You are on page 1of 11

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Finding NEMO (novel electromaterial muscle oscillator): a polypyrrole powered robotic fish
with real-time wireless speed and directional control

This content has been downloaded from IOPscience. Please scroll down to see the full text.
2009 Smart Mater. Struct. 18 095009
(http://iopscience.iop.org/0964-1726/18/9/095009)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 150.216.68.200
This content was downloaded on 27/10/2014 at 15:49

Please note that terms and conditions apply.

IOP PUBLISHING

SMART MATERIALS AND STRUCTURES

Smart Mater. Struct. 18 (2009) 095009 (10pp)

doi:10.1088/0964-1726/18/9/095009

Finding NEMO (novel electromaterial


muscle oscillator): a polypyrrole powered
robotic fish with real-time wireless speed
and directional control
Scott McGovern1,2, Gursel Alici2,3,5 , Van-Tan Truong4 and
Geoffrey Spinks2,3
1

Intelligent Polymer Research Institute, AIIM Building Innovation Campus, University of


Wollongong, Wollongong NSW, Australia
2
ARC Centre of Excellence for Electromaterials Science, AIIM Building Innovation Campus,
University of Wollongong, Wollongong NSW, Australia
3
School of Mechanical, Materials and Mechatronic Engineering, University of Wollongong,
Wollongong NSW, Australia
4
Maritime Platforms Division, Defence Science and Technology Organisation, Australia

Received 16 December 2008, in final form 17 April 2009


Published 1 July 2009
Online at stacks.iop.org/SMS/18/095009
Abstract
This paper presents the development of an autonomously powered and controlled robotic fish
that incorporates an active flexural joint tail fin, activated through conducting polymer actuators
based on polypyrrole (PPy). The novel electromaterial muscle oscillator (NEMO) tail fin
assembly on the fish could be controlled wirelessly in real time by varying the frequency and
duty cycle of the voltage signal supplied to the PPy bending-type actuators. Directional control
was achieved by altering the duty cycle of the voltage input to the NEMO tail fin, which shifted
the axis of oscillation and enabled turning of the robotic fish. At low speeds, the robotic fish had
a turning circle as small as 15 cm (or 1.1 body lengths) in radius.
The highest speed of the fish robot was estimated to be approximately 33 mm s1 (or 0.25
body lengths s1 ) and was achieved with a flapping frequency of 0.60.8 Hz which also
corresponded with the most hydrodynamically efficient mode for tail fin operation. This speed
is approximately ten times faster than those for any previously reported artificial muscle based
device that also offers real-time speed and directional control. This study contributes to
previously published studies on bio-inspired functional devices, demonstrating that
electroactive polymer actuators can be real alternatives to conventional means of actuation such
as electric motors.
(Some figures in this article are in colour only in the electronic version)

(AUVs) in the form of miniature submarines already exist,


a recent review of their performance [1] has highlighted
their lack of low speed manoeuvring capabilities. Highly
manoeuvrable AUVs capable of hovering/station keeping and
small radius turning at low speed are seen as ideal for
underwater inspections, particularly in confined spaces or
difficult to access areas.
Designing AUVs that mimic the swimming action of fish
is one obvious way to improve their performance, as fish
show remarkable swimming abilities [2]. The wide range of

1. Introduction
Mobile platforms offer improved versatility for sensing
systems involved with environmental monitoring, pollution
detection, video mapping, surveillance and other such tasks.
For operation within aquatic environments, highly mobile
swimming robots are an obvious form that can enable high
mobility and versatility. While autonomous undersea vehicles
5 Author to whom any correspondence should be addressed.

0964-1726/09/095009+10$30.00

2009 IOP Publishing Ltd Printed in the UK

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

Table 1. Summary of performance of actuator-driven robotic fish.


Description
Konkuk University:
ceramic piezo actuator
with rack and pinion
mechanism that flaps a
caudal fin (various shapes)
Michigan State University:
IPMC operated caudal fin

Top speed
1

2.5 cm s (0.09 body


lengths s1 )

0.63 cm s1 (0.027 body


length s1 )
0.52 cm s1 (0.12 body
Kagawa University: a
tethered system using two lengths s1 )
IPMC bending actuators
(as parallel tails)
Hankuk Aviation
University: untethered
system propelled with one
IPMC bender
Auckland University:
untethered system
propelled with one
PPy.DBS bender

Dimensions/(mass)

Drive voltage

Reference

27 cm 5.0 cm 6.5 cm
(550 g)

300 V (p-p) at 0.9 Hz


(for max speed)

[16]

23 cm 13 cm 6.5 cm
(295 g)

3.3 V (p-p) at 2 Hz

[11]

4.5 cm 1.0 cm in
diameter (0.76 g)

5 V (p-p) at 1 Hz

[17]

2.36 cm s1 (0.245 body


lengths s1 )

9.6 cm 2.4 cm 2.5 cm 5 V (p-p) at 4 Hz


(16.2 g)

[9]

0.2 cm s1 (0.016 body


lengths s1 )

12.2 cm 3.5 cm 1.0 cm 1.6 V (p-p) at 1 Hz

[12]

voltages, which can be an advantage for small mobile devices


as they can be powered directly from batteries.
A small number of prototype fish robot devices
have already been demonstrated using artificial muscle
actuators [712] and their performances are summarized in
table 1. Most focus to date has been directed at producing
untethered systems that can achieve appreciable speeds. The
fastest speed reported to date is 2.5 cm s1 (or 0.245 body
length s1 ), which is much lower than fish of similar size (for
example, the 140 g bluegill sunfish swims at 24 cm s1 [13],
while the fastest swimming fish achieve up to 15 m s1 [14]).
One prominent example of an artificial muscle powered robotic
fish was a novelty aquarium product developed by Eamex
Corporation in Japan [15]. The small artificial fish were
neutrally buoyant and slowly moved about the aquarium
propelled by IPMC actuated tails. The system was wireless and
power was delivered to the fish via means of electromagnetic
induction. Again, fish swimming speed was low and there was
no means provided for controlling the fish direction.
Most previous robotic fish using polymer actuators use a
single caudal (tail) fin to provide propulsion, which is known
as the ostraciiform swimming mode. However, fish propulsion
involves a varied combination of movement of both the body
and/or the caudal [35, 2]. Undulatory swimming utilizes
movements of the body to develop a wave displacement over
the entire or part of the fishs length to generate forward
propulsion. In oscillatory swimming, the caudal fin moves
about its base without this body wave formation, and displaces
water generating forward thrust (figure 1). In terms of speed,
manoeuvrability, acceleration and swimming efficiency, no
one method of swimming excels in all of these areas. For
example, acceleration and speed increases can be obtained
over the anguilliform method of swimming by better utilizing
the propulsive element of the caudal fin in the carangiform
and thunniform modes of swimming, however, this benefit is
achieved at the sacrifice of manoeuvrability. The ostraciiform

fish shapes and sizes highlight the many variables that affect
fish swimming performance. Body size and shape; tail fin
size and shape; size and placement of pectoral and dorsal
fin(s) all affect the swimming speed and manoeuvrability.
Different species of fish also produce propulsion through
different combinations of movement of both the body and/or
the caudal (tail) fin [35, 2]. While still the subject of ongoing research, it is clear that mimicking fish hydrodynamics
requires a high degree of freedom of the flow control surfaces.
In aquatic animals this control is afforded by their muscular
system. For example, six major muscle groups control
the pectoral fin on fish [6]. The same degree of freedom
cannot be achieved in AUVs using conventional motors
because of size limitations. Bandyopadhyays analysis of fish
hydrodynamics [1] concluded that mimicking the performance
of fish was most feasible by adopting multiple artificial muscle
actuators with neuroscience based control.
Artificial muscles, or actuator materials, are attractive
for mobile robotics for a variety of reasons. As mentioned,
actuator materials offer the ability to generate fish-like
movements by using multiple, small actuators to fine-tune body
and fin movements. Because of their simple structure, artificial
muscles also provide the possibility of scaling down the size
of the AUV. Small sized AUVs would be capable of entering
small spaces, expanding their surveillance capabilities to areas
such as pipe and tank inspections. Actuator materials can
also provide silent operation, which can be important to avoid
detection.
While a large number of actuating materials exist, we
favour the bending-type actuators that can be directly coupled
to a tail or pectoral fin to produce a flapping action without any
other mechanical mechanism. Such a simple design will aid in
the future further miniaturization of the device. Bending-type
polymer actuators are available from piezoelectric materials,
ionic-polymer metal composites (IPMCs) and conducting
polymers (CPs). The latter two materials operate at low
2

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

Figure 1. Various modes of swimming that are related to BCF propulsion [4].

electrochemically deposited onto either side of a platinumcoated poly(vinylidene fluoride) membrane (PVDF) 100 m
thick with a 0.45 m pore size (Immobilon Millipore)
to manufacture a stand-alone actuator with a tri-layer
configuration [7, 25, 24]. The advantage of this configuration
is that both the working and counter electrodes (consisting of
each of the individual PPy layers) are self-contained within the
system and form the boundaries of the electrochemical cell
(the PVDF separator). When fully wetted with electrolyte,
these actuators may work stand-alone in air or other media and
minimize the overall size of the device as the PVDF separator
is paper thin (100 m) and acts as a reservoir to hold the
electrolyte.
Extremely fast actuators have been realized with the use
of bis(trifluoromethane sulfonimide) (TFSI) as dopant for the
PPy [23, 24]. To form the PPy.TFSI, pyrrole monomer was
electropolymerized onto sputter-coated PVDF with an applied
current density of 0.1 mA cm2 for 12 h at a temperature
of 33 C from a propylene carbonate (PC) solution that
contained 0.1 M pyrrole monomer, 0.1 M lithium TFSI and
1 w/w% water.
One design challenge encountered was to make a fish body
to house the electronics that was waterproof but also enabled
ready access to the electronics and battery. Fully encapsulated
systems provide good water-proofing, but do not allow easy
access to the electronics for replacement of parts or batteries.
A simple solution used in our prototype system, was to house
the control unit and the battery in a 25 ml syringe that was
cut down and adapted to facilitate a tail fin for propulsion and
pectoral and dorsal fins for stability. Further, the control unit
and the battery are deliberately placed at the lower section of
the syringe to ensure that the centre of gravity of the prototype
was below the centre of buoyancy to provide inherent stability,
preventing any tilt unconditionally, during operation in liquid.
The end of the syringe was plugged and the fixed pectoral
and dorsal fins were attached with epoxy adhesive. Output
wires to power the actuators were attached to the electronics
board and passed through a small hole in the fish body, which
was also plugged with epoxy. The length of these wires
was made to enable complete removal of the electronics and
battery compartment from the fish body for any necessary later
adaptations. A photo of the apparatus may be seen in figure 2.
The complete fish prototype was 20 mm in diameter 125 mm
in length and weighed 16.2 g.
The wireless capacity of the robot prototype was provided
through a SCTX2B/RX2FS transmitter and receiver unit
chips (IPS Japan). These units were contained in a board
that produced a pulse width modulated constant 1 V

mode has the lowest complexity (utilizing purely oscillatory


motions) [18] and as such is the easiest method of propulsion
to mechanically implement and thus mimic in robotic fish
designs.
Fish swimming efficiency is often assessed in terms of
the dimensionless Strouhal number ( St ) [19] that expresses the
relationship between tail beat amplitude (a ) and frequency (n )
to the forward velocity (v ):

St =

na
.
v

(1)

The product of tail beat amplitude and frequency is directly


related to the thrust force generated by the tail oscillation, so
the Strouhal number reflects how the tail thrust is converted to
forward movement of the fish. Studies of various fish species
has shown that the Strouhal number falls within the range
0.2 < St < 0.4 [5]. Only one of the previously reported
IPMC-fish prototypes quoted a Strouhal number, which was
in the range 0.81.6. These high values suggest that further
improvements in actuator performance or fish body/tail design
are needed to match the efficiency and speeds of real fish.
Significant improvements in CP actuator performance
have been reported recently, and these new formulations
now offer the possibility for improving the speed and
manoeuvrability of robotic fish. Kaneto and co-workers have
shown that certain dopant ions and solvents enable very large
linear actuator strains (up to 40%) to be produced in the
length direction of films [2024]. These same formulations
have also been used in bending-type CP actuators and were
shown to increase the speed of response and amplitude of
bending [24]. The aim of the present study was to use these
high speed polypyrrole CP bending actuators in an untethered
fish swimming device and to assess the speed and efficiency.

2. Experimental details
The biorobotic fish design was based on previously reported
systems and used a single caudal fin for propulsion and
steering. While probably not the optimal design to achieve
the desired speed and manoeuvrability, this design allows
a direct comparison between a polypyrrole (PPy) propelled
system with the previously reported CP and IPMC actuated
ostraciiform swimming fish robots. The system used in the
present study was designed to be simple to construct and
operate. In particular, the ability to readily change batteries
and actuator elements was incorporated into the fish design.
Bending-type PPy actuators were fabricated as previously
described [24]. Polypyrrole (PPy) conducting polymer was
3

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

Figure 2. Photograph of the robotic fish prototype.

power supply to the tri-layer polymer actuators such that the


frequency of forward and backward bending of the actuator
(causing a flapping of the tail) could be achieved with
manipulation of the forward and backward controls on the
external transmitter unit (figure 3).
The final design challenge was the need to easily replace
the polymer actuators if they became damaged during the
testing phase. In our previous study [7], silver paste with a
protective epoxy adhesive was used to provide connections to
the polymer actuators through platinum wires. That previous
system was non-serviceable making actuator replacement
slow. For the current design, we introduce small (5 mm
diameter, 1.5 mm thick) 40 neodymium magnets with a
nickel coating that were soldered to the circuits output wires
to act as a removable clip to both hold the actuators in
place and provide the electrical connection. The neodymium
magnets are extremely powerful and can enable a strong
but detachable gripping force to the actuators and the nickel
coating helped provide corrosion resistance between the
actuator and conducting surface. The magnets were housed
in cavities in the adapted plunger of the syringe.
This connection system was used to attach the novel
electromaterial oscillator (NEMO) tail fin (used to propel the
robotic fish). The tail was constructed from a sheet of high
impact polystyrene approximately 150 m thick. The fin
material was chosen for its low weight and high stiffness
(Youngs modulus 2 GPa) and a simple truncated shape
was chosen to simulate more closely a natural fin shape seen
in aquatic environments (figure 4). Two actuators were cut

Figure 4. Schematic showing the fish fin dimensions and actuator


placement on the NEMO.

25 mm long and 3 mm wide and were attached to the tail


fin with Kapton tape (3 M ). The actuators were positioned
on the fin with a spacing of 10 mm, allowing for a 10 mm
overlap for attachment to both the fin and the magnet clamps.
As such, the NEMO was attached to the fish body via the
magnet attachment to the actuators. This design allowed for
free bending of a 5 mm section within the centre of each
actuator upon application of electrical stimuli. Future studies
will consider in more detail the effects of tail shape and size,
and actuator placement.

3. Results
The operation of this NEMO structure on the fish body was
investigated first in air and later in water to determine the effect
that the frequency of oscillation has on the sweep angle travel.
These studies were used for the analysis of the resulting speed
and manoeuvrability of the fish.
3.1. Operation of the NEMO in air
Operation of the NEMO tail fin in air was extremely responsive
and a large deflection of the fin tip was achieved upon

Figure 3. Electronics design of the fish robot prototype.

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

where n = natural frequency, E = elastic modulus of the


composite actuator structure, I = area moment of inertia of
the composite actuator structure, m b = mass of the NEMO
actuators, m = mass of the NEMO tailfin and L = total
NEMO length (25 mm + 5 mm).
From this relation it can be seen that a tuning of the
frequency in the NEMO may be possible by changing the
mass and length of the attached tail fin. If the fin was made
either shorter or lighter, an increase in the natural frequency
should be observed. The effect of such geometry modifications
on the net thrust force generated is the subject of on-going
investigations [26].

was due to a solvent exchange between the PC electrolyte


within the PVDF separator membrane and water within the
test bath, affecting such factors as the ionic conductivity and
stiffness of the actuator.
Various encapsulants were investigated in an attempt to
minimize or eliminate the electrolyte leaching with varied
success. For example, encapsulation of the actuators within
adhesive Scotch tape stopped this degradation, but significantly
increased the stiffness of the tri-layer system and dramatically
reduced the tip displacement amplitude. The best system
identified to date was petroleum jelly applied to the outside of
the actuators. This material formed a barrier coating between
the PC electrolyte and water without unduly increasing the
stiffness of the actuator and reducing the oscillation amplitude.
The coating enabled a stable performance with little or no
degradation within the first 10 min of operation, after which
a slow reduction in the oscillation amplitude occurred. The
search for longer lasting, flexible encapsulants is the subject
of on-going research. As such, further investigations into the
liquid dampening of the NEMO were undertaken in a reservoir
containing a solution that was the same as the electrolyte within
the actuatorsi.e. 0.1 M Li TFSI in PC, to ensure that a
more accurate representation of the true oscillation amplitude
response to frequency could be obtained.
The frequency response of the NEMO actuated fin in PC is
shown in figure 5. It can be seen that operation in PC gave tip
amplitudes significantly lower at all frequencies compared with
operation in air. The resisting force of the surrounding liquid
significantly slows the bending of the tail fin so that smaller
fin tip amplitudes were produced. The natural frequency of the
NEMO was an order of magnitude lower when operated in PC
compared to air, such that the greatest amplitude of oscillation
was at approximately 0.4 Hz in PC. There was a very fast decay
in the magnitude of the fin tip oscillation as the frequency
was increased such that when operating at 2 Hz the peak to
peak fin tip oscillation was only 2 mm. The properties of the
surrounding liquid can be important factors to consider when
attempting to predict the effect of damping on the frequency
response of the NEMO. As the density of propylene carbonate
is higher than water at 1.2 g ml1 , this will have a damping
effect on the frequency response of the fin. The higher is the
damping effect, the smaller is the resonant frequency of the fin
assemblyNEMO. This follows that the resonant frequency
can be shifted towards the left direction. Further, an increased
damping effect will decrease the magnitude of the response at
the resonance frequency. On the other hand, the viscosity of
propylene carbonate (2.5 cP) [27] is only slightly higher than
that of water (1.0 cP) and is not expected to impart significant
operational changes in the fin response.

3.2. Operation of the NEMO in liquid

3.3. Operation of the robotic fish

The NEMO tail fin was immersed in water to see the effect that
liquid dampening had on the oscillation amplitude of the tail
fin at different operation frequencies. It was observed on the
initial attempt that the tail fin oscillation amplitude diminished
very rapidly such that there was little or no movement after
approximately 60 s operation. This degradation in oscillation

Investigations were undertaken into the effectiveness of the


NEMO for propulsion of the fish. A NEMO powered robotic
fish prototype was placed in a tank 40 cm 60 cm that
was filled with water to a depth of approximately 10 cm
(figure 6). Petroleum jelly was used as an encapsulant for
the actuators and all measurements were taken within the first

Figure 5. Peak to peak fin tip oscillation amplitude of the NEMO


operating in air and when immersed in propylene carbonate (PC).

excitation with the square wave voltage signal. For example,


the application of a 1 V 1 Hz square wave signal was
seen to generate a peak to peak fin tip oscillation amplitude
of approximately 35 mm. The largest oscillation amplitudes
occurred at excitations using the lowest test frequencies (0.5
1 Hz), as shown in figure 5. The fin tip amplitude tended
to decrease when operated at higher operating frequencies,
however a distinct peak in oscillation amplitude was observed
at approximately 4.5 Hz (figure 5).
At the peak frequency the oscillation amplitude of
approximately 35 mm is equivalent to that obtained when the
device is operating at 1 Hz. This phenomenon relates to the
resonance effect that occurs when the system is operated at its
natural frequency. Previous studies of tri-layer PPy bending
actuators of similar length gave a resonance peak at 42 Hz [24].
The natural frequency of the NEMO (polymer actuators + fin)
is expressed as:

3E I

n =  33 
(2)
mb + m L3
140

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

Figure 6. A NEMO powered robotic fish being tested.

(a)

(b)

Figure 7. Typical bending of the tri-layer actuators that occurs in (a)


normal bending conditions under low load, and (b) under high loads
at large tip displacements.

few minutes of operation to ensure that degradation in actuator


performance due to leaching did not affect the results. The fish
was sufficiently buoyant that the dorsal fin extended above the
water surface. The entire NEMO fin was submerged.
The aim of the fish tests was to determine the operating
frequency that generates maximum forward speed. It was
found that operation at low frequencies (0.2 Hz or less)
caused the actuator movement on the NEMO to become
disabled. That is, the actuators bent in one direction but
were unable to bend back in the other direction. A similar
phenomenon was encountered by Alici et al [7] where a
curling of the actuators was seen to occur under heavy load.
When the tri-layer actuator movement was hindered by a load,
yet the voltage was still applied, the continued expansion of
the polymer initiated a bending about the longitudinal axis
(figure 7(b)). This curling had the effect of changing the
flat plate design of the tri-layer actuator to be more tubular,
significantly increasing the rigidity of the structure. The
increased rigidity meant that the actuator could not bend back
in the other direction when the voltage was reversed until
the film had uncurled. The onset of these curling processes
significantly reduced the frequency of the tail fin oscillation.
The actuator curling was seen to occur with oscillation
frequencies 0.2 Hz, and as such, an oscillation frequency
of 0.3 Hz was set as the minimum frequency for operation of
the fish device. Under such conditions the problem of curling
was eliminated, and traditional bending of the actuators was
reinstated (figure 7(a)).
Several different frequency regimes 0.3 Hz were
investigated. Low tail fin frequencies of approximately 0.3
0.4 Hz (enabling large flapping amplitudes) were utilized to
obtain the maximum initial thrust. If once the fish began to
move, the frequency of oscillation was increased to 0.75 Hz
(after 3 or 4 full cycles), the fish was able to obtain a top speed
of approximately 3.3 cm s1 (or 0.25 times its body length per
second) after 30 cm of travel. With this regime, the cruising
speed is higher than all other previously described polymeractuator-propelled fish robots.
Interestingly, the highest cruising speeds were obtained
by operating the NEMO tail fin at frequencies higher than
that which produced the peak flap amplitude, as previously
determined (figure 5). At lower frequencies, when the tail
fin oscillation was highest it was found that the fish nose
turned excessively, thereby limiting the forward speed. The

propulsion produced by the tail fin must overcome the drag


acting on the fish, which can be expressed by:

FD = 12 CD SV 2

(3)

where: FD = drag force, CD = co-efficient of drag, =


density of the fluid, S = frontal surface area of the body and
V = velocity of the body through the fluid. Oscillation of
the tail will cause the fish to accelerate until the point that its
velocity causes the drag force to equal the thrust force. At
this point, the fish will no longer accelerate, but will cruise at
a constant velocity. While the magnitude of the thrust force
will be determined by the amplitude of tail oscillation, so will
the direction of the thrust (figure 8). Large tail fin amplitudes
produced a turning moment on the fish body since the thrust
force contains a larger component in the transverse direction to
travel (x -direction in figure 8). As the fish nose turns (figure 9),
the effective frontal surface area ( S ) exposed to the liquid is
increased, which in turn causes the drag force to increase. As
such, the maximum cruise speed obtainable is not as high for
operation at low fin frequencies as for high fin frequencies
where the nose did not oscillate as much.
3.4. Directional control of the robotic fish
The transverse force described above could be used for
deliberate turning of the fish. In particular, the direction of
travel could be controlled by altering the duty cycle of the
voltage input to the NEMO. With an even oscillation about
the fish neutral axis (50% duty cycle), there is a net forward
travel without turning. By changing the axis of oscillation,
it was found that after each complete oscillation cycle, the
resultant thrust force of the tail fin will contain a net x
component that causes a moment on the fish body enabling
turning (figure 8(C)). Changing the duty cycle of the voltage
supply to the NEMO (giving a longer time at +1 V than at
1 V) produces a greater bending of the tri-layer actuators to
one side, and in turn shifts the axis of oscillation away from the
neutral axis of the fish body.
The true responsiveness with turning of the fish was
difficult to quantify within the small confines of the testing
6

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

Figure 8. Schematic diagram showing typical oscillation amplitudes of the NEMO at both (A) low and (B) high frequencies, and (C) with a
change in the axis of oscillation that can enable turning.

Figure 9. Top view schematic illustration showing how the NEMO tail fin oscillations cause a turning of the nose of the fish.

tank. It was found that turning could be initiated with


application of only small changes in the duty cycle to shift
the axis of oscillation. For example, it was estimated that
the turning circle of the fish at its top speed had a radius of
approximately 30 cm (or 2.2 body lengths) by using a duty
cycle of approximately 65% for 3 cycles. In this example, the
fish was returned to forward propulsion without any turning
by utilization of a duty cycle of 35% for a further 3 cycles.
Tighter turns could be achieved at slower speeds by utilizing
a larger duty cycle of approximately 75%, again for only
2 or 3 cycles. The tightest turning circle had a radius of
approximately 15 cm (1.1 body lengths) and was enabled at
a forward speed of 10 mm s1 . While these turning radii are
larger than found in some fish species (<0.1 body length [1]),
improvements in the robotic fish performance are possible by
adopting active pectoral fins and/or by mimicking anguilliform
swimming styles.

equation (3). The drag coefficient depends on the shape of the


fin (i.e. aspect ratio), and the Reynoldss number ( Re), which
is the ratio of the inertial forces to viscous forces. Re is a
non-dimensional index indicating the relative contributions of
inertial and viscous forces to the resultant force acting on a
propulsion element and is given by:

Re =

LV

(4)

where , L , and are the density of the fluid, the body


length of the swimming device, speed of the device relative
to the fluid, and dynamic viscosity of the fluid, respectively.
Obviously, the inertial forces dominate the hydrodynamics of a
swimming device when the Reynoldss number is much greater
than 1 ( Re  1). Likewise, as the size of the object decreases
or when accelerating from rest, the viscous forces dominate the
movement of the device and Re  1. At such low Reynolds
numbers, the drag coefficient is significantly high [28]. For our
swimming device with a body length of 125 mm moving with a
uniform velocity of 33 mm s1 in the water with a temperature
of 15 C, the Reynolds number is approximately 3570.
The configuration of the caudal fin and its attachment to
the rigid body are shown in figure 10. It is assumed that the
device generates a rectilinear motion. The total drag force

4. Discussion
4.1. Force analysis of swimming device
The drag force is fundamentally a flow loss, which acts to
retard the movement of a swimming device, and is given by
7

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

CD for an oscillating plate as 0.75 [28], then the form drag


f
acting on the fin is estimated as Fdrag
= 0.0693 mN. The drag
force acting on the NEMO was again verified using ANSYS
(figure 11(b)) and including skin friction drag, the maximum
drag was found to be 0.0782 mN. Hence, the total drag force
acting on the fish travelling at the top speed is calculated as the
sum of the body drag and the fin drag to be 0.581 mN.
Two polymer actuators are used to generate the necessary
thrust force to propel the device. The force generated by
a bending-type polymer actuator with the dimensions of
25 mm 3 mm 0.16 mm is approximately 1.3 mN under
1 V [33, 34]. The total force generated by two polymer
actuators is F = 2.6 mN. This force acts at the tip of the
polymer actuators and the reflection of this force onto the
centre of the caudal fin is F = ( L+L L ) F = 2.229 mN. This
force acts perpendicularly onto the centre of mass of the fin, as
depicted in figure 10, and it follows that the net thrust force
for propulsion is Fthrust = 2.229 sin( 2 ) = 0.577 mN.
This estimated thrust force matches very closely with the total
estimated drag force calculated at a velocity of 33 mm s1 .
The analysis successfully predicts the cruising behaviour of the
device and highlights that in order to increase the swimming
speed, either the force generated by the actuators should be
increased or the total drag force acting on the device should be
decreased. Both are part of on-going investigations.
These force analysis results are in agreement with our
previous research on experimental and theoretical performance
characterization of polymer actuators, and show that polymer
actuators can supply sufficient force to activate functional
devices [30, 31, 25].

Figure 10. Configuration of the body, actuators, and fin of the


b
f
swimming device (not to scale). Fdrag
, Fdrag
, and Fthrust are the body
drag force, fin and actuator drag force, and the thrust force acting on
the fin, respectively. The actuator force F is reflected on the
geometric centre of the fin as F . L and L are 5 and 12.5 mm,
respectively.

acting on the device consists of the body drag force and the
caudal fin drag force, where each drag force component is a
summation of the form drag (due to the shape of the object) and
skin friction drag (that is related to the viscous forces acting on
the surface of the object). If we approximate the body shape
to that of a cylinder, the form drag can be easily determined
for forward motion without turning of the nose by using
equation (3), and taking the drag coefficient CD 0.81 as that
of a cylinder in axial flow [29]. The form drag force may be
estimated as 0.138 mN. However, determination of the frontal
surface area upon turning of the nose of the fish (that occurs
during normal swimming) is more difficult calculation that is
made easier using finite element packages such as ANSYS.
Likewise, ANSYS may also be used to easily calculate the skin
friction drag acting on the surface at a given speed. The fish
body was generated in ANSYS with a mesh size of 3 mm and
the drag force acting on the body was calculated for movement
at its maximum velocity of 33 mm s1 (figure 11(a)).
Using ANSYS, the form drag acting on the nose was
calculated as 0.146 mN and the total body drag force (including
skin friction drag) was estimated to be 0.303 mN. However,
when the swimming device turns to one side by 15 , this total
drag force increases to 0.503 mN. Thus the orientation of the
swimming body relative to the flow direction increases the drag
force significantly. The maximum drag force acting on the
b
= 0.503 mN.
rigid body is Fdrag
The form drag force acting on the caudal fin and the
polymer actuators may also be calculated using equation (3).
The bending angle ( 2 ) of the fin was approximately 15 during
the 33 mm s1 movement of the device in the water. This
makes the frontal surface area of the fin and the two polymer
actuators = 169.52 mm2 . If we take the drag coefficient

4.2. Mechanical efficiency of the NEMO device


The mechanical efficiency of the NEMO device to produce
forward propulsion has been evaluated by the Strouhal number,
given by equation (1). The Strouhal number represents the
ratio of unsteady to steady forces generated in the wake left
behind bodies moving through fluids. St indicates how often
the reverse Karman vortices (that produce thrust) are being
generated within the wake of the flapping tail fin, and how
close they are to each other. Too low a Strouhal number
(<0.2) relates to a loosely organized thrust wake that generates
very low, and sometimes negative thrust. At higher St values,
decaying interactions between each of the vortices leads to
inefficient swimming [18]. The optimal range for the Strouhal
number to be in for the most efficient modes of swimming is
between 0.25 and 0.4 [32].
For the NEMO-propelled fish prototype, the maximum
swimming speed of 3.3 cm s1 was produced at a operating
frequency of 0.70.8 Hz. Using the fin tip amplitude (a )
measured at different tailbeat frequencies (n ) in PC electrolyte
and given in figure 5, the Strouhal number for the maximum
speed was approximately 0.28. This St value falls within the
optimal range for efficient swimming, indicating that the tail
and body shapes were close to optimal. However, the fish
design did not allow for increases in speed because of the
complication arising from the fish nose turning. Although
a numerically larger value of na was produced by operating
8

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

b
f
Figure 11. Configuration of the body, actuators, and fin of the swimming device (not to scale). Fdrag
, Fdrag
, and Fthrust are the body drag force,
fin and actuator drag force, and the thrust force acting on the fin, respectively. The actuator force F is reflected on the geometric centre of the
fin as F . L and L are 5 and 12.5 mm, respectively.

While the prototype used was based on previously


reported ostraciiform swimming robots powered by a single
caudal fin, some aspects of this design were observed to limit
swimming speed. In particular, it was observed that large tip
amplitudes caused excessive turning of the fish nose away from
the desired direction of travel. Nose turning increased the drag
force and slowed fish speed. Hence, operating at larger tail
amplitudes was counter-productive.
A major limitation of the NEMO powered tail fin was
the deterioration in its performance when immersed in water.
Better encapsulating materials for the actuators are needed that
do not unduly stiffen the actuator and limit its bending. Further
optimizations in the fin design and shape would also improve
the thrust force generated from the NEMO and a streamlining
of the body shape would reduce drag on the system, allowing a
much greater net propulsive force leading to faster movement.
Miniaturization of the electronics can reduce weight in the
device, improving acceleration rates and/or enabling the fish
robot to carry other electronics including sensors. Such a
device could see wide spread incorporation into autonomous
environmental sensing devices.

Table 2. Experimentally determined fin displacement amplitudes (a )


obtained at different operating frequencies (n ) and frequency
amplitude product (na ), which is related to thrust force.
Frequency (n )

Tail amplitude (a )

na

0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0

28.5
29.5
24.5
17.0
13.5
11.0
8.5
7.5

8.6
11.8
12.3
10.2
9.5
8.8
7.7
7.5

at lower frequencies (table 2), the fish forward speed was


noticeably lower than when operated at 0.70.8 Hz.
Thus, the Strouhal number at these frequencies would be
outside the optimal range, due to the increased drag force
associated with the turning of the fish nose. To convert
the larger actuator movements produced at lower operating
frequencies, it is probable that the fish design should be
modified away from the single caudal fin ostraciiform style of
swimming and towards the thunniform style characteristic of
fast swimming fish.

Acknowledgments

5. Conclusions

The authors would like to thank Drs Stephen John and


Yanzhe Wu for their support in engineering design and
development. Rahim Mutlu is also thanked for the ANSYS
analysis. The authors would also like to acknowledge
the financial support from the ARC Centre of Excellence
for Electromaterials Science (CE0561616) and the Defence
Science and Technology Organisation (DSTO).

A prototype robotic fish device powered using polypyrrole


artificial muscles and that embodies autonomous real-time
control over the devices speed and direction of swimming
has been demonstrated. The study has concentrated on the
characterization of a novel electromaterial muscle oscillator
(NEMO) tail fin propulsor, used to generate forward movement
on the robotic fish device. The maximum forward speed of
3.3 cm s1 is the fastest reported speed for robotic fish powered
by polymer actuators. The smallest turning radius of the
robotic fish was 15 cm, or 1.1 body lengths. At its maximum
speed, the robotic fish operated at a Strouhal number of 0.28,
which is within the optimum range identified from studies of
several fish species.

References
[1] Banyopadhyay P R 2005 Trends in biorobotic autonomous
undersea vehicles IEEE J. Ocean. Eng. 30 10939
[2] Videler J J 1993 Fish Swimming (London: Chapman and Hall)
[3] Colgate E E and Lynch K M 2004 Mechanics and control of
swimming: a review IEEE J. Ocean. Eng. 29 66072

Smart Mater. Struct. 18 (2009) 095009

S McGovern et al

[4] Lindsey C C 1978 Form, Function and Locomotary Habits in


Fish (Fish Physiology vol VII Locomotion) (New York:
Academic)
[5] Triantafyllou G S, Triantafyllou M S and Grosenbauch M A
1993 Optimal thrust development in oscillating foils with
application to fish propulsion J. Fluid Struct. 7 20524
[6] Thorsen D H and Westneat M W 2005 Diversity of pectoral fin
structure and function in fishes with labriform propulsion
J. Morphol. 263 13350
[7] Alici G, Spinks G, Huynh N N, Sarmadi L and Minato R 2007
Establishment of a biomimetic device based on tri-layer
polymer actuatorspropulsion fins Bioinspir. Biomim.
2 S1830
[8] Kamamichi N, Yamakita M, Asaka K and Luo Z-W 2006 A
snake-like swimming robot using IPMC actuator/sensor
Proc. 2006 IEEE Int. Conf. on Robotics and Automation
pp 18127
[9] Kim B, Kim D-H, Jung J and Park J-O 2005 A biomimetic
undulatory tadpole robot using ionic polymer-metal
composite actuators Smart Mater. Struct. 14 157985
[10] Rossiter J, Stoimenov B, Nakabo Y and Mukai T 2006
Three-phase control for miniaturization of a snake-like
swimming robot Proc. IEEE Int. Conf. on Robotics and
Biomimetics
[11] Tan X, Kim D, Usher N, Laboy D, Jackson J, Kapetanovic A,
Rapai J, Sabadus B and Zhou X 2006 An autonomous
robotic fish for mobile sensing Proc. Int. Conf. on Intelligent
Robots and Systems pp 54249
[12] Wang H, Tjahyono S S, MacDonald B, Kilmartin P A,
Travas-Sejdic J and Kiefer R 2007 Robotic fish based on a
polymer actuator Aust. Conf. on Robotics and Automation
[13] Jones E A, Lucey K S and Ellerby D J 2007 Efficiency of
labriform swimming in the bluegill sunfish J. Exp. Biol.
210 34229
[14] Iosilevskii G and Weihs D 2008 Speed limits on swimming of
fishes and cetaceans J. R. Soc. Interface 5 32938
[15] May 2008 http://www.eamex.co.jp/product e.html#hobby.
[16] Wiguna T, Park H C, Heo S and Goo N S 2007 Experimental
parametric study of a biomimetic fish robot actuated by
piezoelectric actuators Proc. SPIE 6525 65250R
[17] Gou S, Fukuda T and Asaka K 2003 A new type of fish-like
underwater micro-robot IEEE/ASME Trans. Mechatron.
8 13641
[18] Sfakiotakis M, Lane D M and Davies J B C 1999 Review of
fish swimming modes for aquatic locomotion IEEE J.
Ocean. Eng. 24 237711
[19] Taylor G K, Nudds R L and Thomas A L R 2003 Flying and
swimming animals cruise at a Strouhal number tuned for
high power efficiency Nature 425 70711

[20] Hara S, Zama T, Takashima W and Kaneto K 2004 Artificial


muscles based on polypyrrole actuators with large strain and
stress induced electrically Polym. J. 36 15161
[21] Hara S, Zama T, Takashima W and Kaneto K 2004 TFSI-doped
polypyrrole actuator with 26% strain J. Mater. Chem.
14 15167
[22] Hara S, Zama T, Takashima W and Kaneto K 2004 Gel-like
polypyrrole based artificial muscles with extremely large
strain Polym. J. 36 9336
[23] Hara S, Zama T, Takashima W and Kaneto K 2005
Free-standing polypyrrole actuators with response rate of
10.8% s-1 Synth. Met. 149 199201
[24] Wu Y, Alici G, Spinks G M and Wallace G G 2006 Fast trilayer
polypyrrole bending actuators for high speed applications
Synth. Met. 156 101722
[25] Minato R, Alc G, McGovern S and Spinks G 2007 Tri-layer
conducting polymer actuators with variable dimensions
EAPAD: Proc. SPIE Electroactive Polymer Actuators and
Devices p 6524
[26] John S W, Alici G and Cook C D 2008 Validation of a resonant
frequency model for polypyrrole tri-layer actuators
IEEE/ASME Trans. Mechatron. 13 4019
[27] Muzikar J, Goor T v d, Gas B and Kenndler E 2002 Propylene
carbonate as a nonaqueous solvent for capillary
electrophoresis: mobility and ionization constant of aliphatic
amines Anal. Chem. 74 42833
[28] Cengel Y A and Cimbala J M 2006 Fluid Mechanics:
Fundamentals and Applications 1st SI edn (New York:
McGraw-Hill)
[29] Hoerner S F 1964 Fluid Dynamic Drag (New York: Hoerner
Fluid Dynamics)
[30] Alici G and Huynh N N 2007 Performance quantification of
conducting polymer actuators for real applications: a
microgripping system IEEE/ASME Trans. Mechatron.
12 7384
[31] Metz P, Alici G and Spinks G M 2006 A finite element model
for bending behaviour of conducting polymer
electromechanical actuators Sensors Actuators A
130/131 111
[32] Triantafyllou M S and Triantafyllou G S 1995 An efficient
swimming machine Sci. Am. 272 6471
[33] Alici G and Gunderson D 2009 A bio-inspired robotic
locomotion system based on conducting polymer actuators
Proc. IEEE/ASME Int. Conf. on Advanced Intelligent
Mechatronics (Singapore, 1417 July 2009)
[34] Alici G and Huynh N N 2006 Predicting force output of trilayer
polymer actuators Sensors Actuators A 132 61625

10