheel for moving on land. The robot
with “Omni-Paddle” can realize an effective movement on the
diameter of the nozzle border of ground and water. AQUA series robots are the most
duty factor of the gait-timing sequences attractive ones among amphibious robots [6]. AQUA is a
rotation angle of the horizontal motor cockroach-inspired hexapod robot with splendid land
walking velocity of the robot performance. AQUA's legs are interchangeable: flippers for
swimming and for limited walking along the beach, and
step size of the robot rubber-treaded legs for walking. AmphiHex-I with a
frequency of one step cycle transformable fin-leg composite propulsion mechanism can
pass through rough land, soft substrate and water
I. INTRODUCTION simultaneously [7].
Compared with underwater robots, amphibious robots are
Amphibians, as the most primitive terrestrial vertebrates,
able to move from the water to the ground without manpower,
experienced a long natural selection. They are transitional
and vice versa. Furthermore, each amphibious robot has its
animals from aquatic to terrestrial in the history of vertebrate
own characteristics and advantages. Wheeled robots have
evolutionary. Amphibians possess great abilities to adapt to
good performance on even ground, while tracked and legged
various complex environments, on land and underwater.
robots have good mobility on rough terrain. Compared with
Researches of the amphibious robots inspired by amphibians
screw propellers, undulatory and oscillatory propulsion with
have been focused on by researchers around the world. With a
lower environmental disturbance can also achieve high
high adaptive faculty, amphibious robots can be used in broad
efficiency and maneuverability. Some robots utilize two sets
applications including topography surveys, water-quality
of propulsion mechanism for terrestrial and aquatic motions, The amphibious robot consists of a sealed transparent
which leads to a heavy body. To simplify the structure, robots upper hemispheroid, two transparent quarter spherical shells
like ACM-R5, Whegs and AQUA2 use composite propulsion that can be opened, a plastic plate for carrying the micro-
mechanism to move in amphibious environments. However, it robots, and four actuating units, which are fastened to a plastic
is still problematic for these amphibious robots to move in circular plate, as shown in Fig. 2. The two quarter-spherical
confined spaces. In order to improve mobility on the complex shells are controlled by two servomotors to open the shells by
terrains, there are no manipulators carried on the existing rotating them through an angle of 90º. The control circuits,
amphibious robots for on-land and underwater operations. power supply, and sensors are placed in the sealed upper
Moreover, it is difficult for these amphibious robots to achieve hemispheroid, which is waterproof. The space in the lower
accurate control of position underwater through swimming hemisphere is connected to the outside environment via gaps
motion. Once there are continuous currents under the sea, on the hull. The plastic plate and actuating system are installed
amphibious robots without legs cannot keep themselves still in the lower hemisphere. Each actuating unit consists of a
for precise manipulation. water-jet propeller and two servomotors, which are
To overcome the limitation of the previous amphibious perpendicular to each other, and can realize two degrees of
robots, the purpose of this study is to design and develop a freedom. The holes on the quarter spherical shell provide the
novel amphibious spherical robot with transformable space for the water-jet propellers to rotate to allow the robot to
composite propulsion mechanisms, which can control and maneuver when it is closed. The diameter of the upper
carry micro-robots. The composite mechanism is designed to hemisphere is 234 mm, and the diameter of the lower
switch between water-jet propeller and leg. A sphere provides hemisphere is 250 mm. The height of the actuating unit in the
maximum internal space, and symmetry provides advantages standing state is 108 mm, and it is 85 mm long.
of flexibility both underwater and on land [8], [9]. With the B. Opening Mechanism
ability of carrying micro-robots actuated by smart actuators
Figure 3 shows the opening mechanism of the amphibious
[10], [17], spherical robot can be employed to collect samples
robot. The quarter spherical shells are mounted on two parallel
or carry out water-quality monitoring in a confined underwater
axles, which are fixed to the upper hemisphere at a distance
space [11], [18]. Micro-robot can be used as a manipulator to
for independent control. From Fig. 3, we can see that this
grip objects underwater or a monitor to monitor the water
structure avoids collisions between the two quarter-spherical
quality and so on. Finally, experiments on various terrains and
shells and the upper hemisphere. The two axles are actuated
underwater are conducted to evaluate the walking motion
by two JR DS3836 servomotors.
performance and underwater performance, and micro-robot
deployment experiment is also carried out. C. Fixture Mechanism of the Micro-robots
The remainder of this paper is organized as follows. In A fixture mechanism was developed to carry the micro-
section II, we describe the mechanical design of the spherical robots and deploy them. To simplify the control system, we
robot and introduce the opening and fixture mechanisms. In propose a mechanism with no actuating parts, as shown in Fig.
section III, we introduce the propulsion mechanism for 4. It is a block that can move up and down freely on a plate.
movement underwater and on land and the control system. When the amphibious robot walks on land or swims
The development of a prototype and three walking gaits are underwater, the block falls down to fix the position of the
described, and the results of on-land experiments to evaluate micro-robots by gravity. When deploying the micro-robots,
the walking stability and velocity are presented in section VI. the four actuating units rotate upwards to allow the robots to
Section V describes the results of underwater thrust and be supported by the plate, and push the block up so that the
velocity experiments to evaluate the underwater performance. micro-robots can leave the amphibious robot.
The feasibility of the fixture and deployment mechanism is
proved in section VI. Section VII concludes the paper.
II. MECHANICAL DESIGN
A. General Design
The design concept of the amphibious spherical robot is
illustrated in Fig. 1. With transformable composite propulsion
mechanisms, the robot has two actuating modes: quadruped (a)
walking mode and water-jet propulsion mode. The amphibious
robot is capable of movement from the water to the ground
without manpower, and vice versa. Underwater micro-robots
with flexible movement and compact structure are carried on
the amphibious robot. The amphibious robot can provide
power and send signals to micro-robots with cables. Micro-
robots can be controlled to collect underwater objects as a
(b)
manipulator or monitor underwater environment in a restricted Fig. 1 Amphibious robot configuration. (a) Spherical robot moving close to
space. the target. (b) Micro-robots conducting tasks in restricted spaces.
 water-jet propeller is controlled to move vertically. Each unit
has two degrees of freedom.
A. Quadruped Actuation System
The quadruped actuation was built to realize walking
motion [12], [13]. Two servomotors, which are located on the
surface of the plastic circular plate outside the upper
hemisphere, are used to control the spherical shells.
To identify the appropriate servomotors for the legs of the
robot, we carried out force analysis calculations. Figure 5
shows the force analysis of the robot on land. In the standing
state, each vertical servomotor provides a torque . In the
walking state, each vertical servomotor of the stationary legs
Fig. 2 Structure of the amphibious spherical robot. provides a torque and each horizontal servomotor of the
moving legs provides a torque . Here, we assume that at a
given moment, the supporting force on each leg in contact
with the ground is the same. Considering that the torque on
each servomotor cannot exceed the rated torque , we arrive
at the following relationships:
(1)

(2)

(3)
Fig. 3 Opening mechanism of the amphibious spherical robot. The condition of frictional motion of each leg without
longitudinal and transversal slipping is:
(4)
where and are the normal forces exerted on the
surfaces, is the force due to friction, is the coefficient of
friction of the contact surface, is the moment of the arm of
the vertical motor, and is the moment arm of the horizontal
motor.
(5)
(a)
The normal force varies depending on the walking
gait. The force is the minimum force that supports the
robot; when robot stands with four legs, this is given by (1).
When the robot walks on land, there are times when only two
or three legs are in contact with the ground. The supporting
force is the maximum supporting force. We chose a
servomotor that allows (5) to hold.

(b)
Fig. 4 Fixture mechanism of the micro-robots: (a) locked state and (b) free
state.

III. ACTUATION AND CONTROL SYSTEM

There are two actuation systems: a quadruped walking
actuation system and a water-jet actuation system. The
actuation system consists of four main units. Each unit
includes a water-jet propeller, two servomotors, and a
stainless-steel stand. The motor connected to the circular plate Fig. 5 Force analysis on land: (a) standing and (b) walking.
is controlled to move horizontally, and the motor fixed on the
 Assuming that the flow is incompressible, according to
the equation of continuity, we know that the volume of
incoming flow must be equal to the outlet flow, as shown in
(9).
(9)

where: the fluid density and are equal to , the cross-

section of the nozzle is equal to .
Hence, we arrive at the following relationship:
(10)

The thrust of the water-jet propeller in two directions can

be calculated as (11) and (12) respectively.
Fig. 6 Force analysis on one actuating unit during (a) horizontal motion, (11)
(b) floating motion, and (c) sinking motion. The blue arrows indicate the
water-jet direction.
(12)
B. Water-Jet Actuation System
The robot is actuated by water-jet propellers when By substituting (8) in (11) and (12) respectively, the thrust
underwater [14]. By controlling the angles of the servomotors, generated by water-jet propeller can be computed as (13) and
the spray angle of the water-jet propellers can be varied to (14).
control the motion of the robot. A force analysis was carried (13)
out to assess the requirements of the servomotors on the
actuator units. Figure 6 shows the force analysis for one
actuator unit during horizontal motion, floating motion, and (14)
sinking motion. By changing the angle of the vertical motor,
thrust can be realized in all directions. The torque of the motor
can be calculated as follows:
(6)

(7)

where , , and are the thrusts generated by the

water-jet propeller, , , and are the torque on the
servomotor, and is the moment arm of the vertical motor.
Fig. 7 Fluid model of water-jet propeller.
Appropriate servomotors were chosen by considering the
maximum required torque.
To achieve the dynamic model of water-jet propeller, we
build a fluid model of water-jet propeller by reference to a
previous method [9], [15], as shown in Fig. 7. The shaft of the
motor, on which four blades are fixed, is perpendicular to the
nozzle. Due to the small diameter of the nozzle, we can ignore
the velocity difference across the nozzle. So we consider the
axis flow velocity as a linear combination of incoming flow
velocity and the central flow velocity . In Fig. 7, is the
rotational velocity of motor shaft, is velocity of ambient
flow, is velocity of incoming flow, is velocity of outlet
flow, is incoming angle of ambient flow and is diameter
of the nozzle. Finally, the axis flow velocity can be obtained:
Fig. 8 Control system.

C. Sensing Mechanism
(8)
The amphibious robot can detect and avoid obstacles
around it using proximity sensors. The spherical symmetry
allows the robot to move in any direction by spinning prior to
forward or backward motion. The sensors should therefore be relative phases of the legs for the first gait are shown in
placed symmetrically at the bottom of the upper hemisphere Fig. 11(a).
next to the transparent hull. Eight Sharp GP2Y0A21YK
infrared proximity sensors were used.
Because of the amphibious requirements, the robot must
be able to determine whether it is on land or underwater. A
Panasonic ADPW11 pressure sensor that can measure the air
pressure and water pressure is located on the robot, fixed to
the plastic circular plate.
D. Electrical System and Power Supply
The control center of the spherical robot is shown in
Fig. 8. It is based on an AVR ATMEGA2560 micro-
controller, and uses ten PWM channels to control the eight
servomotors to drive the robot and the two servomotors on the (a) (b)
upper hemisphere to open and close the two quarter-spherical Fig. 9 Prototype amphibious robot in (a) quadruped walking mode, and (b)
shells. Furthermore, eight input/output ports are used to water-jet propulsion mode.
control the four water-jet propellers. Using two data
transmission ports, we utilized analog-to-digital conversion so
that the microcontroller could receive and transmit data that
control the infrared proximity sensors and pressure sensors,
thereby realizing closed-loop control. The control system also
included a radio frequency (RF) remote control module. Thus, (a)
a further four input/output ports were connected to the receiver
module with four channels to control remotely the movement
of the robot.
The power supply used three batteries: one
6TNH22A/8.4V battery to provide power to the AVR micro-
controller, and two YBP216BE/7.4V batteries to provide (b) (c)
power to the ten servomotors and the four water-jet propellers. Fig. 10 Event sequences of one gait cycle for the three different gaits: (a) Gait
1, (b) Gait 2, and (c) Gait 3. The legs are labeled as follows: left fore (LF),
IV. ON-LAND EXPERIMENTS AND ANALYSIS right fore (RF), left hind (LH), and right hind (RH). The blue bars indicate
that the legs are in contact with the ground.
A. Prototype Robot
As shown in Fig. 9, a prototype amphibious spherical
robot was constructed with two actuating modes. The robot
consisted of two main parts: the upper hemisphere and two
transparent quarter-spherical shells. The actuating system and
the plate for the micro-robots were located in the lower
hemisphere. We chose to use HS-5086WP servomotors,
supplied by the Hitec Company, which are waterproof. The
water-jet propellers were not waterproof, and were supplied by
Raboesch Company. The total mass of the robot is 2.1 kg. (a) (b)
Fig. 11 Relative phases of one cycle for the different gaits (bottom view): (a)
B. Gait Characterization Gait 1 and Gait 2, and (b) Gait 3.
As shown in Figs. 10(b) and 10(c), for the second and
To adapt to different environments, quadruped robots with third gaits, the front leg is lifted while the ipsilateral hind leg
two degrees of freedom in each leg may use a number of is set down, which can achieve a good stability margin. The
different gaits [16]. Three gaits have been implemented in the duty factor of the second gait is β = 0.75, which achieves the
spherical robot. The first gait (Gait 1) is a statically stable maximum contact time with the ground. The duty factor
regular symmetric gait, in which at least three legs are contact decreases with increasing speed. For the third gait, we have
with the ground at any one time. Accordingly, the gait event β = 0.67, and there are times when the robot has only two legs
sequence and the gait timing can be defined by the duty factor in contact with the ground.
β and the relative phase of the left-hind leg . The relative The velocity of the robot is related to the step size and
phases of all the legs are defined so that = 0 and = frequency of the gait cycle, i.e.,
0.5; the right-hind (RH) leg has a phase difference of 0.5
relative to the left-hind (LH) leg. Figure 10(a) shows the event (15)
sequences of the first gait with a duty factor of β = 0.8. The
where is the velocity of the robot, is the step size of the where is the duty factor of the gait-timing sequences, and
robot, is the frequency of one step cycle. The step size is is the rotation angle of the horizontal motor.
proportional to the angle of rotation of the leg, which is the Substituting (16) into (15), we can achieve the following:
given by the angle of rotation of the horizontal motor and the
duty factor of the gait-timing sequence, i.e., (17)
(16) which is related to the frequency and the duty factor of the gait
cycle.

(a) (b)

(c) (d)

(e) (f)

Fig. 12 Vibration amplitude during walking and rotating motions: (a) walking with Gait 1, (b) rotation with Gait 1, (c) walking with Gait 2, (d) rotation with Gait
2, (e) walking with Gait 3, and (f) rotation with Gait 3. The blue and red points correspond to vibration data in the pitch and roll directions, respectively.

(a) (b)
Fig. 13 Experimental results during (a) walking motion, and (b) rotating motion. The blue, red, and green points correspond to Gaits 1, 2, and 3, respectively.
C. Gait Stability Experiments
The gait stability and velocity are two principal elements
that are used to evaluate the motion of a robot. We tested the
stability of the robot using an MTx sensor to detect vibration
during the gait cycle in walking and rotating motions. The
robot was programmed to walk for 1 m and then rotate 180º
on a tiled floor. Figure 12 shows the peak-to-peak amplitude
of vibration of the robot’s body during walking and rotation.
The vibration during walking in the two directions decreased (a)
as the frequency increased for all three gait cycles. Under a
relatively high frequency, the robot could attain a stable state.
With Gait 1 and Gait 2, the vibration amplitude changed little
as the frequency of the gait cycle increased while walking and
rotation. However, with Gait 3, the variation in the vibration
as a function of the frequency of the gait cycle was
considerable during rotation, which means that the rotating
motion was unstable. Gait 1 provided the most stable motion;
however at high frequencies, Gait 3 was the most stable in the
pitch direction.
(b)
With Gait 3, there is a portion of the gait cycle where the
robot has only two legs in contact with the ground, and so
there is an inherent instability, which can lead to vibration and
may be expected to change with the frequency of the gait
cycle. The instantaneous driving force applied to the robot
when the legs make contact with the ground may also generate
vibration. The amplitude of this vibration is expected to
change little with frequency.
The walking and rotational velocities of the robot are
shown in Fig. 13 as a function of the frequency of the gait
cycle for the three different gaits. The velocity initially (c)
increased as a function of frequency, and then decreased
towards zero. The same is true of the rotational velocity. At
relatively low frequencies, the velocities of all three gaits were
approximately equal. At higher frequencies, motion with
Gait 3 was the fastest, and with Gait 1 was the slowest. The
rotational velocity was only affected by the frequency of the
gait cycle, and was approximately the same for the three
different gait cycles. Gait 1 was the most stable and Gait 3 was
the least stable during rotation.

(d)
Fig. 15 Velocity on different terrains: (a) tiled floor, (b) asphalt road (c) brick
road, and (d) on sand. The blue, red, and green points correspond to Gaits 1, 2,
and 3, respectively.

Fig. 14 Walking experiments on different terrains: (a) asphalt road, (b) Fig. 16 Theoretical walking velocity.
cement floor, (c) brick floor, (d) sand, and (e) grass.
D. Experiments on Different Terrains contact time of Gait 3 was the shortest (i.e., it had the smallest
duty factor), and so the actuating ability of the servomotor
For practical applications, it is important that the robot is
could be exploited at high frequencies.
able to negotiate a variety of terrains. We carried out walking
experiments on an asphalt road, a cement floor, and a brick
road, and on sand and grass. As shown in Fig. 14, these
terrains can be characterized by different coefficients of
friction and different roughnesses. The adaptability to these
environments illustrates that the robot can negotiate a range of
on-land environments. In addition, we evaluated the walking
ability on a slope.
Figure 15 shows the velocity of the robot on the different
terrains, including the tiled floor in the laboratory. To evaluate
(a)
the on-land performance, we measured the walking velocity at
different frequencies with the three different gaits. The curves
show that as the frequency increased, the walking velocity
increased at first, and then decreased to zero for each gait and
on each terrain. At relatively low frequencies, the velocities of
the robot on each terrain were approximately equal for all
three gaits.
Figure 15(a) shows the velocity of the robot on different
terrains. The tiled floor was the smoothest surface, and the
robot could move faster on it than on the other surfaces. At
low frequencies, the velocity of the robot was almost the same (b)
on the different surfaces. As the frequency increased, the
rougher the surface was, the more rapidly the velocity
decreased as a function of the frequency.
Figure 16 shows the theoretical walking speed of the three
gaits calculated using (17). The theoretical walking speed is
proportional to the frequency, and at a given frequency, the
walking speed is inversely proportional to the duty factor.
However, the experimental data shows a non-linear
relationship between the velocity and frequency. At high
(c)
frequencies, the speed was not inversely proportional to the
duty factor of the gait-timing sequences, as shown in
Figs. 15(c) and (d).
The velocity of the robot is related to the step size and the
frequency of the gait cycle, which follows from (15). At the
same gait and frequency, the velocity varied between the
different terrains. Hence, the different terrains affected the
step size of the robot during the gait cycle.
The step size of the robot can be calculated from (16).
The step size is dependent on the angle of rotation of the leg,
(d)
θ, which depends on the angle of rotation of the horizontal
motor, which was set to 60°. With = 8.5 cm and 60º of motor Fig. 17 Step size on different terrains: (a) tiled floor, (b) asphalt road, (c) brick
rotation, the step size given by (16) is 8.898 / β cm. However, road, and (d) on sand.
the experimentally measured step sizes shown in Fig. 17
depended on the terrain. From these results, at low
frequencies, the step size increased as a function of the
frequency of the gait cycle, and then decreased towards zero at
high frequencies. This is because there are three losses
affecting the step size: the first is the error caused by the
limited response time of the servomotors, which is the main
loss at high frequencies; the second is the slip between the foot
and the surface during each step; and the third is due to
vibration. When walking forward, the step size of Gait 3
varied the least with the frequency of the gait cycle, and the Fig. 18 Climbing experiments on a slope with different inclination angles.
step size of Gait 1 varied the most. This was because the
 The measured speed was almost equal to the theoretical jet propeller in two situations are shown in Fig. 20. The thrust
value at low frequencies. The two values diverged as the increases as the duty of PWM signals increases. From the
frequency increased. At high frequencies, the rotation angle results, the water-jet propeller in Situation 2 shows the better
error due to the response time of the servomotors affected not performance than Situation 1. A maximum thrust of 99.5 mN
only the horizontal motion, but also the vertical motion, which in Situation 1 and a maximum thrust of 125.8 mN in Situation
led to a decrease in the height that the legs were lifted from the 2 were achieved at a duty of 100%.
ground. The rougher the surface was, the more the drop in the The experimental setup for measuring the horizontal
leg height affected the speed of the robot. As shown in Fig. 17, thrust was designed by employing leverage principle, as
at high frequencies, the step size on the roughest ground was shown in Fig. 21. An electronic scale was used to measure the
the shortest, and that on the smooth surfaces was the longest. thrust generated by the water-jet propellers of the robot.
The performance of the robot was also evaluated on a Assume that the length of two levers are and respectively,
slope. The robot was programmed to climb a slope consisting and the difference between the two readings displayed on the
of a wooden surface with a variable incline. The robot was electronic scale is . The horizontal thrust can be calculated
programmed to move with all three gaits. To simplify the according to (18).
experiments, a fixed frequency of 0.63 Hz was used for Gait 1,
0.78 Hz for Gait 2, and 1.04 Hz for Gait 3. A crucial aspect of
climbing a slope is to maintain balance, which means that the
robot’s center of gravity must remain inside a polygon formed
by the supporting legs.
Figure 18 shows the velocity as a function of the incline
for the three different gates. Gait 3 resulted in faster motion on
a gentle slope; however, as the incline increased, the velocity
of the robot decreased to zero more rapidly than the other
gaits. Gait 2 showed the best performance on a steep slope,
allowing the robot to climb an inclination of 8°. Fig. 19 Experimental setup for thrust measurement of a water-jet propeller in
two situations. Blue arrows indicate the thrust direction.

V. UNDERWATER EXPERIMENTS AND ANALYSIS

A. Underwater Thrust Experiments

To implement a controllable thrust of the water-jet
propeller, the rotational velocity of the motor was controlled
to change in a range as the duty of PWM signals changed.
According to (13) and (14), we know that the thrust is related
to the rotational velocity which varies depending on the duty
of PWM signals. For the purpose of the thrust control,
experiments of thrust measurement of a single water-jet
propeller were conducted to get the relationship between the
Fig. 20 Results of the average thrust of a water-jet propeller.
duty of PWM signals and the thrust of the water-jet propeller.
Additionally, experiments of horizontal thrust and vertical
thrust measurement of the robot as a function of the duty were
carried out. All experiments were repeated ten times at a set of
control signals to achieve an average thrust.
The experimental setup, as shown in Fig. 19, was
designed to measure the thrust of the water-jet propeller
according to leverage principle. The propeller can generate
two kinds of thrust: thrust generated by the flow passing
through the long nozzle (Situation 1) and the flow passing
through the nozzle on the opposite direction (Situation 2). An
Fig. 21 Experimental setup for horizontal thrust measurement. Blue arrows
electronic scale was used to measure the thrust. Assume that indicate the thrust direction. Blue triangles indicate applied actuating units.
the length of two levers are and respectively, and the
difference between the two readings displayed on the There are two kinds of actuating modes while moving
electronic scale is . We calculated the thrust of the water-jet forward and backward. The first actuating mode of the robot is
propeller according to the following: to use two actuating units opposite to the moving direction.
The thrust (in Situation 1) generated by the flow passing
(18) through the long nozzle is the main thrust of the robot. For the
second actuating mode, the robot is actuated by the remaining
The results of the average thrust measurement of a water- two actuating units. The thrust (in Situation 2) produced by the
flow passing through the nozzle on the opposite direction is Figure 24 shows the results of the average vertical thrust
the assist thrust of the robot. In order to gain a larger thrust, as a function of the duty. From the results of two graphs, the
four actuating units work at the same time to get both the main thrust increases in terms of the duty of PWM signals, and the
thrust and the assist thrust. actual thrust is close to the ideal thrust due to less thrust loss.
Figure 22 shows the results of the average horizontal A maximum upward thrust of 333.2 mN and a maximum
thrust as a function of the duty. From the results of two graphs, downward thrust of 362.6 mN were achieved. The direction of
the thrust increases in terms of the duty of PWM signals. A the buoyance of the robot leads to the difference between the
maximum main thrust of 120 mN and a maximum assist thrust two kinds of thrust. The experiments of both horizontal thrust
of 60 mN, which is approximately half of the main thrust, measurement and vertical thrust measurement proved that
were achieved. The big difference between ideal thrust and PWM control method is effective for underwater thrust control.
actual thrust is caused by the thrust loss. The assist thrust of
the robot is relatively small, because that the two nozzles are
deployed inside the robot, which causes part of the thrust to be
offset by other components. At a duty of 100%, a maximum
actual thrust of 180 mN was achieved with all the actuating
units working.
Figure 23 shows the designed experimental setup for
measuring the vertical thrust based on leverage principle. The
electronic scale was also used to measure the thrust of the
robot. Assume that the length of two levers are and
respectively, and the difference between the two readings Fig. 23 Experimental setup for vertical thrust measurement. Blue arrows
displayed on the electronic scale is . The upward thrust and indicate the thrust direction. Blue triangles indicate the applied actuating
downward thrust of the robot can be computed by (18). units.
In the experiments of vertical thrust measurement, four
actuating units were actuated to develop vertical thrust,
including upward thrust for floating upward and downward
thrust for sinking down. Compared with horizontal movement,
actuating four water-jet propellers together can gain a
relatively large actuating force while vertical movement.

(a) Upward thrust

(a) Main thrust

(b) Downward thrust

Fig. 24 Results of average vertical thrust.

B. Underwater Velocity Experiments

There are two states for the amphibious robot while
underwater, which are the semi-submerged state and
submerged state. The semi-submerged state is a state that all
(b) Assist thrust necessary loadings are equipped and no additional weight is
Fig. 22 Results of average horizontal thrust. added on the robot. And the submerged state is a state that
additional weight is added on the robot to keep it suspend in
water. In the semi-submerged state, the robot can float on
water with approximately one-third of the body over the water
surface.
The underwater experiments of the robot in the semi-
submerged state were carried out to evaluate the underwater
velocity performance of the robot. All experiments were
repeated ten times at a set of control signals to achieve an
average velocity. With the advantage of symmetry, the
spherical robot shows the same performance while moving
forward and backward or rotating clockwise and
anticlockwise. Thus, the robot was programmed to move
forward for 1m and then rotate 360º in a pool. The
displacement distance and the rotation angle were measured Fig. 26 Experimental results during underwater rotating motion in the semi-
by a ruler and a protractor respectively. By using the timer to submerged state.
measure the movement time, the average velocity of the robot The robot was also programmed to rise and sink for 1m in
was calculated. In order to make an overall analysis of the a pool. We measured the time and then calculated the average
robot, the experiments of surge motion of the robot actuated rising velocity and sinking velocity. From the experimental
by two main actuating units and four actuating units were results, a maximum rising velocity of 8.4 cm/s and a sinking
conducted respectively. velocity of 8.4 cm/s were achieved.
Figure 25 shows the experimental results of the robot
driven by different actuating units. The surge velocity VI. MICRO-ROBOT DEPLOYMENT EXPERIMENT
increased smoothly as the duty of PWM signals increased. The
blue line in the graph shows the surge velocities of robot as a A. Prototype Underwater Micro-robot
function of the duty of PWM signals under the control of two
main water-jet propellers. And the red line shows the An insect-inspired underwater micro-robot was proposed
velocities of robot actuated by all the water-jet propellers, to be carried on the amphibious robot. Figure 27 shows the
which can provide a maximum thrust combining of main proposed structure and the prototype of the wired micro-robot,
thrust and assist thrust. From the results, we got a maximum which is driven by ICPF (Ionic Conducting Polymer Film)
surge velocity of 16.1 cm/s under the control of four water-jet actuators. It consists of six legs, two fingers, and a caudal fin.
propellers. The results also indicate that the velocity difference Two inside legs are called supporters; other four legs are
in the two kinds of actuating modes is relatively small. Under called drivers. The micro-robot with nine ICPF actuators can
a maximum duty of PWM signals, the velocity difference was perform walking, rotating, floating, swimming and grasping
only 2.4 cm/s. However, the energy consumption of a four- motions. Nine actuators are all 15 mm long, 3 mm wide and
propeller-driven robot was considerable. In order to realize a 0.2 mm thick. The total size of the micro-robot is 65 mm long
high efficiency of the robot, we chose the two-propeller-driven (including the two fingers and the tail), 52 mm wide and 21
mode during the surge motion. So, we got a maximum mm high. The wired micro-robot can draw energy from an
velocity of 13.7 cm/s under such a mode. external power supply carried on the amphibious robot and
Figure 26 shows experimental results of the robot while can be controlled by the micro-controller on the robot. In order
rotation. As we know, the drag torque of a spherical robot is to fix the micro-robot on the amphibious robot with the fixture
relatively small while rotation. Thus the robot can achieve a mechanism, two holes are bored on the body of the micro-
relatively high rotational velocity. Under a duty of 100%, we robot.
got a maximal rotational velocity of 64.3 º/s. With a high B. Micro-robot Deployment Experiment
rotation performance, the robot is capable of flexible
locomotion in the complicated environments. The micro-robot can be carried and deployed by the
amphibious robot with a simple fixture mechanism, as shown
in Fig. 4. The mechanism has no actuating parts, which
consists of a block that can move up and down freely on a
plate in the lower hemisphere. The block on the amphibious
robot drops down by gravity to fix the position of the micro-
robot, during walking on land or swimming underwater. When
the amphibious robot deploys the micro-robot, the four
actuating units rotate upwards to allow the robots to be
supported by the plate, and push the block up so that the
micro-robot can leave the amphibious robot. To prove the
feasibility of the fixture and deployment mechanism, the
micro-robot deployment experiment was conducted on the
Fig. 25 Experimental results during underwater surge motion in the semi- bottom of the pool, as shown in Fig. 28.
submerged state.
 surge velocity of 16.1 cm/s and rotational velocity of 64.3 º/s
were obtained, and a maximum rising velocity of 8.4 cm/s and
a sinking velocity of 8.4 cm/s were achieved. Finally, the
micro-robot deployment experiment was carried out to prove
the feasibility of the fixture and deployment mechanism.
Future work mainly involves the improvement of the on-
land and underwater performance, the implementation of the
communication between the amphibious robot and the multi-
micro robots, and the development of the micro-robot
receiving mechanism.
Fig. 27 Underwater micro-robot: (a) proposed structure and (b) prototype. ACKNOWLEDGMENT
This research was supported by the Kagawa University
Characteristic Prior Research Fund 2012.

REFERENCES
[1] S. Hirose, H. Yamada, “Snake-like robots [Tutorial]”, IEEE Robotics &
Automation Magazine, Vol. 16, pp. 88–98, 2009.
[2] A. S. Boxerbaum, P. Werk, R. D. Quinn, R. Vaidyanathan, “Design of
an autonomous amphibious robot for surf zone operation: Part I-
Mechanical design for multi-mode mobility”, Proceedings of the IEEE
International Conference on Advanced Intelligent Mechatronics, pp.
1459–1464, 2005.
[3] J. Yu, R. Ding, Q. Yang, M. Tan and J. Zhang, “Amphibious Pattern
Design of a Robotic Fish with Wheel-propeller-fin Mechanisms”,
Journal of Field Robotics, pp. 1-15, 2013.
[4] A. J. Ijspeert, A. Crespi, D. Ryczko, J. M. Cabelguen, “From swimming
to walking with a salamander robot driven by a spinal cord model”,
Science, Vol. 315, pp. 1416–1420, 2007.
[5] K. Tadakuma, R. Tadakuma, M. Aigo, M. Shimojo, M. Higashimori, M.
Kaneko, “"Omni-Paddle": Amphibious Spherical Rotary Paddle
Mechanism”, Proceedings of 2011 IEEE International Conference on
Robotics and Automation, pp.5056-5062, 2012.
[6] G. Dudek, P. Giguere, C. Prahacs, S. Saunderson, J. Sattar, L. Torres-
Fig. 28 Micro-robot deployment experiment: (a) initial position, (b)
Mendez, M. Jenkin, A. German, A. Hogue, A. Ripsman, J. Zacher, E.
encountering a confined space, (c) deploying the micro-robot and (d) micro-
Milios, H. Liu, and P. Zhang, M. Buehler, C. Georgiades, “AQUA: An
robot enters the confined space.
Amphibious Autonomous Robot”, Computer, Vol. 40, No. 1, pp. 46-53,
2007.
VII. CONCLUSIONS [7] S. Zhang, X. Liang, L. Xu, M. Xu, “Initial Development of a Novel
Amphibious Robot with Transformable Fin-Leg Composite Propulsion
We have described a spherical robot configuration for Mechanisms”, Journal of Bionic Engineering, Vol. 10, No. 4, pp. 434-
amphibious applications. The amphibious robot can move 445, 2013.
[8] R. Chase, A. Pandya, “A Review of Active Mechanical Driving
with a relatively high velocity and for a relatively long period
Principles of Spherical Robots”, Robotics, pp. 1 – 21, 2012.
of time on land and underwater. Furthermore, by carrying the [9] X. Lin, S. Guo, K. Tanaka, and S. Hata, “Development and Evaluation
micro-robot, the robot can be employed to do the sample of a Vectored Water-jet-based Spherical Underwater Vehicle”,
collections and monitoring in a constricted underwater space. INFORMATION: An International Interdisciplinary Journal, Vol. 13,
No. 6, pp. 1985-1998, 2010.
The gait stability experiments and velocity experiments
[10] L. Shi, S. Guo, M. Li, S. Mao, N. Xiao, B. Gao, Z. Song, K. Asaka,“A
on different terrains were carried on to evaluate the on-land Novel Soft Biomimetic Microrobot with Two Motion Attitudes”,
performance of the amphibious robot. At relatively high Sensors, Vol. 12, No. 12,pp. 16732-16758, 2012.
frequencies, the robot could attain a stable gait when walking [11] S.-C. Yu, J. Yuh, J. Kim, “Armless underwater manipulation using a
small deployable agent vehicle connected by a smart cable”, Ocean
forward. The robot exhibited the best performance on the tiled
Engineering, pp. 149-159, 2013.
floor. At a frequency of 3.33 Hz with Gait 3, we achieved a [12] S. Guo, M. Li, L. Shi, S. Mao, C. Yue, “Performance Evaluation on
maximum velocity of 22.5 cm/s, and at a frequency of 1.56 Hz Land of an Amphibious Spherical Mother Robot.” Proceedings of 2013
with Gait 2, we achieved a maximum rotational velocity of ICME International Conference on Complex Medical Engineering
(ICME CME 2013), pp.602-607, 2013.
71.29 °/s. When climbing, Gait 2 had the best performance,
[13] S. Guo, M. Li, C. Yue, “Performance Evaluation on Land of an
and the robot was able to climb a slope with an incline of 8°. Amphibious Spherical Mother Robot in Different Terrains”,
In addition, underwater thrust and velocity experiments in Proceedings of 2013 IEEE International Conference on Mechatronics
the semi-submerged state were conducted to evaluate the and Automation, pp. 1173-1178, 2013.
[14] S. Guo, M. Li, C. Yue, "Underwater Performance Evaluation of an
underwater performance. Under a duty of 100%, we got a
Amphibious Spherical Mother Robot ", Proceeding of 2013 IEEE
maximum thrust of 180 mN in horizontal direction, and a International Conference on Information and Automation, pp.1038-
maximum upward thrust of 333.2 mN and a downward thrust 1043, Yinchuan, China, August 26-28, 2013.
of 362.6 mN in vertical direction. Furthermore, a maximum [15] J. Kim, W. Chung, “Accurate and Practical Thruster Modeling for
Underwater Vehicles”, Ocean Engineering, Vol. 33, pp. 566-585, 2006.
[16] C. P. Santos, V. Matos, “Gait transition and modulation in a quadruped
robot: A brainstem-like modulation approach”, Robotics and
Autonomous Systems, Vol. 59, pp.620-634, 2011.
[17] P. Brunetto, L. Fortuna, S. Graziani, S. Strazzeri, “A model of ionic
polymer–metal composite actuators in underwater operations”, Journal
of Smart Material and Structures, Vol. 17, No. 2, pp. 025-029, 2008.
[18] Toshiba 4-legged nuclear plant inspection robot (2012). [Online].
Available: http://www.exponentialtimes.net/videos/toshiba-4-legged-
nuclear-plant-inspection-robot/
[19] U. A. Korde, “Study of a jet-propulsion method for an underwater
vehicle”, Ocean Engineering, Vol.31, No.10, pp.1205-1218, 2004.
[20] C. P. Santos, V. Matos, “CPG modulation for navigation and
omnidirectional quadruped locomotion”, Robotics and Autonomous
Systems, Vol. 60, No. 6, pp. 912-927, 2012.
[21] S. Yu, S. Ma, B. Li, Y. Wang, “An amphibious snake-like robot with
terrestrial and aquatic gaits”, 2011 IEEE International Conference on
Robotics and Automation (ICRA), pp. 2960 - 2961, 2011.
[22] C. Zhou, K. Low, “Better Endurance and Load Capacity: An Improved
Design of Manta Ray Robot (RoMan-II)”, Journal of Bionic
Engineering, Vol.7, Supplement, pp. 137-144, 2010.
[23] D. R. Yoerger, J. GSlotine Cooke, J.E. J. “The influence of thruster
dynamics on underwater vehicle behavior and their incorporation into
control system design”, IEEE Journal of Ocean Engineering, Vol.15,
No.3, pp. 167-178, 2009.
[24] W. Zhou, W. Li, “Micro ICPF actuators for aqueous sensing and
manipulation”, Sensors and Actuators A: Physical, Vol. 114, No. 2-3,
pp. 406-412, 2004.
[25] S. Kim, I. Lee and Y. Kim, “Performance enhancement of IPMC
actuator by plasma surface treatment,” Journal of Smart Material and
Structures, Vol. 16, pp.N6-N11, 2007.
[26] D. Lee, G. Kim, D. Kim, H. Myung, H. Choi, “Vision-based object
detection and tracking for autonomous navigation of underwater
robots.”, Ocean Engineering, Vol. 48, pp. 59–68, 2012.
[27] Y. Li, S. Staicu, "Inverse dynamics of a 3-PRC parallel kinematic
machine", Nonlinear Dynamics, Vol. 67, No. 2, pp. 1031-1041, 2012.
Biography of each author

Maoxun Li received the B.S. degree in electrical engineering and

automation from Harbin Engineering University, China, in 2011, and
the M.S. degree in intelligent machine system from Kagawa
University, Japan, in 2014, respectively. Currently, she is a Ph.D.
candidate in intelligent machine at Kagawa University, Japan. She
researches on underwater microrobot powered by smart actuators,
such as IPMC (ionic polymer metal composite) actuators, and
amphibious spherical robot for industry and biomedical applications.
Her current research interests include amphibious spherical robot and
father-son robot system. She has published about 15 refereed journal and conference papers in the
recent years.

Shuxiang Guo (S’93-M’95-SM’03) received the B.S. and the M.S.

degrees in mechanical engineering from the Changchun Institute
of Optics and Fine Mechanics, Changchun, China, in 1983 and in
1986, respectively, and the Ph.D. degree in mechano-informatics
and systems from Nagoya University, Nagoya, Japan, in 1995.
In1995, he was a Faculty Member at Mie University, Mie, Japan
and in 1998 at Kagawa University, Kagawa, Japan. Currently, he
is a Professor with the Department of Intelligent Mechanical
System Engineering at Kagawa University. He has published
about 270 refereed journal and conference papers. His current research interests include micro
robotics and mechatronics, micro robotics system for minimal invasive surgery, micro catheter
system, micro pump, and smart material (SMA, ICPF) based on actuators.
Dr. Guo received research awards from the Tokai Section of the Japan Society of Mechanical
Engineers (JSME), the Tokai Science and Technology Foundation, and the Best Paper Award of
the IS International Conference, Best Paper award of the 2003 International Conference on
Control Science and Technology and Best Conference Paper Award of IEEE ROBIO2004, in
1997, in 1998, in 2000, in 2003 and in 2004, respectively.
 Hideyuki Hirata received the B.S. and the M.S. degrees in
mechanical engineering from Yokohama National University, Japan,
in 1981 and in 1983, respectively, and the Ph.D. degree from Tokyo
Institute of Technology, Japan, in 1992. Currently, he is a Professor
with the Department of Intelligent Mechanical System Engineering
at Kagawa University. He has published about 32 refereed journal
and conference papers. His current research interests include
material strength, material design and micro devices based on
computer simulation technology.

Hidenori Ishihara received the B.S., the M.S. and Ph.D. degrees in
electrical machinery from Nagoya University, Nagoya, Japan, in 1991,
in 1993 and in 1995, respectively. Currently, he is an Associate
Professor with the Department of Intelligent Mechanical System
Engineering at Kagawa University. He has published about 81
refereed journal and conference papers. His current research interests
include intellectualization and functionalization of robotics.
Research Highlights

 We describe an amphibious spherical robot, which is capable of motion on land, as well as

underwater.
 The amphibious robot has two actuation systems: a quadruped walking actuation system and a
water-jet actuation system.
 The amphibious robot can move with a relatively high velocity and for a relatively long period
of time on land and underwater.
 The amphibious robot can carry the micro-robot, which is used as a manipulator underwater.