IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO.
1, JANUARY 2019
A Study on Development of a Hybrid
Aerial/Terrestrial Robot System for
Avoiding Ground Obstacles by Flight
Chinthaka Premachandra, Member, IEEE, Masahiro Otsuka, Ryo Gohara, Takao Ninomiya, and
Kiyotaka Kato Member, IEEE
Abstract—To date, many studies related to robots have been
performed around the world. Many of these studies have assumed
operation at locations where entry is difficult, such as disaster
sites, and have focused on various terrestrial robots, such as
snake-like, humanoid, spider-type, and wheeled units. Another
area of active research in recent years has been aerial robots with
small helicopters for operation indoors and outdoors. However,
less research has been performed on robots that operate both
on the ground and in the air. Accordingly, in this paper, we
propose a hybrid aerial/terrestrial robot system. The proposed
robot system was developed by equipping a quadcopter with
a mechanism for ground movement. It does not use power
dedicated to ground movement, and instead uses the flight
mechanism of the quadcopter to achieve ground movement as
well. Furthermore, we addressed the issue of obstacle avoidance
as part of studies on autonomous control. Thus, we found that
autonomous control of ground movement and flight was possible
for the hybrid aerial/terrestrial robot system, as was autonomous
obstacle avoidance by flight when an obstacle appeared during
ground movement.
Index Terms—Ground movement/flight control, hybrid aeri-
al/terrestrial robot, obstacle avoidance by flight, obstacle recog-
nition.
I. I NTRODUCTION
H ERETOFORE, much research has been conducted
around the world on ground mobile robots, such as those
intended for use at disaster sites where entry of a human is
impossible. Examples include robots that can traverse rough
terrain by using multiple legs in the manner of a spider
or centipede [1]−[3], robots that can move under various
conditions in the manner of a snake [4]−[6], robots that
Manuscript received December 4, 2017; revised March 22, 2018; accepted
May 17, 2018. Recommended by Associate Editor Xiaoming Hu. (Corre-
sponding author: Chinthaka Premachandra.)
Citation: C. Premachandra, M. Otsuka, R. Gohara, T. Ninomiya, and K.
Kato, “A study on development of a hybrid aerial/terrestrial robot system for
avoiding ground obstacles by flight,” IEEE/CAA J. Autom. Sinica, vol. 6, no.
1, pp. 327−336, Jan. 2019.
C. Premachandra is with the Department of Electronic Engineering, Gradu-
ate School of Engineering, Shibaura Institute of Technology, 135-8548, Tokyo,
Japan (e-mail: chintaka@sic.shibaura-it.ac.jp).
M. Otsuka, R. Gohara, and T. Ninomiya are with the Department of
Electrical Engineering, Graduate School of Engineering, Tokyo University
of Science, Tokyo 135-8548, Japan (e-mail: watashidesu1209@gmail.com;
j4311045@ed.tus.ac.jp; reqza.t.n.a.s@gmail.com).
K. Kato is with the Department of Electrical Engineering, Graduate School
of Engineering, Tokyo University of Science, Tokyo 135-8548, Japan (e-mail:
kato@ee.kag.tus.ac.jp).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JAS.2018.7511258
Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
327
perform investigations in indoor spaces contaminated with
chemicals where entry by humans is impossible [7], [8], and
robots that attach to a wall and travel thereupon [9]−[11].
Research on bipedal robots [12]−[14] is also an area of active
investigation. Cameras, GPS systems, ultrasonic sensors, and
the like are being used for behavioral environment recognition
and self-position estimation for such robots [15]−[18].
In recent years, various lines of research have been pur-
sued in regard to aerial robots, such as those using small
helicopters or multirotors. Applications of small aerial robots
are anticipated for indoor and outdoor information gathering,
conveyance, rescue tasks, and other purposes. Like ground
mobile robots, aerial robots must recognize their movement
environment and estimate self-position by means such as cam-
eras, GPS systems, and ultrasonic sensors. For aerial robots,
the negative effects of additional weight can be substantial.
Furthermore, mounting areas are smaller, so devices to be
installed on the robot must be smaller and lighter compared to
those installed on ground mobile robots. Various applications
of aerial robots are being investigated, including teleoperation
[19], [20], movement to a defined destination [21], obstacle
avoidance [22], [23], route planning [24], [25], intrusion detec-
tion [26] etc.. By using some of these capabilities, technology
is now being established for navigation and mapping by an
aerial robot equipped with an ultrasonic device or an infrared
laser device [27]−[29].
The advantages of ground-mobile robots include the ability
to perform work in a stable state, ease of work in tight
locations, and easy operation. The advantages of aerial robots
include the ability to move quickly without being affected
by the terrain. Conversely, disadvantages of ground mobile
robots include the difficulty of arriving quickly at a site when
there are locations with poor footing as a result of rubble or
cracks in the ground. Disadvantages of aerial robots include
the difficulty of performing inspections under the ground or
within rubble, the difficulty of working in close contact with
an object, and susceptibility to wind. Therefore, we propose a
hybrid aerial/terrestrial robot system for operation both on the
ground and in the air. A concept image of this system is shown
in Fig. 1. The robot combines an aerial robot and a ground
mobile robot, enabling compensation for the aforementioned
disadvantages while combining their advantages. Applications
are anticipated in a wider range than has been possible for
conventional robots, for example, in large-area soil investi-
gation, searching through rubble during a disaster, large-area
328 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019

cleaning, and selection of a movement method according to during autonomous ground movement and flight. The useful-
wind speed. ness of this means of obstacle avoidance was confirmed.
The remainder of this paper is organized as follows. In
Section II, the robot system that we developed is described. In
Section III, main operation of the robot is described. Section
IV present automatic avoidance by flight of obstacles appear-
ing during ground movement. Then, in Section V, experiments
are described. Section VI gives our conclusions.

II. D EVELOPMENT OF THE H YBRID ROBOT

A. Configuration of the Chassis
An image of the developed hybrid aerial/terrestrial robot
system is shown in Fig. 2. The robot system is equipped
with a mechanism that enables movement of the quadcopter
on the ground. Aerial robots have strict limitations on the
weight of equipment that can be mounted, and therefore,
mounted hardware must be small and lightweight, and should
be miniaturized to the fullest extent possible. As shown in
Fig. 3, the robot system is equipped with omni wheels, in
which barrel-shaped small wheels are placed on the perimeter
of a large wheel, thus allowing 360o movement. The positional
relationship between the axle and the ground plane is uniform,
thus facilitating control. The dampers for the shafts of the
Fig. 1. Concept of the hybrid aerial/terrestrial robot system. chassis are shown in Fig. 4. By using dampers, shocks to the
chassis are suppressed.
In the development of our robot system, the weight of the
chassis and expandability were considered, and a method was
adopted in which the power of a quadcopter was used for
ground movement. For this hybrid aerial/terrestrial robot, the
quadcopter was equipped with a mechanism enabling ground
movement, and designed for durability by use of dampers
and aluminum. Furthermore, as part of the control of ground
movement and flight for the robot, studies were performed
on automatic obstacle avoidance by flight for obstacles that
appeared during ground movement. Robots capable of both Fig. 2. Hybrid aerial/terrestrial robot system.
ground movement and flight have been developed previously
[30], [31]; however, these robots required manual maneuver-
ing, and almost no research has been done on robots capable
of both ground movement and flight with a complete on-
board embedded control system. In the obstacle avoidance
presented here, obstacles that appear during ground movement
are avoided by flight. Very few works have been previously
performed a basic study of hybrid aerial/terrestrial robot
systems [32], [33]. In these, movement and flight have been
controlled entirely by external hardware for environmental
Fig. 3. Omni wheel.
recognition and control. However, when the hardware is placed
externally, operation is restricted to the installation site and
versatility is impaired. Accordingly, in this study, the hardware
for autonomous control of ground movement and flight is
mounted on the chassis. Furthermore, the chassis configuration
of previous machines was examined, and this robot was retrofit
with shock-resistant material, wheels with higher mobility, and
other improvements. In obstacle avoidance, the presence of
an obstacle was recognized on the basis of an image from a
small camera, and if an obstacle was present, it was avoided by
flight. For the robot, experiments were performed on avoidance
by flight of obstacles that appear during movement, including Fig. 4. Damper.

Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
PREMACHANDRA et al.: A STUDY ON DEVELOPMENT OF A HYBRID AERIAL/TERRESTRIAL ROBOT SYSTEM FOR · · ·
To maintain axle stability, the framework of aluminum
shown in Fig. 5 was sandwiched between the shafts and the
wheels. The framework allows mounting of hardware and a
battery, and is provided with fins to promote lift generation.
Furthermore, plastic screws were used that are designed to
break in the event of strong shock to the chassis, thus reducing
its effects on the robot.
Fig. 5. Framework for stability maintaining.
Here, recognition of the behavioral environment of the robot
system was performed using environmental recognition by
image processing. Therefore, the front part of the chassis
was equipped with a small camera and IMAPCAR2 (inte-
grated memory array processor for CAR2) as a small image-
processing unit. A microcomputer used for determining control
quantities and the like was also mounted on the front part of
the chassis. Details of these hardware units mounted on the
front part of the chassis are given in Section IV.
1) Selection of the Rotor Blades and Motor
When mounting the flight mechanism, it is necessary to se-
lect a mechanism appropriate for the weight of the chassis. The
flight mechanism consists of rotor blades and a motor, which
were selected by calculating static thrust (i.e., the weight that
can be lifted at liftoff). For the chassis configuration used here,
if the static thrust exceeds the weight of the chassis, flight is
possible. The mass corresponding to the static thrust is given
by the following formula.
T = (D/10)3 × (P/10) × (N/1000)2 × 22.
This formula is based on the static thrust defined by some
studies in the literature [34]. Here, T represents the static
thrust in grams (g), D represents the diameter of the rotor
blades in inches (in), P represents the pitch of the rotor
blades in inches (in), and N represents the rotation speed
of the motor in revolutions per minute (rpm). In this study,
this formula and other factors were considered in selecting
the rotor blades and motor. Rotor blades with a diameter of 8
in. and a pitch of 4.5 in were used. A motor of 1000 kv was
used, where kv represents the speed of the motor (rpm) per 1 V.
Assuming voltage applied to the motor to be 12 V, the mass
corresponding to the static thrust generated by the selected
flight mechanism is 730 g. The chassis is equipped with four
rotors, giving a total mass of 2920 g for the generated static
Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
329
thrust. The rotor blades, motor, chassis, and battery together
total 1750 g in weight, making it is possible to achieve liftoff
with an output of approximately 60 %; a value sufficient for
the purpose of the present research.
B. Flight
Flight of the hybrid aerial/terrestrial robot system was
performed in a manner similar to a quad copter flight, using
throttle, elevator, aileron, and rudder operations. An overview
of these four actions is shown in Table I and Fig. 6.
TABLE I
D ESCRIPTION OF ACTIONS FOR F LIGHT
Name Action
Throttle Up and down
Elevator Forward and backward
Aileron Left and right
Rudder Rotation
Fig. 6. Details of flight actions. (a) Throttle. (b) Elevator. (c) Aileron. (d)
Rudder.
Throttle in Fig. 6 (a) is an operation that uniformly increases
(1)
or decreases the output of all four motors. If the magnitude
of the throttle does not exceed a certain value, then flight
is impossible; therefore, when flying, throttle action is always
implemented. Elevator in Fig. 6 (b) and aileron in Fig. 6 (c) are
operations that cause the output of two motors on one side of
the robot to be greater than the output of the other motors, thus
inclining the chassis and causing it to move to the front, back,
left, or right. Rudder control increases the output of one of
two pairs of motors, where the motors are located diagonally
to each, thus rotating the chassis to the left or right by the
generated torque.
C. Ground Movement
A schematic diagram of ground movement is shown in
Fig. 7. Similar to flight mechanism, ground movement is
330 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019

possible using the motors and the rotor blades for power. In
multirotor flight, aileron and elevator operations incline the
chassis, and by acquiring thrust in a lateral direction, the
chassis moves. Furthermore, the rudder operation performs
rotation by using the torque produced by the motors. In
ground movement, the chassis is made to move forward and
backward by elevator operations, to move left and right by
aileron operations, and to rotate to the left and right using
rudder operations. When moving to the front, back, left, or
right, as shown in Fig. 7, it is necessary for lateral thrust to
be greater than downward thrust. Therefore, the aileron or
elevator operation is set so that the chassis inclines at a 45o
angle. The throttle operation is set so that output is at such
a level that the chassis lifts off the ground, so that sufficient
lateral thrust is acquired. In this manner, by integrating the
power for flight and ground movement, weight is reduced and
expandability is increased.

Fig. 8. System configuration for manual operation.

Fig. 7. Ground movement. A. Manual Operation

III. O PERATION OF THE ROBOT
For this robot system, autonomous movement is presumed;
however, manual operation is also accommodated. In con-
sideration of managing the chassis, was introduced in order
to perform start up, shut down, and other such operations
safely. Furthermore, additional management of safety aspects
was also considered, such as autonomous movement under
conditions where a person may be subject to harm or pre-
vention of autonomous movement in a location that is not
an anticipated site. Fig. 8 shows the system configuration for
manual operation of the chassis. The system consists of a
transmitter for a radio-controlled toy, a receiver, a microcom-
puter, and a control board. The transmitter is a controller
used for general multirotors and the like, and the receiver
is a module for receiving signals from the transmitter. The
receiver is connected in series to the microcomputer. The
microcomputer generates and receives various signals, and
communicates with the control board, the receiver, and the
hardware for environmental recognition. The control board
transmits to the motors signals corresponding to those acquired
from the microcomputer and controls the chassis. A detailed
description of the microcomputer and control board is given in
Section IV. General manual operation and switching between
autonomous operation and manual operation are performed by
operation of the transmitter.
Manual operation of the chassis has the same specification
as operation of a general quadcopter. Sticks and switches on
the transmitter are operated, and by adjusting the amount of
motion as described in Section II, the chassis is operated. In
order to accommodate autonomous operation, the robot system
transmits signals from the receiver via the microcomputer
to the control board as shown in Fig. 8; however, a general
quadcopter directly transmits signals from the receiver to the
control board. In order to make the communication method
similar to that of a general quadcopter, the robot has been
configured so that signals similar to the those from the receiver
are transmitted from the microcomputer to the control board
during manual operation.
B. Switching Between Autonomous Operation and Manual
Operation
In autonomous operation, the transmitter is not operated and
the robot system is operated using only commands from the
microcomputer corresponding to the results of environmental
recognition. Switching between autonomous operation and
manual operation is performed by operation of the transmitter.
Communication from the receiver to the microcomputer then
uses pulse position modulation (PPM) signals. Table II shows
channel correspondence for PPM signals, and Fig. 9 shows an
overview of the PPM signals. As the figure shows, the PPM
signals consist of eight channels and a dummy pulse, for which

Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
PREMACHANDRA et al.: A STUDY ON DEVELOPMENT OF A HYBRID AERIAL/TERRESTRIAL ROBOT SYSTEM FOR · · · 331

Fig. 9. PPM signals.

TABLE II
autonomous operation, the transmitter is not used for oper-
C HANNEL A SSIGNMENTS OF PPM S IGNALS
ation; instead, operation of the chassis is performed using
Channel Action only commands from the microcomputer using the results of
CH1 Already used environmental recognition. Among the hardware constituting
CH2 Already used the control command unit in Fig. 10, the R8C/25 microcom-
CH3 Already used puter and the control board in Fig. 8 are the same. The ground
CH4 Already used movement/flight control system used here is constituted by an
CH5 Switch to auto or manual obstacle recognition unit that detects obstacles and a control
CH6 Unused command unit that executes command and control of the
CH7 Unused chassis.
CH8 Unused
Dummy pulse Pulse for periodic tuning

control can be performed by varying each channel width.

Here, switching between manual operation and autonomous
operation is performed by varying the pulse width of channel
5; channels 1 to 4 and channels 6 to 8 are not used. Channels
1 to 4 are not used in operation of a general quadcopter, and
therefore, the channels cannot be used for switching between
manual operation and autonomous operation. The pulse width
of the PPM signals is varied by operation of the transmitter.
Furthermore, the PPM signals are transmitted in cycles of 22
ms, and maintain their period by means of a dummy pulse.

IV. O BSTACLE AVOIDANCE BY F LIGHT

The proposed hybrid aerial/terrestrial robot system is as-
sumed to work in areas such as disaster sites, where humans
cannot easily enter. In such locations, there are environments
outside the effective range of radio. Performing remote opera-
tion in such environments is extremely difficult, and therefore,
when robots are made to operate in these environments,
autonomous control is essential. Here, as a basic study on the
control of ground movement and flight, avoidance of obstacles
appearing during ground movement by flight was investigated.
When an obstacle appears in front of the chassis during
movement, the object is automatically recognized by image
Fig. 10. System configuration for autonomous operation.
processing and is automatically avoided by the robot taking
flight. This obstacle avoidance is assumed as an avoidance
B. Obstacle Recognition
method when the obstacle appears during work operations
during ground movement, or when avoidance by ground Obstacle recognition is performed using a small monocular
movement is impossible or takes a significant amount of time. camera and the IMAPCAR2 image-processing unit. Fig. 11
shows the external appearance of the IMAPCAR2 and the
monocular camera, which are connected together with pins
A. Configuration for Obstacle Avoidance and mounted on the front part of the chassis. Furthermore, the
The system configuration for obstacle avoidance is shown IMAPCAR2 is connected in series to the microcomputer of
in Fig. 10. The transmitter and receiver shown in Fig. 8 are the control command unit. Characteristics of the IMAPCAR2
not used; instead, an obstacle recognition unit is used. In include a high-performance parallel processor with a high

Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
332
image processing capability, while retaining a small size and
being lightweight. Image processing performed by software,
and its power consumption does not exceed 2W. From these
characteristics, we determined that it is a suitable processor for
mounting hardware onto the chassis and performing obstacle
avoidance by image processing. Furthermore, for IMAPCAR2,
by developing an algorithm in a form suitable for parallel pro-
cessing, it is possible to perform high-speed image processing
[35]−[39].
A schematic diagram of obstacle recognition is shown in
Fig. 12. For recognition of an obstacle, background subtraction
of the previous frame from the current frame is used. Attention
is set on a pixel (x, y) in an image; treating pixels in the
current frame as I(x, y), and pixels in the previous frame as
Ib (x, y), the difference Id (x, y) for the pixel (x, y) is express
as follows:
Id (x, y) = |I(x, y) − Ib (x, y)|.
Fig. 11. Camera and IMAPCAR2.
Fig. 12. Image processing for object recognition.
The difference Id (x, y) is binarized, as in (7), on the basis
of T , which is a threshold for binarization:
½ [37].
255, if |Id (x, y) > T
Ir (x, y) = (3)
0, otherwise.
The result of image summing for all the Ir (x, y) pixels is
the processing result. A threshold is set for determining the
presence or absence of an obstacle. If the processing result
exceeds the threshold, an obstacle is judged to exist; otherwise,
Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019
no obstacle is judged to exist. If the value of the difference
exceeds the threshold, an obstacle is recognized to exist in
the direction of travel, and obstacle avoidance by flight is
performed. Otherwise, ground movement continues.
C. Control Command Unit
The command control unit communicates with the re-
ceiver when manual operation is performed, and with the
IMAPCAR2 when autonomous operation is performed. During
manual operation, it is possible to switch between manual
operation and autonomous operation on the basis of a signal
from the receiver. During autonomous operation, according to
image processing results from the IMAPCAR2, the microcom-
puter generates and transmits signals, controlling the chassis.
The control command unit is constituted by the control board
and the R8C/25, which is the microcomputer. The external
(2) appearance of the R8C/25 is shown in Fig. 13. The R8C/25 is
small and has performance capable of sufficiently processing
the control required in this research. Some key features of
R8C/25 includes; 8-bit multifunction timer with 8-bit pre
scaler (Timer RA and RB): 2 channels, input capture/output
compare timer (Timer RC): 16-bit × 1 channel, real-time clock
timer with compare match function (Timer RE): 1 channel.
The control board transmits signals corresponding to those
acquired from the microcomputer to the motors to perform
control of the chassis.
Fig. 13. R8C/25.
1) Communication From IMAPCAR2 to the Microcomputer
The IMAPCAR2 and the microcomputer have different
communication bit rates, and so one unit of information must
be separated across a number of sessions and communicated.
Therefore, for communication from the IMAPCAR2 to the
microcomputer, 4-line bus communication is employed. Here,
two hardware units are separated into a master side and a
slave side, and communication is performed using four lines.
In performing communication, the microcomputer was treated
as the master side, and the IMAPCAR2 as the slave side [34],
2) Transmission of Commands From the Microcomputer to
the Control Board
Transmission of a command from the microcomputer to the
control board is performed in the cases of both manual and
autonomous operation. During manual operation, a command
signal similar to a command from the receiver is generated
PREMACHANDRA et al.: A STUDY ON DEVELOPMENT OF A HYBRID AERIAL/TERRESTRIAL ROBOT SYSTEM FOR · · · 333

by the microcomputer, and transmitted to the control board.
During autonomous operation, on the basis of information
from the IMAPCAR2, a signal corresponding to the presence
or absence of an obstacle is generated by the microcomputer,
and transmitted to the control board. For signals sent from
the microcomputer to the control board, four pulse width
modulation (PWM) signals are employed. Examples of PWM
signals are shown in Fig. 14.
operation of the transmitter. During autonomous operation, the
robot determines the presence or absence of an obstacle; if
no obstacle is detected, then the robot automatically performs
movement on the ground, and if and obstacle is detected, it
automatically avoids the obstacle by flight. In this paper, the
obstacles approximately appear in between 30 cm and 2 m,
away from the robot and can be avoided by flight. Only the
obstacles within the field of view of the camera are detected.
The field of view is 120 degrees. In addition, the camera frame
rate is: 30 fps, and the image size: VGA.

Fig. 14. Examples of PWM signals. (a) Minimum-width (1.02 ms) PWM
signal. (b) Maximum-width (2.02 ms) PWM signal.

Each of the four PWM signals corresponds to one of the

four actions in Table I, and by modifying the pulse width
between 1.02 ms to 2.02 ms, the parameter of each is modified
so that if there is no obstacle, control for ground movement is
performed, and if there is an obstacle, obstacle avoidance by
flight is performed. Settings for the parameters are illustrated
in (4)−(7):
Ai = 200(W − 1.02) − 100 (4)
El = 200(W − 1.02) − 100 (5)
T h = 100(W − 0.02) − 100 (6)
Ru = 200(W − 1.02) − 100 (7)
where, W is the pulse width in milliseconds, Ai is the
parameter for the aileron, El is the parameter for the elevator,
and T h is the parameter for the throttle. The pulse width varies
between 1.02 and 2.02 ms, and the pulse period is 22 ms. For
the aileron, if the value of the parameter is positive, movement
is to the right, and if it is negative, movement is to the left. As
the absolute value of the parameter for the aileron is increased,
the inclination of the chassis increases. For the elevator, if the Fig. 15. Flowchart for control system.
parameter is positive, the chassis moves backward, and if it
is negative, it moves forward. As the absolute value of the
parameter for the elevator is increased, the inclination of the
chassis increases. The throttle parameter cannot take a negative
value, and as its value increases, the output increases. For the
rudder, if the value of the parameter is negative, the chassis
rotates to the right, and if it is positive, the chassis rotates to
the left. As the absolute value of the parameter for the rudder
is increased, the speed of rotation increases. The pulse width Fig. 16. Experimental scenario.
in this study was set on the basis of experience during the
experiments.
The experiments were performed for a scenario like the
V. E XPERIMENTS one shown in Fig. 16. In the course of ground movement
by the hybrid aerial/terrestrial robot system, an obstacle was
A. Methods manually caused to appear in front of the chassis to test
Fig. 15 shows the flow of an experiment on the obsta- whether the chassis could avoid the obstacle by flight and land.
cle avoidance system. The system, from the perspective of The obstacle was approximately the same height as the robot,
safety, allows for switching between manual operation and assuming a case in which avoidance by ground movement was
autonomous operation, and this switching is performed by impossible. Parameter settings for movement on the ground

Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
334 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019

are listed in Table III, and parameter settings for obstacle flight. We plan to work on this part in the future. In addition,
avoidance by flight are listed in Table IV. The values in the as another future direction, we plan to widen the obstacle
tables were determined by (4) to (7), representing the values detection range of the robot by installing an omnidirectional
of parameters subsequent to starting each of the actions of camera on the robot.
ground movement and obstacle avoidance by flight. These
values cannot be set without taking into consideration detailed
data on the shape, weight, traveling performance, and the like
of the chassis. Therefore, the values were set on the basis of
experience acquired from the experiments.
TABLE III
PARAMETERS FOR G ROUND M OVEMENT

Start to 0.5 s 0.5−0.55 s

Aileron −1 −1
Elevator −18 0
Throttle 62 38
Rudder −1 −1

TABLE IV
PARAMETERS FOR O BSTACLE AVOIDANCE BY F LIGHT

Start to 0.5 s 0.8−1.3 s 1.3−2.3 s

Aileron −6 −6 −6
Elevator 0 −4 0
Throttle 74 74 50
Rudder 0 0 0

The obstacle was a bar that was manually placed in front Fig. 17. Obstacle avoidance by flight. (a)−(c) Ground movement. (d)−(f)
of the chassis. Furthermore, ordinances and safety issues for Obstacle avoidance by flight. (g) Landing.
the area were taken into consideration in performing the
experiments in a gymnasium, which is an indoor environment.

B. Avoidance of Ground Obstacles by Flight

Figs. 17 and 18 show images in which the robot avoids an
obstacle by flight. In Fig. 17, ground movement was performed
in the order of (a) to (c), obstacle avoidance by flight was
performed in the order of (d) to (f), and landing was performed
in (g). In Fig. 18, ground movement was performed in the order
of (a) to (b), obstacle avoidance by flight was performed in the
order of (d) to (f), and landing was performed in (g). As these
figures show, the robot avoided the obstacle by flight when it
appeared (please see the submitted video).

VI. C ONCLUSION
A hybrid aerial/terrestrial robot system was introduced and
a basic autonomous control system is installed on it. The
results of this study show that it was possible to manufacture a
robot system that performs both ground movement and flight.
Furthermore, by mounting a basic autonomous control system,
it was possible to control the chassis by autonomous control as
to create a more practical robot. A topic for future investigation
is use of a more advanced image-processing algorithm for
measuring the distance to an obstacle, its size, and its shape
after mounting an improved system for obstacle avoidance.
In this study, we have not proposed a mechanism to search Fig. 18. Obstacle avoidance by flight (alternate view). (a), (b) Ground
the ground condition before it lands on the ground after a movement. (d)−(f) Obstacle avoidance by flight. (g) Landing.

Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
PREMACHANDRA et al.: A STUDY ON DEVELOPMENT OF A HYBRID AERIAL/TERRESTRIAL ROBOT SYSTEM FOR · · ·
R EFERENCES
[1] G. Takeo, T. Takubo, K. Ohara, Y. Mae, and T. Arai, “Rotational
operation of polygonal prism by multi-legged robot,” in Proc. IEEE
Int. Conf. Mechatronics and Automation, Changchun, China, 2009, pp.
2677−2682.
[2] G. Takeo, T. Takubo, K. Ohara, Y. Mae, and T. Arai, “ Rotation control
of polygonal prism by multi-legged robot,” in Proc. 11th IEEE Int.
Workshop on Advanced Motion Control, Nagaoka, Niigata, Japan, 2010,
pp. 601−606.
[3] S. Inagaki, T. Niwa, and T. Suzuki, “Follow-the-contact-point gait
control of centipede-like multi-legged robot to navigate and walk on
uneven terrain,” in Proc. IEEE/RSJ Int. Conf. Intelligent Robots and
Systems, Taiwan, China, 2010, pp. 5341−5346.
[4] Z. Yang, K. Ito, K. Hirotsune, K. Saijo, and A. Gofuku, “A mechanical
intelligence in assisting the navigation by a force feedback steering
wheel for a snake rescue robot,” in Proc. 13th IEEE Int. Workshop
on Robot and Human Interactive Communication, Kurashiki, Okayama,
Japan, 2004, pp. 113−118.
[5] J. Y. Gao, X. S. Gao, W. Zhu, J. G. Zhu, and B. Y. Wei, “Design
and research of a new structure rescue snake robot with all body
drive system,” in Proc. IEEE Int. Conf. Mechatronics and Automation,
Takamatsu, Japan, 2008, pp. 119−124.
[6] P. Polchankajorn and T. Maneewarn, “Helical controller for modular
snake robot with non-holonomic constraint,” in Proc. 8th Electrical En-
gineering/Electronics, Computer, Telecommunications and Information
Technology (ECTI-CON), Khon Kaen, Thailand, 2011, pp. 589−592.
[7] A. Gargade, D. Tambuskar, and G. Thokal, “Modelling and Analysis of
Pipe Inspection Robot,” Int. J. Emerg. Technol. Adv. Eng., vol. 3, no. 5,
pp. 120−126, May 2013.
[8] G. J. Bruemmer, C. W. Nielsen, and D. I. Gertman, “How training and
experience affect the benefits of autonomy in a dirty-bomb experiment,”
in Proc. 3rd ACM/IEEE Int. Conf. Human-Robot Interaction (HRI),
Amsterdam, Netherlands, 2008, pp. 161−168.
[9] P. Liljeback, I. U. Haugstuen, K. Y. Pettersen, “Path Following Control
of Planar Snake Robots Using a Cascaded Approach,” IEEE Transac-
tions on Control Systems Technology, vol. 20, no. 1, pp. 111−126, Feb.
2012.
[10] S. Q. Wu, L. J. Wu, and T. Liu, “Design of a sliding wall climbing
robot with a novel negative adsorption device,” in Proc. 8th Int. Conf.
Ubiquitous Robots and Ambient Intelligence, Incheon, South Korea,
2011, pp. 97−100.
[11] A. Sekhar R, A. Mary, S. N. Raju, A. G. Ravi, V. Sharma, and G. Bala
“A novel design technique to develop a low cost and highly stable wall
climbing robot,” in Proc. 4th Int. Conf. Intelligent Systems, Modelling
and Simulation, Bangkok, Thailand, 2013, pp. 360−363.
[12] T. Takuma, K. Hosoda, and M. Asada, “Walking stabilization of biped
with pneumatic actuators against terrain changes,” in Proc. IEEE/RSJ Int.
Conf. Intelligent Robots and Systems, Edmonton, Alta, Canada, 2005,
pp. 4095−4100.
[13] T. Sugihara, “Standing stabilizability and stepping maneuver in planar
bipedalism based on the best COM-ZMP regulator,” in Proc. IEEE Int.
Conf. Robotics and Automation, Kobe, Japan, 2009, pp. 1966−1971.
[14] J. Z. Zhao and H. H. Ju, “Detail analysis on human bipedalism and a
natural gait generation method for biped robots,” in Proc. IEEE Int. Conf.
Mechatronics and Automation, Takamatsu, Japan, 2013, pp. 873−878.
[15] Y. S. Suh, S. K. Park, D. N. Kim, and K. H. Jo, “Remote control of a
moving robot using the virtual link,” in Proc. IEEE Int. Conf. Robotics
and Automation, Roma, Italy, 2007, pp. 2343−2348.
[16] C. Lee, J. Park, W. Bahn, T. Kim, T. Lee, M. M. Shaikh, K. Kim, and
D. Cho, “Vision tracking of a moving robot from a second moving robot
using both relative and absolute position referencing methods,” in Proc.
37th Ann. Conf. IEEE Industrial Electronics Society, Melbourne, VIC,
Australia, 2011, pp. 325−330.
[17] Q. Y. Li and P. L. Yang, “Keep up with me: A Gesture guided moving
robot with microsoft kinect,” in Proc. 10th IEEE Int. Conf. Mobile Ad-
Hoc and Sensor Systems, Hangzhou, China, 2013, pp. 435−436.
[18] C. R. Torres, G. L. Torres, L. E. B. da Silva, and J. M. Abe, “Intelligent
system of paraconsistent logic to control autonomous moving robots,” in
Proc. 32nd Ann. Conf. IEEE Industrial Electronics, Paris, France, 2006,
pp. 4009−4013.
[19] X. L. Hou, R. Mahony, and F. Schill, “Comparative study of haptic
interfaces for bilateral teleoperation of VTOL aerial robots,” IEEE Trans.
Syst. Man Cybernet.: Syst., vol. 46, no. 10 pp. 1352−1363, Oct. 2016.
[20] X. L. Hou and R. Mahony, “Dynamic kinesthetic boundary for haptic
teleoperation of VTOL aerial robots in complex environments,” IEEE
Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
335
Trans. Syst. Man Cybernet.: Syst., vol. 46, no. 5, pp. 694−705, May
2016.
[21] A. Ruangwiset, “Path generation for ground target tracking of airplane-
typed UAV,” in Proc. IEEE In. Conf. Robotics and Biomimetics,
Bangkok, Thailand, 2008, pp. 1354−1358.
[22] J. C. Zufferey and D. Floreano, “Toward 30-gram autonomous indoor
aircraft: Vision-based obstacle avoidance and altitude control,” in Proc.
IEEE Int. Conf. Robotics and Automation, Barcelona, Spain, 2005, pp.
2594−2599.
[23] N. Gageik, P. Benz, and S. Montenegro, “Obstacle detection and
collision avoidance for a UAV with complementary low-cost sensors,”
IEEE Access, vol. 3, pp. 599−609, Jun. 2015.
[24] Y. G. Fu, M. Y. Ding, C. P. Zhou, and H. P. Hu, “Route planning
for unmanned aerial vehicle (UAV) on the sea using hybrid differential
evolution and quantum-behaved particle swarm optimization,” IEEE
Trans. Syst. Man Cybernet.: Syst., vol. 43, no. 6, pp. 1451−1465, Nov.
2013.
[25] K. Dorling, J. Heinrichs, G. G. Messier, and S. Magierowski, “Vehicle
routing problems for drone delivery,” IEEE Trans. Syst. Man Cybernet.:
Syst., vol. 47, no. 1, pp. 70−85, Jan. 2017.
[26] R. Mitchell and I. R. Chen, “Adaptive intrusion detection of malicious
unmanned air vehicles using behavior rule specifications,” IEEE Trans.
Syst. Man Cybernet.: Syst., vol. 44, no. 5, pp. 593−604, May 2014.
[27] K. Cho, J. Shin, M. S. Kang, W. Shon, and S. Park, “Indoor flying robot
control and 3D indoor localization system,” in Proc. 11th WSEAS Int.
Conf. Automatic Control, Modelling and Simulation, Istanbul, Turkey,
2013, pp. 524−528.
[28] C. R. Yuan, F. Recktenwald, and H. Mallot, “Visual steering of UAV in
unknown environments,” in Proc. IEEE/RSJ Int. Conf. Intelligent Robots
and Systems, St. Louis, MO, USA, 2009, pp. 3906−3911.
[29] K. Schmid, T. Tomic, F. Ruess, H. Hirschmiiler, and M. Suppa, “Stereo
vision based indoor/outdoor navigation for flying robots,” in Proc.
IEEE/RSJ Int. Conf. Intelligent Robots and Systems, Tokyo, Japan, 2013,
pp. 3955−3962.
[30] A. Kossett and N. Papanikolopoulos, “A robust miniature robot design
for land/air hybrid locomotion,” in Proc. IEEE Int. Conf. Robotics and
Automation, Shanghai, China, 2011, pp. 4595−4600.
[31] K. Kawasaki, M. J. Zhao, K. Okada, and M. Inada, “MUWA: Multi-field
universal wheel for air-land vehicle with quad variable-pitch propellers,”
in Proc. IEEE/RSJ Int. Conf. Intelligent Robots and Systems, Tokyo,
Japan, 2013, pp. 1880−1885.
[32] M. Ootsuka, C. Premachandra, and K. Kato, “Development of an air-
ground operational robot and its fundamental controlling approach,” in
Proc. Joint 7th Int. Conf. Soft Computing and Intelligent Systems (SCIS)
and 15th Int. Symp. Advanced Intelligent Systems (ISIS) , Kitakyushu,
Japan, 2014, pp. 1470−1474.
[33] C. Premachandra and M. Otsuka, “Development of hybrid aerial/ter-
restrial robot system and its automation,” in Proc. IEEE Int. Systems
Engineering Symp., Vienna, Austria, 2017.
[34] M. T. Chen, “Static thrust measurement for propeller-driven light air-
craft,” in Proc. Int. Conf. Computer Application and System Modeling,
Taiyuan, China, 2012, pp. 650−652.
[35] R. Gohara, C. Premachandra, and K. Kato, “Smooth automatic vehi-
cle stopping control system for unexpected obstacles,” in Proc. 10th
Asia-Pacific Symp. Information and Telecommunication Technologies,
Colombo, Sri Lanka, 2015, pp. 103−105.
[36] Y. Okamoto, C. Premachandra, and K. Kato, “A study on computational
time reduction of road obstacle detection by parallel image processor,”
J. Adv. Comput. Intell. Intell. Informat., vol. 18. no. 5, pp. 849−855,
Aug. 2014.
[37] Y. Okamoto, C. Premachandra, and K. Kato, “Parallel processing based
speed-up road obstacle detection algorithm using discriminant analysis,”
in Proc. 14th Int. Symp. Advanced Intelligent Systems, Daejeon, Korea,
2013.
[38] R. Gohara, C. Premachandra, and K. Kato, “A study on smooth
automatic vehicle stopping control for suddenly-appeared obstacles,”
in Proc. IEEE Int. Conf. Vehicular Electronics and Safety, Yokohama,
Japan, 2015, pp. 86−90.
[39] C. Premachandra, Y. Okamoto, and K. Kato, “High performance em-
bedding environment for reacting suddenly appeared road obstacles,” in
Proc. IEEE Int. Conf. Robotics and Biomimetics, Bali, Indonesia, 2014,
pp. 2394−2397.
336
Chinthaka Premachandra received the B.Sc. and
M.Sc. degrees from Mie University, Tsu, Japan in
2006 and 2008 respectively, and the Ph.D. degree
from the Nagoya University, Nagoya, Japan, in
2011.
From 2012 to 2015, Dr. Premachandra was an
Assistant Professor in the Department of Electrical
Engineering, Faculty of Engineering, Tokyo Univer-
sity of Science, Tokyo, Japan. From 2016 to 2017, he
was an Assistant in the Department of Electronic En-
gineering, School of Engineering, Shibaura Institute
of Technology, Tokyo, Japan. In 2018, he was promoted to Associate Professor
in the Department of Electronic Engineering, Graduate School of Engineering,
Shibaura Institute of Technology. In addition, he is the manager of Image
Processing and Robotic Lab at the same department. His lab conducts research
in two main fields: image processing and robotics. Former research includes
computer vision, pattern recognition, speed up image processing, and camera
based intelligent transportation systems, while latter fields include terrestrial
robotic systems, flying robotic systems, and integration of terrestrial robot and
flying robot.
He is the member of IEEE, IEICE (Japan), SICE (Japan) and SOFT (Japan).
Masahiro Otsuka received the B.Sc. and M.Sc. de-
grees from the Tokyo University of Science, Tokyo,
Japan in 2014 and 2016 respectively. He is currently
working in Daifuku Inc., Japan as an Engineer.
His research interests include image processing and
aerial robotics.
cal control, geometric modeling, virtual reality and computer graphics.
Authorized licensed use limited to: VIT University. Downloaded on November 26,2020 at 09:52:19 UTC from IEEE Xplore. Restrictions apply.
IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019
Ryo Gohara received the B.S degree in electri-
cal engineering from Tokyo University of Science,
Tokyo, Japan, in 2015.
From 2016 to 2017, he was a M.S student in De-
partment of Electrical Engineering, Graduate School
of Engineering, Tokyo University of Science, Tokyo,
Japan. His research interests include robotics and
intelligent transport systems.
Takao Ninomiya received the B.S degree in electri-
cal engineering from Tokyo University of Science,
Tokyo, Japan, in 2016.
From 2017 to 2018, he was a M.S degree in De-
partment of Electrical Engineering, Graduate School
of Engineering, Tokyo University of Science, Tokyo,
Japan. His research interests include image process-
ing and aerial robotics.
Kiyotaka Kato (M’10) received the B.S. degree
from Waseda University, Tokyo, Japan in 1978.
From 1978 to 2000, he was a Researcher in Mit-
subishi Electric Cooperation, Japan. From 2000, He
has been Professor in the Department of Electri-
cal Engineering, Graduate School of Engineering,
Tokyo University of Science, Tokyo, Japan.
He is member of IEEE, The Japan Society for
Precision Engineering and The Society of Instrument
and Control Engineers, Japan. His research interests
include aerial robotics, factory automation, numeri-