IEEE Robotics & Automation Magazine - September 2023 - 21

For example, in a balloon, air supports compression, and the
membrane supports tension. This structural scheme allows the
shape-changing abilities of muscular hydrostats like tongues
and tentacles. Muscular hydrostats are composed mostly of
water, which supports compressive loads [4]. Cell membranes
and muscle fibers support the tensile loads necessary to keep
the water contained. This method of bearing forces means that
muscular hydrostats are analogous to tensegrity structures,
despite looking very different than the rod-and-cable robots
and so on associated with that name. (For a recent review of
tensegrity robots, see [5]. For a nonfluidic robotic analogy to a
muscular hydrostat, see [6].)
Tensegrity structures contain only elements purely in tension
or purely in compression [7]. The stiffness and shape of
the structures can be changed via the loading of the elements.
A low-energy method for causing shape change is to apply the
type of load to an element that it is not suited to support and
allow it to buckle, flex, or flow. For example, a fluid can be
used to support compressive loads but not shear loads: even
a small application of force in shear will cause it to dramatically
change shape.
Pneumatic soft robots such as fluidic elastomer actuators
(FEAs), when actuated, are approximately tensegrity structures
with a fluid-compressive element [8]. FEA-based robots passively
deform due to environmental forces, however, rather than
actively changing between separately functional forms, and so
are not shape-changing robots in the sense used here.
In this article, we show how shape-changing robots can be
created using fluid-based tensegrity-type structures. We provide
three design principles which not only allow shape change
but also offer significant improvement on one of the key limitations
of soft robots: current soft robots are severely constrained
in mobile applications because they have low efficiencies
(commonly below 0.1% for mobile soft robots) [9]. Untethered
pressure sources are either slow (diaphragm pumps) or low
capacity (compressed air) [10]. Because of this, either speed or
operational lifetime must be sacrificed when untethered. Much
of this inefficiency is due to the actuation method in FEAs.
Motion is caused by changing the volume of the fluid to stretch
the membrane into a new shape, rather than being caused by
the tensile element (like muscle fiber). These motions require
a lot of material deformation that does not do useful work, and
it is not possible to change stiffness separately from causing
motion. Our design principles address these problems.
We first present our three " tensegristat " design principles (the
" Concept of Tensegristats " section), then we explore the benefits
of these design principles for untethering and shape change with
controlled demonstrations (the " Untethered Pneumatic Actuation
Versus Untethered Tendon Actuation " and " StretchingBased
Shape Change Versus Buckling-Based Shape Change "
sections). Next we present a full-amphibious robot that morphs
from a form with legs to a form with a tail to show that the
tensegristat principles can be used to create robots that change
between different functional forms (the " Tensegristat Amphibian
Robot " section). Finally, we discuss the possibility of scaling
tensegristat robots (the " Scaling " section).
CONCEPT OF TENSEGRISTATS
We propose fluid-membrane tensegrity structures made with
the following design principles: 1) actuation will be caused by
tendons driven by electric motors, 2) both shape change and
motion will rely primarily on loadings that do not work
against the strengths of the structural elements (e.g., buckling
rather than stretching), and 3) the total mass of fluid in the
system will stay constant.
We call this type of system a tensegristat as it is inspired
by tensegrity structures and the mechanics of muscular hydrostats.
These principles will not only allow large-scale shape
changes, they will also allow better untethered speed and efficiency
relative to other soft robots, and stiffness modulation
separate from motion. The only robot we are aware of in the
literature that uses a combination of tendon and pneumatic
drive to create a soft robot untethered from an air compressor
is in [11]. The article, however, still relies on stretching for
motion and does not explore mobility or shape change.
Figure 1(a) shows the concept of a shape-changing tensegristat
robot. Figure 1(a) at left shows an inflated tentacle made
of a thin, nearly inextensible membrane. The tentacle can be
actuated (swung from side to side) using tendons along its
edges and can be retracted by pulling on a tendon attached
to its tip. To morph from a tentacle configuration to a gripper,
the central tendon of the gripper is lengthened and the central
tendon of the tentacle is shortened, transferring fluid from the
tentacle to the gripping appendage.
Extensible membranes and compressible fluids act like
springs in the system, whereas inextensible membranes and
incompressible fluids transmit forces without spring-like
behavior. If the fluid is compressible, the volumes of different
forms may vary, with lower-volume shapes having higher
stiffness [like the stub shown at the top of Figure 1(a)] Energy
is required to compress the fluid as the robot switches between
forms, but not to hold it in a form; the retracted appendages are
prevented from popping out under pressure by the unpowered
holding torque of the motor.
UNTETHERED PNEUMATIC ACTUATION VERSUS
UNTETHERED TENDON ACTUATION
Tensegristat designs should improve the tradeoff between operational
lifetime and speed for untethered soft robots. To test this
hypothesis in one controlled situation, we compared the speed
and efficiency of a generalized form of tensegristat to a common
pneumatic FEA architecture. We created " finger " actuators
of the same length, the first as would be found in a
tendon-actuated tensegristat, and the second actuated pneumatically,
as shown in Figure 1(b). We added the same mass (m) on
a hook at the end of each and found the time and energy
required to lift the mass a set height. When the mass reached its
end position [height (h)], it completed a circuit that cut off the
measurement. We ran seven trials for each finger.
TENSEGRISTAT FINGER DESIGN
We printed the tensegristat finger using a Digital Light Synthesis
3D printer (M1; Carbon, Inc.), out of a silicone urethane
SEPTEMBER 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
21
IEEE Robotics & Automation Magazine - September 2023

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