2013 IEEE/RSJ International Conference on
Intelligent Robots and Systems (IROS)
November 3-7, 2013. Tokyo, Japan
Dynamic Surface Grasping with Directional Adhesion
Elliot W. Hawkes, David L. Christensen, Eric V. Eason, Matthew A. Estrada, Matthew Heverly
Evan Hilgemann, Hao Jiang, Morgan T. Pope, Aaron Parness and Mark R. Cutkosky
Abstract— Dynamic surface grasping is applicable to landing
of micro air vehicles (MAVs) and to grappling objects in space.
In both applications, the grasper must absorb the kinetic
energy of a moving object and provide secure attachment
to a surface using, for example, gecko-inspired directional
adhesives. Functional principles of dynamic surface grasping
are presented, and two prototype grasper designs are discussed.
Computer simulation and physical testing confirms the expected
relationships concerning (i) the alignment of the grasper at
initial contact, (ii) the absorption of energy during collision
and rebound, and (iii) the force limits of synthetic directional
adhesives.
I. I NTRODUCTION
The ability to grasp flat or gently curved surfaces repeatably and releasably has several compelling robotic applications including the perching of micro air vehicles (MAVs) on
walls or ceilings (Fig. 1) and the grappling of orbital debris in
space. In both applications, low attachment and detachment
forces are required; however an important difference is that
MAVs are low mass and high velocity whereas orbital debris
typically has a larger mass and lower relative velocity.
Directional, gecko-inspired adhesives are suitable for these
applications because they require little energy for attachment
and detachment, work on many surfaces, can undergo many
attach/release cycles [1], and can be scaled to either small
or large applications [2]. Because the adhesives rely only
on van der Waals forces to stick, they are compatible with
spaceflight applications where nearly all pressure-sensitive
adhesives are prohibited because of radiation, temperature
and outgassing issues. Further, many of these applications
require the ability to attach and release with low force,
making gecko-inspired adhesives particularly appropriate.
The work presented here builds upon prior work on climbing robots, perching MAVs, and gecko-inspired adhesives.
Unlike a robot climbing a wall, which can control the
position, orientation and contact forces of its feet (e.g. [3],
[4]), the applications considered here involve either a grasper
or a target that is in free flight. The entire collision event
typically lasts less than 0.1 s between initial contact and
equilibrium (Fig. 2). As in other recent work [2], the devices
use directional adhesives mounted to arrays of rigid tiles,
E. W. Hawkes, D. L. Christensen, M. A. Estrada, H. Jiang, M. T. Pope,
and M. R. Cutkosky are with the Dept. of Mechanical Engineering, Stanford
University, Stanford, CA 94305, USA ewhawkes@stanford.edu
E. V. Eason is with the Dept. of Applied Physics, Stanford University,
Stanford, CA 94305, USA easone@stanford.edu
M. Heverly, E. Hilgemann and A. Parness are with the Robotic Vehicles and Manipulators Group, NASA JPL, Pasadena, CA 91109, USA
Aaron.Parness@jpl.nasa.gov
978-1-4673-6357-0/13/$31.00 ©2013 European
Union
Fig. 1. Quadrotor Micro Air Vehicle hanging from a glass surface using
the directional adhesive Collapsing Truss Grasper (Sec. III-A).
loaded with central tendons. This scheme ensures that the adhesive area is loaded evenly and no moments are transferred
from the device, which would cause stress concentrations
and premature failure.
Previous work on MAVs that can perch on walls and other
flat surfaces has exploited spines or arrays of spines [5], [6],
sticky materials [7], and dry adhesives [8]. The present work
is aimed at small rotorcraft that can fly at several meters per
second and takes advantage of directional adhesives capable
of sticking and releasing rapidly and with very low effort
[1], [9]. Work on the control of MAVs in confined spaces
(e.g. [10], [11], [12], [13]) is also relevant for establishing
the range of velocities and orientations that may be expected
at contact.
The problem of space debris is increasingly of concern to
space agencies around the world. There are currently over
1500 rocket bodies and over 10,000 other debris objects in
Earth orbit [14]. In 2007, a piece of debris collided with
an active communication satellite causing a total loss worth
many millions of dollars in damage. With the mechanisms
presented here, “non-cooperative” targets can be acquired,
in contrast to previous systems which have relied on preinstalled grapple features on cooperative targets. For example
the arms on the Space Shuttle and the International Space
Station. Similarly, the Orbital Express mission demonstrated
docking with a cooperative target [15], [16]. The FREND
arm is planned for use aboard the DARPA Phoenix mission
and is expected to grasp a non-cooperative target using
a Marman Clamp, a fixed hard point on the side of the
spacecraft [17]. The devices presented here do not require
specialized fixtures and can attach to flat or gently curved
smooth surfaces including solar panels and the sides of
spacecraft, fuel tanks, etc. They have the potential to simplify
orbital debris clearance, making it more robust and less
5487
B. Rebound Mitigation
The remaining kinetic energy of the grasper must be
absorbed during the collision or during rebound (Fig. 3, B).
The maximum energy that can be absorbed is limited by the
size of the device and the energy absorbing force. The energy
absorbing force is itself limited. During collision, it must not
damage the device; and during rebound, it must not exceed
the adhesion limits of the adhesive tiles.
C. Adhesive Loading
Fig. 2. Left: Illustration of the Collapsing Truss Grasper on a quadrotor
grasping a surface. Right: Screenshots from a high speed video.
reliant on precision sensing and navigation.
The following section of this paper presents a set of
functional principles for grasping surfaces under dynamic
conditions, when either the grasper or target is in free
flight. Next, prototype designs embodying these principles
are described. Modeling and testing results show that these
designs are capable of absorbing collision energy and using
it to align the surfaces and apply loads to the directional
adhesives, causing them to attach without bouncing away.
The paper concludes with a discussion of ongoing work
to incorporate these prototypes into MAVs and into space
grappling devices for environment testing to simulate orbital
conditions [18].
Unlike pressure sensitive adhesives, directional adhesives
are not sensitive to normal preload [19]: simply pressing
them into the surface will not make them stick. Directional
adhesives produce negligible adhesion unless shear force is
applied in the correct direction to turn the adhesive “ON”
(Fig. 3, C). In order to support normal loads without shear,
the grasper must use multiple tiles of directional adhesive
which are loaded with internal shear forces in opposing
directions.
With an appropriate mechanism, the energy of the collision
can be exploited to passively create these forces and turn
the adhesives “ON” at the appropriate time. Excessive shear
force will cause the directional adhesives to fail, so the
mechanism must ensure the shear force lies within acceptable
limits. The excess energy must be dissipated or stored
elsewhere. Alternatively, the forces may be produced by an
active mechanism. All adhesive tiles must be aligned and in
contact with the surface before the adhesives are loaded, so
an active mechanism must have accurate sensing to ensure
correct timing.
II. F UNCTIONAL P RINCIPLES
Grasping a surface dynamically requires several properties
for the gripper, whether for perching MAVs on a surface in
Earth’s gravity or grappling a target in space. This section
generalizes the problem of dynamic surface grasping and
describes several functional principles that must be embodied
by a gripper using directional adhesives.
A. Dynamic Passive Alignment
When the grasper first makes contact with the surface, it
is unlikely that the adhesive tiles will be aligned. Hence the
grasper must compensate for misalignment before or during
the collision (Fig. 3, A). A passive alignment system can be
lighter, simpler, and more robust than an actuated system.
For a passive system, it is important that the work required
for alignment is small compared to the grasper’s kinetic
energy in order to prevent rebounding before alignment has
occurred. The system should therefore have low moments of
inertia and rotational stiffnesses.
5488
Fig. 3.
Illustration of the Functional Principles described in Sec. II.
Rebound Spring
Compliant foam joint
D. System Locking
Once the internal shear force has been applied to the
adhesives and as much energy as possible has been absorbed
during the collision, the grasper must enter a locked state to
keep the internal shear forces in place and store the absorbed
energy. (Fig. 3, D). This can be achieved passively using a
ratchet or latch.
Truss Spring
Outrigger Adhesive
Tile Tendon
tile
Latch
Truss Tendon
MAV
Sliding joint
Tile
Support
Foam
E. Resistance to Arbitrary Wrenches
The grasper must be able to support arbitrary wrenches,
i.e. combinations of applied forces and moments (Fig. 3, E).
Ideally, the grasper mechanism should distribute these loads
optimally to limit the maximum force on the adhesive, so that
the grasper’s force limit equals the combined force limits of
the separate individual adhesive tiles.
This is not straightforward because the tiles are initially
misaligned on the surface, and their positions change during
the collision. Therefore, the grasper mechanism must compensate by taking up any slack in the loading tendons, and
it must distribute loads optimally despite this compensation.
F. Releasing the Grasp
For directional adhesives, it is not necessary to apply a
detachment force. When releasing the grasp is desired, a
release mechanism can disengage the system lock to release
the internal shear loads and turn the adhesives “OFF.” This
allows the stored energy, if any, to push the surface and
grasper apart (Fig. 3, F).
III. D ESIGN
Two designs are presented that display the functional
principles of dynamic surface grasping. The first, a collapsing
truss design, is sized for use on a MAV. The second, a
pivoting linkage design, has been sized and fabricated both
for use on a MAV and as a prototype for future use in Earth
orbit to grapple orbital debris.
A. Collapsing Truss Grasper
This grasper design is based on a collapsing truss mechanism (Fig. 4). It is designed as low-mass landing gear (3.5 g)
for a 120 g MAV, and uses 2 adhesive tiles (1×1 cm square).
To decrease the pitch-back moment when the MAV is
attached to a wall, the Collapsing Truss Grasper is designed
to be low profile in the collapsed position. The grasper
is designed in accordance with the functional principles
presented in Sec. II.
The truss is attached to the MAV at its apex by a single
tendon which passes through a compliant foam joint, which
keeps the grasper aligned to the MAV during flight but
allows it to rotate and translate during a collision. Translation
is necessary because one tile of adhesive makes contact
before the other, and the tiles resist sliding. The grasper
uses a set of outriggers to decrease the alignment force
and ensure it is partially aligned before contact (Dynamic
Passive Alignment). A model of this alignment system is
described in Sec. IV-A.
Fig. 4.
Collapsing Truss Grasper. A) Schematic showing functional
components. B) Device in locked state (grasping a surface).
As the truss collapses, the Truss Tendon routed between
the two legs of the truss becomes taut. This pulls the center
of the Tile Tendon against the bottom of the truss, applying
shear forces to the adhesive tiles and turning them “ON.”
The internal shear force is limited by the length of the Truss
Tendon (Adhesive Loading). Energy is absorbed during
the collision by the Truss Spring. When the truss collapses
fully, a latch engages to lock the truss in the collapsed
state (System Locking). If desired, the stiffness of the Truss
Spring can be adjusted to reduce the amount of normal force
required to collapse the truss to 0.3 N (see Fig. 10), which is
the minimum force needed to align the pads to the surface.
Extra energy is absorbed by the Rebound Spring, which is
attached to the tendon through the compliant foam joint
(Rebound Mitigation). This spring is preloaded in order
to keep the truss pulled tight to the MAV and because a
preloaded spring can absorb more energy in this situation
(see Sec. VI-C).
Once the grasper is locked in place, the Tile Tendon
remains under tension and stays at an essentially constant
angle, geometrically defined by the length of the Tile Tendon
and the distance between the tiles. When a large external load
is applied (e.g. wind on the MAV), this load is distributed
between the two tiles and additional tension is applied to
the Tile Tendon, adding more internal shear force, which
produces more adhesion due to the directional nature of the
adhesives (Resistance to Arbitrary Wrenches). The Tile
Tendon angle can be fine-tuned to change the performance
characteristics of the grasper, as described in Sec. IV-B.
B. Pivot Linkage Grasper
The other grasper design uses a pivoting linkage to apply
tension to the Tile Tendons. Unlike the Collapsing Truss
Grasper, the adhesive tiles are loaded with semi-independent
mechanisms, so the Pivot Linkage Grasper can have a larger
number of adhesive tiles. Two versions of this design are
presented, each using 4 adhesive tiles: The MAV Pivot
Linkage Grasper is designed as landing gear for a 120 g
MAV and uses 1×1 cm square adhesive tiles (Fig. 5); and
the Space Pivot Linkage Grasper is designed as a prototype
for grappling operations in Earth orbit and uses 4×4 cm
square adhesive tiles (Fig. 6).
5489
Fig. 5. MAV Pivot Linkage Grasper. A) Schematic showing functional
components. B) Device in locked state (grasping a surface).
to deflect the Tile Support Foam and align the tiles to the
surface. In the Space grasper, the Tendon Springs are linear,
but the Leadscrew allows the grasper to control the tension
as necessary: for example, a lower tension could be used
when grasping a rougher surface to prevent the adhesives
from failing prematurely, but a higher tension could be used
on a smoother surface to increase the grasper’s loadbearing
capacity.
Kinetic energy is absorbed by the Energy Absorbers and
locked in place using ratchets or a Ratcheting Nut. These
ratchet systems may lock at multiple points, which allows
the Pivot Linkage Graspers to absorb a variable amount
of energy during different collisions (unlike the Collapsing
Truss Grasper). In addition, the Energy Absorbers have
nonlinear stiffness to provide maximum deceleration in a
short distance (Sec. VI-C). A rebound spring may be added
to the MAV grasper to absorb additional energy; alternatively,
the Space grasper is intended to be mounted on a compliant
robotic arm which may be used for active rebound mitigation.
IV. M ODELING
A. Collapsing Truss Grasper: Passive Alignment
Fig. 6. Space Pivot Linkage Grasper shown in the locked state (grasping
a surface).
The mechanisms are actuated by pressing the Center Plate
and the Baseplate together. This causes the Tensioning Arms
to rotate around the Pivots and apply force to the Tile
Tendons through the Tendon Springs. The MAV version uses
tendons that pull inwards, crossing under the center of the
Baseplate for compactness, while the Space version uses
tendons that pull outward to enable grasping flexible surfaces
such as thermal blankets.
The MAV Pivot Linkage Grasper uses the energy of
collision to turn “ON” the adhesive tiles. It requires a larger
normal preload force than the Collapsing Truss Grasper to
apply the internal shear forces to the adhesive tiles. This
is partly because it has less mechanical advantage, but also
because the system of 4 tiles is over-constrained and therefore some amount of preload is necessary to deflect the Tile
Support Foam and bring all tiles into contact. Once the tiles
make contact, the Tendon Springs compensate for any initial
misalignment of the adhesive tiles. In the MAV grasper,
the Tendon Springs are preloaded and nonlinear, producing
a nearly constant force over a large range of deflection
to ensure that all 4 tiles are loaded evenly throughout the
collision.
The Space Pivot Linkage Grasper works similarly but can
also load the tendons after only a small collision by turning
its Leadscrew. This actively applies the shear load to the
adhesive tiles, so preload during the collision is only required
In order for a passive alignment strategy to work, the
grasper must be fully aligned to the surface before any
tension is applied to the Tile Tendons. For the Collapsing
Truss Grasper, this means the outriggers must apply enough
force to overcome the rotational inertia before the adhesive
tiles contact the surface, while not applying enough force to
cause the truss to begin to collapse.
In order to obtain a better understanding of the passive
alignment process, the equations of motion of this system
were modeled in Motion Genesis™, using the dimensions
and material properties of the physical Collapsing Truss
Grasper. Simulation results are shown in Fig. 7. For the
combination of initial conditions seen in high speed video
screenshots of Fig. 2, the simulation predicts that the grasper
aligns to the wall before the adhesive tiles contact the surface.
The maximum reaction force at the apex of the truss is about
0.38 N, which is lower than the force required to collapse
the truss, or 0.59 N. Thus the passive alignment process
completes before the truss collapses, which is also verified
by high speed video.
B. Collapsing Truss Grasper: Adhesive Loading
The behavior of a directional adhesive can be described
by a limit curve in force space, which is the locus of normal
and shear stresses that the adhesive can support before failure
[20], [19]. The limit curve of a grasper mechanism, however,
is a different shape than the limit curves of the individual
adhesive tiles (for instance, the grasper can support pure
normal loads while the adhesive tiles cannot).
In the case of the Collapsing Truss Grasper, a simple
model can be created to find the grasper limit curve using
the two adhesive tiles’ limit curves. The model is shown
in Fig. 8a. The load on the tiles is the resultant of the
tension force along each Tile Tendon at angle θ and the
compressive force through each Tile Support Foam piece.
5490
Force (N)
0.4
0.2
0
0
4
(a)
2
Force to collapse truss
1
Force to
align grasper
1
2
3
Time (ms)
4
Normal Force (N)
y (cm)
6
0.6
5
1.6 ms
3.3 ms
5.0 ms
2
θ = 13°
0
−1
Safe Region
(Smax, Nmax)
= (4, 1)
−2
−2
0
2
4
6
x (cm)
8
10
−3
−2
Fig. 7. Wireframe animation of simulated grasper model during passive
alignment with incoming velocity 1 m/s and angular misalignment 6.6°.
Inset: reaction force at apex of grasper (blue). Force to collapse truss shown
for comparison (green).
(a)
θ
(b) Nfoam
Sint
Smax
Smax
θ
Sint =
Nfoam
tan(θ)
Sint
Nfoam
Nmax
Nmax
Normal load
(c)
Smax
Shear load
0
1
2
3
Shear Force (N)
4
5
6
0
-1
Safe Region
-2
-3
-5
θ
−1
1
(b)
Normal Force (N)
0
-4
-3
-2
-1
0
1
Shear Force (N)
2
3
4
5
Fig. 9. A) Measured limit curve of a single 1×1 cm adhesive tile used in
the collapsing truss grasper. Adhesive forces are plotted as negative values.
B) Measured limit curve of the collapsing truss grasper, which uses two
1×1 cm adhesive tiles.
Nmax-Nfoam
V. FABRICATION
2Nmax
Fig. 8. A) Simple model of the collapsing truss grasper. B) Individual
limit curves of the two opposed tiles. C) Predicted combined limit curve of
the collapsing truss grasper.
Since the geometry is fixed (the tendons are inextensible), the
foam pieces produce a constant force Nf oam , and the forces
on the tiles are constrained to lie along a line segment in
force space with angle θ and intercept Nf oam , and which
intersects the limit curve at the point (Smax , Nmax ), as
shown in Fig. 8b.
If the assumption of constant geometry is valid, the combined limit curve of the grasper mechanism is the direct sum
of these line segments, which is a rhombus-shaped region in
force space (Fig. 8c). Along the lower edges of this rhombus,
one of the tiles is at its maximum load (Smax , Nmax ) while
the other is somewhere else on the line segment of possible
loads. The maximum normal load for the grasper is 2Nmax .
Adjacent to the upper edges of the rhombus are regions
in force space where one of the Tile Tendons is no longer
in tension. This does not necessarily mean grasper failure,
but the geometry is no longer expected to be constant so
this simple model is no longer accurate. These regions are
shaded gray in Fig. 8c.
The MAV graspers are fabricated using fiberglass and
acetal laser-cut parts, carbon fiber rods, silicone open-cell
foam, and kevlar braided cord. The Collapsing Truss Grasper
has dimensions 50×20×8 mm in the locked state. The Space
grasper is fabricated using 3-D printed parts (fused filament
fabrication), laser cut acrylic, braided line and other off the
shelf components.
The directional adhesive used in these mechanisms is fabricated by casting PDMS silicone into a mold created using a
photolithographic process [1]. This produces a 300-400 µm
thick film with an array of 80 µm tall angled micro-wedges.
A thin, smooth PDMS film is then deposited on the tips of the
features through a post-treatment process involving dipping
them into uncured PDMS and then pressing them against a
wafer [9], causing a change in shape and surface smoothness
on the engaging surfaces. After post-treatment, the back side
of the film is glued to a fiberglass sheet using RTV silicone
adhesive (Smooth-On Sil-Poxy), and the fiberglass sheet is
then cut into tiles using a laser cutter. Tendons made of kevlar
braided cord are attached to the front center of the tiles and
routed through rectangular cutouts, in a similar design to
adhesive tiles developed previously [2].
VI. R ESULTS
A. Collapsing Truss Grasper: Limit Curves
The limit curve of one of the adhesive tiles used in the
Collapsing Truss Grasper was measured using a motorized
5491
70"
2"
3"
No"Spring"
Linear"Spring"
60"
1.5"
2.5"
Non6linear"
Spring"
0.5"
0"
0"
0.5"
1"
1.5"
2"
Preload"Force"(N)"
2.5"
3"
Fig. 10. Graph showing the amount of adhesion force generated versus
normal preload force applied for the collapsing truss grasper. The adhesion
force is independent of preload beyond some very small threshold.
50"
Force&Applied&(N)&
1"
Rebound&Energy&(mJ)&
Adhesion"Force"(N)"
2.5"
40"
30"
20"
0"
100"
200"
300"
Collision&Energy&(mJ)&
B. Collapsing Truss Grasper: Preload
The maximum adhesion of the Collapsing Truss Grasper
with varying amounts of normal preload was measured by
pressing the grasper into a glass surface using weights of
various sizes, and then pulling the grasper perpendicularly
off the surface and recording the maximum normal force.
The Truss Spring was removed for this test to decrease
the force required to collapse the truss. This removes the
capability of the device to absorb energy during collision,
but this would be acceptable in a low-energy application.
The test data, plotted in Fig. 10, indicate that the normal
preload has no observable effect on the maximum adhesive
load: the grasper can be used with any preload in the
range of 0.3 to 2.4 N with an essentially constant normal
adhesion (2 N). However, a preload smaller than 0.3 N is
not sufficient to engage the mechanism’s latch. The adhesive
tiles themselves appear to be very insensitive to preload, as
has been previously shown [19].
C. MAV graspers: Energy absorption
An experiment was conducted to investigate energy absorption during collision. The MAV Pivot Linkage Grasper
was tested by dropping it onto a surface and measuring the
rebound height (the adhesives were removed for this test).
This experiment was repeated with different incoming kinetic
energies and with three different Energy Absorbing configurations: linear springs with a ratchet, nonlinear springs (near
1.5"
1"
0.5"
10"
0"
positioning stage and a six-axis force-torque transducer,
using methods described in [9]. This limit curve is plotted in
Fig. 9, A. Note that the limit curve goes through the origin,
indicating that no adhesion is produced without shear force.
By drawing a line segment in Fig. 9, A at the angle of the
tendons (measured to be θ = 13 deg), approximate values of
Smax = 4 N and Nmax = 1 N can be determined. Based on
the simple model of Fig. 8, it is predicted that the grasper’s
combined limit curve will have a maximum normal force of
2Nmax = 2 N and a maximum shear force of Smax = 4 N
(or more). Next, the combined limit curve of the grasper was
also measured; the resulting data are shown in Fig. 9, B.
While there is some amount of asymmetry due to variations
between the two adhesive tiles, the maximum normal and
shear forces agree well with the predictions from the simple
model.
2"
400"
0"
0"
10"
20"
30"
Overload&Stretch&Displacement&(mm)&
Fig. 11. Left: Rebound energy vs. collision energy for three different energy
absorbing configurations. Right: Force vs. displacement of the Rebound
Spring.
constant-force) with a ratchet, and no springs. These data are
plotted in Fig. 11, left.
For nearly all incoming kinetic energies tested, the rebound energies of the three configurations were in a specific
order, with the no-spring configuration being highest and
the nonlinear spring being lowest. This indicates that the
nonlinear spring absorbs the most energy during collision of
all the energy absorbing configurations.
To compare this energy with the energy absorbed during
rebound, a force vs. displacement curve was measured for the
Rebound Spring used by the MAV graspers (Fig. 11, right).
Given the maximum normal load the grasper can support
(e.g. 2 N for the Collapsing Truss Grasper), the maximum
permissible rebound energy is the integral of the curve up
to the maximum load (red region in Fig. 11), which is
approximately 29 mJ. This energy is plotted as a red dashed
line in Fig. 11, left. This plot can be used to approximate the
maximum permissible incoming kinetic energy for the three
damping configurations, with the result that the nonlinear
spring allows nearly twice as much kinetic energy as the
linear spring and 4 times as much as no spring.
D. Pivot Linkage Grasper: Scalability and Dynamic Capture
The scalability of grasper designs was investigated by
testing the Space Pivot Linkage Grasper at higher loads. Pull
tests were performed on a variety of surfaces using a digital
force gauge. Successful grasping was demonstrated on a variety of surfaces including satellite solar panels. A maximum
normal adhesive force of >60 N was demonstrated. Tests
were also performed to verify the load-sharing function of the
tendons with carefully selected tiles. The normal adhesion
of the grasper was found to be four times as large as the
lowest-performing tile. When a set of opposing tiles were
intentionally separated from the surface, hard stops in the
mechanism allowed the remaining two tiles to demonstrate
a contact strength of twice the lowest-performing tile.
Dynamic capture tests were performed with a human
holding the grasper system in place of the FREND robotic
5492
arm that will be used in future testing [17]. Using a simulated
piece of debris (a 33 kg foot locker with wheels and several space-like surfaces mounted), capture experiments were
performed at a variety of relative velocities and spin rates.
The maximum successful capture had relative motion of over
2 m/s and a spin rate of >75 deg/s. This demonstration is
included in the paper’s accompanying video.
VII. C ONCLUSION
The surface grasper designs developed here are practical
and effective, as seen in the demonstration video accompanying this paper. Their performance in real-world tests is in
line with expectations, and their behavior matches well with
simple models. The functional principles presented in Sec. II
will be useful to inform grasper designs in the future. New
designs or iterations of the present designs can enable flying
robots to land on arbitrary surfaces and spacecraft to perform
orbital debris cleanup, as well as any other application where
sticking and un-sticking to a smooth surface is desired.
VIII. F UTURE W ORK
While there has been initial success grasping a surface
during flight with a controlled vehicle using a MAV grasper,
it remains to be done to probe the landing envelope to learn
what combination of incoming velocities and misalignments
are acceptable. Beyond this, automating the grasp release and
achieving a transition back to flight will be a future goal.
Future experiments with space graspers are planned for a
large air bearing surface in the next several months. Using a
robotic arm, the grasper mechanism will be integrated into a
realistic grappling scenario that mimics the microgravity environment in Earth’s orbit. Characterization of the grasper’s
performance over a range of spin and tumble rates will be
performed using a Vicon motion capture system. This venue
will also allow system integration with impedance control on
the arm and several computer vision algorithms.
IX. ACKNOWLEDGMENTS
Research was carried out, in part, at the Jet Propulsion
Lab, California Institute of Technology, under contract with
the National Aeronautics and Space Administration. Government Sponsorship Acknowledged. The views expressed are
those of the author and do not reflect the official policy or
position of the Department of Defense or the U.S. Government. Approved for Public Release, Distribution Unlimited.
The work of E. W. Hawkes, M. A. Estrada, and E. V. Eason
was supported by the National Science Foundation Graduate
Research Fellowship Program. E. V. Eason is additionally
supported by the Stanford Graduate Fellowship Program and
the Hertz Foundation.
[2] E. Hawkes, E. Eason, A. Asbeck, and M. Cutkosky, “The gecko’s toe:
Scaling directional adhesives for climbing applications,” Mechatronics,
IEEE/ASME Transactions on, vol. 18, no. 2, pp. 518–526, 2013.
[3] S. Kim, M. Spenko, S. Trujillo, B. Heyneman, D. Santos, and
M. Cutkosky, “Smooth Vertical Surface Climbing With Directional
Adhesion,” IEEE Trans. on Robotics, vol. 24, no. 1, pp. 65–74, 2008.
[4] M. Murphy, C. Kute, Y. Menguc, and M. Sitti, “Waalbot II: Adhesion
Recovery and Improved Performance of a Climbing Robot using
Fibrillar Adhesives,” The International Journal of Robotics Research,
vol. 30, no. 1, pp. 118–133, 2011.
[5] M. Kovac, J. M. Germann, C. Hurzeler, R. Siegwart, and D. Floreano,
“A perching mechanism for micro aerial vehicles,” Journal of MicroNano Mechatronics, vol. 5, pp. 77–91, 2009.
[6] A. Lussier-Desbiens, A. T. Asbeck, and M. R. Cutkosky, “Landing,
perching and taking off from vertical surfaces,” IJRR, vol. 30, no. 3,
pp. 355–370, 2011.
[7] M. L. Anderson, C. J. Perry, B. M. Hua, D. S. Olsen, J. R. Parcus,
K. M. Pederson, and D. D. Jensen, “The sticky-pad plane and other
innovative concepts for perching uavs,” in 47th AIAA Aerospace
Sciences Meeting, 2009.
[8] L. Daler, A. Klaptocz, A. Briod, M. Sitti, and D. Floreano, “A perching
mechanism for flying robots using a fibre-based adhesive,” in IEEE
ICRA, 2013.
[9] P. Day, E. Eason, N. Esparza, D. Christensen, and M. Cutkosky,
“Micro-wedge machining for the manufacture of directional dry adhesives,” Journal of Micro and Nano-Manufacturing, 2013 (in press).
[10] N. Michael, D. Mellinger, Q. Lindsey, and V. Kumar, “The GRASP
multiple micro UAV testbed,” IEEE Robotics and Automation Magazine, vol. 17, pp. 56–65, September 2010.
[11] D. Mellinger, N. Michael, and V. Kumar, “Trajectory generation
and control for precise aggressive maneuvers with quadrotors,” The
International Journal of Robotics Research, vol. 31, no. 5, pp. 664–
674, April 2012.
[12] A. M. Hyslop and J. S. Humbert, “Autonomous Navigation in 3-D
Urban Environments Using Wide-Field Integration of Optic Flow,”
AIAA Journal of Guidance, Control, and Dynamics, vol. 33, no. 1, pp.
147–159, 2010.
[13] E. Glassman, A. Lussier-Desbiens, M. Tobenkin, M. Cutkosky, and
R. Tedrake, “Region of Attraction Estimation for a Perching Aircraft:
A Lyapunov Method Exploiting Barrier Certificates,” IEEE International Conference on Robotics and Automation, pp. 2235–2242, May
2012.
[14] “Orbital debris quarterly news,” NASA Orbital Debris Program Office,
pp. 1–10, Apr 2009.
[15] D. A. Whelan, E. A. Adler, S. B. Wilson III, and G. M. Roesler,
Jr., “DARPA orbital express program: effecting a revolution in spacebased systems,” SPIE Small Payloads in Space, vol. 4136, pp. 48–56,
2000.
[16] R. Friend, “Orbital express program summary and mission overview,”
Proc. of SPIE, Sensors and Systems for Space Applications II, April
2008.
[17] T. Debus and S. Dougherty, “Overview and perrformance of the
front-end robotics enabling near-term demonstratiion (FREND) robotic
arm,” AIAA Aerospace Conference, pp. 2009–1870, 2009.
[18] A. Parness, T. Hilgendorf, P. Daniel, M. Frost, V. White, and
B. Kennedy, “Controllable on-off adhesion for earth orbit grappling
applications,” IEEE Aerosapce Conference, 2013.
[19] D. Santos, M. Spenko, A. Parness, S. Kim, and M. Cutkosky, “Directional adhesion for climbing: theoretical and practical considerations,”
Journal of Adhesion Science and Technology, vol. 21, no. 12-13, pp.
1317–1341, 2007.
[20] K. Autumn, A. Dittmore, D. Santos, M. Spenko, and M. Cutkosky,
“Frictional adhesion: a new angle on gecko attachment,” Journal of
Experimental Biology, vol. 209, no. 18, pp. 3569–3579, 2006.
R EFERENCES
[1] 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,” J. Royal Society, Interface, vol. 6, no. 41, pp. 1223–1232,
Mar 2009.
5493