Bone-Free: Soft Mechanics for Adaptive Locomotion

Abstract

Muscular hydrostats (such as mollusks), and fluid-filled animals (such as annelids), can exploit their constant-volume tissues to transfer forces and displacements in predictable ways, much as articulated animals use hinges and levers. Although larval insects contain pressurized fluids, they also have internal air tubes that are compressible and, as a result, they have more uncontrolled degrees of freedom. Therefore, the mechanisms by which larval insects control their movements are expected to reveal useful strategies for designing soft biomimetic robots. Using caterpillars as a tractable model system, it is now possible to identify the biomechanical and neural strategies for controlling movements in such highly deformable animals. For example, the tobacco hornworm, Manduca sexta, can stiffen its body by increasing muscular tension (and therefore body pressure) but the internal cavity (hemocoel) is not iso-barometric, nor is pressure used to directly control the movements of its limbs. Instead, fluid and tissues flow within the hemocoel and the body is soft and flexible to conform to the substrate. Even the gut contributes to the biomechanics of locomotion; it is decoupled from the movements of the body wall and slides forward within the body cavity at the start of each step. During crawling the body is kept in tension for part of the stride and compressive forces are exerted on the substrate along the axis of the caterpillar, thereby using the environment as a skeleton. The timing of muscular activity suggests that crawling is coordinated by proleg-retractor motoneurons and that the large segmental muscles produce anterograde waves of lifting that do not require precise timing. This strategy produces a robust form of locomotion in which the kinematics changes little with orientation. In different species of caterpillar, the presence of prolegs on particular body segments is related to alternative kinematics such as "inching." This suggests a mechanism for the evolution of different gaits through changes in the usage of prolegs, rather than, through extensive alterations in the motor program controlling the body wall. Some of these findings are being used to design and test novel control-strategies for highly deformable robots. These "softworm" devices are providing new insights into the challenges faced by any soft animal navigating in a terrestrial environment.

Full text

SYMPOSIUM
Bone-Free: Soft Mechanics for Adaptive Locomotion
B. A. Trimmer
1,
* and Huai-ti Lin
†
*Department of Biology, School of Arts and Sciences, Tufts University, 200 Boston Avenue, Suite 2600, Medford, MA
02155, USA;
†
Howard Hughes Medical Institute, Janelia Farm, Ashburn, VA, USA
From the symposium ‘‘Terrestrial Locomotion: Where Do We Stand, Where Are We Going?’’ presented at the annual
meeting of the Society for Integrative and Comparative Biology, January 3–7, 2014 at Austin, Texas.
1
E-mail: barry.trimmer@tufts.edu
Synopsis Muscular hydrostats (such as mollusks), and fluid-filled animals (such as annelids), can exploit their constant-
volume tissues to transfer forces and displacements in predictable ways, much as articulated animals use hinges and
levers. Although larval insects contain pressurized fluids, they also have internal air tubes that are compressible and, as a
result, they have more uncontrolled degrees of freedom. Therefore, the mechanisms by which larval insects control their
movements are expected to reveal useful strategies for designing soft biomimetic robots. Using caterpillars as a tractable
model system, it is now possible to identify the biomechanical and neural strategies for controlling movements in such
highly deformable animals. For example, the tobacco hornworm, Manduca sexta, can stiffen its body by increasing
muscular tension (and therefore body pressure) but the internal cavity (hemocoel) is not iso-barometric, nor is pressure
used to directly control the movements of its limbs. Instead, fluid and tissues flow within the hemocoel and the body is
soft and flexible to conform to the substrate. Even the gut contributes to the biomechanics of locomotion; it is decoupled
from the movements of the body wall and slides forward within the body cavity at the start of each step. During crawling
the body is kept in tension for part of the stride and compressive forces are exerted on the substrate along the axis of the
caterpillar, thereby using the environment as a skeleton. The timing of muscular activity suggests that crawling is
coordinated by proleg-retractor motoneurons and that the large segmental muscles produce anterograde waves of lifting
that do not require precise timing. This strategy produces a robust form of locomotion in which the kinematics changes
little with orientation. In different species of caterpillar, the presence of prolegs on particular body segments is related to
alternative kinematics such as ‘‘inching.’’ This suggests a mechanism for the evolution of different gaits through changes
in the usage of prolegs, rather than, through extensive alterations in the motor program controlling the body wall. Some
of these findings are being used to design and test novel control-strategies for highly deformable robots. These
‘‘softworm’’ devices are providing new insights into the challenges faced by any soft animal navigating in a terrestrial
environment.
Introduction
Until recently, research on biomechanics of soft-
bodied animals has focused on the use of hydrostat-
ics to control movements and to provide rigidity
(Skierczynski et al. 1996;Quillin 1998;1999;Van
Leeuwen et al. 2000;Accoto et al. 2004). Muscular
hydrostats (such as mollusks), and segmented, fluid-
filled animals (such as annelids), can exploit their
constant-volume tissues to transfer forces and
displacements in predictable ways, much as articu-
lated animals use hinges and levers. In simple cases
such as worms, the circumferential and longitudinal
muscles perform the same antagonistic functions as
do the flexors and extensors of vertebrates
(Wainwright 1988;Vogel 2003). Although larval
insects contain pressurized fluids, they also have in-
ternal air tubes (trachea) that are compressible and,
as a result, they have more uncontrolled degrees of
freedom. Furthermore, larval insects have layers of
muscles oriented in multiple planes and intercon-
nected though the soft body wall. These animals
cannot encode movements (or forces) using a
simple relationship between pressure and volume.
Movements in response to neural commands are
also affected by the pseudoelastic properties of tissues
and by deformations caused by environmental forces.
Integrative and Comparative Biology
Integrative and Comparative Biology, pp. 1–14
doi:10.1093/icb/icu076 Society for Integrative and Comparative Biology
ßThe Author 2014. Published by Oxford University Press on behalf of the Society for Integrative and Comparative Biology. All rights reserved.
For permissions please email: journals.permissions@oup.com.
Integrative and Comparative Biology Advance Access published June 18, 2014
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

The main challenge for researchers studying locomo-
tion in insect larvae is to understand how such soft
animals coordinate the many degrees of mechanical
freedom and how they translate muscular forces into
useful displacements. Without a stiff skeleton, mus-
cles tend to deform the body rather than transmit
forces for effective movement.
This problem has been increasingly recognized
within the context of ‘‘neuromechanics,’’ an inte-
grated approach to understanding motor control as
the interplay between neural commands and biome-
chanics. The physical plant (an animal’s body)
does not simply carry out detailed behavioral in-
structions from the central nervous system, but in-
stead has an integral role in movement-control. This
concept (called embodiment or morphological com-
putation) has been put forward in many contexts,
notably in the fields of artificial intelligence (Pfeifer
et al. 2005,2007) and neuroethology (Chiel and Beer
1997;Nishikawa et al. 2007;Chiel et al. 2009).
Embodiment is an extremely useful framework for
investigating the control of complex behavior.
However, it demands a detailed knowledge of an
animal’s anatomy, material properties of its tissues,
body kinematics, dynamics, and neural signaling. It
is rarely possible for all of these parameters to be
collected with sufficient resolution in a single
model system to predict how an animal will respond
to environmental challenges. The problem is com-
pounded in the study of soft animals in which
tissue mechanics play such an important part, or in
which the neural components are relatively inacces-
sible (cephalopods being a notable example) (Wells
1976;Gutfreund et al. 1998;Matzner et al. 2000;
Sumbre et al. 2005).
Caterpillars are now emerging as a tractable model
system for identifying some of the biomechanical
and neural strategies for controlling movements in
highly deformable animals. Caterpillars are extremely
robust to experimental manipulation and they have a
large number of discrete identifiable muscles. A key
technical advantage is that each muscle is innervated
by a single (occasionally two) excitatory motoneu-
ron(s) (Levine and Truman 1985). With the recent
development of flexible, implantable, microelectrode
arrays (Metallo et al. 2011), it is possible to record
ongoing motor activity with single-neuron resolu-
tion. The mechanical properties of key tissues, in-
cluding muscles and body wall, have been
characterized in detail (Dorfmann et al. 2007,2008;
Lin et al. 2009,2011b;Paetsch et al. 2012) and both
kinematics (Trimmer and Issberner 2007;van
Griethuijsen and Trimmer 2009), and ground-
reaction forces (GRFs) have been measured in
freely moving caterpillars (Lin and Trimmer
2010a). This article discusses some of these find-
ings and relates them to other forms of locomotion
and their application to the development of soft
robots.
We show that in the tobacco hornworm caterpil-
lar, Manduca sexta, crawling relies on periods in
which the body is kept in tension and compressive
forces are exerted on the substrate along the axis of
the caterpillar. When using this strategy, a caterpillar
cannot crawl on substrates softer than itself but it
can conform more closely to the surface and
remain flexible. We have called this strategy an
‘‘environmental-skeleton’’ and it could occur in
other animals in similar environmental niches (Lin
and Trimmer 2010b). We also show that electrical
activity in the abdominal muscles produces broad
anterograde waves of movement that do not require
precise neural timing. This strategy produces an
extremely robust form of locomotion in which the
kinematics changes very little with orientation or
other disturbances. Crawling is primarily coordinated
by the activity of the proleg-retractor motoneurons.
In different species of caterpillar, the presence or
absence of prolegs on particular body segments is
closely related to the overall kinematics of
locomotion, suggesting that changes in mechanical
interactions might have played a major part in the
evolution of different gaits. We predict that proleg
gripping plays an essential role in controlling the
transition from crawling to inching gaits. Some of
these findings are now being used to design and
test novel control-strategies for highly deformable
robots. These ‘‘softworm’’ devices are also providing
new insights into the challenges faced by any soft
animal navigating in a terrestrial environment.
A definition of soft
‘‘Soft’’ is not a rigorously defined material property.
Unlike stiffness (expressed as Young’s modulus) and
hardness (measured by indentation and other stan-
dardized tests), soft is an elastic property associated
with large deformations. Interestingly, this property
is dependent on scale; steel beams will undergo large
deformations when very large forces are applied but
steel is not typically considered to be soft. A good
working description is that soft materials deform
greatly under the loads they normally encounter.
For a given structure this can be formally described
by measuring the specific stiffness (force normalized
by the weight and length changes normalized to the
structural dimensions) (Lin et al. 2011b).
2B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

Anatomy and hydraulics of caterpillars:
Contrast with worms and mollusks
Control of hydrostatic movement
Although the biomechanics of terrestrial soft-bodied
animals are often discussed as a common group, the
body plans of worms, mollusks, and caterpillars
differ in significant ways. Worms have a fluid-filled
coelom that resembles a constant-volume hydrostatic
cylinder. Contraction of circumferential muscles re-
duces the radius and increases the length of a body
segment whereas contraction of longitudinal muscles
reverses this effect. The relationship between diame-
ter and length, and the effect of anisotropic materials
in the body wall, has been the subject of numerous
studies (Wainwright 1988;Wilson et al. 1991;Quillin
1998,1999). The initial aspect ratio can also be
exploited to vary the speed at which limbs can be
extended and withdrawn (Kier and Smith 1985;
Smith and Kier 1989;Kier and Leeuwen 1997).
Although terrestrial mollusks do not usually have a
fluid-filled cavity surrounded by muscle, they share
many of the biomechanical features of worms.
Densely packed muscular tissues squeeze against
one another, thereby operating as constant-volume
muscular hydrostats. Despite their obvious differ-
ences from animals with skeletons, the control of
movements, both in worms and in muscular hydro-
stats, fundamentally resembles that of animals with
articulated skeletons. By pressurizing the body cavity
and tissues, worms and mollusks increase their effec-
tive body stiffness to help support their weight and
exert propulsive forces on the substrate.
Furthermore, by controlling pressure in different
fluid/tissue compartments, hydrostatic animals can
exchange forces and displacements in much the
same way that animals with stiff skeletons exploit
joints and levers. When force is applied to liquid
in a narrow cylinder, it can exert greater force in a
connected cylinder of larger diameter (a mechanical
advantage) at the expense of displacement (Pascal’s
principle, the same mechanism as a hydraulic jack).
This simple exchange of force and displacement is
analogous to muscles acting on joints to create lever-
age, and it can be similarly controlled through antag-
onistic muscles (circumferential and longitudinal).
Body muscles of caterpillars
In contrast to worms, caterpillars do not have mus-
cles arranged in these antagonistic groups; they have
longitudinal and oblique, but no circumferential,
muscles (Fig. 1A) (Snodgrass 1961;Randall 1968;
Belton 1969;Tsujimura 1983;Eaton 1988). There
are small muscle fibers with different orientations
but these are associated with folds near the spiracle
and prolegs. There are two major muscle layers; the
one adjacent to the body wall is called the external
layer and the deeper one is called the internal layer.
Barth proposed that external muscles control body
turgor and internal ones generate major movements
(Barth 1937) and this is commonly restated (Levine
and Truman 1985). Barth based his assertion on the
arrangement of smaller external muscles that connect
the inter-segmental folds with the intra-segmental
body wall. He proposed that these muscles resist out-
ward bulging when the large internal muscles con-
tract. This is a reasonable suggestion but there is no
evidence indicating that the internal and external
muscles are functionally distinct. The lack of obvious
muscle antagonism has interesting consequences for
Fig. 1 (A) The major muscles in a representative body segment
(A4) of Manduca. Ventral (V) and dorsal (D) muscles are named
as internal (I) or external (E) and medial (M), lateral (L), or
oblique (O). TP-tergo-pleural. (B) The overall anatomy of the
nerve cord with cerebral ganglion (CG), subosophogeal (SOG),
thoracic (T1-3), abdominal (A1-6), and terminal ganglia. Ganglion
A3 is shown in detail with major nerves and connectives labeled.
Caterpillar locomotion 3
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

the control of larval movements: the forces generated
by muscles are distributed to other muscles and body
tissues in a complex fashion through changes in
muscle stiffness and shifts in the relative positions
of the attachment-points of muscles. This might be
viewed as a dynamic skeleton capable of producing
movement both through direct actuation and
through changes in stiffness (Sato et al. 2011,2012).
Caterpillars’ tissue and the flow of fluid
In Manduca, the body is not divided by septa so
fluid and tissues can move back and forth between
segments. This is most obvious during simple crawl-
ing movements when the gut slides forward in syn-
chrony with the swing phase of the terminal prolegs.
It can be displaced as much as a segment before the
body wall and other tissues ‘‘catch-up’’ during suc-
cessive proleg steps. This was first reported using
synchrotron-sourced X-rays to visualize movements
of internal tissues in Manduca (Simon et al. 2010b)
but it can also be observed directly in the transparent
larvae of the Brazilian skipper, Calpodes sp. (Fig. 2).
It is thought that this movement of the gut contrib-
utes to locomotion by shifting the center of mass
forwards and also by acting as a ‘‘visceral piston’’
to help extend the anterior segments and thorax
forwards (Simon et al. 2010b). Presumably
the movements of the gut also result in a flow
of hemolymph that contributes to local differences in
pressure throughout the body cavity (see below).
Although caterpillars pressurize their hemocoel
and stiffen it by increasing muscular tension, they
do not use pressure to control their limbs or to
drive locomotion. This was demonstrated in
Manduca by implanting a pressure sensor at the
base of the prolegs and examining how pressure
changed with the proleg’s movements. Surprisingly,
extension (and adduction) of the proleg was more
likely to correlate with a pressure drop than with an
increase, suggesting that pressure changes resulted
from proleg movement instead of the other way
round (Mezoff et al. 2004). We have also used
these sensors to measure changes in pressure
during crawling and find that it fluctuates between
1 and 2 kPa but is not correlated very well with the
step cycle. By implanting two sensors that simulta-
neously record pressure in different parts of the
hemocoel it has been found that Manduca is not
iso-barometric; pressure fluctuations vary with loca-
tion (Fig. 3). These pressure differences within the
body cavity must be associated with fluid flow.
Presumably, as the gut and other tissue move back
Fig. 2 The body wall of Calpodes sp. is transparent and internal
tissues such as the Malpighian tubules (MT) and gut-associated
trachea (GAT) can be seen moving forward in synchrony with the
swing phase of the terminal prolegs. This occurs before the rest
of the body is in motion. Reference lines are drawn showing the
initial anterior margin of the MT and the motionless A4 proleg.
Fig. 3 (A) Pressure changes can be measured at multiple places
inside the hemocoel of Manduca.(B) Hemocoel pressure is
comparatively low and changes do not correlate very well with
crawling movements. The proleg’s and thoracic leg’s swing phases
are shown (cross-hatched boxes) for two successive crawls along
with the corresponding variations in pressure in the posterior
and anterior parts of the abdomen. These pressure differences
imply that fluid moves throughout the body cavity. Details of the
pressure sensors, their calibration, and the procedures for
implanting catheters are provided in Mezoff et al. (2004).
4B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

and forth, hemolymph is displaced in the narrow
space between the gut and the body wall to create
complex flows and pressure differentials.
It appears that caterpillars have decoupled
motion-control from changes in body pressure and
instead remain relatively soft and able to conform to
complex substrates. Although it is speculative, it
is possible that this decoupling is related to the
air-filled trachea that are characteristics of insects
(Harrison et al. 2013) but not of worms or mollusks.
Because air is compressible, the trachea are compli-
ant and tend to reduce the effectiveness and preci-
sion of the transfer of hydraulic force. Furthermore,
the spiracles open from time to time and exchange
air with the environment so the body is not neces-
sarily a fixed volume. We have verified this by
observing Manduca crawling underwater: air bubbles
expand and contract as air is expelled and drawn
back through the spiracles during normal crawling.
In addition to the problems caused by compress-
ible compartments, using hydrostatic pressure to
regulate body stiffness can be energetically expensive.
Pressurization is most effective for small spherical or
cylindrical shapes. We have shown that the bending
stiffness of an anisotropic cylinder (resembling
Manduca’s body wall) increases with pressure,
radius, and wall-thickness but this reaches an asymp-
totic limit (Lin et al. 2011b). Furthermore, when this
bending stiffness is expressed as mass-specific stiff-
ness, it is found that, at any given pressure, small
caterpillars are much stiffer relative to their own
body weight than are large caterpillars. This could
partially explain why small, slender caterpillars are
more likely to cantilever their bodies than are
large, rotund animals.
Kinematics of crawling
The most detailed kinematic analysis of caterpillars’
locomotion has focused on crawling in the tobacco
hornworm M. sexta (Trimmer and Issberner 2007).
All the movements of the prolegs and thoracic legs
are bilaterally synchronized. Each cycle of a crawl
begins with the terminal prolegs releasing from the
substrate and moving forward (swing phase) to
re-establish contact (the start of stance phase) at
approximately the former position of the prolegs in
abdominal segment 6 (A6). This wave of movement
progresses anteriorly, with each proleg taking one
step in each crawl cycle. Although other stepping
patterns can occur, the most common sequence of
proleg timing is for all mid-body prolegs (P6 to A3)
to successively enter swing phase and then to enter
stance phase in the same sequence. The thoracic
legs also take part in the anterograde sequence al-
though they may also take ‘‘extra’’ steps, or make
out-of-sequence movements, as they re-grip the sub-
strate (Johnston and Levine 1996b). Unlike the limbs
of most land animals, the mid-body prolegs do not
rotate in the plane of forward motion but instead are
lifted by the dorsal flexion of the body segment and
then placed back on the substrate. By tracking the
origin (close to the spiracle), and insertion (at the tip
of the proleg), of the mid-body proleg-retractor
muscles, it can be seen that the points remain ap-
proximately parallel and at a fixed distance from one
another; the prolegs appear to act as support struts
(Fig. 4). In contrast, the terminal proleg shortens
considerably during each cycle (Fig. 4) and the tip
angles posteriorly at the end of stance and rotates
into a forward-facing position as it enters swing
phase (Trimmer and Issberner 2007). This rotation
resembles the leg-motion of a running biped
(an inverted spring-loaded pendulum) but the termi-
nal proleg is actually exerting very different forces
(see below). Another important finding is that the
mid-body prolegs do not retract very much (if at all)
when they enter swing phase. Instead, forward move-
ment of each body segment precedes release of the
grip, thereby causing the proleg to stretch slightly
before the retractor muscles release the crochets
and the proleg recoils to its pre-stretch dimensions
(Belanger and Trimmer 2000).
Fig. 4 The upper and lower tracks of proleg-retractor muscles
are traced during a crawl. (A) In mid-body segment A3, the
proleg does not shorten (tracks are parallel) and the angle
changes very little. (B) In contrast, the terminal prolegs shorten
and extend each cycle and are held at a pronounced angle during
stance. Crawling kinematics are quantified in detail (Trimmer and
Issberner 2007).
Caterpillar locomotion 5
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

Interestingly, each mid-body segment shortens
during the swing phase of its associated proleg but
the diameter of the segment changes very little. Most
of the shortening is due to folding and unfolding at
the inter-segment boundaries rather than to stretch-
ing of the body wall itself (Trimmer and Issberner
2007). Lifting and moving forward of the segment
are in phase with one another and there is a 30–
358delay in phase between successive body segments.
During horizontal, upright, crawling the vertical
displacement and horizontal velocity are in phase
with one another, consistent with the interchange
of kinetic and gravitational potential energy, presum-
ably through the storage and release of elastic energy.
Dynamics of crawling
To understand how these proleg movements contrib-
ute to forward movement, it is necessary to measure
the ground forces both in the direction of travel
(axial) and orthogonal to the plane of travel
(normal). We designed and built a multi-beam
array of sensors to measure these GRFs at several
contact points simultaneously (Lin and Trimmer
2012). A major challenge was to devise a system
that could support the considerable mass of a cater-
pillar while capturing the very small induced forces
of a slowly moving multi-legged animal (Lin and
Trimmer 2010a). Two remarkable findings are that:
(1) the prolegs release from the substrate so effec-
tively they do not lift the surface (no negative
normal GRF) and (2) for a substantial proportion
of the crawl cycle the posterior prolegs resist forward
motion (Fig. 5; negative axial forces). In contrast to
the limbs of terrestrial vertebrates, the prolegs do not
generate much thrust but instead drag as the head
and thorax move forward. This creates internal
tension in the abdomen that is released when the
prolegs retract their crochets. The effect of this ten-
sion-based strategy is that compressive forces are
transmitted through the substrate instead of through
the body tissues. Manduca, and presumably other
crawling caterpillars, cannot locomote on substrates
softer than themselves (Lin and Trimmer 2010b). Of
course this is not the only way caterpillars move.
Manduca can also pressurize itself enough to canti-
lever its entire body in space, and held only by the
terminal prolegs. Furthermore, the biomechanics of
inching and hybrid inching/crawling gaits have not
been closely examined.
Body coordination, neural and muscular
coordination, and timing of the prolegs’
movements
Organization of motor control
The classical studies of caterpillar behavior by Kopec,
von Holst, and Barth early in the twentieth century
(Kopec 1919;Holst 1934;Barth 1937) concentrated
on the role of body-wall musculature and the neural
control of stepping patterns. In the absence of phar-
macological or electrophysiological methods, these
studies relied on anatomical and surgical techniques.
It was found that the overall tonus of the body de-
pends on intact nerves to the musculature; cutting
these nerves resulted in bulging of the body wall
because of the loss of tension. Cutting one side of
the connective (see Fig. 1) between T2 and T3 shifted
the timing of the left and right stepping motions
posterior to the cut with the cut side phase-delayed
relative to the intact side. Although this suggests an-
terior control of R–L synchrony, it was also noted
that the overall muscle tone was weaker on the cut
side, perhaps accounting for the lag in phase of
movements (Kopec 1919;Holst 1934). Severing the
connectives between the first thoracic and subeso-
phageal ganglion prevented spontaneous coordinated
crawling and cutting the connectives between the
thoracic and abdominal ganglia prevented spontane-
ous crawling in the segments posterior to the cut.
Remarkably, when all the peripheral nerves in
selected body segments were cut, leaving the
connectives intact, anterograde waves were
propagated forward from the terminal segment and
emerged from the paralyzed body segments as if the
activity had passed unaffected through them (Holst
1934;Dominick and Truman 1986a). This suggests
that natural body-movements and sensory feedback
are not essential for the anterograde propagation of
the crawling motor program (but presumably they
are important for normal locomotion).
Fig. 5 Summary of the proleg’s GRFs during Manduca’s upright
horizontal crawling. Each trace represents the forces exerted by
the indicated proleg during a single step-cycle. The upper lines
are normal forces (in the direction of gravity) dominated by
Manduca’s weight. The lower lines filled are axial forces, upward
causing propulsion and downward creating drag. The drag forces
over time have been filled in to show how the posterior seg-
ments resist forward movement during stance. More details are
available in the paper by Lin and Trimmer (2010a).
6B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

Motor programs
On the basis of his observations, Barth proposed a
model of caterpillar locomotion in which the dorsal
and ventral muscles in each segment contracted
alternately to produce the anterograde wave of
dorsal bending (Barth 1937). This was presumed to
be assisted by small lateral muscles oriented more or
less dorso-ventrally. His model has been the most
commonly depicted description of caterpillars’ crawl-
ing. Although he provided no supportive evidence,
the prodigious insect anatomist R.E. Snodgrass
(1961) dismissed the peristaltic model of Barth and
instead emphasized the re-lengthening of body seg-
ments by the elasticity and unfolding of the body
wall as a critical means of propulsion. Through the
use of nerve and electromyogenic recordings, it is
now possible to examine these two models of cater-
pillar locomotion. In the wandering stage of fifth
instar Manduca, shortly before pupation, crawling
is accompanied by a segmentally stereotyped and bi-
laterally symmetrical pattern of muscle-activation
that moves from posterior to anterior (Dominick
and Truman 1986a). In these studies, activity in
the dorsal longitudinal muscles preceded that in
the ventral longitudinal muscles, such that the
ventral muscles in a given segment contracted
synchronously with the dorsal muscles two segments
anteriorly (Dominick and Truman 1986a), appar-
ently supporting Barth’s proposed crawling mecha-
nism. Burrowing was characterized by simultaneous
activation of ventral and dorsal muscles in all the
abdominal body segments (Dominick and Truman
1986a). Other studies of Manduca’s stepping patterns
concentrated on the thoracic legs and their changes
during development. This work includes
Electromyography (EMG) analysis of the levator
and depressor muscles (Johnston and Levine
1996b). Rhythmic motor patterns can be evoked in
isolated nerve cords using bath-applied muscarinic
agonists. These patterns resemble the anterograde
wave of activation seen in natural crawling but the
duration of the cycle is four times slower (Johnston
and Levine 1996a). These en-passant nerve recordings
include many motor units and the contribution of
different motoneurons has not been determined.
More recently, using a combination of bipolar
wire electrodes (Simon et al. 2010a) and flexible
multi-electrode arrays (Metallo et al. 2011), the
coordination of EMG activity in identified muscles
has been correlated with body-movements in more
detail. These experiments verified the anterograde
wave of activation of the muscles but, in contrast
to previous studies, found that dorsal, ventral, and
oblique muscles in each body segment are co-active
(Fig. 6A and B). In fact, activity of some ventral
muscles is so prolonged that muscles in segments
A6 and A3 are co-active for 50% the duration of
their cycles. This means that the large intra-
segmental muscles are active in at least four abdom-
inal segments simultaneously (Fig. 6A and D).
Therefore, crawling consists of a broad wave of
motor activity moving forwards through successive
delays in phase. Despite considerable variation in
the onset and cessation of motor activity in each
segment, crawling movements are very similar cycle
to cycle. The environmental-skeleton makes this
possible because the large longitudinal muscles can
pre-tension each segment and forward movement is
produced by releasing the proleg’s grip in successive
segments. This new model implies that the timing of
the proleg’s release of grip is the most precise and
critical aspect of caterpillars’ locomotion, a theme
that will be addressed again in the context of
different gaits.
The coordination of muscle-activation has been
confirmed with single-neuron resolution using
multi-electrode EMG arrays. Because larval
Manduca muscles are innervated by one (occasion-
ally two) motoneurons (there are no common inhib-
itors or fast or slow motoneurons), it is possible to
measure the exact spike-patterns driving individual
muscles during natural behaviors (Metallo et al.
2011). This provides an unparalleled opportunity to
study the neuromechanics of locomotion by soft-
bodied animals. An example can be seen in studies
of Manduca’s muscle performance measured using
work-loop analysis (Woods et al. 2008). Both the
strain cycle of an individual muscle, and the pattern
and timing of neural activation, can be recorded in
freely behaving animals. The muscle can then be dis-
sected with its associated nerve and mounted in a
computer-controlled ergometer that replays the nat-
ural strain cycle while measuring the applied force.
The work-loop of a muscle carrying out its normal
function can be simulated by stimulating the nerve
using the timing and pattern of nerve impulses that
were recorded during free behavior. For a large intra-
segmental muscle, such as VIL, the natural work-
loop resembles a figure-of-eight, dissipating energy
at large strains and performing positive work when
the muscle shortens (Fig. 7B). The proportions of
these positive and negative work cycles can be
changed by varying the timing and duration of the
stimulus and presumably by changing the frequency
or pattern of nerve impulses. This represents a
real-world transfer of information into mechanical
work that, in turn, makes an excellent case-study
Caterpillar locomotion 7
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

of morphological computation (embodiment)
(Pfeifer and Iida 2005;Pfeifer et al. 2007).
In contrast to the detailed studies of efferent ac-
tivity, the central pathways controlling crawling
remain enigmatic. Surgical lesions of the peripheral
nerves and connectives show that muscles in each
segment are controlled by local ganglia, but proper
coordination of crawling requires inter-segmental in-
formation (Holst 1934). The normal initiation of
crawling requires the subosophogeal ganglion to be
intact and the brain (most likely the central com-
plex), either inhibits, or excites, the initiation of
crawling, depending on the stage of development
(Dominick and Truman 1986a,1986b). Rhythmic
motor activity can be evoked from isolated nerve
cords by the muscarinic agonist pilocarpine
(Johnston and Levine 1996a). This anterograde
wave of nerve activity has been interpreted as ‘‘fictive
crawling’’ driven by a central pattern generator and
indeed, the sequence of activation of motor neurons
in the thorax resembles that seen in EMG recordings.
However, there are substantial differences in the
cycle-time and coordination of natural motor pat-
terns compared with fictive crawling. In our experi-
ments, these motor patterns are also very unstable
and transient. An alternative explanation is that
motor programs produced by isolated nerve cords
arise from activity in dominant neural connections
but they do not necessarily reflect the normal mech-
anisms controlling crawling. The results also imply
that sensory feedback is an essential coordinating in-
fluence for normal crawling (as it is in Drosophila
Fig. 6 (A) During crawling, waves of muscle-activation (red/dark shaded area) move forward and encompass at least four body
segments in each step-cycle. The approximate sequence of movements during activation of the proleg’s retraction is also illustrated
(green/pale shaded). (B) Within each segment dorsal (DIM), ventral (VIL), and oblique (VEO) muscles are co-active much of the time.
(C) Examples of EMG recordings from muscles DIM and VIL in segment A4 (upper two traces) and from VIL in segments A4 and A6 of
a different animal (lower two traces). (D) Although the onset and end of EMG activity is phase-delayed between segments A6 and A3,
muscles in each segment are co-active for 50% of their cycle leading to the broad activation seen in A. In (B) and (D), the times of EMG
onset, peak activity, and offset are plotted relative to the time and duration of the swing phase of the A3 proleg (gray bar). Values are
means (s.e.m.) for six animals (5–10 steps each), adapted from Simon et al. (2010a).
8B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

larvae) (Hughes and Thomas 2007;Song et al. 2007).
During Manduca’s fictive crawling, rhythmic activity
in the abdominal ganglia excites interneurons in the
suboesophageal ganglion, which in turn produce
rhythmic drive to the neuromodulatory unpaired
median neurons (Johnston and Levine 1996a;
Johnston et al. 1999). It is not known whether this
circuit is involved in normal crawling.
Sensory feedback and integration
The signaling responses of Manduca’s stretch-recep-
tor organs are not well-matched to a role in loco-
motion, nor are they essential for normal crawling
(Simon and Trimmer 2009) so it is likely that either
tactile hairs (Zacharuk and Shields 1991) or subder-
mal multi-dendritic (MD) neurons (Grueber et al.
2001) provide feedback for crawling and initiate
changes in gait on different substrates. It has been
demonstrated that stimulation of sensory hairs at the
tip of the proleg can produce an assistance reflex
(Belanger et al. 2000) during Manduca’s crawling
and that these hairs also help negotiate irregularities
in the surface. After prolegs on abdominal segment
A3 have encountered a small obstacle in their path,
the rate of proleg-lift in subsequent body segments is
increased in anticipation of stepping onto the obsta-
cle (van Griethuijsen and Trimmer 2010). This
anticipation requires a transient ‘‘memory’’ and
transfer of information along the nerve cord. The
behavior is supplemented by local touch-sensing
that can be manipulated by cutting the hairs. The
role of the MD neurons in Manduca is unknown
but in Drosophila larvae, specific groups of MD neu-
rons are important for normal locomotion (Fox et al.
2006;Hughes and Thomas 2007;Song et al. 2007;
Heckscher et al. 2012). It is expected that MD
neurons play a similar role in caterpillars by encod-
ing local forces, displacements, and vibrations for
mechanical feedback.
Other inching, hybrid, and crawling gaits
and the role of proleg timing
Although it has been studied in detail, crawling is
only one of a variety of caterpillars’ gaits. Manduca
itself uses several different stepping patterns, particu-
larly when climbing vertically (C. Metallo and B. A.
Trimmer, submitted for publication). The most
common alternative gait in other species is inching,
exemplified by the geometrids. Inching has the
longest possible length of step so, for a given cycle
frequency, it is the fastest gait. However, because
much of the body is lifted away from the substrate
during the swing phase, inching relies on a stable
grip by the posterior prolegs and the thorax, each
of which must alternately support the entire weight
of the caterpillar. As the mid-body segments lift away
from the substrate, they create lateral instability,
thereby further decreasing the safety factor. Inching
also requires that the body bends sharply, sometimes
turning as much as 1808within a few body segments.
This would be very difficult to achieve with a stiff or
highly pressurized body and it also favors long aspect
ratios. Although most inching caterpillars are rela-
tively thin and long, there are a surprising number
of inching species with relatively thick bodies (e.g.,
Fig. 8).
The gaits of different caterpillar species appear to
correlate with the number and arrangement of
prolegs. The ancestral lepidopterans probably had
four pairs of abdominal prolegs (A3–A6) and one
pair of anal prolegs (Hinton 1955) but reduction
of the prolegs has occurred in many groups, leading
to a diversity of arrangements (Bitsch 2012). For ex-
ample, some Noctuidae have reduced A3 prolegs and
Fig. 7 (A) An isolated VIL muscle from Manduca is shortened
and lengthened, using both a range and strain rate typically seen
during crawling. The force required to do this in passive muscle is
shown (passive force, blue). During stimulation of the VIL motor
nerve (20 Hz) additional force is generated (active force, red).
(B) Plots of muscle force (F) against length (L) reveal complex
responses that depend on the timing of the stimulus (top) and its
duration (bottom). Clockwise loops dissipate work; counter-
clockwise perform active work. Both loops can occur in one
strain cycle. Adapted from Woods et al. (2008).
Caterpillar locomotion 9
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

terminal prolegs, and the subfamily Catocalinae have
even lost the A4 prolegs. Most strikingly, the
Geometridae (loopers) use the thoracic legs together
with the A6 and anal prolegs to crawl by inching
(Wagner 2005). Indeed, proleg development appears
to be quite labile and can be genetically suppressed
or induced in any given segment (Warren et al. 1994;
Suzuki and Palopoli 2001). The relationship between
gaits, and the arrangements of prolegs, has interest-
ing repercussions for motor control. From kinemat-
ics, it appears that crawling and inching mainly differ
in the ‘‘wavelength’’ of each step but the discovery of
broad, multi-segment, activation of muscles during
crawling suggests another possibility. Once longitu-
dinal muscles have been activated to tension the
body, the sequence of subsequent movement is lar-
gely determined by the timing of the release of grip.
Therefore, inching, and hybrids of inching and
crawling, could be based on the same anterograde
wave of muscle-activation but differ in attachment
by the prolegs and in body mechanics. If a proleg
is missing, or fails to grip, the most likely outcome is
an upward bending of that body segment. This could
be the genesis of inching forms of locomotion.
In addition to basic locomotion, some caterpillars
have evolved special adaptations including ballistic
rolling (Brackenbury 1997,1999) and silk-climbing
(Brackenbury 1996;Sugiura and Yamazaki 2006).
There are no neural or mechanical data on these
behaviors but the kinematics have been charaterized
in detail. Ballistic rolling in Pleurotya ruralis results
from the continuation of a ‘‘backward gallop’’ in
which the entire body becomes flexed and forms a
wheel that rolls backwards. In other species, such as
Cacoecimorpha pronubana and members of the
Crambidae family (Lin et al. 2011a), this process is
suffciently rapid that the whole body is flicked into
the air. The resultant movements appear uncon-
trolled and unstable but function as a startling
escape maneuver. Silk-climbing typically involves
swinging the body back and forth while winding
threads onto the thoracic legs using alternate left
and right movements. Some species, especially
those that construct silk nests, can also use looping
or wavelike movements of the abdomen to help pull
themselves up a silk life-line (Sugiura and Yamazaki
2006). Presumably this process is very different from
climbing on a stiff substrate because compressive
forces cannot be applied along the silk thread (Lin
and Trimmer 2010b).
Control-strategies for highly deformable
robots
As an alternative approach to understanding cater-
pillar locomotion, we have developed a series of
robots made from soft materials (Lin et al. 2011a,
2013;Kim et al. 2013;Umedachi et al. 2013). These
devices are monolithic elastomeric structures that
can be actuated using ‘‘artificial muscles’’ (Fig. 9).
Each device is designed using CAD software and
Fig. 8 Examples of caterpillars’ gaits. Each series of images show a complete cycle of horizontal upright locomotion: (A)Sphacelodes sp.
uses an inching pattern, (B)Selenisa sp. and (C)Gonodonta sp. appear to combine crawling and inching, (D) an unknown species using a
pattern of crawl similar to that of Manduca. All images are from wild-caught species in Costa Rica.
10 B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

then cast from silicone materials in 3-D-printed
molds or printed directly using ultraviolet-cured
elastomers. We have chosen shape-memory alloy
(SMA) coils as the primary actuators because, like
muscles, they are soft, passively-elastic linear actua-
tors that generate force by contraction. Also, like
muscles, they must be re-extended with an opposing
force. By taking into account the transition in mate-
rial properties as they are activated, and the geomet-
ric effects of coiling, it is possible to design SMA
coils for different applications (An et al. 2012).
Each coil used in our devices is only a fraction of
a millimeter in diameter and formed from a thin
nickel titanium wire. In its resting state the coil is
easily stretched by low loads but when heated, using
a 200-mA current, it undergoes a crystalline trans-
formation to create forces as high as 0.4 N and
strains as high as 200%. Using pulse-width or fre-
quency-modulated current pulses, it is possible to
control the power applied to a coil to produce
muscle-like responses (Lin et al. 2011a). The maxi-
mum practical duration of a cycle is approximately 1
s (very similar to the ventral internal muscle of
Manduca). Rather than attempt to copy the detailed
anatomy of a caterpillar or its thousands of muscles,
we have simplified the robot’s design to capture sev-
eral mechanical features expected to be important for
caterpillar locomotion. This includes making the
body from an elastomer so that it is very soft and
attaching internal soft actuators that are arranged
axially to deform different parts of the body, much
as longitudinal muscles shorten the abdomen of
Manduca. More details of this design process are
available in Lin et al. (2013).
Most devices have between two and four length-
wise SMA coils arranged in anterior–posterior and
left–right configurations. We have designed passive
gripping systems based on deployable sticky pads
(Lin et al. 2013) or by simple orientation-dependent
changes in materials in contact with the surface
(Umedachi et al. 2013). Gaits can be generated by
scaling the power and timing of the commands. For
example, a robot with a loose-crawling gait switches
to a fast inching gait if the pattern of the gait is
temporally compressed to reduce the difference in
phase between posterior and anterior flexion. We
have not yet tested the effects of the timing of the
grip but our observations of caterpillars suggest that
this could be a very effective way of controlling tran-
sitions in gait. Interestingly, by further increasing the
actuator power, it is possible to tip the robot over
into a forward tumble with sufficient momentum to
produce ballistic rolling (Lin et al. 2011a). This
movement resembles the escape responses of many
small caterpillars (see above), including Pleuroptya
ruralis (Brackenbury 1999), and several caterpillars
from the family Crambidae, that flex into a wheel
shape and can even flick into the air when startled
(see above) (Lin et al. 2011a). On a flat, level surface,
this mode of locomotion boosts the immediate speed
over 10-fold. Soft robots provide a novel opportunity
to better understand the mechanics of this sort of
motion and also to explore how it is controlled
and how it could have evolved from existing motor
strategies.
Major questions remaining
Although there have been significant advances in our
understanding of caterpillars’ locomotion, there is
still no general conceptual framework describing
the mechanics of the locomotion of soft-bodied
animals, or its control (Trimmer 2013). This is a
major challenge that needs to be addressed so that
the relationship between neural commands and
body-movements can be described and predicted.
Not only will this help our understanding of how
animal behavior has evolved but also will provide
tools for engineering better devices. This, in turn,
may prove useful for incorporating soft materials
into safe, adaptable, assistive robots. Such machines
will find uses wherever there are close interactions
with humans or the environment.
More specific questions remain about caterpillars’
locomotion. In particular, how do caterpillars
Fig. 9 Soft caterpillar-like robots. (A) GoQBot is the prototype
device fabricated with a composite body consisting of several
mixtures of silicone rubbers. Five infrared LEDs were bonded on
the left side of the body for kinematical tracking. All the signal
wires were gathered into a 50-cm tether braid extending from
the right side of the robot around the center of mass. Two-SMA
coils were threaded through the entire length of the robot, side
by side. From Lin et al. (2011a) (B) InchBot V is a radio-
controlled robot that can inch and crawl. (C) The Softworm
robots are a series of devices 3-D-printed in soft elastic polymer
to explore the role of material properties in robot control.
Caterpillar locomotion 11
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

transition from the environmental-skeleton strategy
during crawling, to a stiffened hydrostatic skeleton
for cantilevering behavior? During pressurization,
what are the repercussions for gas exchange and do
caterpillars actively regulate spiracles both for
mechanical and for respiratory functions? Finally,
little is known about the way that sensory informa-
tion is used to control the locomotion of soft-bodied
animals. The response properties of stretch-receptor
organs in Manduca seem better suited to measuring
the long-term changes in shape and size, but do they
also contribute proprioceptive information for loco-
motion? It seems more likely that mechanosensors
(both filiform hairs and MD neurons) are important
for detecting and adapting to the environment but
how such information is integrated and used for
motor control is the subject of future research.
Acknowledgments
The authors would like to thank Drs Takuya
Umedachi and Vishesh Vikas for their permission
to use Fig. 9C and all the other colleagues on the
Tufts ChemBot team for their constructive feedback
during regular project meetings.
Funding
This work was supported by the National Science
Foundation IOS-1050908 [to B.A.T.]; and the
Defense Advanced Research Projects Agency
ChemBot project [W911SR-08-C-0012 to B.A.T.
and Dr David Kaplan]. Support for participation in
this symposium was provided by the Company of
Biologists and the Society for Integrative and
Comparative Biology (Divisions of Vertebrate
Morphology, Comparative Biomechanics,
Neurobiology, and Animal Behavior).
References
Accoto D, Castrataro P, Dario P. 2004. Biomechanical analysis
of Oligochaeta crawling. J Theor Biol 230:49–55.
An S-M, Ryu J, Cho M, Cho K-J. 2012. Engineering design
framework for a shape memory alloy coil spring actuator
using a static two-state model. Smart Mater Struct 21:16.
Barth R. 1937. Muskulatur und Bewegungsart der Raupen.
Zool Jahrb Anat 62:507–66.
Belanger JH, Bender KJ, Trimmer BA. 2000. Context-depen-
dency of a limb-withdrawal reflex in the caterpillar
Manduca sexta. J Comp Physiol A 186:1041–8.
Belanger JH, Trimmer BA. 2000. Combined kinematic and
electromyographic analyses of proleg function during
crawling by the caterpillar Manduca sexta. J Comp
Physiol A 186:1031–9.
Belton P. 1969. Innervation and neural excitation of ventral
muscle fibres of the larva of the waxmoth, Galleria mello-
nella. J Insect Physiol 15:731–4. IN1–IN2, 735–41.
Bitsch J. 2012. The controversial origin of the abdominal ap-
pendage-like processes in immature insects: are they true
segmental appendages or secondary outgrowths?
(Arthropoda hexapoda). J Morphol 273:919–31.
Brackenbury J. 1996. Novel locomotory mechanism in cater-
pillars: life-line climbing in Epinotia abbreviana
(Tortricidae) and Yponomeuta padella (Ypononmeutidae).
Physiol Entomol 21:7–14.
Brackenbury J. 1997. Caterpillar kinematics. Nature 390:453.
Brackenbury J. 1999. Fast locomotion in caterpillars. J Insect
Physiol 45:525–33.
Chiel HJ, Beer RD. 1997. The brain has a body: adaptive
behavior emerges from interactions of nervous system,
body and environment. Trends Neurosci 20:553–7.
Chiel HJ, Ting LH, Ekeberg O, Hartmann MJZ. 2009. The
brain in its body: motor control and sensing in a biome-
chanical context. J Neurosci 29:12807–14.
Dominick OS, Truman JW. 1986a. The physiology of
wandering behaviour in Manduca sexta. III. Organization
of wandering behaviour in the larval nervous system. J Exp
Biol 121:115–32.
Dominick OS, Truman JW. 1986b. The physiology of
wandering behaviour in Manduca sexta. IV. Hormonal in-
duction of wandering behaviour from the isolated nervous
system. J Exp Biol 121:133–51.
Dorfmann A, Trimmer BA, Woods WA Jr. 2007. A constitu-
tive model for muscle properties in a soft-bodied arthro-
pod. J R Soc Interface 4:257–69.
Dorfmann AL, Woods WA, Trimmer BA. 2008. Muscle per-
formance in a soft-bodied terrestrial crawler: constitutive
modelling of strain-rate dependency. J R Soc Interface
5:349–62.
Eaton JL. 1988. Lepidopteran anatomy. New York (NY): John
Wiley and Sons.
Fox LE, Soll DR, Wu C-F. 2006. Coordination and modula-
tion of locomotion pattern generators in Drosophila larvae:
effects of altered biogenic amine levels by the tyramine beta
hydroxlyase mutation. J Neurosci 26:1486–98.
Grueber WB, Graubard K, Truman JW. 2001. Tiling of the
body wall by multidendritic sensory neurons in Manduca
sexta. J Comp Neurol 440:271–83.
Gutfreund Y, Flash T, Fiorito G, Hochner B. 1998. Patterns of
arm muscle activation involved in octopus reaching move-
ments. J Neurosci 18:5976–87.
Harrison JF, Waters JS, Cease AJ, VandenBrooks JM,
Callier V, Klok CJ, Shaffer K, Socha JJ. 2013. How locusts
breathe. Physiology 28:18–27.
Heckscher ES, Lockery SR, Doe CQ. 2012. Characterization of
Drosophila larval crawling at the level of organism, segment,
and somatic body wall musculature. J Neurosci
32:12460–71.
Hinton HE. 1955. On the structure, function and distribution
of the prolegs of the panorpoidea, with a criticism of the
Berlese–Imms theory. Trans R Ent Soc Lond B
106:455–541.
Holst EV. 1934. Motorische und tonischeErregung und ihr
Bahnenverlauf bei Lepidopteranlarven. Z Vergl Physiol
21:395–414.
Hughes CL, Thomas JB. 2007. A sensory feedback circuit co-
ordinates muscle activity in Drosophila. Mol Cell Neurosci
35:383–96.
12 B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

Johnston RM, Consoulas C, Pfluger H, Levine RB. 1999.
Patterned activation of unpaired median neurons during
fictive crawling in Manduca sexta larvae. J Exp Biol
202(Pt 2):103–13.
Johnston RM, Levine RB. 1996a. Crawling motor patterns
induced by pilocarpine in isolated larval nerve cords of
Manduca sexta. J Neurophysiol 76:3178–95.
Johnston RM, Levine RB. 1996b. Locomotory behavior in the
hawkmoth Manduca sexta: kinematic and electromyo-
graphic analyses of the thoracic legs in larvae and adults.
J Exp Biol 199(Pt 4):759–74.
Kier W, Leeuwen J. 1997. A kinematic analysis of tentacle
extension in the squid Loligo pealei. J Exp Biol 200(Pt
1):41–53.
Kier WM, Smith KK. 1985. Tongues, tentacles and trunks—
the biomechanics of movement in muscular-hydrostats.
Zool J Linn Soc 83:307–24.
Kim S, Laschi C, Trimmer B. 2013. Soft robotics: a
bioinspired evolution in robotics. Trends Biotechnol
31:287–94.
Kopec S. 1919. Lokalizationsversuche an Zentralen
Nervensysyteme der Raupen und Falter. Zool Jahrb
Physiol 36:453–502.
Levine RB, Truman JW. 1985. Dendritic reorganization of
abdominal motoneurons during metamorphosis of the
moth, Manduca sexta. J Neurosci 5:2424–31.
Lin H-T, Leisk G, Trimmer BA. 2011a. GoQBot: a caterpillar-
inspired soft-bodied rolling robot. Bioinspir Biomim
6:026007–21.
Lin H-T, Leisk GG, Trimmer BA. 2013. Soft robots in space: a
perspective for soft robotics. In: Barker DJ, editor. Acta
Futura. ESTEC, PPC-PF, 2201 AZ, Noordwijk, The
Netherlands: European Space Agency, Advanced Concepts
Team. p. 69–79.
Lin H-T, Trimmer BA. 2010a. The substrate as a skeleton:
ground reaction forces from a soft-bodied legged animal.
J Exp Biol 213(Pt 7):1133–42.
Lin HT, Dorfmann AL, Trimmer BA. 2009. Soft-cuticle
biomechanics: a constitutive model of anisotropy for cater-
pillar integument. J Theor Biol 256:447–57.
Lin HT, Slate DJ, Paetsch CR, Dorfmann AL, Trimmer BA.
2011b. Scaling of caterpillar body properties and its biome-
chanical implications for the use of a hydrostatic skeleton.
J Exp Biol 214(Pt 7):1194–204.
Lin HT, Trimmer BA. 2010b. Caterpillars use the substrate as
their external skeleton: a behavior confirmation. Commun
Integr Biol 3:471–4.
Lin HT, Trimmer BA. 2012. A new bi-axial cantilever beam
design for biomechanics force measurements. J Biomech
45:2310–4.
Matzner H, Gutfreund Y, Hochner B. 2000. Neuromuscular
system of the flexible arm of the octopus: physiological
characterization. J Neurophysiol 83:1315–28.
Metallo C, White RD, Trimmer BA. 2011. Flexible parylene-
based microelectrode arrays for high resolution EMG
recordings in freely moving small animals. J Neurosci
Methods 195:176–84.
Mezoff S, Papastathis N, Takesian A, Trimmer BA. 2004. The
biomechanical and neural control of hydrostatic limb
movements in Manduca sexta. J Exp Biol 207(Pt
17):3043–53.
Nishikawa K, Biewener AA, Aerts P, Ahn AN, Chiel HJ,
Daley MA, Daniel TL, Full RJ, Hale ME, Hedrick TL,
et al. 2007. Neuromechanics: an integrative approach for
understanding motor control. Integr Comp Biol 47:1–39.
Paetsch C, Trimmer BA, Dorfmann A. 2012. A constitutive
model for active–passive transition of muscle fibers. Int J
Nonlinear Mech 47:377–87.
Pfeifer R, Iida F. 2005. Morphological computation: connect-
ing body, brain and environment. Jpn Sci Mon 58:48–54.
Pfeifer R, Iida F, Bongard J. 2005. New robotics: design prin-
ciples for intelligent systems. Artif Life 11:99–120.
Pfeifer R, Lungarella M, Iida F. 2007. Self-organization, em-
bodiment, and biologically inspired robotics. Science
318:1088.
Quillin KJ. 1998. Ontogenetic scaling of hydrostatic skeletons:
geometric, static stress and dynamic stress scaling of the
earthworm Lumbricus terrestris. J Exp Biol 201(Pt
12):1871–83.
Quillin KJ. 1999. Kinematic scaling of locomotion by hydro-
static animals: ontogeny of peristaltic crawling by the earth-
worm Lumbricus terrestris. J Exp Biol 202(Pt 6):661–74.
Randall WC. 1968. Anatomical changes in the neuromuscular
complex of the proleg of Galleria mellonella (L.) (Lepidoptera:
Pyralidiadae) during metamorphosis. 125:105–27.
Sato T, Kano T, Ishiguro A. 2011. On the applicability of the
decentralized control mechanism extracted from the true
slime mold: a robotic case study with a serpentine robot.
Bioinspir Biomim 6:026006.
Sato T, Kano T, Ishiguro A. 2012. A decentralized control
scheme for an effective coordination of phasic and tonic
control in a snake-like robot. Bioinspir Biomim 7:016005.
Simon MA, Fusillo SJ, Colman K, Trimmer BA. 2010a. Motor
patterns associated with crawling in a soft-bodied arthro-
pod. J Exp Biol 213(Pt 13):2303–9.
Simon MA, Trimmer BA. 2009. Movement encoding by a
stretch receptor in the soft-bodied caterpillar, Manduca
sexta. J Exp Biol 212(Pt 7):1021–31.
Simon MA, Woods WA Jr, Serebrenik YV, Simon SM, van
Griethuijsen LI, Socha JJ, Lee WK, Trimmer BA. 2010b.
Visceral-locomotory pistoning in crawling caterpillars.
Curr Biol 20:1458–63.
Skierczynski BA, Wilson RJ, Kristan WB Jr, Skalak R. 1996. A
model of the hydrostatic skeleton of the leech. J Theor Biol
181:329–42.
Smith KK, Kier WM. 1989. Trunks, tongues, and tentacles—
moving with skeletons of muscle. Am Sci 77:29–35.
Snodgrass RE. 1961. The caterpillar and the butterfly.
Smithson Misc Collect 143:51.
Song W, Onishi M, Jan LY, Jan YN. 2007. Peripheral multi-
dendritic sensory neurons are necessary for rhythmic loco-
motion behavior in Drosophila larvae. Proc Natl Acad Sci
USA 104:5199–204.
Sugiura S, Yamazaki K. 2006. The role of silk threads as life-
lines for caterpillars: pattern and significance of lifeline-
climbing behaviour. Ecol Entomol 31:52–7.
Sumbre G, Fiorito G, Flash T, Hochner B. 2005.
Neurobiology: motor control of flexible octopus arms.
Nature 433:595–6.
Suzuki Y, Palopoli MF. 2001. Evolution of insect abdominal
appendages: are prolegs homologous or convergent traits?
Dev Genes Evol 211:486–92.
Caterpillar locomotion 13
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from

Trimmer B. 2013. Soft robots. Curr Biol 23:R639–41.
Trimmer B, Issberner J. 2007. Kinematics of soft-bodied,
legged locomotion in Manduca sexta larvae. Biol Bull
212:130–42.
Tsujimura H. 1983. Anatomy of external structure and mus-
culature of abdominal segments in the larva of the lepidop-
teran Pieris rapae crucivora. J Morphol 177:191–203.
Umedachi T, Vikas V, Trimmer B. 2013. Highly deformable
3-D printed soft robot generating inching and crawling
locomotions with variable friction legs. 2013 IEEE/RSJ
International Conference on Intelligent Robots and
Systems. Tokyo, Japan: IEEE. p. 4590–5.
van Griethuijsen LI, Trimmer BA. 2009. Kinematics of hori-
zontal and vertical caterpillar crawling. J Exp Biol 212(Pt
10):1455–62.
van Griethuijsen LI, Trimmer BA. 2010. Caterpillar crawling
over irregular terrain: anticipation and local sensing. J
Comp Physiol A Neuroethol Sens Neural Behav Physiol
196:397–406.
Van Leeuwen JL, De Groot JH, Kier WM. 2000. Evolutionary
mechanics of protrusible tentacles and tongues. Neth J Zool
50:113–39.
Vogel S. 2003. Comparative biomechanics: life’s physical
world. Princeton (NJ): Princeton University Press.
Wagner DL. 2005. Caterpillars of eastern North America: a
guide to identification and natural history. Princeton:
Princeton University Press.
Wainwright SA. 1988. Axis and circumference. Cambridge
(MA): Harvard University Press.
Warren RW, Nagy L, Selegue J, Gates J, Carroll S. 1994.
Evolution of homeotic gene regulation and function in
flies and butterflies. Nature 372:458–61.
Wells MJ. 1976. Octopus: physiology and behaviour of an
advanced invertebrate. New York (NY): Chapman and
Hall, John Wiley & Sons.
Wilson JF, Mahajan U, Wainwright SA, Croner LJ. 1991. A
continuum model of elephant trunks. J Biomech Eng
113:79–84.
Woods WA Jr, Fusillo SJ, Trimmer BA. 2008. Dynamic prop-
erties of a locomotory muscle of the tobacco hornworm
Manduca sexta during strain cycling and simulated natural
crawling. J Exp Biol 211(Pt 6):873–82.
Zacharuk RY, Shields VD. 1991. Sensilla of immature insects.
Annu Rev Entomol 36:331–54.
14 B. A. Trimmer and H.-t. Lin
at Tufts University on June 19, 2014http://icb.oxfordjournals.org/Downloaded from