Sunteți pe pagina 1din 8

Modeling of the Supporting Legs for

Designing Biomimetic Water Strider Robots


Yun Seong Song

Steve H. Suhr

Metin Sitti

Department of
Mechanical Engineering
Carnegie Mellon University
Pittsburgh, Pennsylvania 15213
Email: yssong@cmu.edu

Department of
Mechanical Engineering
Carnegie Mellon University
Pittsburgh, Pennsylvania 15213
Email: ssuhr80@hotmail.com

Department of
Mechanical Engineering and
Robotics Institute
Carnegie Mellon University
Pittsburgh, Pennsylvania 15213
Email: sitti@cmu.edu

Abstract Recent studies on the insect water strider showed


that the insect heavily relies on surface tension force to stay afloat.
Inspired by this insect, water strider robots have been developed
using the same locomotive principles as the insect. This paper
focuses on numerically modeling the supporting legs of the insect
and the robots. The rigid-leg model as well as the compliantleg model is developed using numerical approaches, under an
assumption made on the water surface breaking condition. The
effect of different leg material and geometry are discussed. It is
R
shown through simulations that four 7 cm-long Teflon
coated
compliant supporting legs with optimized shapes can lift up to
4.3 grams (0.15 g/cm), while an actual prototype carried 3.7
grams. Another prototype using twelve of these legs successfully
lifted 9.3 grams. Experiments show that the analyses capture
the important features of the supporting legs. The design rules
proposed in this paper will be useful in understanding the insect
statics and also the robotic water strider supporting leg design.
This study will allow a heavier robot to be used for education,
entertainment or environment monitoring purposes.

I. I NTRODUCTION
Adapting sub-optimized working principles of the biological
systems to synthetic technologies is one of the issues in
engineering design. Particularly in robotic applications, biologically inspired systems can prove highly efficient, as can
be seen from wall-climbing gecko robot [1] to underwater fishlike robots [2]. Many robots adapt designs inspired by insects,
such as the 6-legged robots similar to cockroaches [3]-[5], 8legged robot [6], and the micromechanical flying robot [7].
Recently, the unique characteristics of the locomotion of the
insect water strider have been studied, including the dynamic
locomotion behaviors of the insect [8] and the hydrophobicity
of their legs [9]. This insect maintains its body on water almost
only using the surface tension force, and can propel itself at
the peak speed of 1.5 m/s. Hu et al. [8] proposed a robot
mimicking this insect that uses elastic energy storage, and Suhr
et al. [10] built a more robotic water strider robot that uses
R
four Teflon
coated supporting legs.
Dynamic analysis on the robotic water strider robot is
present [8], [10], but a detailed analysis on the supporting
legs of these robots is yet to be proposed. Since both the insect
and the robot rely heavily on the surface tension force of the
water to gain lift force, detailed analysis on the statics on water

Fig. 1. (a) A photo of the water strider, and (b) the robotic water strider. A,
B, C and D in (b) are the supporting legs. Other labels in (b) are explained
in [10].

surface can provide useful information on the characteristics


of the insect, which can be applied to the design of the
robot. One important feature of the robot (and the insect)
is its loading capacity, and a static model of the legs would
provide a way to calculate the maximum weight a robot can
carry. In this paper, a numerical approach to understand the
physics between the supporting legs and the water surface is
proposed. Starting from the Young-Laplace equation and some
assumptions, characteristics of various kinds of supporting legs
are analyzed.
II. L EG M ODELING
As aforementioned, it is important to know how much
weight the robot can actually carry. And if this information

TABLE I
D EFINITION OF PARAMETERS IN F IG . 3

Parameters/Labels

Fig. 2.

A 3-D illustration of a rigid cylinder partially submerged in water.

Notation

Typical value

Contact angle

0 180

Submerge angle

0 180

Length of the cylinder

10 50 mm

Radius of the cylinder

0 1 mm

Density of water

1000 kg/m3

x0 (= R sin )

0 R mm

Area representing the


buoyancy force

N/A

Area representing the


surface tension force

N/A

Buoyancy force

fB

0 10 mN

Surface tension force

fT

0 10 mN

Water take-off point

point on the interface, P , can be written as


Fig. 3. 2-D model of the problem with depth L into y direction. Parameters
are defined in Table I.

can be predicted given the design parameters of the supporting


legs, more rigorous planning of the robot fabrication can be
performed. In this section, a numerical model of the water
surface near a partially submerged rigid cylinder is proposed.
From this, another numerical model regarding a long flexible
supporting leg is developed.
A. Problem Definition
A long and thin rigid cylinder that represents the supporting
leg of the water strider, made of a material with contact angle
c with water, is partially submerged on water as shown in
Fig. 2. The axis of symmetry of the cylinder is horizontal
to the undisturbed surface of the water. Assuming there are
no complex water surface behaviors at the two ends of the
cylinder, the problem can be understood as a 2-D problem
with a depth of L, which is the length of the cylinder (Fig. 3).
Parameters in Fig. 3 are explained in Table I.
The area A in Fig. 3 is proportional to the buoyancy force
acting on the cylinder, and the area B is proportional to the
vertical component of the surface tension force [11]. The
weight of the water volume, bound above by the z = 0
surface and below by either the water-air interface or the watercylinder interface, equals the total lift force the cylinder body
experiences. The ratio between the area A and B represents the
magnitude of the lift force from the buoyancy effect compared
to that from the surface tension force.
The shape of the air-water interface can be calculated using
the Young-Laplace equation. Then, by integrating this interface
profile along the x-axis, the total lift force can be calculated.
1) Governing Equation: From Young-Laplace equation, the
difference of pressure across the water-air interface on a given

P = (

1
1
+
),
R1
R2

(1)

where is the coefficient of surface tension (0.072 N/m), and


R1 and R2 are the principle radii of curvature of the surface at
the point. In this specific case of a long rigid cylinder (R2 =
), (1) can be specified as

gh(x) =

d2

h(x)
2

dx
3 ,
!

2 2

d
1+
h(x)
dx

(2)

where h(x) is the surface profile of the water-air interface


as shown in Fig. 3, is the density of water, and g is the
gravitational constant. That is, equation z = h(x) is the
description of the water-air interface profile. The boundary
conditions (BCs) for h(x) are
dh
(x0 ) = tan(c + )
(3)
dx
h() = 0
(4)
dh
() = 0
(5)
dx
where x0 = R sin is where the water, air and the rigid
cylinder all meet together. Since the number of BCs needed
to solve (2) is two, one of the conditions must be removed.
In the analysis, the third BC, (5), was ignored. It can be seen
later that the solution to (2) also satisfies (5).
To solve the problem defined by (2)-(5), the value of must
be known to completely define the BC. First, assume that the
value of to be given, and then (2) is solved. The increasing
value of implies that the cylinder is pushed deeper toward the
z direction, given a specific value of c . The exact relation
between and the depth of the cylinder is irrelevant to this
problem and is not dealt with.

2) Surface Breaking Condition: The formulation of the


problem is sufficient by (2)-(5), but there is still one issue
to be resolved; how far can we descend the cylinder before it
breaks the water surface? In other words, what is the surface
breaking condition? Only when this question is answered can
the maximum lift force be calculated. Equation (2) does not
reveal any other behavior of water to trigger the breaking of the
surface. But the first BC, (3), can provide some information;
if the slope of h(x) at x0 is infinite, there is no solution to the
problem. Further more, in this case, vertical component of the
surface tension force is in its maximum and cannot withstand
any further increase in payload. For the case of c 90 , this
never happens, unless the cylinder is above the water surface.
But when c > 90 , the slope at x0 becomes infinite for
= 270 c . This specific value of is called crit in later
sections. For cases when c 90 , the cylinder is completely
surrounded by water when = 180 , and therefore the water
surface must break. Thus, crit = 180 if c 90 .
The analyses in the following sections are based on this
assumption that = crit is the condition for water surface
breaking.

Fig. 4. Water-air interface profile when c = 112 The circles represent the
cross-section of the supporting legs (radius of 165 m).

B. Rigid Leg Modeling


Assuming the supporting legs to be very rigid and stiff, the
maximum lift force can be computed using the following two
steps.
1) Coarse Solution: For sufficiently smaller that crit ,
we assume that
dh
(x) = s = constant.
(6)
dx
Then (2) becomes
3

g(1 + s2 ) 2
d2
h(x) =
h(x).
2
dx

(7)

Fig. 5.

Water-air interface profile for various c and crit .

The solution to this simpler problem is


1

h(x, , c ) = B(, c ) e c x ,

(8)

where
1

B(, c ) = c e( c R sin ) tan(c + )


and

s
c=

1
.
g (1 + s2 ) 32

(9)

(10)

Since this solution is only an approximation, it does not


describe the water-air interface correctly. However, this coarse
solution serves as an initial guess to the numerical solution
later.
2) Numerical Solution: The exact solution to (2) is solved
using Matlab, given two initial conditions of (3) and (4), and
the initial guess of (8). To implement (4), the infinity term
must be replaced by a specific distance that is sufficiently far
from x = 0. From eye-observation, the maximum width of
the water dimple was around 0.02 m, thus 0.01 m wide from
the axis of symmetry. Here, the infinity in (4) is replaced by
the specific distance of 0.02 m, or 2 cm, which is twice the

maximum size of the dimple. This value can be bigger, making


the solution more exact, but then the solution fails to capture
the profile of the water-air interface when crit .
Fig. 4 shows the numerically solved water-air interface
profile transition when a rigid cylinder with c = 112 is
lowered on to the water surface. The result is consistent with
eye-observations. The water surface is not yet broken if the
assumption on the water breaking condition made earlier is
correct. For the case of crit , the total lift force was 3.05
mN, and the depth of the center of the cylinder was 3.85 mm.
The radius of the cylinder is 165 m, which is the radius of
the wires used in the actual robot.
Fig. 5 shows the surface profile for various c when
crit . When c > 90 , the interface profile looks almost
identical regardless of the value of c . This result implies that
the maximum lift force will be almost the same if c > 90 ,
since most of the lift force come from the surface tension
force. For the case of c < 90 , the lower the contact angle
is (or the more hydrophilic the leg material is), the lower the
maximum lift force is. For a 2 cm-long cylinder with c = 60 ,

the maximum lift force was calculated to be 2.57 mN.


The numerical solution verified the results of [11] that the
weight of the water that would fill the volume B in Fig. 3
would equal the vertical component
of the surface tension
R
force, fT . The value of g x0 h(x) dx was very close to the
value of sin(tan1 ( dh
when c > 90
dx (x0 ))). Specifically,
R
and crit , the value of g x0 h(x) dx almost exactly
equaled the surface tension coefficient of 0.072 N/m.
It was calculated by observing h(x) that when two supporting legs that are parallel to each other are pushed down
until just before they break the water surface, they lose some
lift force due to the water dimples created by each legs
overlapping with each other between the legs. For two plain
stainless steel legs, they have to be at least 21.0 mm apart in
order to lose no more than 1% of the possible maximum lift
force. To conserve at least 95% of the lift force, the legs have
R
to be 12.4 mm apart. For the Teflon
coated legs, they have to
be at least 21.3 mm apart to conserve 99%, or 12.6 mm apart to
conserve 95% of the possible maximum lift force (Table II).
All these results are assuming only two legs in parallel. If
there are more than two legs in parallel, the percentage of the
lift force conserved drops down due to more than two water
dimples overlapping with one another.
All these results assumes a perfectly rigid cylinder with
no considerations of interface profile at the two ends of the
cylinder. In actual case, there exists some end-effects that may
or may not be important. Since the legs of the robot are long
and thin, these effects are neglected in this simulation. The
sharp edges at the ends of the cylinder may break the water
surface even before the condition crit is satisfied. In the
actual experiment, the ends of the cylinder are bent upwards
so that the geometry of the leg at the ends is round enough.
3) Experiments: Fig. 6 shows the numerically solved c
vs maximum lift force relation, for the geometry of specimen
shown in the inset, using the experiment setup as shown in
TABLE II
E FFECT OF THE S PACING B ETWEEN T WO S UPPORTING L EGS

Leg Material

Spacing

Conserved Lift Force

Stainless steel

21.0 mm

99%

Stainless steel

12.4 mm

95%

R
Teflon

21.3 mm

99%

R
Teflon

12.6 mm

95%

TABLE III
E XPERIMENTAL DATA

Leg Material

Stainless steel

50 70

R
Teflon

FluorothaneTM MP

110

Maximum Lift Force [mN]


(Numerical Estimation)
2.6 (2.3 3.0)

120

3.1 (3.6)

140 150

3.2 (3.7)

Fig. 6. Numerically solved maximum lift forces of the specimen in the inset.

Fig. 7. Photo of the experiment setup using an xyz stage and a load cell,
for lift force measurement for the specimen geometry given in Fig. 6 inset.

Fig. 7. The angled ends of the specimen are taken account


into the model in such a way that they are made of cylinders
of length dx with different depths of their center into the
water. Actual experiments were performed using three differR
ent leg materials; stainless steel wire, Teflon
coated wire
TM
and Fluorothane MP (Cytronics Corp., model 1.00) coated
R
wire. The stainless steel wire and the Teflon
coated wire
TM
had the same radius, but the Fluorothane MP coated wire
was expected to have a larger radius because of the manuallyperformed coating. Table III shows the experimental data for
the three specimen. They are pushed down onto the water
surface automatically with the speed of 0.3 mm/sec, using a
desktop computer. The numerical solution captures the general
trend of the experimental data fairly well, but has 10-15%
errors for the c > 90 cases. This may be due to the fact that
the numerical solution assumes a perfectly static case, whereas
the experiments are performed at a certain speed. Also, the
model assumes a perfectly rigid cylinder, whereas the actual
specimen is not so rigid. The angled ends of the specimen
might have forced the water surface to break more easily.
The numerical solution, along with the experimental results,
suggests that the maximum possible lift force does not differ
much if the leg materials are hydrophobic. That is, any
material hydrophobic enough can make a good supporting leg.

center of the rigid cylinder is also calculated, and many data


points describing the depth versus the lift force relation can
be obtained. These data points were used to fit a curve, f (z),
R
that captures this relation. For example, for Teflon
coated
wire with the radius of 165 m and the contact angle of 112 ,
the lift force per millimeter can be approximated as
f (z)
Fig. 8.
i < j.

1.1488 103 z 3 + 4.3505 102 z 2


5.5535 102 z + 3.1837 106 , (11)

Compliant supporting leg divided into N segments. |fi | > |fj | if

Fig. 9.

Parameters used for determining the flexibility matrix, F.

C. Compliant Leg Modeling


From the analysis in the previous section, the theoretical
amount of lift force for a rigid cylinder body can now be
numerically solved. Applying the result, the robot with four
5 cm-long supporting legs (only 4 cm of each legs actually
R
touching the water) coated with Teflon
will experience
maximum lift force of 2.44 grams. However, experiments
showed that the robot can actually hold 1.82 grams including
its body weight [10]. The difference in the values may be
due to the leg being compliant, instead of it being completely
rigid. Fig. 8 shows the compliant supporting leg pushing down
the water surface. Since the amount of lift force per unit
length is a function of the depth of the cylinder, the end-tip
of the supporting leg experiences less lift force per unit length
compared to the part of the leg that is closer to the base. To
find more realistic value of total lift force, this difference in
lift forces along the length must be taken into account.
1) Numerical Solution: To find the exact deformed shape
of the compliant supporting leg, the leg is divided into N
equal-length segments along the length, as shown in Fig. 8.
Each segment experience some lift forces, which are assumed
to be point forces acting at the center of the segments (fi s
in Fig. 8). These point loads are all the external forces that
deform the leg. Each segment is assumed to be rigid enough so
that it can be assumed as a cylinder whose axis of symmetry
is parallel to the undisturbed surface of the water (z = 0
surface). The depth of the position where the load is applied
determines the magnitude of the point load. Given the radius R
and the contact angle c of the segments, the exact theoretical
relation between the lift force per unit length of the cylinder
and the depth of the center of the cylinder can be known from
the results of the previous section. That is, when calculating
the magnitude of the lift force given , the depth of the

where z is the depth of the center of the cylinder in meters,


and f (z) is in Newtons. The coordinate system is shown in
Fig. 8, which is consistent with Fig. 2.
Now, let the fixed value z0 be the depth of the base of the
supporting leg, and set an alternative xyz 0 coordinate system
at the base of the supporting leg (Fig. 8). The alternative coordinate system is a translation of the original coordinate system
0
T
by z0 in z direction. Define a vector z0 = [z10 z20 zN
+1 ] ,
0
where zi is the depth of the point where fi is applied, in xyz 0
coordinate system. Then z0 describes the depth profile of the
supporting leg. Let the vector g(z0 ) = [f1 f2 fn ]T =
0
T
[f (z10 z0 ) f (z20 z0 ) f (zN
represent the
+1 z0 )]
point forces that the segments experience. Then the problem
of solving for z0 can be formulated as
z0 = l F g(z0 )

(12)

where l is the length of each segment and F is the flexibility


matrix of the leg, whose n-th row describes the depth profile of
the leg (in xyz 0 coordinate) when a unit force is applied only
to the position where fn would be applied. All the elements
in the first row and the first column of F are zero, since the
position where f1 is applied is the base of the supporting leg
(Fig. 8). (The position of the base of the leg does not change
regardless of the magnitude of any vertical force applied to
the point.) Having all the elements in the first column of this
matrix to be zero allows the rows of the flexibility matrix
to be superimposed and still have the first element of the
superimposed vector to be zero. The flexibility matrix, F, is
determined by

x2j

if 0 xj aj ,

6EI (3ai xj )
(13)
Fi,j =

(3xj ai )
if aj xj L,
6EI
where E is the elastic modulus of the leg material, I is the
moment of inertia of the cross section about the x-axis, L is the
length of the supporting leg, ai is the distance between the base
of the supporting leg and the point where fi is applied to, and
xj is the distance between the base of the supporting leg and
the point where fj would be applied to (Fig. 9). Note that since
x1 = a1 = 0, Fi,1 and F1,j are all zero, as intended. Equation
(13) is derived assuming small deflection of the leg. This is a
feasible assumption since the length of the leg, which is greater
than 4 cm, is more than 10 times bigger than the possible
maximum depth of the base of the leg, which is approximately
R
3.8 mm for Teflon
coated wires.

Fig. 10. Numerically calculated profile of the compliant supporting leg


R
made of Teflon
coated stainless wire, when z0 = 3.85 mm. The z 0 -axis is
magnified.

Applying the initial guess of z0 = [0 0 0]T , (12) is


solved. Fig. 10 shows the bended supporting leg profile derived
from solving (12). The leg material is stainless steel coated
R
with Teflon
, and the initial depth of the base is 3.85 mm,
R
which is the maximum depth possible for Teflon
coated leg,
estimated in the previous section. When the length of the leg is
60 mm, the lift force in this case is calculated to be 6.88 mN,
or 0.70 grams. Same calculation process was applied for plain
stainless steel legs of same dimensions, and the maximum lift
force was estimated to be 6.43 mN, which is not much of a
loss in the loading capacity. Even though plain stainless steel
has smaller contact angle and is considered hydrophilic, the
R
coated
moment of inertia is far greater than that of the Teflon
stainless steel wire of the same diameter, making the leg
stiffer to increase the maximum payload. For four 4 cm-long
R
Teflon
coated supporting legs, the maximum payload was
expected to be 2.32 grams according to this model, which
is about 5% less than the model assuming perfectly rigid
cylinder.
Fig. 11 shows the numerically calculated payloads (lift force
minus the weight of the leg) for various lengths of the
supporting legs with the radius of 165 m unless otherwise
specified. The maximum lift force for the compliant legs are

Fig. 11. Numerically calculated payloads for legs with different geometry
and materials.

R
Fig. 12. Photo of the prototype robot with sixteen 4 cm-long Teflon
coated
straight supporting legs whose bases are 4 mm apart, carrying a 5 gram weight.
The body is made of the carbon fiber sheet.

capped at around 6.9 mN. When the lengths of the legs are
less than 4 cm, no significant difference is present between
the rigid leg and the compliant leg. However, when the length
of the leg is greater than 5 cm, the difference increases and
the maximum lift force for the compliant leg saturates. This
result suggests that the supporting legs should be no longer
than 5 cm in order for them to be most effective. It would be
better to have large number of short legs instead of having a
few long legs. However, if the radius of the legs becomes 0.2
mm, the supporting legs become stiffer and they can support
more weight even when they are longer than 5 cm. The weight
added due to the greater radius was proved to be insignificant.
2) Experiments: As mentioned earlier, the maximum lift
force for the robot with four 4 cm-long supporting legs was
1.82 grams [10], which is around 78% of the numerical
estimation of 2.32 grams. Fig. 12 shows the sixteen-legged
R
prototype robot with 4 cm-long Teflon
coated legs. The
maximum lift force in this case was 6.31 grams, or 61.8 mN.
Using the results from the previous sections, it is predicted
that this prototype can hold 92.8 mN if the legs are spaced far
enough. When the legs are 4 mm apart at the bases and 16 mm
apart at the ends as in this case, it is roughly estimated that
the prototype can hold up to 7.27 grams. The estimation was
done in the following steps; the legs with 4 mm spacing can
only hold 23.1% of the maximum lift force, whereas if they
are 10 mm apart, they can hold roughly 92% of the maximum
lift force. If they are 16 mm apart, as the tips of these legs are,
then they can hold almost 100% of the estimated lift force.
Now since the sixteen supporting legs are 4 mm apart only at
the base and 16 mm apart at the ends, the prototype would
carry roughly 0.25 (100 + 92) + 0.25 (92 + 23.1) = 76.8% of
the possible lift force of 92.8 mN, which is calculated to be
71.3 mN, or 7.27 grams. Applying this result, the experimental
value is showing 86.8% of the numerical estimation. More
deliberate estimation of the possible lift force is left for future
research.
The difference between the numerical estimation and experimental result may be explained in a few ways. Firstly, there

Fig. 13.
Numerically calculated profile of a 7 cm-long supporting leg
experiencing maximum lift forces at each segments.

Fig. 14. Numerical simulation of the optimized supporting leg straightening


up as it is pushed down to the water surface.

are fabrication errors; all sixteen of the legs must be identical


to one another so that they all touch the water at the same time
and break the water surface at the same time. If the legs were
not consistent, one of the legs would break the surface before
others take their possible maximum lift force, reducing the
possible overall maximum lift force. The more you have the
legs, the more this is likely to happen. And some misconduct
of the experiment could also affect the result. When placing a
weight on the test robot, the water surface is suddenly pushed
down and the water surface may become easier to break due
to this dynamic effect.
It is clear that all these sources of errors can also exist in
the actual robot. The robot will be able to carry more weight
in the future if these issues are resolved.
It was observed that when a 7 cm-long leg was fully loaded,
the end tip of the leg was outside of water. At these ends, the
water surface was bended up toward the +z direction, thus
pulling the leg downwards. It is a good idea to design the legs
such that this never happens.

with four legs was able to carry 3.7 grams including the
body weight, which is 86% of the theoretical maximum of
4.3 grams. Seeing that a 7 cm-long compliant leg without
any initial bending can support less than 7 mN (or 0.7
grams) from Fig. 11, this optimizing of the leg improved
the maximum payload by 32%. (Theoretically, it would be
53.6% improvement.) The prototype with twelve legs withheld
9.3 grams, which is only 71% of the expectation, but still
better than the expected performance of the legs without initial
bendings (which is 8.4 grams). It is therefore concluded that
longer legs can be used and still be effective if they are
optimized to maximize performance.
When these optimized legs are loaded far less than the
maximum possible payload, only the ends of the supporting
legs were touching the water surface, just like the insect water
strider.

D. Optimized Legs
Fig. 11 suggests that compliant legs lose loading capacity
at longer lengths. This is because when the leg is pushed
down onto the water surface, it is bended upwards due to
compliance. If the legs could be initially bent downwards
to overcome for this upward bending which decreases the
payload, then compliant legs can be longer and still be
effective.
1) Profile Calculation: The best possible scenario is to have
a compliant leg that straightens up at the maximum loading.
Fig. 13 shows the bended profile of a compliant leg (7 cmlong) when each of the segments (as described in previous
sections) are experiencing z direction forces of 1.5 mN/cm
(or 0.15 g/cm), which is the maximum possible lift force
R
calculated for Teflon
. A leg with this profile is expected to
be straightened up when pushed down onto the water surface,
just before breaking it. Fig. 14 shows numerically estimated
profile change as this leg is being pushed down onto the water
surface.
2) Experiment: Fig. 15 shows the optimized 7 cm-long
R
Teflon
coated stainless steel wire, whose profile is as shown
in Fig. 13. When pushed down to the limit onto the water
surface, the leg straightened up with the uniform depth as
intended, as shown in Fig. 16. Prototype robots were built
using four and twelve of these legs to test the loading capacity.
The one with twelve legs is shown in Fig. 17. The prototype

III. D ISCUSSIONS
From the numerical analyses in the previous sections, a
specific design rules for the supporting leg of the water strider
robot can be summarized as follows:
1) Maximum lift force is almost linearly proportional to the
length of the legs, as long as they are shorter than 4 cm

Fig. 15.

R
Photo of the 7 cm-long optimized Teflon
coated leg.

R
Fig. 16. Photo of the 7 cm-long optimized Teflon
coated leg, straightened
up just before breaking the water surface.

stands up vertical ( dh
dx (x0 ) = ), it can be asserted that this
assumption made earlier is correct.
IV. C ONCLUSION

Fig. 17. Photo of the prototype robot with twelve optimized 7 cm-long
R
Teflon
coated legs, carrying 8.3 gram payload (total of 9.3 gram lift force).
R
for stainless steel or Teflon
coated stainless steel wire
with the radius of 165 m. To use legs longer than 7
cm, optimizing their profile is necessary. In both cases,
one can expect 0.15 grams of payload per centimeter for
R
the Teflon
coated legs (Fig. 11).
2) Number of supporting legs are directly proportional to
the maximum lift force, as long as the legs are not so
close to one other. The minimum distance between the
R
neighboring supporting legs coated with Teflon
, must
be at least 10.2 mm apart so that the dimples they create
would not overlap too much and lose possible maximum
lift force (Table II).
3) Leg material having c much greater than 90 is desired,
but when c > 120 , the advantage diminishes (Fig. 6).
4) The end tip of the supporting legs should be bent
upwards so to prevent any singularity of geometry,
which may cause early breaking of the water surface.
5) The radius of the supporting leg is insignificant to the
magnitude of the surface tension force. However, the
smaller radius of the leg is better for a surface tension
dominated support and to minimize overall weight of
the robot.
Abiding by these rules and reducing the fabrication error
may increase the loading capacity of the robot by 20 30%.
Throughout the analyses, the numerically calculated maximum lift forces are greater than the experimental results by
1030%. Considering the errors in the fabrication process and
experiments, these differences are not surprising. Therefore,
it can be concluded that the numerical models capture the
physics behind the supporting legs and the water surface reasonably well. Since all the analyses starts from the assumption
that the water surface breaks when the water-air interface

Motivated by the recent studies on the insect water strider,


a numerical method to evaluate the synthetic supporting legs
is proposed. Under an assumption on the water-breaking
condition, supporting legs of different materials and shapes
are analyzed. It is shown that a robot with four 7 cm-long
R
Teflon
coated supporting legs with optimized shapes would
have loading capacity of 4.3 grams (0.15 g/cm), including its
own body weight. The design rules proposed in the previous
section can be used to design future robots. And the analyses
procedure can be helpful in predicting the maximum loading
capacity of the robot, or in designing a better geometry or
properties of the legs. In the future, dynamic analysis on
the actuating legs and their behavior on the water surface
would also be realized. When all the physics of the robot are
analyzed, the robot can potentially be used for the purposes
of entertainment, education or environmental monitoring on
water surfaces.
R EFERENCES
[1] C. Menon, M. Murphy, and M. Sitti, Gecko Inspired Surface Climbing
Robots, Proceedings of the IEEE International Conference on Robotics
and Biomimetics (ROBIO), pp. 431-436, 2004.
[2] F. Chiu, J. Guo, J. Chen, and Y. Lin, Dynamic Characteristic of a
Biomimetic Underwater Vehicle, Proceedings of the IEEE International
Symposium on Underwater Technology, pp. 172-177, 2002.
[3] J. E. Clark, J. G. Cham, S. A. Bailey, E. M. Froehlich, P. K. Nahata,
R. J. Full, and M. R. Cutkosky, Biomimetic Design and Fabrication
of a Hexapedal Running Robot, Proceedings of IEEE International
Conference on Robotics and Automation, Vol. 4, pp. 3643-3649, 2001.
[4] N. Hasegawa and H. Kazerooni, Biomimetic Small Walking Machine,
Proceedings of the IEEE/ASME International Conference on Advanced
Intelligent Mechatronics, pp. 971-979, 2001.
[5] U. Saranli, M. Buehler, and D. E. Koditschek, Design, Modeling and
Preliminary Control of a Compliant Hexapod Robot, Proceedings of the
IEEE International Conference on Robotics and Automation, pp. 25892596, 2000.
[6] D. Spenneberg, K. McCullough, and F. Kirchner, Stability of Walking
in a Multilegged Robot Suffering Leg Loss, Proceedings of the IEEE
International Conference on Robotics and Automation, pp. 2159-2164,
2004.
[7] R. S. Fearing, K. H. Chiang, M. H. Dickinson, D. L. Pick, M. Sitti,
and J. Yan, Wing Transmission for a Micromechanical Flying Insect,
Proceedings of the IEEE International Conference on Robotics and
Automation, pp. 1509-1516, 2000.
[8] D. L. Hu, B. Chan, and J. W. M. Bush, The Hydrodynamics of Water
Strider Locomotion, Nature, Vol. 424, pp. 663-666, 2003.
[9] X. Gao and L. Jiang, Water-repellent Legs of Water Striders, Nature,
Vol. 432, pp. 36, 2004.
[10] S. H. Suhr, Y. S. Song, S. J. Lee, and M. Sitti, Biologically Inspired
Miniature Water Strider Robot, Proceedings of the Robotics: Science and
Systems I, pp. 319-325, 2005.
[11] J. B. Keller, Surface Tension Force on a Partially Submerged Body,
Physics of Fluids, Vol. 10, pp. 3009-3010, 1998.

S-ar putea să vă placă și