A Tip Mount for Transporting Sensors

and Tools using Soft Growing Robots

Sang-Goo Jeong1,2∗ , Margaret M. Coad1∗ , Laura H. Blumenschein1 , Ming Luo1 ,
Usman Mehmood2 , Ji Hun Kim2 , Allison M. Okamura1 , and Jee-Hwan Ryu3

Abstract— Pneumatically operated soft growing robots that

extend via tip eversion are well-suited for navigation in confined
spaces. Adding the ability to interact with the environment
using sensors and tools attached to the robot tip would greatly
arXiv:1912.08297v3 [cs.RO] 1 Aug 2020

enhance the usefulness of these robots for exploration in the

field. However, because the material at the tip of the robot
body continually changes as the robot grows and retracts, it is
challenging to keep sensors and tools attached to the robot tip
during actuation and environment interaction. In this paper,
we analyze previous designs for mounting to the tip of soft
growing robots, and we present a novel device that successfully
remains attached to the robot tip while providing a mounting
point for sensors and tools. Our tip mount incorporates and
builds on our previous work on a device to retract the robot
without undesired buckling of its body. Using our tip mount,
we demonstrate two new soft growing robot capabilities: (1)
pulling on the environment while retracting, and (2) retrieving
and delivering objects. Finally, we discuss the limitations of
our design and opportunities for improvement in future soft
growing robot tip mounts.

I. I NTRODUCTION
Many potential robotic applications require the transport
of sensors and tools through confined spaces to explore
and interact with the environment. Continuum robots have
particular strengths for these types of applications due to
their ability both to pass through small apertures and to
support their body weight to rise up over obstacles. For
example, a camera-equipped snake robot [1] was deployed
within a collapsed building for search and rescue after
the 2017 Mexico City earthquake. Additionally, a gripper-
equipped snake-like mobile robot [2] has been demonstrated
for grasping and retrieving objects and turning a valve in a Fig. 1. (a) Challenges related to mounting to the tip of soft growing robots.
(left) The material at the robot tip changes during growth and retraction,
mock disaster scenario. so anything affixed to the tip material will not remain at the tip. (middle)
Another type of continuum robot that is well-suited for The inner material (the “tail”) moves at twice the speed of the robot tip
relative to the base, so anything packaged within the tail will be ejected
navigation in confined spaces is a pneumatically driven soft during growth and engulfed during retraction. (right) When retracted from
the base, soft growing robots often undergo undesired buckling, leading to
This research was partially supported by the project “Toward the Next lack of control of their motion and force. (b) Our current tip mount design
Generation of Robotic Humanitarian Assistance and Disaster Relief: Fun- successfully surmounts these challenges. Our design remains at the robot
damental Enabling Technologies (10069072),” the National Science Foun- tip during growth to arbitrary lengths as well as retraction and incorporates
dation (NSF) under Award 1637446, the Air Force Office of Scientific a device [3] to retract the robot without undesired buckling. Like previous
Research under award FA2386-17-1-4658, Toyota Research Institute (TRI), tip mounts, our design does not interfere with steering of the robot body
and the NSF Graduate Fellowship Program. TRI provided funds to assist using external actuators. Our tip mount can also apply significant pulling
the authors with their research, but this article solely reflects the opinions forces to the environment.
and conclusions of its authors and not TRI or any other Toyota entity.
*These authors contributed equally to this work.
1 Department of Mechanical Engineering, Stanford University, growing robot [4], [5]. Unlike typical snake-like robots, this
Stanford, CA 94305 USA {mmcoad, lblumens, mingluo,
aokamura} @stanford.edu type of robot “grows” from a fixed base by transporting new
2 School of Mechanical Engineering, Korea University of Technology and
material through its body and everting the material at its tip,
Education, Cheonan-si, Republic of Korea {jsg1215z, umehmood, driven by internal air pressure. These robots are particularly
vgty5678} @koreatech.ac.kr
3 Department of Civil and Environmental Engineering, KAIST, Daejeon, good at traversing long, tortuous paths over varied terrain
South Korea. [email protected] due to their ability to grow to arbitrary lengths from a
fixed base [5], [6]; propel their tip forward without relative incorporates a device to retract the soft growing robot with-
movement between their body and the environment [5], [7]; out undesired buckling. Additionally, like other tip mount
conform to the shape of their surroundings via their natural designs, our tip mount allows steering of the robot tip using
softness [8], [9]; shrink in diameter to pass through apertures external actuators such as series pouch motors [6].
smaller than their body diameter [5], [6], [10]; support their
own body weight to move over obstacles and across gaps [5], II. P REVIOUS D ESIGN S UMMARY
[6]; and controllably direct their tip in combination with Motivated by useful tasks that require transport of sensors
growth [5], [6], [11], [12]. and tools at the tip of soft growing robots, four different
Adding the ability to transport sensors and tools at the tip tip mount designs have previously been developed. Here,
of soft growing robots during movement and environment we present a summary of each previous design and discuss
interaction would greatly expand their usefulness in the field. each design’s benefits and limitations related to the goals of
Mounting cameras and other sensors at the tip of the robot (1) remaining at the robot tip during retraction and growth,
would enable information gathering tasks (e.g., collecting (2) transmitting pulling forces to the environment, and (3)
data in an unknown environment), while mounting tools such functioning at an arbitrary robot body length (Fig. 2). We also
as grippers at the tip of the robot would allow delivery and explain the problem of undesired buckling during retraction
retrieval of objects in the environment (e.g., transporting (a limitation of soft growing robots that affects all previous
supplies to a trapped disaster victim or retrieving items from designs) and summarize our previous work on a device to
a confined space) and application of force to the environment retract soft growing robots without undesired buckling, a
(e.g., to turn a valve or open a door). building block towards our current design. Other attributes of
However, mounting to the robot tip in a reliable manner is previous designs, including the weight they add to the robot
a key challenge for soft growing robot design. Unlike many tip and whether they encumber the natural abilities of soft
continuum robots, the material at the tip of the soft growing growing robots to move through constrained environments,
robot continually changes as the robot grows and retracts. are discussed in Section VI.
Thus, anything affixed to the robot body material that is
currently at the robot tip will not remain at the tip as the A. String Mount
robot grows or retracts (Fig. 1(a, left)). As the robot grows, One previous tip mount design was used in [5], [12] to
the current tip material becomes part of the stationary outer transport a camera at the tip of a soft growing robot during
robot body (the “wall”), and as the robot retracts, the tip laboratory demonstrations of the robot’s navigation ability.
material becomes part of the inner robot body (the “tail”), This relatively simple design ties a string to the tip mount
which moves towards the base at twice the speed of the (which can be the sensor or tool itself) and uses the robot tail
robot tip (Fig. 1(a, middle)). Therefore, sensors and tools as a conduit for the string to pass from the robot tip to the
must move relative to the robot body material to remain at base. Thus, the base can be the grounding point for the force
the robot tip and cannot be permanently attached to the body to keep the mount at the tip. However, as the robot body is
wall through simple means like tape or glue. This challenge pressurized, the tail squeezes around the string, forcing the
is shared by other tip-growing robots [13], [14] and everting string to move with the tail. Since the material of the tail
toroidal robots [15], [16]. moves twice as fast as the tip relative to the base, the string
Various designs for tip mounts for soft growing robots and tip mount are ejected from the robot during growth and
have previously been developed [4], [5], [6], [12], [17], engulfed into the body during retraction (Figs. 1(a, middle)
[18] and will be described in detail in Sections II and VI. and 2(a)). In [5], [12], to overcome this issue during growth,
One design [6] carried a camera during deployment of a airflow was added inside the tail, and the string was pulled
soft growing robot for exploration of an archeological site, back from the base to keep the mount at the tip, which
and another design [18] carried a gripper and transported required the tail material to be stored straight rather than on
lightweight objects during an object delivery task. However, a spool. This limits the potential length change and does not
none of the previous designs is able to remain at the robot allow growth to arbitrary lengths. Also, this design can only
tip during retraction as well as growth, transmit significant remain at the robot tip during growth, not retraction, since
pulling forces from the robot body to the environment, and the string can only apply tensile, not compressive forces.
function consistently at an arbitrary robot body length. An One benefit of this design is the physical connection back to
additional shortcoming of all previous tip mount designs is the base through the string, which can be used to transmit
that none of them incorporates a device to retract the soft significant pulling forces to the environment.
growing robot without undesired buckling (and thus lack of
control) of its body (Fig. 1(a, right)), a problem solved in B. Outer Cap Mount
our previous work [3] and described in detail in Section II-E. A second design was used in [6] to transport a camera on a
Improving upon previous work, we present a new tip robot deployed in the field for exploration of an archeological
mount for soft growing robots (Fig. 1(b)) that (1) reliably site. This design is also quite simple and uses a rigid cap that
remains at the robot tip during retraction and growth, (2) fits over the outside of the robot tip and is pushed along by
transmits pulling forces from the robot to the environment, the robot’s growing force. Unlike the string mount design,
(3) functions at an arbitrary robot body length, and (4) this design does not have a direct connection to the robot
 (a) String Mount Design (b) Outer Cap Mount (c) Outer Cap with Reel Mount (d) Magnetic Rings Mount

Wall
Tip
Stopped Tail F

v
Growing 2v

v
Retracting
2v

Fig. 2. Function of previous soft growing robot tip mount designs during growth and retraction. Each design has benefits and at least one limitation. (a)
The string mount [5], [12] implemented on a spooled tail robot gets ejected during growth and engulfed during retraction. (b) The outer cap mount [6]
remains at the robot tip during growth but falls off during retraction. (c) The outer cap with reel mount [4] remains at the robot tip during both growth
and retraction, but the reel at the tip increases in size as the robot body grows, and it may become too large or heavy when the robot reaches a long
length. (d) The magnetic rings mount [17], [18] remains at the robot tip during growth and retraction but is susceptible to falling off due to forces from
the environment.

base and relies on friction between the inside of the cap This design is limited in that, unlike the outer cap design,
and the outside of the robot body wall to keep the mount at it does not function at arbitrary robot body lengths. As the
the tip (Fig. 2(b)). The size of the outer cap can be varied length of the robot increases, the reel inside the mount needs
relative to the robot body size, where smaller caps provide to grow to hold the extra string, limiting the length of the
higher normal and thus frictional forces. robot based on the size of the tip mount.
Unlike the string mount, this design functions for arbitrary
robot body lengths and can be used when the robot is stored D. Magnetic Rings Mount
in a reel at the base, enabling enormous length change from The fourth previous tip mount design was used in [17]
a small initial form factor. However, this design has limited to transport a camera during a laboratory demonstration
ability to remain at the robot tip during retraction. While of growth of a water-filled robot, as well as in [18] to
the friction between the outer cap and the robot body wall transport a gripper for completion of a pick-and-place task
holds the cap on the tip during growth, it does not have in a laboratory environment. This is the first of the tip mount
the same effect during retraction. Instead, the cap remains designs to place part of the mount inside the pressurized area
behind as the robot retracts within it. This design also of the robot body. This design consists of a ring inside the
cannot transmit significant pulling forces to the environment. tip of the robot and another ring (on which the sensor/tool
The frictional forces between the outer cap and the robot is mounted) outside the tip of the robot. The two rings are
body wall determine the largest pulling forces that can be held together, and at the tip, using magnetic rollers that roll
transmitted to the environment, and a deterrent to choosing along the robot body material during growth and retraction,
a design with high frictional forces is the resulting increase allowing the material to pass between the two halves of the
in the pressure required to begin growth of the robot, as mount in a low-friction manner (Fig. 2(d)).
discussed for our current design in Section IV-A. This design is able to remain at the robot tip during
retraction as well as growth, independent of body length,
C. Outer Cap with Reel Mount provided that the outer ring is large enough in diameter that
A third tip mount design was used in [4] to transport a it will not get engulfed into the robot body. However, like the
camera during a laboratory demonstration of robot growth. outer cap mount, the pulling forces that can be transmitted to
This design combines features of the string mount and outer the environment are limited. In this case, the magnetic force
cap mount, with an outer cap (on which the sensor/tool can between the two halves of the mount, limited by the strength
be mounted) containing a motorized reel attached to a string of the magnets used, is the upper bound on pulling forces.
running internal to the tail. During growth, as the string ejects
from the tip, the motor actively reels in the slack, keeping the E. Retraction Without Buckling
cap at the tip, while during retraction, the motor lets out slack A limitation with soft growing robots is the tendency of
to feed the string into the tail (Fig. 2(c)). This successfully their body to undergo undesired buckling when retraction
keeps the mount at the tip during retraction, but does not is attempted by pulling on the tail from the robot base
solve the problem of buckling during retraction. Like with (Fig. 1(a, right)), causing a lack of control of body motion
the string mount, the force to hold the mount at the tip comes and force applied to the environment. This limitation is likely
from the connection of the string back to the base, so the the reason why none of the previous tip mount demonstra-
outer cap stays at the robot tip during retraction and growth, tions showed any significant retraction of the soft growing
and this connection back to the base can be used to transmit robot body. Our previous work [3] analyzes the problem of
significant pulling forces to the environment. undesired buckling during retraction and its dependence on
robot length, curvature, and pressure. In summary, buckling
occurs because the force required to invert the soft robot body
is independent of length, but the force required to buckle
the soft robot body decreases with length. Thus, regardless
of internal pressure and curvature, soft growing robots will
always buckle rather than inverting above a certain length
when the tail is pulled from the base.
Our previous work presents a solution to buckling in the
form of a “retraction device,” which sits at the robot tip and
uses motor-driven rollers to apply force on the tail, grounded
to the robot tip. Applying force from the tip rather than from (a) Tip mount CAD design and prototype

the base makes the effective robot body length zero, enabling
retraction without undesired buckling at any robot length,
pressure, and curvature.
Both retraction devices and tip mounts must sit at the robot
(c) Interchangeable outer cap
tip, so it makes sense to combine them in a tip mount for
tasks that require retraction of the soft growing robot.

III. P ROPOSED T IP M OUNT D ESIGN

Our soft growing robot tip mount design aims to overcome
the limitations of previous designs in completing tasks that Motor
require object retrieval and environment force application.
We present a tip mount that (1) remains at the robot tip Roller
Needle bearing
during growth and retraction, (2) transmits pulling forces
between the robot body and the environment, (3) functions (d) Retraction device
at an arbitrary robot body length, and (4) incorporates a
retraction device to prevent buckling during retraction.
Our design combines concepts from the outer cap de-
sign [6] and the magnetic rollers design [17], [18], and it Bearing
Magnet
incorporates a retraction device [3]. The retraction device sits
inside the pressurized area at the robot tip, similar to the inner (b) Tip mount cross-section (e) Rolling interlock

ring of the magnetic rollers design. An outer cap outside the

Fig. 3. Our current tip mount design. (a) (left) CAD rendering and (right)
robot tip provides a mounting location for sensors and tools. photo of the tip mount. (b) The tip mount consists of three parts: (c) outer
A key improvement of our current design upon the magnetic cap to mount (left) sensors or (right) tools, (d) retraction device including
rollers design is the attachment mechanism between the outer (top) passive rollers to decrease friction with the robot tip and (bottom)
motor-driven rollers to apply retraction forces on the robot tail, and (e) a
and inner parts of the mount. Rather than relying on magnetic rolling interlock using (left) magnets and (right) bearings to hook the inner
force to hold the two parts together, our design employs a and outer parts together.
rolling interlock where the outer cap hooks around the inner
retraction device so that the two pieces cannot physically be retraction device to allow retraction without buckling, and
separated without breaking. Each piece has passive rollers at (3) a rolling interlock to attach them together.
the connection point so that the material of the soft robot a) Outer Cap: The outer cap (Fig. 3(c)) provides a
body wall can pass between them in a low friction manner. mounting location for sensors and tools and a place for the
While this tip mount does not have a physical connection robot body to push on and propel the tip mount forward.
back to the base like the string mount [5], [12] and outer The outer cap design should not impede the function of the
cap with reel mount [13], it can transmit significant forces steering actuators used to direct the tip of the robot. In our
through the rolling interlock, which is grounded to the robot implementation, we used three series pouch motor steering
tip using the retraction device. These forces hold the mount actuators, which shorten when pressurized and are attached
at the robot tip and allow significant pulling forces to be circumferentially around the robot body, as in [6], so we
transmitted to the environment: from the base, along the added three cutouts in the cap to give the actuators room to
robot body, and then through the mount. The following inflate and deflate without affecting the fit on the robot body.
subsections describe in detail the mechanical design and b) Retraction Device: The retraction device (Fig. 3(d))
control of our current tip mount design. is as presented in [3], with stronger motors and a different tip
grounding mechanism. During retraction, two motors (3485,
A. Mechanical Design
Pololu Corporation, Las Vegas, NV) drive rollers coated in
Our current tip mount design (Fig. 3) consists of three high friction material (Non-Slip Reel, Dycem Corporation,
main components: (1) an outer cap for mounting, (2) a Bristol, UK) that squeeze the tail material and pull it toward
the base. Three passive rollers at the top of the device apply a on the joystick, between 2.4 V and 15 V. Using this control
reaction force to the robot tip while allowing the material to method, the soft robot body with our tip mount attached is
move easily around the tip. During growth, the active rollers able to retract at the same maximum speed as growth: 5 cm/s.
are driven to move tail material towards the robot tip, and
the material pushes on the outer cap instead of contacting the IV. C HARACTERIZATION
passive rollers. Aside from the high friction material coating We conducted two experiments to characterize the capa-
the actively driven rollers, all other parts are designed to bilities of a soft growing robot with our tip mount design.
reduce unnecessary friction with the soft robot body material. The first quantifies the effect of the additional friction due to
c) Rolling Interlock: The rolling interlock (Fig. 3(e)) the tip mount on the pressure required to grow the robot. The
consists of three matching sets of roller-magnet units, placed second quantifies the pulling force that can be transmitted to
circumferentially around the base of the outer cap and the the environment. For both experiments, we explore how the
base of the retraction device, such that the series pouch design of different portions of the device affects the robot’s
motors lie in between. Each roller-magnet unit has a passive capabilities. Throughout this paper, the robot body was made
roller with a disk-shaped magnet on either side. Only the using a tube of low-density polyethylene (LDPE) plastic,
rollers contact the wall material, which must slide between a material easy to use for quick prototyping, with inflated
the pairs of disk-shaped magnets (separated by a small space) diameter 8.5 cm and wall thickness 60 µm.
as the robot grows or retracts. The rollers can transmit high
forces across the membrane, and the magnets prevent relative A. Minimum Pressure to Grow
tilting or rotation (and thus separation) between the outer cap The minimum pressure required to begin growth of a soft
and retraction device. We use separate rollers and magnets, as growing robot is an important predictor of its capabilities.
opposed to magnetic rollers, since disk magnets can provide The internal pressure can be set higher than the minimum
higher magnetic force. pressure required to begin growth, and, if the tail is free
to move forward, the additional pressure will either make
B. Control the robot grow faster [5], [7] or apply more force at its
We used the base, joystick, and steering control algorithm tip [4], [19], up to the buckling load of the robot body.
presented in [6] to steer the soft robot body by coordinating The maximum internal pressure is limited by the burst
pressures (0-14 kPa) in the three series pouch motor actuators pressure of the soft robot body. The addition of our tip
placed circumferentially around the body of the robot. Addi- mount adds friction between the body material and the mount
tionally, we developed a method of coordinating the voltages and between the mount and the environment, increasing
sent to the base motor and the retraction device motors to the minimum pressure required to grow. This decreases the
allow growth and retraction without building up slack in the maximum growth speed and the maximum pushing force that
tail or buckling the body. For simplicity, we used open-loop can be applied before bursting the soft robot body.
voltage control of the motors with no position sensing. To understand the effect of our tip mount on the pressure
a) Growth: During growth, the pressure in the soft to grow (and thus the growth speed and pushing force
robot body was set (using the joystick) higher than needed to capability), we conducted growth tests horizontally on a
grow at the desired speed (0-17 kPa), the motor in the base foam board floor with different parts of our device installed.
was backdriven to let out the tail material without building We slowly increased the pressure with slack in the tail and
up slack, and the motors in the retraction device (which are observed the minimum pressure at which growth occurred.
not backdrivable) were controlled to release the material at The results are shown in Fig. 4. The soft robot body
the desired speed. To achieve this, we set the voltage of the without any tip mount (Fig. 4(a)) requires 2 kPa to begin
motor in the robot base to offset static friction in the motor growing, due to the forces needed to turn the soft robot body
(3.5 V), and the voltage of the motors in the retraction device inside out at its tip. The addition of the outer cap (Fig. 4(b))
based on the joystick input to be between 2.4 V and 15 V. increases the required pressure to 3.4 kPa. This represents
Using this control method, the soft robot body with our tip the friction as the wall of the soft robot body slides against
mount attached can grow at a maximum speed of 5 cm/s. the outer cap and the outer cap slides on the floor. Adding
b) Retraction: During retraction, the pressure in the the inner part without the motors and rollers (Fig. 4(c))
soft robot body was set (using the joystick) as low as possible increases the pressure to 6.8 kPa, due to the friction at the
while keeping the robot body pressurized (approximately point of contact between the retraction device and the outer
7 kPa), to allow easy sliding of the robot body material cap, where there is some sliding between the magnets and
between the magnets while limiting the retraction force. The wall material. Finally, with the addition of the motors and
motors in the base were run with enough force to take in the rollers (Fig. 4(d)), the robot still requires 6.8 kPa to grow,
slack in the tail but not to buckle the body, and the motors in since the rollers do not slide on the tail.
the retraction device provided the rest of the necessary force The largest friction increase occurs at the bearing and
to retract. To achieve this, we set the voltage of the motor magnet interface, so improving the design there would have
in the robot base to the highest voltage before the straight the largest impact. Other locations to improve friction are
robot body began to buckle at the retraction pressure (9.4 V), between the tip mount and the environment and between the
and the voltage of the motors in the retraction device based outer cap and the robot wall. The experimentally determined
 in the tip mount, the need to counter the internal pressure of
the soft robot body, and the weight of the tip mount itself
Floor
(0.5 kg), further limit the weight that our tip mount can lift
to 2.5 kg. This is the limiting factor in force that the soft
23
Bursting Pressure : 22.0 kPa growing robot can currently pull, but increasing the force the
Growing Pressure [kPa]

22
21 rollers can apply before slipping would increase this limit.
20 (20.0 kPa) (18.6 kPa) (15.2 kPa) (15.2 kPa)
b) Motor Torque Limit: The motors in the device have
7
6 a gearbox torque limit of 5 kg·cm, so the maximum torque
5
4 that the two motors can withstand together is 10 kg·cm. With
3
2 a roller radius of 3 cm, this results in a motor torque pulling
1
2.0 kPa 3.4 kPa 6.8 kPa 6.8 kPa limit of 3.3 kg. If the roller slip force is increased, the use
(a) No cap (b) Outer cap only (c) Tip mount without motors (d) Full tip mount
of stronger motors could increase the force limit.
c) Device Yielding: We applied an increasing force
Fig. 4. The minimum pressure required to begin growth as compared to the
soft growing robot burst pressure when different portions of our current tip until the tip mount came apart. When a load of 7 kg
mount are attached to the robot body: (a) no tip mount, (b) only the outer was applied, the outer cap flexed and broke, which could
cap, (c) the retraction device and the outer cap are attached, but the motors be improved by reinforcing the design and switching to a
and active rollers are removed, and (d) the full tip mount is attached. The
addition of the tip mount decreases the difference between the minimum stronger material than the 3-D printed PLA used.
pressure to grow and the burst pressure by only 24%, meaning that the d) Material Yielding: To calculate the pulling force
impact on relevant robot capabilities is relatively small. limit due to yielding of the soft robot body material, we first
experimentally determined the yield stress of the material
burst pressure of the soft robot body is 22.0 kPa, so the by increasing the pressure until the robot body burst (P =
additional friction of this device decreases the available range 22.0 kPa). Using the equation for hoop stress in a thin-walled
of pressure above the minimum growth pressure by only 24% cylinder (the highest-stress direction), we can calculate the
(from 20.0 kPa to 15.2 kPa). yield stress of the material:
Pr
B. Maximum Pulling Force σ= , (1)
t
An important goal of our tip mount design is to allow where P is the pressure inside the robot, r is the tube radius
transmission of significant pulling forces to the environment. (r = 4.25 cm), t is the material thickness (t = 60 µm)
Soft growing robots are much weaker in pushing and side and σ is the hoop stress at yield, which was calculated to
loading than traditional engineering materials, so their ability be 15.6 MPa. To reach material yielding with a pulling force
to pull on the environment is vital. The force these robots can instead, we need a force of 25.5 kg. This force limit could be
support in compressive or transverse loading before buckling improved by using a material with a higher yield stress, or by
can be calculated using inflated beam models [20], [21], [22], increasing the cross-sectional area of the body wall, though
[23] and decreases with length. However, the force these this would also affect burst pressure and growth pressure.
robots can pull is length-independent and depends only on Overall, with our tip mount implementation, we are for
the mechanical properties of their body material. Because the first time able to apply significant pulling force to the
soft growing robots can grow to arbitrary lengths, harnessing environment while retracting without buckling. The 2.5 kg of
this pulling capability is key to useful environment interac- pulling force that we are able to transmit to the environment
tion and applications such as turning a valve, opening a door, is only 10% of the 25.5 kg of potential force transmission of
or retrieving items from a confined space. the robot body material. This is encouraging, as it indicates
To quantify the pulling ability of a soft growing robot there is room to better implement our design (e.g., through
with our tip mount, we hung the robot vertically downward rollers capable of transmitting more force before slipping,
and attached a weight to a hook on the tip mount. We then and higher-torque motors) to exert even higher pulling forces
attempted to retract the weight with the robot and increased on the environment.
the weight until failure (2.5 kg) (Fig. 5(a)). The robot’s
pulling force could be limited by the frictional force the V. D EMONSTRATION
retraction device rollers transmit to the tail, the torque of With our current tip mount, soft growing robots can for
the retraction device motors, the breaking strength of the the first time grow to arbitrary lengths and retract without
tip mount material, or the yielding strength of the robot buckling while transporting sensors and tools at the robot
body material. We calculated or measured the pulling limits tip, all while steering with external actuators such as series
based on each of these factors, as described in the following pouch motors. In addition to pulling forces, our tip mount
subsections and shown in Fig. 5(b). can support compression loads, allowing the robot to push
a) Rollers Slipping on Tail: The force applied by the objects up to the buckling load and burst pressure of the
retraction device on the tail of the robot is transmitted at the soft robot body. These capabilities will greatly enhance the
connection point between the motor-driven rollers and the usefulness of soft growing robots for tasks in the field.
tail. The maximum measured force that can be applied by the To demonstrate the usefulness of these capabilities, we
rollers on the tail before slip occurs is 5 kg. Frictional losses show a simple object retrieval and delivery task that involves
 TABLE I
25 C APABILITIES AND L IMITATIONS OF S OFT G ROWING ROBOT T IP

Weight Limit (kg)

20 M OUNT D ESIGNS
15
String Outer Outer Magnetic Current
10 [5], Cap Cap Rings Design
[12] [6] with [17],
5 Reel [18]
2.5 kg 3.3 kg 7 kg 25.5 kg [4]
Ro Mo De Ma Remains at the robot Some- Yes Yes Yes Yes
ller tor vic ter tip during growth? times
sS To eY ial
2.5 kg lipp rqu ield Yie Remains at the robot No No Yes Yes Yes
ing eL ing ldin
on im g tip during retraction?
Ta it Can transmit pulling Yes Some- Yes Some- Yes
il
forces? what what
(a) (b) Functions at arbitrary No Yes No Yes Yes
robot body lengths?
Incorporates retraction No No No No Yes
Fig. 5. Vertical weight pulling limit for a soft growing robot with our tip without buckling?
mount attached, based on various factors. (a) Using our current tip mount,
the soft growing robot can pull a weight of up to 2.5 kg while retracting. (b) Adds minimal weight Yes Yes No Yes No
to the robot tip?
Four factors contribute to the maximum pulling force. The rollers slipping
Avoids sliding relative Yes No No Yes No
on the tail is currently the limiting factor, but all factors could potentially to the environment?
be improved in future designs. Allows body shrinking Yes No No Some- No
through apertures? what

growth, retraction, and steering using a gripper-equipped

robot. In Fig. 6(a and b), the robot grows and steers along The outer cap in our design covers part of the robot
the floor, approaching and grabbing the water bottle. It body wall and, therefore, slides relative to the environment
then retracts without buckling to move around an obstacle during growth and retraction. Thus, unlike with soft growing
(Fig. 6(c)) before growing and steering in the opposite di- robots without a tip mount, where only internal friction limits
rection, to head to the hand of the trapped person (Fig. 6(d)). growth [7], here, friction between the environment and outer
While this demonstration is not necessarily representative cap factors into the force required to grow or retract the
of a real disaster scenario, it showcases the robot capabilities robot. This limitation is shared by the outer cap mount and
to pull/push objects while growing, retracting, and steering the outer cap with reel mount. The magnetic rings and string
through an environment, which are made possible for the mount avoid this limitation by only connecting to the tail or
first time with our tip mount design. the robot tip, and not the wall.
Lastly, the fact that our current tip mount has a rigid outer
VI. D ISCUSSION cap means that it cannot allow the robot body to deform
Unlike the four previous soft growing robot tip mount and pass through apertures smaller than its body diameter,
designs, our current tip mount design (1) remains at the robot another of the natural strengths of soft growing robots. This
tip during growth and retraction, (2) transmits pulling forces limitation is also shared by the outer cap mount and the outer
to the environment, (3) functions at arbitrary robot body cap with reel mount. The magnetic rings mount allows some
lengths, and (4) incorporates retraction without buckling. body shrinking, and the string mount is only limited by the
However, our design has some limitations that are not shared size of the sensor or tool.
by all previous tip mount designs: it (1) adds significant Development of the ideal soft growing robot tip mount
weight to the robot tip, (2) slides relative to the environment, design that meets all of the desired capabilities in Table I is
and (3) does not allow body shrinking through apertures. still an open research question, but different tip mounts could
Table I summarizes the capabilities and limitations of all work for different applications. For some applications, the
five tip mount designs discussed in this paper. limitations of the retraction device and rolling interlock may
The additional weight of our current tip mount (0.5 kg) be worthwhile for the added benefits of controlled retraction
comes primarily from the two motors of the retraction and significant pulling force transmission.
device, though the bearings and magnets that make up the An additional consideration for soft growing robot tip
rolling interlock also contribute. For the soft robot body mounts not discussed thus far is power and signal trans-
in this paper, this additional weight greatly decreases the mission to and/or from the sensor or tool at the robot
maneuverability of the robot tip, making it impossible for tip. For some applications, wireless signal transmission and
the series pouch motor actuators to lift the tip against battery power may work, but for other applications, a wired
gravity, and undercutting the natural ability of soft growing connection is crucial. Various methods of passing a wire
robots to support their own body weight over obstacles and from the robot base to the tip mount have been developed,
across gaps. The limitation of extra weight is shared by the including passing a wire inside the tail [4], [5], [12], inside
outer cap with reel design, which also contains a motor in a self-sealing pocket outside the robot body [6], or entirely
the tip mount, while the string mount, outer cap mount, outside the robot body [18]. The development of methods to
and magnetic rings mount keep the additional weight to a transmit power and signal to the robot tip is an open research
minimum. question.
 Trapped Victim

Water bottle

(a) (b) (c) (d)

Fig. 6. Demonstration of a new capability of object retrieval and delivery made possible by our current tip mount design. (a) In a mock disaster scenario,
the soft growing robot (b) grows and steers to pick up a water bottle, (c) retracts with the water bottle, and (d) grows and steers and places the object in
the trapped victim’s hand.

VII. C ONCLUSION AND F UTURE W ORK [8] J. D. Greer, L. H. Blumenschein, A. M. Okamura, and E. W. Hawkes,
“Obstacle-aided navigation of a soft growing robot,” in IEEE Interna-
We presented a novel tip mount for transporting sensors tional Conference on Robotics and Automation, 2018, pp. 4165–4172.
and tools with soft growing robots that overcomes some [9] D. A. Haggerty, N. D. Naclerio, and E. W. Hawkes, “Characterizing
environmental interactions for soft growing robots,” in IEEE/RSJ
limitations of previous tip mount designs and is able for the International Conference on Intelligent Robots and Systems, 2019, pp.
first time to exert significant pulling force on the environment 3335–3342.
while retracting, as well as to retrieve and deliver objects. [10] T. Nakamura and H. Tsukagoshi, “Soft pneumatic manipulator capable
of sliding under the human body and its application to preventing
We also analyzed the four previous soft growing robot tip bedsores,” in IEEE/ASME International Conference on Advanced
mount designs in comparison with our current design. Intelligent Mechatronics, 2018, pp. 956–961.
Future work on our current tip mount will include the [11] L. H. Blumenschein, N. S. Usevitch, B. Do, E. W. Hawkes, and A. M.
Okamura, “Helical actuation on a soft inflated robot body,” in IEEE
development of a more robust control scheme that adds International Conference on Soft Robotics, 2018, pp. 245–252.
encoders to the retraction device, as well as the development [12] J. D. Greer, T. K. Morimoto, A. M. Okamura, and E. W. Hawkes,
of a wire management scheme for the retraction device “A soft, steerable continuum robot that grows via tip extension,” Soft
Robotics, pp. 95–108, 2019.
wires. We will also improve the tip mount design limita- [13] H. Tsukagoshi, N. Arai, I. Kiryu, and A. Kitagawa, “Tip growing
tions, including reducing weight, reducing sliding along the actuator with the hose-like structure aiming for inspection on narrow
environment, and allowing body shrinking through apertures. terrain.” International Journal of Automation Technology, vol. 5, no. 4,
pp. 516–522, 2011.
With the successful development of methods to apply [14] A. Sadeghi, A. Mondini, and B. Mazzolai, “Toward self-growing soft
force to the environment, we are encouraged to pursue the robots inspired by plant roots and based on additive manufacturing
development of soft growing robots as true manipulators technologies,” Soft Robotics, vol. 4, no. 3, pp. 211–223, 2017.
[15] V. Orekhov, M. Yim, and D. Hong, “Mechanics of a fluid filled
that are able to move payloads with precision through a everting toroidal robot for propulsion and going through a hole,” in
large workspace. This requires development of stronger and ASME International Design Engineering Technical Conferences, 2010,
higher-curvature actuators for soft growing robots, as well pp. 1205–1212.
[16] V. Orekhov, D. W. Hong, and M. Yim, “Actuation mechanisms for bio-
as the ability to control and sense stiffness, shape, and force logically inspired everting toroidal robots,” in IEEE/RSJ International
application. Conference on Intelligent Robots and Systems, 2010, pp. 2535–2536.
[17] J. Luong, P. Glick, A. Ong, M. S. deVries, S. Sandin, E. W. Hawkes,
and M. T. Tolley, “Eversion and retraction of a soft robot towards the
R EFERENCES exploration of coral reefs,” in IEEE International Conference on Soft
[1] J. Whitman, N. Zevallos, M. Travers, and H. Choset, “Snake robot Robotics, 2019, pp. 801–807.
urban search after the 2017 Mexico City earthquake,” IEEE Interna- [18] F. Stroppa, M. Luo, K. Yoshida, M. M. Coad, L. H. Blumenschein,
tional Symposium on Safety, Security, and Rescue Robotics, pp. 1–6, and A. M. Okamura, “Human interface for teleoperated object manip-
2018. ulation with a soft growing robot,” in IEEE International Conference
[2] M. Tanaka, K. Tadakuma, M. Nakajima, and M. Fujita, “Task- on Robotics and Automation, 2020, pp. 726–732.
Space Control of Articulated Mobile Robots with a Soft Gripper for [19] H. Godaba, F. Putzu, T. Abrar, J. Konstantinova, and K. Althoefer,
Operations,” IEEE Transactions on Robotics, vol. 35, no. 1, pp. 135– “Payload capabilities and operational limits of eversion robots,” in
146, 2019. Annual Conference Towards Autonomous Robotic Systems, 2019, pp.
[3] M. M. Coad, R. P. Thomasson, L. H. Blumenschein, N. S. Usevitch, 383–394.
E. W. Hawkes, and A. M. Okamura, “Retraction of soft growing robots [20] W. Fichter, “A theory for inflated thin-wall cylindrical beams,” Na-
without buckling,” IEEE Robotics and Automation Letters, vol. 5, tional Air and Space Administration (NASA) Technical Note, vol. 3466,
no. 2, pp. 2115–2122, 2020. pp. 1–19, 1966.
[4] D. Mishima, T. Aoki, and S. Hirose, “Development of pneumatically [21] A. Le-van and C. Wielgosz, “Bending and buckling of inflatable
controlled expandable arm for search in the environment with tight beams: Some new theoretical results,” Thin-Walled Structures, vol. 43,
access,” in Field and Service Robotics. Springer, 2006, pp. 509–518. no. 8, pp. 1166–1187, 2005.
[22] R. W. Leonard, G. W. Brooks, and H. G. McComb Jr, “Structural
[5] E. W. Hawkes, L. H. Blumenschein, J. D. Greer, and A. M. Okamura,
considerations of inflatable reentry vehicles,” National Air and Space
“A soft robot that navigates its environment through growth,” Science
Administration (NASA) Technical Note, vol. 457, pp. 1–23, 1960.
Robotics, vol. 2, no. 8, p. eaan3028, 2017.
[23] R. Comer and S. Levy, “Deflections of an inflated circular-cylindrical
[6] M. M. Coad, L. H. Blumenschein, S. Cutler, J. A. R. Zepeda, N. D.
cantilever beam,” AIAA Journal, vol. 1, no. 7, pp. 1652–1655, 1963.
Naclerio, H. El-Hussieny, U. Mehmood, J. Ryu, E. W. Hawkes, and
A. M. Okamura, “Vine robots: Design, teleoperation, and deployment
for navigation and exploration,” IEEE Robotics and Automation Mag-
azine, 2020, doi:10.1109/MRA.2019.2947538.
[7] L. H. Blumenschein, A. M. Okamura, and E. W. Hawkes, “Modeling
of bioinspired apical extension in a soft robot,” in Conference on
Biomimetic and Biohybrid Systems. Springer, 2017, pp. 522–531.