IEEE Robotics & Automation Magazine - September 2023 - 23

elastomer (Sil30; Carbon, Inc.). The tensegristat finger was
inflated by a constant pressure source at inflation gauge pressure
( PD ) of 24 kPa to simulate being attached to a pressurized
robot. A tendon pulled by an electric motor actuated the
finger to lift the mass.
PNEUMATIC FEA FINGER DESIGN
We chose a common FEA geometry, molded out of the silicone
elastomer used in [8] (Ecoflex-30, Smooth-On, Inc.),
with a thin, inextensible layer made of carbon veil (Item
1064; Fibre Glast, Inc.). We actuated it using a diaphragm
pump (the most popular choice for untethered FEAs) [10].
CHOICE OF MOTOR AND PUMP
We chose a motor and diaphragm pump of similar size and
cost and ran them at their rated voltages. More details can be
found in the supplementary material (available at https://doi.
org/10.1109/MRA.2022.3204234). As both diaphragm pumps
and tendon actuators are powered by built-in dc motors, our
comparison is really of the efficiency of two different kinds of
transmission under similar size and cost restraints.
RESULTS
Figure 1(b) shows the starting and ending state and power
draw over time for both the tendon and pneumatic actuators.
The average time required to lift the weight for the tendon
actuator was 0.1 s, and for the pneumatic actuator it was 18.2 s:
roughly 180-times longer. The tendon actuator consumed an
average of 0.4 J, and the pneumatic actuator consumed an
average of 14.2 J: roughly 36 times as much energy. These
outcomes can of course be adjusted with design changes. For
example, reducing unnecessary material deformation by using
a double-pleated FEA has been shown to reduce the energy
required to achieve a desired shape with a pneumatic actuator
by 36 times [12]. As this was measured based on shape change
rather than desired work done, it is not a direct comparison,
but it is likely that wider adoption of the double-pleated design
for FEAs would increase their efficiency. Still, the tendondriven
tensegristat has the advantage in speed and ability to
control stiffness separately from actuation.
STRETCHING-BASED SHAPE CHANGE VERSUS
BUCKLING-BASED SHAPE CHANGE
To compare the resistance to shape change between bucklingbased
and stretching structures, we measured the pressure
required versus change in length of appendages [see (Figure
1(c)]. There were three types of samples: 1) everting, to represent
buckling-based shape change; 2) freely stretching; and 3)
stretching, with the shape controlled by inextensible circumferential
fibers. We 3D printed seven samples of each type. The
inextensible fibers for the controlled stretch samples were
made from strips of carbon veil bonded to the appendage surface.
More details can be found in the supplementary material
(available at https://doi.org/10.1109/MRA.2022.3204234).
All the samples started at approximately the same length
(more details can be found in the supplementary material,
available at https://doi.org/10.1109/MRA.2022.3204234), and
we tracked their extension using image processing software
(Tracker Video Analysis and Modeling Tool). We began at a
pressure differential of PD ≈0-kPa gauge pressure and raised it
incrementally until the sample exceeded its balloon instability
[13] and began to freely inflate.
RESULTS
Figure 1(c) shows the extension of the samples. As hypothesized,
the everting samples extended much farther than the
stretching ones and achieved most of this extension at low
pressure ( PD <10 kPa). (After the appendage everted, farther
extension came from stretching).
TENSEGRISTAT AMPHIBIAN ROBOT
We chose an amphibian as our demonstration robot (the
design shown in Figure 2) to highlight the potential of largescale
shape change in a way that is accessible to a wide
audience. Pressurized air fills the body and stiffens extended
appendages. Dc motors pull tendons inside the body to make
the robot walk or swim. When, for example, the robot needs
to switch from having legs to having a tail, the tendon
attached to the tail lengthens, reducing the pressure. Then the
tendons attached to the feet can easily buckle the membrane
and invert or collapse the legs. The main body of the robot is
rigid to simplify assembly for the demonstration. The robot's
length in the swimming configuration is 32 cm; in the walking
configuration, it is approximately 19 cm.
MOTORS, TENDONS, AND CONTROL
We used 298:1 gear ratio high-power dc motors from Pololu.
Figure 3(a) shows how each motor is controlled. For tail actuation,
we designed a ratchet pulley so that we could properly
tension and control two antagonistic tendons with a single
motor. We made the tendons from braided polymer fishing
line. Actuation was controlled by a Teensy 3.2 microcontroller
programmed through a port in the " nose " of the robot.
FLUIDS AND PRESSURIZATION
We chose air as the robot's working fluid because pneumatic
structures have high strength-to-weight ratios. We also used a
small amount of mineral oil inside the robot to lubricate the
membranes, provide waterproofing to the electronics in case
of small leaks, and adjust buoyancy. For consistency, we
began each run by inflating the robot to PD
~30 kPa in the
walking configuration (by injecting pressurized air through a
soft elastomer plug molded into a latticework in the lid). Figure
3(b) shows that pressurization of the legs allows them to
bear the robot's body weight.
MEMBRANE AND MANUFACTURING
It is best to construct everting-type structures from a thin,
flexible material with a high-elastic modulus so that the membranes
can be made very thin and easily bunched or folded,
and that the final stiffness can be tuned by changing the pressure
without greatly changing the shape. The materials we
SEPTEMBER 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
23
https://www.doi.org/10.1109/MRA.2022.3204234 https://doi.org/10.1109/MRA.2022.3204234 https://doi.org/10.1109/MRA.2022.3204234 https://www.doi.org/10.1109/MRA.2022.3204234
IEEE Robotics & Automation Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2023
