Professional Documents
Culture Documents
Abstract We present two foot mechanisms that allow relatively large patches of synthetic fibrillar dry adhesives applied
inexactly by a climbing robot to perform at levels previously obtained only for small samples in precisely aligned and controlled
bench-top tests. The mechanisms are inspired by the structures
found in the toes of the gecko. The first mechanism uses ankles
with roll and yaw flexures and a compliant structure behind the
adhesive material to achieve approximately uniform pressures
under nominal loading conditions on flat and curved surfaces.
The second design uses a tendon-supported structure to achieve
uniform loading and prevent premature peeling failures despite
significant misalignment with a flat wall surface. The two
designs are demonstrated on Stickybot III, an approximately 1
kg climbing robot, and can be scaled to larger areas and loads
by tiling the basic structure.
I. INTRODUCTION
Recently, gecko-like synthetic adhesives (GSA) have been
the subject of extensive research. Much of this interest
is due to the exceptional climbing abilities of the gecko,
which is capable of rapid climbing on rough, smooth, wet,
and even inverted surfaces. To accomplish these feats, the
gecko employs a hierarchical system of lamellae, setae, and
spatulae, -keratin structures on its toe-pads with dimensions on the millimeter, micrometer, and nanometer scale,
respectively [1], [2]. A bibliography of developments in
synthetic adhesives can be found at [3], which reveals steady
improvements in adhesive pressures. Recent efforts also seek
to emulate the directional behavior of the geckos adhesive
system [4], [5], [6], [7], [8], [9], [10] and its ability to
conform to textured surfaces using a hierarchy of compliant
structures [7], [11], [12], [13].
Much less attention, however, has focused on applying
adhesives to climbing. Climbing presents different challenges
than creating samples of material that perform well on a stiff,
calibrated test stage or in carefully controlled hang tests;
alignment, large patch sizes, moments, and peel propagation
are all formidable obstacles. Nonetheless, several groups
have demonstrated robots that climb with micro-structured
dry adhesives [10], [14], [15]. In each case, the robot obtains
much lower adhesive pressures than possible under ideal
conditions. For example, the 0.3 kg Stickybot [10] required
16 cm2 of Directional Polymer Stalks (DPS) in contact,
which could otherwise support up to 5 kg in a controlled
vertical hang test. Similarly, Murphy et al. [14] note that
carrying payloads over 0.5 kg is impractical for the Waalbot
design although the adhesive would permit significantly
higher loads under ideal conditions.
5100
a.
35m
b.
3cm patch of microwedges
perfectly aligned plate and the ability to accommodate misalignment of a suspension when climbing smooth surfaces.
2) Applying Load Evenly: If the alignment problem were
solved such that every fiber was perfectly engaged, the next
task would be to distribute the adhesive load evenly among
all fibers. Stress concentrations must be avoided as they lead
to local areas of failure; the load on the adjacent fibers then
increases and the result is a peeling failure. The tendency
for peeling failures to propagate rapidly across the pad is
particularly severe when the pad is loaded in an applied
force mode, as in the case of a hanging weight or robot,
versus an applied displacement mode, as in the case of
a stiff, controlled testing stage. In the latter case, with a
compliant adhesive system, the failure will progress only
until a new equilibrium is achieved.
When climbing, three conditions of peeling must be
avoided: peeling by backing layer2 bending, peeling by
moments created by a load offset, and peeling by moments
transferred through the foot.
Backing layer bending occurs when the load is applied
to a point on the backing layer, e.g. by a tendon, but the
backing layer is not stiff enough to transfer this force evenly
to the entire area of the pad. If the pad is loaded at the
center, this will cause an outward peeling. If we assume a
load at the center of the pad, then for optimal performance
(all fibers receiving an even load), the center of the backing
layer must deflect less than 35 m away from the wall when
fully loaded to ensure no fiber disengages prematurely.
bending
load
adhesion
Fig. 3: The backing layer will bend when loaded at the center,
but pulled toward the wall by adhesion. This bending can be
modeled as a clamped-fixed beam with even loading.
To calculate the required stiffness of the backing layer, the
structure can be modeled as a beam, loaded towards the wall
by an even adhesive pressure distribution and away from the
wall by a point force at the center (Fig. 3). The maximum
deflection away from the wall is then given by Eq. 1. If the
load, F, is evenly distributed over the entire area, then the
pressure as force per unit length is w = F/2L. Using forces
and sizes from Stickybot III, F = 1N, L = .018m, E = 21GPa
(glass fiber composite), and y = 35 m, the thickness of the
backing needs to be at least 1.1 mm.
wL4
(1)
8EI
The second mode of peeling is due to moments created by
a load offset. Here, the offset d between the load vector and
y=
2 The layer behind the adhesive is termed the backing layer. It can be
rigid as in the tendon-inspired design, or rigid with a layer of compliant
material connecting it to the adhesive, as in the compliant design.
5101
Fluid-filled sac
Pm
Fsin(!")
+
l
!"
Fcos(!")
Lamellae
Lamella
(b)
(c)
B. Bioinspiration
The design begins with bioinspiration from the geckos
foot. To simplify the functional anatomy for climbing in
the gecko, there are four distinct sections of the foot that
play important roles: the adhesive-covered lamellae, the
phalanges, the fluid-filled sac, and the lateral digital tendon
(Fig. 5).
Lamellae- These long, slender flaps each have a small
section of adhesive at their ends, which are free to orient
at a variety of angles since they are supported only by
a thin section of skin.
Phalanges- The bones of the toe provide the pre-load
for pressing the adhesive to the surface before loading.
Fluid-Filled Sac- This layer is between the phalanges
and the lamellae, and transfers the pre-load from the
former to the latter. It acts as a universal joint, allowing
a large range of bone angles to press the pad flat against
the surface. Further, it bears little shear load, allowing
this to fall to the tendon.
Lateral Digital Tendon- This structure acts as the load
bearing member in the finger. Once the phalanges and
sac have preloaded the lamellae, the tendon, which
branches into each individual lamella, transfers the
geckos weight to the adhesive.
III. D ESIGNS
F sin() F cos()dh
Pm =
a
2I
h
F m = Pa a
6dcos() + hsin()
(2a)
(2b)
F0
(1 Fm
)tan()
d
m
=
0
F
m
h
6 Fm
(2c)
5102
adhesive
a.
joint
b.
hole
riser
tendon
A
tendon
c.
process
Section A-A
phalanx
strut
backing
Fig. 6: Top left: Two-axis foot flexure design allows for roll
and yaw misalignments of up to 8 while climbing. Right:
The flexure mechanism. Bottom left: The suspension system
supporting a layer of microwedges (See Fig. 2).
anvil
strut
d.
e.
anvil
skin
f.
5103
A. Gait Tuning
3 The tail is still used on Stickybot III when climbing, to help keep the
body parallel to the wall during the swing phase. However, once feet are
engaged, the tail is not required.
Gait fig
Wall
a. normal
preload
b.
shear
loading
a.
b.
e. swing
e.
d.
c.
c. grip
d. shear
unloading
Side view
Fig. 8: Left: The force trace of a single foot during one gait
cycle of StickybotIII. Right: The programmed gait.
A positive normal preload is created by commanding the
two incoming feet to move into the wall (Fig. 8, a.). Next, the
detaching and attaching feet are pulled toward one another
along the wall to transfer the shear load (Fig. 8, b.). Once the
detaching feet no longer carry a shear load, they are pulled
away from the wall with almost zero detachment force. The
attached feet remain in the grip phase (Fig. 8, c.). They
transfer the shear load back to the now incoming feet, (Fig.
8, d.), then swing to reset (Fig. 8, e.).
5104
F.S.
= compliant foot
= tendon-inspired foot
= stage data
F.S.
5105
load evenly to every fiber, preventing any force concentrations or moments that might load one section of the pad more
than others.
Two bio-inspired foot designs are presented that meet the
above requirements. The second of these creates a limit
surface that matches that produced by a small patch on
a three-axis stage. While shown on Stickybot III using
Stanfords microwedges, this concept is not limited to this
platform or adhesive; any legged climbing device using
a directional adhesive could achieve performance from its
adhesives similar to that found on a stage.
Scaling the size of the operator is the next step for the foot.
As mentioned in Section V-D, the foot could theoretically
scale to a 338 kg robot, while retaining the same safety
factor and ratio of foot area to body mass. Of course, scaling
the feet to this size (from the current two 1.7 cm x 1.7 cm
patches per foot to two 13.6 cm x 13.6 cm patches) may
not be feasible due to the manufacturing tolerances that are
required. For the 13.6 cm x 13.6 cm patch to work, the
backing layer must be flat to within 35 m. However, if this
is not possible, initial experiments with tiling the pads have
proven promising. An 8x8 array of the current feet (or 3x3 of
a slightly larger version) with differential loading to evenly
distribute the forces among the feet could act like many feet
in parallel, all controlled from a single connection point.
VII. ACKNOWLEDGMENTS
This work is supported in part by DARPA-Zman and NSF
NIRT. E. W. Hawkes works with Government support under
and awarded by DoD, Air Force Office of Scientific Research, National Defense Science and Engineering Graduate
(NDSEG) Fellowship, 32 CFR 168a.
R EFERENCES
[1] K. Autumn, Y. A. Liang, S. T. Hsieh, W. Zesch, W. P. Chan, T. W.
Kenny, R. Fearing, and R. J. Full, Adhesive force of a single gecko
foot-hair, Nature, vol. 405, no. 6787, pp. 681685, 2000.
[2] A. P. Russell, Descriptive and functional anatomy of the digital
vascular system of the tokay, gekko gecko, Journal of Morphology,
vol. 169, pp. 293323, 1982.
[3] R. Fearing, Gecko Adhesion Bibliography, 2011 (updated Dec
2010). [Online]. Available: http://robotics.eecs.berkeley.edu/ronf/
Gecko/gecko-biblio.html
[4] D. Sameoto and C. Menon, Direct molding of dry adhesives with
anisotropic peel strength, Journal of Micromechanics and Microengineering, vol. 21, pp. 227681, 2009.
[5] T.-I. Kim, H. E. Jeong, K. Y. Suh, and H. H. Lee, Stooped nanohairs:
Geometry-controllable, unidirectional, reversible, and robust geckolike dry adhesive, Advanced Materials, vol. 21, no. 22, pp. 2276
2281, 2009.
[6] A. Parness, D. Soto, N. Esparza, N. Gravish, M. Wilkinson, K. Autumn, and M. Cutkosky, A microfabricated wedge-shaped adhesive
array displaying gecko-like dynamic adhesion, directionality and long
lifetime, Journal of the Royal Society Interface the Royal Society,
vol. 6, no. 41, pp. 12231232, 2009.
[7] M. Murphy, S. Kim, and M. Sitti, Enhanced adhesion by gecko
inspired hierarchical adhesives, Applied Materials and Interfaces,
vol. 1, no. 4, pp. 849855, 2009.
[8] M. P. Murphy, B. Aksak, and M. Sitti, Gecko inspired directional
and controllable adhesion, Small, 2008.
[9] J. Lee, R. S. Fearing, and K. Komvopoulos, Directional adhesion
of gecko-inspired angled microfiber arrays, Applied Physics Letters,
vol. 93, no. 19, p. 191910, 2008.
5106