An Animatronic System Including Lifelike
Robotic Fish
YUUZI TERADA AND IKUO YAMAMOTO
Invited Paper
This paper provides an outline of a new animatronic system,
based on the technology of the flexible oscillating fin. The oscillating
fin propulsion system was designed and constructed to be combined
with a ship model. The system’s feasibility has been confirmed by
tank tests using the ship model. As a result, several advantages of the
oscillating fin system have been found. A neural network was suc-
cessfully applied for an identification of the ship model dynamics
with the oscillating fin and its effectiveness was confirmed. The an-
imatronic system is a computer-controlled biomechanically engi-
neered model, rapidly gaining popularity throughout the world. We
have developed aquatic robots with oscillating fins for the anima-
tronics system to build a virtual aquarium. We have proposed an
exhibition system for enhancing event spaces that includes an ani-
matronic system for modern-day fish, coelacanths, and Cambrian-
world creatures, able to swim under their own electric power.
Keywords—Animatronic system, fishlike swimming control,
neural network, oscillating fin propulsion, robotic fish.
I. INTRODUCTION
It is well known that marine creatures, such as fish, that
swim using small power (sweetfish) as well as at high speeds
(dolphin: 60 km/h, swordfish: 80 km/h) [1] are superior in
their position-keeping characteristics. These characteristics
as creatures have been of interest to science for a long time,
and much research has been conducted [2]; however, it is
rare to study these characteristics from the viewpoint of en-
gineering [3].
The purpose of this research is a flexible oscillating fin
control system which could be used for the propulsion of ma-
rine vehicles by positively making most of the characteristics
of the flexible part. This method obtains a propulsion force
by oscillating fins equipped to vehicles on the analogy of the
motion of marine creatures.
After the control system for a flexible oscillating fin
propulsion device and the oscillating fin driving device were
Manuscript received May 17, 2003; revised March 3, 2004.
Y. Terada is with Fukui University of Technology, Fukui 910-8505, Japan
(e-mail: terada@ccmails.fukui-ut.ac.jp).
I. Yamamoto is with Mitsubishi Heavy Industries Ltd., Tokyo 108-8215,
Japan (e-mail: ikuo2_yamamoto@mhi.co.jp).
Digital Object Identifier 10.1109/JPROC.2004.835363
0018-9219/04$20.00 © 2004 IEEE
1814
designed and manufactured, a cruising test was performed,
first by a numerical simulation and then with a model ship,
and the fundamental performance has been grasped and
prospects of putting the devices to practical use have been
obtained. Robotic fish for amusement in aquariums, etc.,
have been developed as an applied product.
Advantages of the oscillating fin propulsion system have
been found and products of application have been created by
the research.
II. BASIC OSCILLATING FIN PROPULSION SYSTEM
In many cases, the kinetic parameters of the oscillating fin
cannot be directly detected in control of an oscillating fin,
and there are problems choosing and identifying parameters
to be used for control; a control system able to cope with such
problems should be architected.
In the research, to cope with the above problems, a study
on the application of neural network learning control has
been made using a model ship. The control algorithm is
architected and the control computer software has been
mounted; the cruising test was then conducted in a tank [4].
Fig. 1 shows the outline of the test device for the oscillating
fin propulsion system which has been developed for the basic
tank test.
The neural network learning algorithm has been created
in the control device. It consists of a hierarchy network of
three layers, which are input, middle, and output layers. The
Hess and Smith method [5] has been expanded to a nonsta-
tionary problem; furthermore, a model applying the method
of solving deformation of the wake vortex using the discrete
vortex method [6] has been used, and the I/O variables and
node numbers of the middle layer have been determined by
simulation.
From Fig. 1, the input signals are formed to give the ship
speed, the propulsion thrust, the learning signal, and the
output signal to give the vibrating frequency, phase angle,
sway angle, and yaw angle amplitudes. The node number
of the middle layer was determined to be four from the
PROCEEDINGS OF THE IEEE, VOL. 92, NO. 11, NOVEMBER 2004
Fig. 1. Test device.
Fig. 2. Oscillating fin model.
viewpoint of error energy function and simplification of the
system.
The two-phase control oscillator, ac servo control ampli-
fier, oscillating fin driving device, and small-sized three-
component force block gauge for fluid measurement were
designed and manufactured for this test. The oscillating fin
driving device was designed to be actuated linking sway di-
rection motion with yaw direction motion by mounting the
yaw direction driving device on the sway direction driving
device.
The oscillating fin is actuated by varying the amplitude,
phase difference, and oscillating frequency of sway and yaw
motion. The oscillating fin consists of rigid and flexible parts
and the propulsion efficiency is improved by the flexibility
of the flexible part [4]. The control computer consists of the
neural network software and the command generator. The
command generator gives the command values of the sway
and yaw motion parameters of the oscillating fin. During the
TERADA AND YAMAMOTO: AN ANIMATRONIC SYSTEM INCLUDING LIFELIKE ROBOTIC FISH
optimal adjustment of motion parameters of the oscillating
fin in the cruising test, the neural network gets the learning
data for back propagation. After the network was constructed
by back propagation, the oscillating fin is actuated only by
the neural network control and can self-cruise the vehicle [7].
III. NUMERICAL SIMULATION ANALYSIS OF THE
OSCILLATING FIN PROPULSION SYSTEM
The oscillating fin propulsion system is used as a vehicle’s
actuator. In this paper, this system is applied to a marine ship
as a vehicle.
A ship motion model in surge, sway, and yaw direction was
developed [8], and the model shown in Fig. 2 was constructed
as an oscillating fin motion model [9].
After numerical simulation and the tank test were per-
formed, the oscillating fin propulsion system was loaded on
a model ship 3.5 m in length, 0.5 m in breadth, and 194 kg in
1815
Fig. 3. Comparison between simulation and the experiment.

weight. The oscillating characteristics in the numerical sim-

ulation agreed well with that in the tank tests, as is shown in
Fig. 3.
Through this result, a case study on motion performance
using a simulator, including, for example, a study on instal-
lation of plural oscillating fins, became possible.

IV. EXPERIMENTAL TEST OF THE OSCILLATING FIN

PROPULSION SYSTEM
The flexible oscillating fin propulsion device was pro-
duced experimentally, and only the flexible oscillating fin
propulsion device was independently tested before loading it
onto a model ship to examine the influence of the oscillating
fin shape and the flexible part on propulsion. Fig. 4. Ship model with the oscillating fin propulsion device.
The oscillating fin propulsion device was loaded onto a
model ship and the tank cruising tests were carried out. The
purpose of the tank tests was to grasp the propulsion charac-
teristics as a ship’s actuator and self-cruising capability using
only the neural network.
Fig. 4 shows the model ship with the oscillating fin propul-
sion device. During the cruising test and by using the com-
mand generator, the cruising test learning data for the neural
network control is accumulated and the weights in the neural
network are determined by back propagation.
The oscillating fin driving command signal is given by
the forward operation in the neural network, based on the
target value command, and then the model ship cruises. Fig. 5
shows the results of the self-cruising test by a neural network. Fig. 5. Self-cruising test result by a neural network.
The neural network is effective for the identification of a ship
model for a vehicle equipped with the oscillating fin. Its ap- fin shape. The fish-tail-type fin is found to produce higher
plication for any ship type has been conducted and the effec- power compared with the same area of a rectangular fin.
tiveness was confirmed [9]. The widespread application to
vehicles can be widened by the neural network.
V. LIFELIKE ROBOTIC FISH DEVELOPMENT
Also, transition of the thrust force from a positive to a neg-
ative direction can be conducted smoothly only by changing Fig. 6 shows the principle of the control system based on
the phase angle of the sway and yaw motion. Therefore, it the technical research of the flexible oscillating fin propul-
was found that the transition of the thrust force from progress sion system. We can control the fish by regulating the ampli-
to reverse of the vehicle can be conducted smoothly. The tude, frequency, and phase of the joints of the fin.
maximum positive thrust force can be generated around a Here, sway direction motion of the oscillating fin can be
phase difference of 90 of sway and yaw motion and the max- achieved from the front joint angle of the fish and yaw direc-
imum negative thrust force around 90 . tion motion from the rear joint angle, respectively.
Moreover, propulsive thrust and efficiency can be im- Robotic fish have been developed as an applied product
proved by using a flexible part of the oscillating fin. The of this system. Table 1 shows specifications of a sea bream-
characteristics can be further improved by improving the type robotic fish (weight: 2.6 kg, length: 60 cm) [10] and a

1816 PROCEEDINGS OF THE IEEE, VOL. 92, NO. 11, NOVEMBER 2004
Fig. 6. Principle of control of robotic fish.
Table 1
Specifications of Fish Robots
coelacanth-type robotic fish (weight: 40 kg, length: 1.2 m)
[11].
Batteries and a buoyancy control device are built into the
robotic fish, and three-dimensional movement of the robotic
fish is possible by remote control, using underwater wireless
information communication. It is generally thought that it is
difficult to transmit a signal underwater using the radio wave
because the attenuation of the radio wave underwater is large.
However, it is actually possible to transmit a signal under-
water by using appropriate frequency and modification. The
computer wireless maneuvering control device and noncon-
tact submerged charging equipment as peripheral equipment
have been developed and continuous swimming can be con-
ducted for hours.
Hydrodynamic tests of robotic fish are shown in Fig. 7
as conducted. In the case of sea bream-type robotic fish, the
vortex interval value was 0.26 and approximately near to the
theoretical asymmetric stable vortex interval value (0.28) by
Karman vortex street theory. Also, the Strouhal number was
0.25 and the value was within the measured swimming range
values of real fish (0.25–0.35). The controller was designed
and checked based on such hydrodynamic tests. Advanced
control algorithms [12] were also applied to the robotic fish.
Optimization of the fin shape was conducted by nonlinear
programming and genetic algorithm. Artificial intelligence
(AI) and chaos control were tried to simulate the real fish
maneuvering.
An example is shown in Fig. 7. A very realistic and lifelike
swimming method can be realized by the flexible oscillating
fin propulsion.
TERADA AND YAMAMOTO: AN ANIMATRONIC SYSTEM INCLUDING LIFELIKE ROBOTIC FISH
Fig. 7. Hydrodynamic tests of robotic fish.
Fig. 8. Sea bream-type robotic fish and swimming view.
VI. ANIMATRONIC SYSTEM
This research is extremely important in technology for the
new field of animatronics—a computer-controlled biome-
chanically engineered model, in this case, aquatic creatures.
Animatronics technology is rapidly gaining popularity
throughout the world. It can be applied to create a virtual
aquarium not possible with computer graphics technology
alone. We have developed a method of enhancing event
spaces that included animatronics for modern-day fish,
coelacanths, and Cambrian-world creatures able to swim
under their own electrical power. The concept of the system
is shown in Fig. 9. Spectators will be able to see the various
lifelike fish robots swimming in the tank. By using voice
treatment technology and image recognition technology,
they will use variety variation functions where the fish
answers to a voice or the fish reads a number card.
Operation of our new animatronics system has been
demonstrated to the public at several exhibitions in Japan,
where visitors come to see robotic fish swimming in an
experimental tank. Figs. 8 and 10 show photographs taken
at a public demonstration of a trial system.
It is very difficult to detect the motion of a controlled os-
cillating fin directly; therefore, a control system needs to be
constructed so as to solve the problem of choosing and iden-
tifying the control parameters. Furthermore, unless a math-
ematical model for the oscillating fin is created, designing a
1817
Fig. 9. Concept of the animatronic system.
Fig. 10. Demonstration of the trial system.
model based on the control system becomes too difficult. Var-
ious systems were tried; ultimately, a combination of neural
network, chaos, and other systems has been applied to repro-
duce the motion of the fish naturally and realistically with the
robots. As for the control algorithm, software was made and
added to the control computer device; then a tank cruising
test using software was performed.
The block diagram of the control system is shown in
Fig. 11. Via a computer, supersonic sensors, controllers, and
the oscillating fin generate, adjust, and control the motion of
the robotic fish. The control computer consists of the control
software and a command generator, which produces sway
and yaw motion parameters for the oscillating fin.
1818
Fig. 11. Block diagram of the control system.
VII. EVALUATING REALISM
This animatronics system is based on the “system reality”
model whereby spectators take pleasure in watching the
robotic fish swimming in the tank. The premise is that reality
appeals to human sensitivity. The spectators compare the
known image in their brain with the robotic image before
them and determine the level of realism (and by extension
the level of pleasure). However, if spectators have not seen
real fish swimming, they cannot evaluate the realism of
the robotic fish because they have no strong impression
from which to draw comparisons. On the other hand, if the
spectator has at some point in the past seen the real fish in
action, then their excitement at the sight of the robotic model
will be greater. For this reason, we have chosen to have the
spectators evaluate the realism of an animatronic sea bream,
a species of fish well known to Japanese spectators.
PROCEEDINGS OF THE IEEE, VOL. 92, NO. 11, NOVEMBER 2004
VIII. AIMS OF THE ANIMATRONIC SYSTEM
We have two aims for animatronics systems. The first
is to attempt to create biomechanical systems in harmony
with human beings and nature. As technology matures and
people become more familiar with machines, the demand
for machines which do not appear mechanical will become
stronger. We can see this as computers become less and less
“computer-like” and as demand emerges for increasingly
elaborately designed household appliances which blend
seamlessly into the context of our daily lives. This trend
will push developers to produce more “nonmechanical”
machines and drive industry to develop ever more techno-
logically advanced machinery which puts consumers at ease.
The robotic fish, while covered with scales, is nonetheless
loaded to capacity with new electronic devices. It attracts
people with its likeness to real fish and brings them closer
to technology [13].
The second aim is to arouse public awareness and help
people learn from nature. We may be accustomed to seeing
fish on our dinner plates, but observing them from a different
perspective gives us a new understanding of the exquisite
shape, function, and swimming ability of common marine
creatures. However, no matter how advanced the technolo-
gies used to replicate creatures, the robot will never perform
exactly as live fish. This became crystal clear in the course
of the development process.
We sometimes fall into the illusion of believing that any-
thing can be created with advanced technology. This is true
of control system design and of heavy industry. As we con-
tinue to develop mechanical systems, however, it is important
to take our lessons from natural phenomenon.
IX. PUBLIC REACTION TO THE ROBOTIC FISH
After the initial unveiling of the fish robot, many reporters
requested further information. Reports on the fish robot ap-
peared on TV programs, in newspapers, and even in famous
magazines such as Science and Time. The mass media com-
mented frequently on the lifelikeness of the robot, and one
TV star even said she was made uncomfortable by its lifelike-
ness. The point was to start by introducing a robotic model of
a fish with which the audience was very familiar and to pro-
voke surprise at the realism of its movements. Consequently,
when the unfamiliar robotic coelacanth was presented, the
audience would be able to infer the realism of its movements
based on their experience with the robotic sea bream.
In this way we seek to build acceptance for animatronic
models.
X. CONCLUSION
The outcomes of the study can be summarized as follows.
1) A safe actuator can be created by avoiding dangerous
rolling generated by a screw when the oscillating fin
propulsion system is used.
2) Identification of vehicles with various oscillating fins
by the neural network becomes possible. The gener-
ality of the system’s application is much improved.
TERADA AND YAMAMOTO: AN ANIMATRONIC SYSTEM INCLUDING LIFELIKE ROBOTIC FISH
3) The transition from positive/negative to negative/pos-
itive thrust force can be conducted easily by the
phase angle control of the oscillating fin motion. The
hovering characteristics of vehicles are, therefore,
improved.
4) One fin can control both the thrust force and its direc-
tion simultaneously. Then a compact actuator can be
constructed.
5) The fin’s flexibility can be utilized actively. It is, there-
fore, possible to improve the propulsion performance.
6) Multijoint fin structure has possibilities for further im-
provement of the propulsion performance, and a quiet
and calm motion by a large fin is promising in the fu-
ture by setting up widely the actuator operating area.
7) Realistic movement becomes possible by making the
oscillating fin propulsion control system and buoyancy
control system for robotic fish.
8) Transmission of a radio signal underwater is possible
for robotic fish maneuvering control.
As a result of the research, it was found that the oscillating
fin propulsion device was especially effective for actuators in
a cruising area consisting of muddy or sludgy water, a quiet
cruising demanded area, a slow speed cruising area, and a
hovering area and actuators of robotic fish for amusement at
aquariums, etc. In the research, the test was conducted by the
oscillating fin propulsion device using sway and yaw motion.
It is possible to get propulsion by the same theory even by a
method of using heave and pitch motion.
Robotic fish is the most well-known product of the re-
search. However, the research outcomes can be widely ap-
plied to various vehicles such as ships, underwater vehicles,
leisure equipment, and underwater monitoring vehicles.
As one of many application targets for research on
robotic fish, we proposed exhibition systems for the most
creative event hall, including animatronic systems such as
modern-day fish, coelacanths, and Cambrian-world crea-
tures able to swim under their own electrical power.
We expect that in the future the new system will be in-
cluded in plans of new theme parks or aquariums around the
world. We are gratified that children will experience control
engineering or bioscience firsthand as they witness the im-
pressive robotic fish swimming in the tank.
ACKNOWLEDGMENT
The authors would like to thank members of Nagasaki
and Hiroshima Research and Development Center of Mit-
subishi Heavy Industries Ltd., Tokyo, Japan, and Ryomei En-
gineering Co., Ltd., Hiroshima, Japan, for their cooperation
during the research.
REFERENCES
[1] Yoshida, Fune no Kagaku (Science of Ships). Tokyo, Japan: Ko-
dansha, 1976.
[2] H. Hertel, Biology and Technology. New York: Reinbold, 1966, pp.
110–190.
[3] Isiki et al., “Research on oscillating fin propulsion,” J. Naval Archi-
tects Jpn., vol. 642, 1983.
[4] I. Yamamoto and Y. Terada, “Research on an oscillating fin propul-
sion control system,” in Proc. IEEE OCEANS 1993, vol. 3, pp.
III/259–III/263.
1819
 [5] J. Hess et al., “Calculation of potential flow about arbitrary bodies,”
Prog. Aeronaut. Sci., vol. 8, pp. 1–138, 1996.
[6] A. Kubota et al., “Study on propulsion by partially elastic oscillating
foil,” J. Naval Architects Jpn., no. 156, pp. 95–105, 1985.
[7] I. Yamamoto and Y. Terada, “Development research of oscillating
fin propulsion system,” in Proc. Annu. Conf. Soc. Instrument and
Control Engineers, 1994, pp. 37–38.
[8] , “Design of a maneuvering control system for a ship,” Trans.
West-Jpn. Soc. Naval Architects, no. 83, pp. 96–100, 1994.
[9] I. Yamamoto, Y. Terada, T. Nagamatsu, and Y. Imaizumi, “Propul-
sion system with flexible/rigid oscillating fin,” IEEE J. Oceanic Eng.,
vol. 20, pp. 23–30, Jan. 1995.
[10] Y. Terada and I. Yamamoto, “Development of oscillating fin propul-
sion system and its application to ship and artificial fish,” Tech. Rev.
Mitsubishi Heavy Ind., Ltd., vol. 36, no. 3, pp. 84–88, 1999.
[11] Y. Terada, “A trial for animatronic system including aquatic robots,”
J. Robot. Soc. Jpn., vol. 18, no. 2, pp. 195–197, 2000.
[12] I. Yamamoto, “Robust and nonlinear control,” Int. J. Int. Fed. Au-
tomat. Control, vol. 11, no. 13, 2000.
[13] P. Menzel and F. D’Aluisio, Robo Sapiens. Cambridge, MA: MIT
Press, 2000.
Ikui Yamamoto received the B.E. degree in
aeronautic engineering, the M.Eng.Sc. degree in
applied mechanics, and the Dr. Eng. degree from
Kyushu University, Fukuoka, Japan, in 1982,
1984, and 1994, respectively.
He has worked for the System Engineering
Department and the Nagasaki Research and
Development Center of Mitsubishi Heavy In-
dustries, Ltd., Tokyo, Japan, since 1984. He
was with the Electronics Division as a Project
Manager of Oceanic Engineering and a Senior
Researcher of Control System Engineering from 1992 to 2002. Since 2003,
he has been a Engineering Manager of the Technical Planning Department
at Mitsubishi headquarters. His current research interests are in vehicle con-
trol, platform control, robot control, wave control, and computer-integrated
manufacturing (CIM) technologies.

Yuuzi Terada received the B.E. degree in

applied physics and the Dr. Eng. degree from
Osaka University, Osaka, Japan, in 1970 and
1998, respectively.
He has worked for the Hiroshima Research
and Development Center, Takasago Research
and Development Center, System Engineering
Department and Electronics Research, and
Development Department at the Technical Head-
quarters of Mitsubishi Heavy Industries Ltd.
from 1970 to 1998. He was General Manager
of the Electronics Research and Development Department and a technical
advisor to the Hiroshima Research and Development Center of Mitsubishi
Heavy Industries from 1998 to 2001. Since 2002, he has been a Professor
with the Faculty of Engineering, Fukui University of Technology, Fukui,
Japan. His current research interests are in control systems design and
analysis of robot, vehicle and ship mechanical systems.
Prof. Terada is a Member of the Society of Instrument and Control Engi-
neers of Japan (SICE), the Japan Society of Mechanical Engineers (JSME),
the Institute of Systems Control and Information Engineers (ISCIE), and the
Japan Institution of Marine Engineering.

1820 PROCEEDINGS OF THE IEEE, VOL. 92, NO. 11, NOVEMBER 2004