2013 IEEE International Conference on Rehabilitation Robotics June 24-26, 2013 Seattle, Washington USA

Biologically-inspired Soft Exosuit

Alan T. Asbeck, Robert J. Dyer, Arnar F. Larusson, Conor J. Walsh

School of Engineering and Applied Sciences
The Wyss Institute for Biologically Inspired Engineering
Harvard University
Cambridge, MA 02138 USA
walsh@seas.harvard.edu

Abstract— In this paper, we present the design and evaluation

of a novel soft cable-driven exosuit that can apply forces to the
body to assist walking. Unlike traditional exoskeletons which
contain rigid framing elements, the soft exosuit is worn like
clothing, yet can generate moments at the ankle and hip with
magnitudes of 18% and 30% of those naturally generated by the
body during walking, respectively. Our design uses geared
motors to pull on Bowden cables connected to the suit near the
ankle. The suit has the advantages over a traditional exoskeleton
in that the wearer's joints are unconstrained by external rigid
structures, and the worn part of the suit is extremely light, which
minimizes the suit's unintentional interference with the body's
natural biomechanics. However, a soft suit presents challenges
related to actuation force transfer and control, since the body is
compliant and cannot support large pressures comfortably. We
discuss the design of the suit and actuation system, including
principles by which soft suits can transfer force to the body
effectively and the biological inspiration for the design. For a soft
exosuit, an important design parameter is the combined effective
stiffness of the suit and its interface to the wearer. We
characterize the exosuit’s effective stiffness, and present
preliminary results from it generating assistive torques to a Fig. 1. Left, prototype exosuit described in this paper. Right, illustration
subject during walking. We envision such an exosuit having showing the main components.
broad applicability for assisting healthy individuals as well as
those with muscle weakness.
wearer, but also provide a flexible structure so that assistive
torques can be applied to the biological joints [19]. Fig. 1
Keywords—exosuit; walking; wearable robot, soft robot; shows the design of such an “exosuit” that is the focus of this
exoskeleton paper.
The optimum form and function of an exoskeleton will be
I. INTRODUCTION closely tied to its intended application; however, there are two
main challenges that apply to the design of these systems. The
Robotic exoskeletons have been developed for a large first major challenge relates to the exoskeleton imposing
number of applications, with the goal of assisting or enhancing kinematic restrictions on the wearer’s natural degrees of
human activities. For the lower extremity, devices have been freedom. By having a parallel robotic mechanism, care has to
developed to apply assistive torques to the biological joints to be taken to ensure alignment between the robotic joints and
augment healthy individuals or assist those with disabilities to those of the biological limb. This is challenging because the
walk [1-10]; assist with load carriage by providing a parallel biological joints do not have a fixed center of rotation [20].
path to transfer load to the ground [11-14], thus off-loading the Misalignment with the wearer's kinematics will cause forces or
wearer and bypassing their musculature; or finally, provide torques that resist the wearer's motion, increasing the effort
gait retraining or rehabilitation for those with disabilities [15- required to move, altering natural gait patterns and decreasing
18]. These systems all are based on the principle of having a the exoskeleton's efficacy [21]. To address this, self-aligning
robotic mechanism (with rigid links, joints, and actuators) that mechanisms have been proposed [22, 23]; however, this results
runs in parallel with the biological limb. Typically, the links of in an overall increased system mass. An approach often taken
these exoskeletons are connected to the wearer at a few to reduce the challenge associated with matching the
locations with a waist belt, leg strapping, or foot attachment. In kinematics of the exoskeleton and wearer is to have a device
our recent work we have been exploring the use of soft flexible that provides assistance at a single joint only. Most of these
materials as an alternative means to not only interface to the systems have focused on the ankle [6, 7, 24] or the knee [4] as

978-1-4673-6024-1/13/$31.00 ©2013 IEEE

they can be approximated as pin joints in the sagittal plane; This paper focuses on the initial development of a
however, alignment with the joint remains important. biologically-inspired soft exosuit designed to create joint
torques on the wearer during walking. A battery-powered
The second main challenge associated with traditional portable prototype system was demonstrated using a motor-
exoskeletons is that they are relatively heavy. While actuation actuated cable-drive system. Preliminary human studies
can help accelerate and decelerate this inertia, often the wearer indicate that it is possible to apply significant forces and
must assist with this. This effect is often an increased effort for torques with a soft suit, without causing discomfort or injury.
the wearer and this is most significant when the mass is located
distally near the foot and ankle [25]. For instance, a mass of
4kg near the wearer’s center of mass has been shown to be II. BIOLOGICALLY-INSPIRED SOFT EXOSUIT
associated with a 7.6% increase in metabolic effort, but this To achieve a lightweight and efficient exosuit, there is
increases to 34% for the same amount attached at both of the much insight to be gained through understanding how humans
feet. The mass of a heavy exoskeleton leads to large power walk. By designing the architecture of the exosuit to mimic
requirements to power it which in turns leads to needing large some of that of the biological limb, it is hypothesized that a
payloads of batteries or other energy storage means in order to more transparent, safe, and effective design can be achieved.
achieve a mobile system. To address the power requirement
issue, a number of groups have looked at using passive and A. Motivation
quasi-passive systems that can exploit the natural generation The goal for the exosuit is to assist with forward propulsion
and absorption of energy during the gait [8, 9]. during walking on level ground at 1.25 m/s by applying
Despite the drawbacks associated with restricting appropriate assistance at the ankle, knee, and hip joints.
kinematics and added inertia, a number of exoskeletons have Specifically, the goal is to duplicate the forces generated by
demonstrated the ability to augment or assist the wearer. In two muscles and tendons in the biological leg. In principle, if the
human subjects studies with a leg exoskeleton for load suit generates forces on the body that mirror what the muscles
carrying, significant off-loading was demonstrated [12, 26]; do, the muscles may be required to do less work, thus reducing
however, at the cost of an increased metabolic effort to walk. If the wearer’s metabolic cost of transport. Human walking is
such systems can be improved to the point where they are very efficient due to the tuned limb lengths, stiffnesses, and
neutral in terms of metabolic effect, the benefits of off-loading muscle sizes in the leg. The system can be well-modeled as a
the skeletal and musculature could reduce the risk of injury. A passive dynamic walking system requiring minimal energy
number of exoskeletons have also recently demonstrated that input [29], and these principles have been implemented in
the metabolic cost of the wearer can be reduced under certain robots that demonstrated a similar cost of transport to human
conditions. A single degree of freedom tethered pneumatic locomotion [30].
exoskeleton for the ankle demonstrated that at fast walking
(1.75 m/s), metabolic effort could be reduced by 13.8% [7].
Recently, a squat-assistance exoskeleton was shown to reduce
the wearer's metabolic effort [27] and an exoskeleton in
parallel with the legs was shown to reduce the metabolic effort
required to hop in place [28]. However, it remains an open
research question if a net metabolic reduction can be achieved
for walking or running with a non-tethered portable system.
In contrast to traditional exoskeleton systems to augment
human capability, a new paradigm is to use soft clothing-like
"exosuits" to achieve many of the same objectives [19]. In this
case, the suit does not contain any rigid elements supporting
compressive loads, so the wearer's bone structure must sustain
all the compressive forces normally encountered by the body
plus the forces generated by the exosuit.
Exosuits offer a number of benefits as compared to
traditional exoskeletons. The suits themselves, composed
primarily of fabrics, can be significantly lighter than an
exoskeleton frame or linkage system, leading to very low
inertias and lower cost of transporting the suit mass. Since the
suit does not contain a rigid frame, it also provides minimal
restrictions to the wearer's natural kinematics, avoiding
problems relating to joint misalignment. Further, control for Fig. 2. Diagram of when the muscles in the leg are active. Our suit is active
the system can be less precise as the inherent compliance in the for the duration specified by the shaded region, to assist ankle plantar flexion,
knee flexion, and hip flexion. The gait percentages correspond to the
system can make it more forgiving. Finally, the suit can also following periods of the gait: 0-12%: Loading Response; 12-31%: Mid-
provide torques on multiple joints simultaneously but having it stance; 31-50%: Terminal stance; 50-62%: Pre-swing; 62-75%: Initial swing;
span multiple joints, which may assist with reducing the total 75-87%: Mid-swing; 87-100%: Terminal swing.
number of actuators.
 During normal walking, power is expended by the body a small moment flexing the knee due to their attaching above
primarily at the transitions of support from one leg to the other the knee joint on strap (6), and the knee being bent slightly
and this power is provided largely by the hip and ankle. During during the time in the gait when the cable is actuated. On a
stance, the calf muscles (gastrocnemius, soleus, etc.) contract typical subject, the moment arms about the hip, knee, and ankle
isometrically, which stretches the Achilles tendon due to the are 8cm, 1cm, and 8cm, respectively.
natural motion of the body falling forward. Subsequently, these
muscles contract concentrically and the Achilles tendon
recoils, giving a large positive power burst from 40-60% of the
gait cycle, and providing 0.39 J/kg of energy to help redirect
the body's momentum and prepare for the opposite leg's
touchdown [31]. This creates a moment about the foot which
propels the body upward and forwards. The gastrocnemius
muscle also flexes the knee during this time, permitting the
foot to clear the ground for swing. The activity of the muscles
in the leg as a function of the gait cycle is shown in Fig. 2.
The hip provides a smaller power burst and a period of
power absorption during the transition between legs. The
power absorption occurs from 35-50% of the gait cycle,
absorbing 0.12 J/kg to help stop the body's falling forward, and
occurs through the extension of the tendons at the front of the
hip. The power burst occurs from 50-75% of the gait cycle,
providing 0.14 J/kg of energy to help swing the leg [31]. This
is accomplished by the contraction of the muscles at the front
of the thigh (rectus femoris, adductor longus, adductor magnus,
etc.). The exosuit is designed to assist the power absorption
through passively extending, as well as assisting the muscle
contractions by actuating. To aid both the hip and the ankle, the
Fig. 3. Left, several views of the forces in the suit as it is actuated. Dotted
suit is primarily actuated from 40-60% of the gait cycle, shown lines indicate forces in straps on the obscured side of the leg. The cable sheath
in Fig. 2 by the shaded region, then is made slack otherwise. is attached to and pulls down at the bottom of straps (7), and pulls up at the
back of the boot.
B. Design Overview
A key feature of the exosuit is that forces are generated in
Based on this understanding of the biomechanics of the webbing both passively due to the natural kinematics of
walking, an exosuit consisting of various fabrics sewn together walking and actively from the Bowden cable contracting; thus
in a conformal form factor was designed. Fig. 3 shows a mimicking the action of the biological limb. Referring toFig. 2,
diagram of the forces in the suit when tension is applied to it it can be seen that the hip flexor muscles are active when the
with a cable close to the ankle. The numbers referred to in the hip is maximally extended. Due to the fact that the suit is
text are labels in the figure. The exosuit attaches around the located primarily at the front of the hip, it will become taut and
waist (1) and above the knee (6), and transfers force between thus generate a resistive moment as the hip extends maximally,
the back of the calf and the waist through a series of webbing absorbing power which can later be returned. The suit tension
straps (2-5), (7). The webbing attaches to the sheath of a is enhanced when the cable is actuated in a manner similar to
Bowden cable at the bottom of strap (7), which is located at the how biological muscle stretches pre-stressed tendons. When
back of the calf. The cable itself extends downward from this the hip flexes, the suit becomes slack and is transparent to the
sheath termination point to a small webbing strap coming up wearer. Similarly, the suit tenses when the ankle is dorsiflexed,
from the heel, where it is attached to the wearer's shoe. When and actuating the cable provides additional force. The suit
the Bowden cable is actuated, it pulls these two points together. becomes slack when the ankle is plantarflexed. Due to this
Thus, the back of the ankle is pulled upward and the bottom of construction, the suit can be made to apply beneficial moments
the exosuit is pulled downward during actuation. The exosuit purely passively during walking if it is adjusted properly.
then transfers the force up to the wearer's waist, so the pelvis
bone is pulled downward. The skeletal structure of the wearer
then transfers this downward force back to the ankle joint and C. Suit Design Principles
to the ground through the foot. The suit was designed to create a path to transfer loads
between the ankle and the pelvis. As such, high effective
The actuation applied by the cable creates tension in the exosuit stiffness (resultant of fabric and the series compliance
suit that creates moments about the ankle, knee, and hip. A due to the interface to the wearer) is required in order to
moment is generated about the ankle because the webbing on transfer power to the body efficiently from the actuator instead
the back of the heel, pulling upward, is several centimeters of putting energy primarily into stretching the suit. Moreover,
away from the ankle pivot point, which experiences the high suit stiffness is beneficial because it corresponds to lower
downward force through the bone structure. Moments are suit displacements during actuation, and thus less risk of
generated about the hip joint since the majority of the straps (2- chafing. If the stiffness is high enough, the suit designer can
4) are positioned in front of the hip. The lower straps (7) create place other elastic elements in series with the suit, thereby
having control over the effective exosuit stiffness. The suit the pelvis becomes larger in the downward direction; this
construction follows several principles to provide a wedge effect limit suit motion. With the thigh muscle and fat
comfortable, functional device. initially pulled slightly (1cm) vertically, the skin can displace
down a larger distance downward before it reaches its travel
1) Terminate on bony parts of the body limit and leads to suit slippage or chafing. On the waist, this
First, the top of the suit is terminated on the pelvis because preloading compresses the skin and fat against the pelvis
it is a bony part of the body with a relatively thin layer of skin somewhat and prevents the suit from riding upward on the
coverage. When the skin is compressed against the bone pelvis, thereby increasing the resulting stiffness.
during the suit's operation, the displacement is low as
compared to other parts of the body covered by thicker muscle 4) Minimize normal pressures on the body
or fat, leading to high stiffness. The shins and shoulders also Fourth, the suit makes use of wide straps throughout to
have thin layers of muscle over the bone, but these were not minimize normal pressures on the body and minimize
chosen because pulling on the shoulders during preliminary displacement. The skin can accommodate a certain amount of
human subjects experiments was uncomfortable and forced normal pressure before discomfort or lack of blood flow
the wearer to hunch over, and the shin orientation makes it occurs. Estimates of the maximum comfortable normal
difficult to create the upward forces needed at the heel. pressure are typically around 0.5 N/cm2 [32, 33].
Furthermore, the body will compress if forces are applied
On the suit, strap (5) transfers loads to the suit from the
normally, such as occurs over the thigh with our suit as straps
opposite iliac crest of the pelvis bone. This geometry is (2)-(5) push inward against it (see Fig. 3 (center)). To
important to prevent the suit from sagging. The center area of maximize suit stiffness, this displacement must be minimized,
the abdomen does not have any rigid elements to support which can be achieved by having wide contact areas. The suit
vertical loads, so if strap (5) connected to the center of the stiffness also increases if it is actuated while the underlying
waist belt, it would pull down several centimeters during muscles are tensed, which causes them to become stiffer and
loading. compress less. Ideally, the suit will assist muscles that are
2) Load the body normally as much as possible firing, which will result in this behavior. At the back of the
Second, the bulk of the load is transferred to the pelvis to calf, the straps (7) extend low on the leg to bypass the bulk of
load the body normal to the skin and bones as much as the calf muscle. Having the straps higher results in the calf
possible. Shear forces against the skin will cause slipping of compressing when the suit is actuated, which is uncomfortable
the suit and chafing if the friction force with the skin is and results in a lower suit stiffness.
exceeded [32]. It is impossible to avoid shear forces entirely 5) Other factors
throughout the suit, as vertical loads at the bottom of the suit In addition to minimizing normal pressures from wide
will cause some displacement along the length of the leg. straps, pressures can be distributed through equalizing the
Some displacement is possible, though, without ill effect if tension in various parts of the suit. During the suit donning
there is a large amount of tissue over the bone, and the skin is process, straps (2)-(5) are adjusted iteratively until they have
not stretched taut. In this scenario, the skin and underlying around equal tension, which we have found maximizes both
muscle and fat can move back and forth over the bone under comfort and performance.
relatively low forces where the thicker the layer of tissue, the
more displacement is possible. For example, when a person is Finally, the suit should not restrict muscles from expanding
standing vertically, the skin can deflect up or down several during operation. In our suit, the knee strap is made from a
centimeters at the thigh without discomfort. The exosuit stretch cotton material to permit the thigh muscles to bulge
displaces less than 3 cm at the thigh under loads of up to 200N when they contract. Our experiments have shown that a non-
at the ankle, and so does not slip or cause chafing there. The extensible strap is uncomfortable there.
lack of slipping is aided by the normal force against the skin
from the knee strap (6), which increases the friction force D. Suit Implementation
needed there before slipping occurs. Lower on the suit, some Three pictures of the front, side, and back of the exosuit are
motion of the webbing does occur, around the knee and over shown in Fig. 4. As shown in the front view, the suit begins
the calf. In these regions, the normal force is small and the with a waist belt (1) which sits on the iliac crests of the pelvis.
motion is not problematic. Further protection from chafing can The waist belt is constructed of rip-stop nylon and tensioned by
be accomplished by wearing a close-fitting knee-length two polyester 1" webbing straps that are above and below the
garment under the suit that the suit can slide over. iliac crest. The top strap supports loads while the bottom strap
3) Preload the suit against the body provides snugness and prevents lateral motion. This top strap
The lack of slipping at the thigh is also facilitated by a supports almost the entire downward force of the suit, which is
third principle; preloading the suit against the body. When then spread out across the leg through the rip-stop nylon and
donning the suit, first the knee straps (6) are tightened and the straps attached to it.
belt (1) put around the waist. Then, the vertical straps between Attached to the waist belt are four 2" polyester straps per
the knee straps and waist belt (2-5) are tightened. This pulls up leg, (2)-(5), extending down to a strap above the knee (6) that
on the knee straps and down on the waist belt when standing is positioned with the bottom edge 3-4 cm above the top of the
vertically, causing the thigh skin to displace 1-2cm upward. patella. Straps (2)-(5) are sewn onto the waist belt, loop
This preloading is possible because the thigh expands in through slides (buckles) on the knee strap, and then attach back
diameter going upward from the knee, while the iliac crest of on themselves (2,4,5) or to the waist belt (3) with 2"-wide
industrial strength hook Velcro. This hook Velcro mated with and their respective masses in Table I. The system was not
standard 2" loop Velcro. Although the industrial strength hook optimized for mass or volume, but rather designed to provide
Velcro performs better than a standard hook, it still displaces flexibility to explore a range of gait assistance strategies.
several millimeters in shear under loads due to the length of the
loops. Using the Velcro on the vertical webbing straps, the suit B. Cable-actuation unit
can be adjusted to fit individuals from 1.75 to 2.00 meters tall. The Bowden cable ends are pulled by actuator units, and
the cable then passes through a force sensor module and
continues down the leg. Each actuator unit consists of a motor
and 111:1 gearbox connected to a secondary shaft via 1:1
gearing. The secondary shaft contains a pulley (radius=2.45cm)
capable of winding up the cable one rotation, which
corresponds to 15cm of cable travel. This range allows for
some pre-tensioning of the cable to ensure a snug fit before it
pulls to provide assistance during walking (corresponding to a
typical cable travel of 5-6cm).
TABLE I. MOBILE ACTUATION SYSTEM COMPONENTS
Component Mass
Batteries to power motors
1.97 kg
5000 mAh of LiPo batteries to power system for 4 hours
Batteries to power computer
0.42 kg
4000 mAh 14.8V LiPo battery to power system for 5 hours
Motor Controllers (2)
0.86 kg
Copley Accelnet ADP-090-36
Direct Drive Actuator Units (2)
Maxon EC 4-pole 30 200W, 24V motor with 111:1 2.27 kg
gearbox, connected spool with radius 2.45 cm
PC/104 Computer
0.53 kg
Diamond Systems Aurora PC/104
Fig. 4. Several views of the exosuit. Labels correspond to those in Fig. 3. Frame
2.72 kg
ALICE pack frame, aluminum and acrylic structure
Strap (2) holds up the center of the knee strap (strap 5). Strap Force sensor modules (2)
0.87 kg
(4) provides a direct path from the calf straps (7) to the side of Phidgets 3135 50kg Micro Load Cell, Futek CSG110 amp.
the waist belt, and holds up the outside edge of the knee strap. Other
1.00 kg
Connectors, switches, etc.
Straps (3) and (5) support the inside edge of the knee strap,
with strap (3) crossing the front of the thigh and strap (5) going Total 10.64 kg
behind the leg to support it evenly. Adjusting the relative
tension of these straps allows the wearer to control the position
of the knee strap rotationally around the leg. Straps (3,4,5) are
sewn together where they intersect on the side of the hip, to
distribute the forces from each of them. Strap (3) comes down
from the side of the waist, loops through the buckle on the knee
strap, then continues vertically to terminate on strap (5).

III. PORTABLE ACTUATION UNIT

The Bowden cable extends up the wearer's leg and to an
actuator module carried on the wearer's back. This module
contains Maxon motors, Copley motor controllers, a PC/104
computer, and batteries. Just below the motors is another
module which contains a sensor to measure the cable tension.
At the ankle and foot, the webbing behind the heel connects to
webbing straps attached to the back of the boot heel.
Footswitches are placed inside the shoes to measure heel strike
and toe off. The mass of the cloth exosuit is 1.07 kg, the
Fig. 5. Left, mobile actuation system used with the exosuit with PC/104 and
footswitches have a combined mass of 0.44 kg, and the Copley Accelnet motor controllers highlighted. Right, side view of system
actuator/control system and backpack frame have a mass of showing actuator and force sensor module.
10.64 kg, for a total system mass of 12.15 kg.
C. Force-sensing unit
A. System overview
Below the actuator unit, along the cable, is a force sensing
A diagram of the initial portable actuator and control unit. This contains an idler pulley connected to one end of a
hardware is shown in Fig. 5, with a detailed list of components cantilever-style force sensor. The cable glances off the pulley,
changing angle by θ=8 degrees on each side and causing a
force on the pulley. The force sensor has strain gages, and
measures the pulley force as it causes shear in the force sensor.
The force sensor is located next to the motor to reduce wiring
down the body. This location means that the force at the ankle
is masked by the efficiency of the Bowden cables, but with
high efficiency cables (Nokon Part# KON05020), a good
estimate of the force at the ankle is still achieved.

IV. EXOSUIT EVALUATION

The system was designed to be intrinsically safe in several
ways. The maximum force that the suit can apply to the
person, under normal operation, is <30% of the maximum
ankle moment present during normal walking. The travel
range of the cables was limited mechanically by the spool and
Bowden cable ends, and the velocity was limited by the no-
load speed of the motor followed by a large gear reduction.
Finally, in case of undesired operation, the stoppers (ferrules)
at the ends of the cable act as mechanical fuses, coming off
consistently at 600-650N and preventing any further force
transfer to the wearer.
A. Suit stiffness characterization
The suit stiffness was characterized by positioning a
subject in two poses, tensioning and releasing the suit, and
simultaneously recording the forces and displacement. Forces
were recorded by a load cell at the ankle (Omega LC201).
Loading and unloading were performed with a ramp profile,
over a period of 3 seconds each, and the displacement of the
motor was recorded. The results are shown in Fig. 6.
As seen in Fig. 6, the suit exhibits a nonlinear stiffness and
significant hysteresis. The suit stiffness increases to 7350 N/m
or 5400 N/m depending on the subject’s pose (stepping
Fig. 6. Top, stiffness curves for the exosuit, when the wearer is positioned
forward or standing upright with legs parallel). As seen in the with both legs together (blue curve) and one leg 60cm in front of the other
lower portion of the figure, with the cable retracted from 1 cm (red curve). Bottom, a depiction of how much the suit and boot attachments
to 9 cm, the suit itself displaces 4.7cm downward while the displace during actuation. An additional load cell is used at the heel to
boot attachment displaces 2.0cm upward and the cable system measure the forces.
stretches 1.0 cm.
During walking, forces and displacements of the cable were
In these trials, the energy contributed to the suit during recorded. A graph of the cable displacement and force as a
loading was 7.9J and 12.0J for the legs parallel and stepping function of gait percentage are shown in Fig. 7. In this figure,
forward cases, respectively. The hysteresis in the suit caused the force is observed to begin increasing at around 22% in the
59.7% and 53.8% of the energy in the suit to be lost for each gait cycle, although the cable is still at zero displacement. This
test actuation cycle, respectively. occurs because the suit naturally tensions due to the motion of
the ankle and hip. Once the cable is actuated, here at 30% of
B. Human walking trials the gait cycle, the force increases rapidly with cable
For human subjects experiments, a preliminary control system displacement. During walking trials, the stiffness of the suit
was implemented that could function during steady state was observed to be close to 225N/5.5cm = 4090N/m, which
walking. A position-controlled trajectory for the cable was corresponds closely to the stiffness measured in the angled-leg
programmed, which was transformed into a force at the ankle standing test with a displacement of 5.5cm and small preload.
through the compliance of the suit. The position trajectory was The power input and output of the system were measured,
pre-computed as a function of gait percentage based on the as well. For each leg, the motor used 17.6 Watts on average
joint angles of the normal walking cycle and an estimated suit over a gait cycle of 1.16 seconds, for a total energy of 20.4
stiffness, to generate an estimated force at the ankle. The Joules. The direct drive actuator unit output 12.3 Joules, and
controller was then tuned in realtime while a person was the ankle received 7.0 Joules for an overall efficiency of
wearing the system to correct for modeling errors. The 39.7%. As seen in Fig. 8, the power at the ankle is 56.4% of
controller is triggered from a signal from the footswitch the power at the output of the actuator unit. The losses come
indicating that heelstrike has occurred, and then the position from the Bowden cables, which were inefficient due to the
trajectory is played back as a function of time. cables being worn and bent.
 The moment arms of the suit at the ankle, knee, and hip
were 8cm in front of the hip, 1cm behind the knee joint, and
8cm behind the ankle, respectively. With a force of 200N, for
example, the suit will generate moments of 16 N-m at the hip
and ankle, and a moment of less than 2 N-m at the knee.
Compared to normal walking data [31], this corresponds to
12% of the ankle moment and 20% of the hip moment. In other
trials, the suit has been able to produce peak forces of over
300N—corresponding to 18% of the ankle moment and 30% of
the hip moment.

I. CONCLUSIONS AND FUTURE WORK

In conclusion, we have demonstrated a soft exosuit capable
of applying forces to the body during walking. The suit can
apply moments to the hip and ankle of at least 18% of the Fig. 7. Cable position and force at the ankle while walking seven steps. The
normal human walking moments, enough to be noticeable to actual cable position is computed by the motor controller, taking into account
the wearer. Our experiences have shown that if the timing of acceleration limits of the motor.
actuation is adjusted poorly, the wearer can feel the suit
resisting them, and one or more of their muscles tire quickly.
If the timing and suit tension are adjusted well, the suit can feel
like it is helping the wearer, lifting the heel and helping the leg
swing. More studies need to be performed to determine the
optimal timing for suit actuation and the accompanying best
suit architecture.
The suit is extremely light, minimizing distal mass, and
adds negligible inertia to the legs. The suit feels comfortable
and does not cause chafing, even though it moves with the skin
a small amount during actuation. The suit also does not
constrain any of the degrees of freedom in the legs, and permits
the wearer to move through their full range of motion.
The current suit stiffness is relatively low, which sets
bounds on the forces that can be applied for a given motor
power. It also is nonlinear, increasing substantially after several Fig. 8. Powers input to motor, output from direct drive actuator unit, and
applied to the ankle during a walking trial.
centimeters of displacement. As such, control schemes can
benefit from fast motor motions to remove slack in the system,
followed by power transfer at higher stiffnesses. The maximum ACKNOWLEDGEMENT
forces that the human body can tolerate via the suit are
This material is based upon work supported by the
unknown; measurement of the pressures applied by the suit is
needed. One risk of high forces is that they create increased Defense Advanced Research Projects Agency (DARPA),
suit displacements, which will lead to chafing. Based on our Warrior Web Program (Contract No. W911QX-12-C-0084).
experience with the suit, we estimate that chafing will become The views and conclusions contained in this document are
a problem at forces above 250-300N, unless suit stiffness can those of the authors and should not be interpreted as
be increased further. representing the official policies, either expressly or implied,
of DARPA or the U.S.Government.
There are many areas for future improvement of the
system. Measurements must be made of the suit's benefit, This work was also partially funded by the Wyss Institute
including metabolic and electromyography measurements. for Biologically Inspired Engineering at Harvard University.
The actuator unit must be substantially lightened to create a The authors would like to thank Leia Stirling, Ken Holt and
unit that does not increase the wearer's metabolic rate due to Mike Mogenson for their input during this project.
the additional load, and must be packaged into a smaller form
factor. We estimate that the entire system can be constructed REFERENCES
with a mass of only 5kg, including batteries. Finally, further
tuning of the control system is needed to understand how it can [1] A. M. Dollar and H. Herr, "Lower extremity exoskeletons and active
benefit the body most. Additional work is needed to determine orthoses: Challenges and state-of-the-art," Robotics, IEEE Transactions
the most effective strategies for aiding human locomotion, on, vol. 24, pp. 144-158, 2008.
including schemes that assist the load-acceptance phase of the [2] H. Herr, "Exoskeletons and orthoses: classification, design challenges
gait, instead of push-off as currently. and future directions," Journal of NeuroEngineering and rehabilitation,
vol. 6, p. 21, 2009.
[3] H. Kawamoto, S. Lee, S. Kanbe, and Y. Sankai, "Power assist method [18] K. A. Shorter, G. F. Kogler, E. Loth, W. K. Durfee, and E. T. Hsiao-
for HAL-3 using EMG-based feedback controller," in Systems, Man and Wecksler, "A portable powered ankle-foot orthosis for rehabilitation,"
Cybernetics, 2003. IEEE International Conference on, 2003, pp. 1648- Journal of Rehabilitation Research & Development, Accepted, 2010.
1653. [19] M. Wehner, B. Quinlivan, P. M. Aubin, E. Martinez-Villalpando, M.
[4] J. E. Pratt, B. T. Krupp, C. J. Morse, and S. H. Collins, "The RoboKnee: Bauman, L. Stirling, K. Holt, R. Wood, and C. Walsh, "Design and
an exoskeleton for enhancing strength and endurance during walking," Evaluation of a Lightweight Soft Exosuit for Gait Assistance," in IEEE
in Robotics and Automation, 2004. Proceedings. ICRA'04. 2004 IEEE International Conference on Robotics and Automation, 2013.
International Conference on, 2004, pp. 2430-2435. [20] J. Perry and J. R. Davids, "Gait analysis: normal and pathological
[5] H. A. Quintero, R. J. Farris, C. Hartigan, I. Clesson, and M. Goldfarb, function," Journal of Pediatric Orthopaedics, vol. 12, p. 815, 1992.
"A Powered Lower Limb Orthosis for Providing Legged Mobility in [21] A. Schiele, "An explicit model to predict and interpret constraint force
Paraplegic Individuals," Topics in spinal cord injury rehabilitation, vol. creation in pHRI with exoskeletons," in Robotics and Automation, 2008.
17, pp. 25-33, 2011. ICRA 2008. IEEE International Conference on, 2008, pp. 1324-1330.
[6] G. S. Sawicki and D. P. Ferris, "A pneumatically powered knee-ankle- [22] A. H. Stienen, E. E. Hekman, F. C. Van Der Helm, and H. Van Der
foot orthosis (KAFO) with myoelectric activation and inhibition," Kooij, "Self-aligning exoskeleton axes through decoupling of joint
Journal of NeuroEngineering and rehabilitation, vol. 6, p. 23, 2009. rotations and translations," Robotics, IEEE Transactions on, vol. 25, pp.
[7] G. S. Sawicki and D. P. Ferris, "Powered ankle exoskeletons reveal the 628-633, 2009.
metabolic cost of plantar flexor mechanical work during walking with [23] M. A. Ergin and V. Patoglu, "A self-adjusting knee exoskeleton for
longer steps at constant step frequency," Journal of Experimental robot-assisted treatment of knee injuries," in Intelligent Robots and
Biology, vol. 212, pp. 21-31, 2009. Systems (IROS), 2011 IEEE/RSJ International Conference on, 2011, pp.
[8] W. van Dijk, H. van der Kooij, and E. Hekman, "A passive exoskeleton 4917-4922.
with artificial tendons: Design and experimental evaluation," in [24] M. B. Wiggin, G. S. Sawicki, and S. H. Collins, "An exoskeleton using
Rehabilitation Robotics (ICORR), 2011 IEEE International Conference controlled energy storage and release to aid ankle propulsion," in
on, 2011, pp. 1-6. Rehabilitation Robotics (ICORR), 2011 IEEE International Conference
[9] M. Goldfarb and W. K. Durfee, "Design of a controlled-brake orthosis on, 2011, pp. 1-5.
for FES-aided gait," Rehabilitation Engineering, IEEE Transactions on, [25] R. C. Browning, J. R. Modica, R. Kram, and A. Goswami, "The effects
vol. 4, pp. 13-24, 1996. of adding mass to the legs on the energetics and biomechanics of
[10] P. D. Neuhaus, J. H. Noorden, T. J. Craig, T. Torres, J. Kirschbaum, and walking," Medicine and science in sports and exercise, vol. 39, p. 515,
J. E. Pratt, "Design and evaluation of Mina: A robotic orthosis for 2007.
paraplegics," in Rehabilitation Robotics (ICORR), 2011 IEEE [26] K. N. Gregorczyk, J. P. Obusek, L. Hasselquist, J. M. S. Bensel, K.
International Conference on, 2011, pp. 1-8. Carolyn, D. Gutekunst, and P. Frykman, "The effects of a lower body
[11] H. Kazerooni and R. Steger, "The Berkeley Lower Extremity exoskeleton load carriage assistive device on oxygen consumption and
Exoskeleton," Journal of Dynamic Systems, Measurement, and Control, kinematics during walking with loads," DTIC Document2006.
vol. 128, p. 14, 2006. [27] A. Gams, T. Petric, T. Debevec, and J. Babic, "Effects of robotic knee-
[12] C. J. Walsh, K. Endo, and H. Herr, "A quasi-passive leg exoskeleton for exoskeleton on human energy expenditure," 2013.
load-carrying augmentation," International Journal of Humanoid [28] A. M. Grabowski and H. M. Herr, "Leg exoskeleton reduces the
Robotics, vol. 4, pp. 487-506, 2007. metabolic cost of human hopping," Journal of Applied Physiology, vol.
[13] M. Wehner, D. Rempel, and H. Kazerooni, "Lower Extremity 107, pp. 670-678, 2009.
Exoskeleton Reduces Back Forces in Lifting," 2009. [29] T. McGeer, "Passive dynamic walking," The International Journal of
[14] E. Garcia, J. M. Sater, and J. Main, "Exoskeletons for human Robotics Research, vol. 9, pp. 62-82, 1990.
performance augmentation (EHPA): A program summary," Journal- [30] S. Collins, A. Ruina, R. Tedrake, and M. Wisse, "Efficient bipedal
Robotics Society Of Japan, vol. 20, pp. 44-48, 2002. robots based on passive-dynamic walkers," Science, vol. 307, pp. 1082-
[15] S. K. Banala, S. K. Agrawal, and J. P. Scholz, "Active Leg Exoskeleton 1085, 2005.
(ALEX) for gait rehabilitation of motor-impaired patients," in [31] J. J. Eng and D. A. Winter, "Kinetic analysis of the lower limbs during
Rehabilitation Robotics, 2007. ICORR 2007. IEEE 10th International walking: what information can be gained from a three-dimensional
Conference on, 2007, pp. 401-407. model?," Journal of Biomechanics, vol. 28, pp. 753-758, 1995.
[16] J. F. Veneman, R. Kruidhof, E. E. G. Hekman, R. Ekkelenkamp, E. H. [32] J. Cool, "Biomechanics of orthoses for the subluxed shoulder,"
F. Van Asseldonk, and H. van der Kooij, "Design and evaluation of the Prosthetics and Orthotics international, vol. 13, pp. 90-96, 1989.
LOPES exoskeleton robot for interactive gait rehabilitation," Neural [33] G. Holloway, C. H. Daly, D. Kennedy, and J. Chimoskey, "Effects of
Systems and Rehabilitation Engineering, IEEE Transactions on, vol. 15, external pressure loading on human skin blood flow measured by 133Xe
pp. 379-386, 2007. clearance," Journal of Applied Physiology, vol. 40, pp. 597-600, 1976.
[17] S. Jezernik, G. Colombo, T. Keller, H. Frueh, and M. Morari, "Robotic
orthosis Lokomat: a rehabilitation and research tool," Neuromodulation:
Technology at the Neural Interface, vol. 6, pp. 108-115, 2003.