Toward A Soft Robotic Ankle-Foot Orthosis (SR-AFO) Exosuit for
Human Locomotion: Preliminary Results in Late Stance Plantarﬂexion
Carly M. Thalman, Student Member, IEEE, Tiffany Hertzell, and Hyunglae Lee∗,Member, IEEE
Abstract— This paper presents the design of a soft robotic
ankle-foot orthosis (SR-AFO) exosuit to aid in plantarﬂexion
for gait rehabilitation in individuals who suffer from irregular
gaits due to stroke or other injuries. The SR-AFO exosuit is
a sock-like garment fabricated from compliant fabrics. The
SR-AFO exosuit aids in late stance of the walking gait in
plantarﬂexion by contracting the actuator to pull the posterior
end of the foot upward. This helps to reduce the muscle effort
of the user during plantarﬂexion. The addition of a second
actuator shows a 45.3% increase to 13.51 ±0.31 kg payload
capacity. The actuators are oriented at an optimal angle of 5◦
to produce the highest pulling force. Three healthy participants
are evaluated during walking trials with and without SR-AFO
exosuit assistance while ankle angle and muscle activity are
monitored. The gastrocnemius (GA) and soleus (SOL) muscle
activity during late stance is reduced by 13.4% and 16.6%
respectively. Tibialis anterior (TA) increases slightly during
swing most likely due to the hysteresis in the system deﬂating
during that window. The ankle range of motion remains within
natural walking limitations and plantarﬂexion angle increases
when the SR-AFO exosuit is active.
Keywords - Soft Robotics, Wearable Robots, Assistive Robots,
The ankle joint is responsible for 45% of the power behind
human locomotion, and plantarﬂexion is a critical motion
throughout the entire gait cycle . Body propulsion during
a forward gait requires a propulsive force that pushes off of
the ground and creates forward motion. Ankle plantarﬂexion
produces propulsion by using ankle muscles to push off the
ground during late stance phase , . The soleus (SOL)
and gastrocnemius (GA) muscles provide vertical support
during single-leg stance . During mid single-leg stance,
the SOL and GA have an opposite energetic effect on the
leg and trunk of the body to ensure support and forward
progression of the leg and trunk . Proper plantarﬂexion
ensures stable weight transfer and drives critical lever-force
behind the ﬁnal rocker action transitioning to pre-swing
. Individuals suffering from hemiparesis after a stroke,
paralysis of pretibial muscles, or ﬁxed plantarﬂexion will
often experience a lack of shock absorption and loss of
deﬁnitive heel strike . This can cause a loss of the ﬁnal
rocker action needed to propel the foot forward for toe-off
to transition to pre-swing , . As a result, various forms
Carly M. Thalman, Tiffany Hertzell, and Hyunglae Lee are with the
Ira A. Fulton Schools of Engineering, Arizona State University, AZ, USA.
Fig. 1. An illustration of the SR-AFO Exosuit for plantarﬂexion is shown
in (a), while (b) shows the posterior view of the actual device being worn
by a user.
of gait abnormalities arise that can cause further injury, pain,
or risk of trips and falls . The aforementioned afﬂictions
to gait allude to reasons why a passive rigid orthotic device
may not be beneﬁcial in all cases. In typical rehabilitative
environments, treatments can be labor intensive for both
the therapists and patients performing the training. For gait
rehabilitation, therapies often require manually moving the
lower limbs and torso of the patient  or rigid AFOs
which create a ﬁxed ankle angle that can cause further
gait abnormalities . To eliminate some limitations of
traditional AFOs, a range of wearable robotic solutions
has been proposed to provide more dynamic and tailored
methods of assistance to a wide range of users –.
Robotics designed for lower-extremity assistance has been
shown to assist in restoring natural gait patterns through
assistive actuation and controlled perturbations , .
However, adding additional weight to the foot increases
the inertia that can cause issues with balance, fatigue and
stride length . Soft robotics is a rapidly growing ﬁeld
in addressing needs in human assistance, speciﬁcally related
to post-stroke therapy and rehabilitation strategies for the
lower body , . Some beneﬁts of soft robotic solutions
for ankle rehabilitation are lightweight materials and reduced
issues with joint alignment due to increased compliance ,
In previous work, a soft robotic ankle-foot orthosis (SR-
AFO) exosuit was developed to address the issue of foot
drop using soft pneumatic actuators to assist dorsiﬂexion,
inversion and eversion (IE) . The sock-like design was
implemented to go over the user’s athletic shoe and is
2020 3rd IEEE International Conference on Soft Robotics (RoboSoft)
Yale University, USA
978-1-7281-6569-1/20/$31.00 ©2020 IEEE 801
Fig. 2. (a) A user with the donned SR-AFO Exosuit inactive and deﬂated,
(b) the SR-AFO Exosuit when active and inﬂated.
Fig. 3. Simpliﬁed diagram showing the lever force produced to assist in
plantarﬂexion when going from inactive in (a) to active and inﬂated in (b),
while (c) shows the plantarﬂexion actuator post inﬂation
inspired by thin ﬁlm contracting actuator design. Further
evaluation of the IE actuators was completed . This
paper presents new developments to the SR-AFO exosuit,
including the addition of new actuators placed to assist
plantarﬂexion, a more robust sensor placement, and a smaller,
more reﬁned design of the control belt. Section II discusses
the design of the new SR-AFO exosuit, the addition of
the new plantarﬂexion actuators, and the characterization
and experimental testing of the plantarﬂexion actuators is
explored. Section III introduces the control strategy for
the exosuit, user testing, investigating the range of motion
(ROM) of the user with and without the SR-AFO exosuit, as
well as an electromyography (EMG) study of the muscles
used in plantarﬂexion at the ankle joint during walking,
and the antagonistic muscles for dorsiﬂexion. Section IV
includes a discussion of the overall performance of the SR-
II. DESIGN AND CHARACTERIZATION OF THE SR-AFO
A. Design Overview
The SR-AFO exosuit aims to assist ankle plantarﬂexion by
utilizing the intrinsic compliance of the fabric-based design
paired with pneumatic actuators anchored at speciﬁc points
on the body. The SR-AFO exosuit has a low proﬁle and ﬁts
tightly around the user’s foot and lower leg (Fig. 2). The low
proﬁle makes the wearable robot less cumbersome and cre-
ates an ease of application in different rehabilitative settings.
The lightweight materials prevent unwanted alterations to the
user’s kinematics during natural gait from becoming strewn
by restricted motion or added weight . The actuator
runs along the back of the user’s calf and pulls the heel
upward, producing a negative angle from a neutral position.
FUNCTIONAL REQUIREMENTS,CONSIDERATIONS,AND CONSTRAINTS
Design Requirements Characteristics
Design Considerations and Criteria 
Soft, compliant material Neoprene, Spandex and Nylon
Low proﬁle ∼5mm
Easy don/doff ≤30 s
Light weight ≤200 g
Motion and Force Considerations , –
Support Plantarﬂexion 1.6 Nm/kg
PF Muscle Assistance GA, SOL muscles
Assists Pre-Swing Active during late stance
Minimum ROM ≥30◦Dorsiﬂexion-Plantarﬂexion (DP)
Controls Criteria , 
Perturbation Timing + 10% natural weight on toe
Gait Cycle % Assisted 40% to 60% of gait
Pressure Threshold 100 kPa
Walk at a natural pace ≥1.0 m/s
The controls and hardware for the SR-AFO exosuit are not
carried or worn by the user. The maximum force output of
the actuator is modeled similarly to previous work , 
using the simple notion shown in Fig 3(a-b) to get torque:
τ=FLf oot (1)
where τis the assistive torque applied by the actuator, Fis
the uniaxial tensile force during contraction and Lf oot is the
distance from the center of the ankle joint to the posterior end
of the foot. In previous work , the soft fabric actuators are
modeled and tested in ideal conditions and ﬁxed boundary
conditions through ﬁnite element analysis (FEA) and testing
on a universal testing machine using rigid vice grips in
perfectly ﬁxed positioning. The contraction percentage (ε)
of the actuator was optimized via FEA using the following
L(x1) = (n−1)d+nx1+2d,(3)
where the deﬂated chamber length Liis equal to x1, which is
the initial length of each chamber, Lfis the ﬁnal length after
inﬂation, dis the distance between each chamber created by
the heat seal, and nis the total number of chambers. For
this work, 8 chambers have been selected to achieve desired
contraction. As contraction is a function of overall length,
the selected length was sufﬁcient to pull the ankle to the
1) Design Considerations and Criteria: A complete list
of all considered design requirements and characteristics is
outlined in Table I. The sock-like SR-AFO exosuit design
allows the materials to lay close to the net-shape of the
user. The SR-AFO exosuit has a proﬁle of 5 mm thickness,
adding minimal bulk to the leg with layers comprised of
neoprene and spandex. The materials and design of the SR-
AFO exosuit make it easy for the user to don/doff. The SR-
AFO exosuit is donned from the back of the leg, slipping
the heel into an opening in the back of the SR-AFO exosuit.
Straps with Velcro secure the SR-AFO exosuit on the top
of the foot, around the ankle, and below the knee (Fig.
2). The dofﬁng process is quick and easy, only requiring
the user to undo the Velcro straps and removing the SR-
AFO exosuit from the back. The actuators also have a low
proﬁle when active and inactive. The fabrication of the soft
actuators uses a thermoplastic polyurethane (TPU) coated
nylon fabric (200 Denier Rockywoods Fabrics) which is
thermally bonded with a 2 mm heat impulse sealer (AIE-500
2mm Impulse Sealer, American International Electric INC,
CA) which applies uniform heat and pressure to the seam to
create an air-tight seal .
2) Motion and Force Considerations: The tensile force
of actuation will match a percentage of assistance, which is
typically measured at 1.6 Nm/kg for plantarﬂexion torque
. The SR-AFO exosuit will assist the primary plantar
ﬂexor muscles, GA and SOL, to show a reduction of muscle
effort during late stance. The range of motion (ROM) for
plantarﬂexion and dorsiﬂexion is 30◦and 20◦, respectively.
The actuator is pulled tight to lay ﬂat along the calf but does
not restrict dorsiﬂexion. This ensures the actuator has ample
tension to contract and produce plantarﬂexion when walking
ensues. The tension of the actuator is critical to the amount
of pulling force it provides to the ankle; this topic will be
discussed in later sections.
3) Control Criteria: A tether connects the SR-AFO exo-
suit to a control box containing all electro-pneumatic hard-
ware. The SR-AFO exosuit allows the user to walk at their
natural pace while the actuators engage just before the foot
pushes off of the ground. This point varies across participants
but the timing coincides with the natural pace of the user.
The actuators inﬂate from 40% to 60% of the gait. This range
was chosen because it is during this point in the gait cycle
where the GA and SOL muscles are activated to produce
plantarﬂexion. A pressure threshold of 100 kPa is chosen to
reduce risk of mechanical failure due to the cyclical nature of
operation. The natural walking pace of a healthy participant
typically starts around 1 m/s, so the SR-AFO exosuit should
operate within this range.
B. Soft Actuator Design Principles
While preliminary modeling in ideal conditions was re-
quired for actuator design optimization in geometric pattern
determinations, it is not enough to assume that the actuators
will perform and yield the same force output when worn
on a person and afﬁxed to inextensible fabric anchoring
points on the SR-AFO exosuit. By allowing the actuator to
interface with a fabric anchor, it is assumed that a more
realistic force output can be measured to represent the
behavior of the actuator when worn on the SR-AFO exosuit.
The soft actuator is evaluated following the same testing
protocol used in testing Mckibben actuators  , as
the primary function is consistent between the two designs:
uniaxial contraction. The experimental setup for the quasi-
static isometric tests for varying angles consists of a vice
clamp, load cell, and a custom housing structure. The vice
clamp is ﬁxed to the bottom of the housing unit and the load
cell is supported at the top at a ﬁxed height.
Fig. 4. Single actuator testing overview and setup is shown in (a), and
double actuators in parallel (b). (c) Double actuators at varying angle, α.
Fig. 5. (a) Single actuator testing at varying contractions and pressure.
(b) Double actuator testing at varying angles at constant 100 kPa at
varying contractions. (c) Double actuator testing at varying contractions
C. Quasi-Static Isometric Test Conditions
A single actuator is evaluated in this setup to characterize
the contracting and uniaxial tension forces produced with
a fabric anchoring point (Fig. 4a). The actuator is held at
a ﬁxed length for 0%, 10%, 20%, and 30% contraction.
Pressure is increased from 0 kPa to 100 kPa in increments of
10 kPa and the resulting payload is recorded across a total
of three trials for each pressure and contraction ratio (Fig.
5a). A linear trend in force output is observed past 50 kPa in
each displacement evaluated and a maximum payload at 100
kPa. While previous work analyzed a single actuator, this
paper investigates how parallel actuators contribute to force
output and inﬂation time. To determine the resultant force
difference, two actuators are placed in parallel (Fig. 4b).
The comparison of a single actuator and parallel actuators is
observed at a ﬁxed 0% contraction to track maximum force
output for 0 kPa to 100 kPa in increments of 10 kPa. A single
actuator results in a maximum payload capacity of 9.3 ±0.15
kg whereas two actuators in parallel in the same orientation
and test condition result in a 45.3% increase to 13.51 ±0.31
kg payload capacity (Fig. 5b). The two actuators are tested in
parallel to each other and used the same experimental setup
as the single actuator. The pressure is varied from 0 kPa to
100 kPa at increments of 10 kPa. This test is repeated at
different contraction percentages; 0%, 10%, 20%, and 30%
contraction. The maximum payload observed is of 13.52 kg
and occurs at 0% contraction.
1) Constant Pressure: To verify which angle the two
actuators perform best at, the angle between the two actuators
is varied while maintaining a constant pressure (100 kPa) and
recording the tensile force produced as seen in Fig. 4c. The
two actuators are ﬁxed to a load cell at the same contact
Fig. 6. Time delay in actuation to the target pressure for inﬂation (a) and
deﬂation (b) a with a single actuator and two parallel actuators.
Fig. 7. Time delay in resultant force from actuation to the target pressure
of 100 kPa for a ﬁxed rise of (a) 0.1 sec, (b) 0.2 sec, (c) 0.3 sec, and (d)
0.4 sec for a period of 1 sec. The force output over four pressure levels is
shown in (e) for a period of 1 sec and rise time of 0.3 sec.
point, while the distance between the actuators is varied at
the bottom to provide different angles. Initially, the actuator
angle is set to 0◦, where one actuator is positioned vertically
parallel to the adjacent actuator. The angle is adjusted in
increasing increments of 5◦from a parallel vertical position
until a total angle of 30◦between the two actuators is
obtained. This process is repeated with the actuator being
held at a ﬁxed contraction percentage of 0%, 10%, 20%,
and 30% from its initial length. This allows the force output
of the actuators to be analyzed quasi-statically as both
a function of the placement angle and displacement. The
results of the testing can be seen in (Fig. 5c), which show
that varying the actuator angle is relatively insigniﬁcant.
2) Constant Angle: The experimental setup for the quasi-
static isometric test for varying pressure uses the same
housing unit as the quasi-static isometric tests for varying
angles. A vice clamp is used to ﬁx the actuators in place
at the bottom of the housing unit. The angle between the
two actuators is ﬁxed at 5◦, which is the angle determined
to have the greatest force by the previous quasi-static test
for each level of contraction: 0%, 10%, 20%, and 30% (Fig.
5c). The two actuators are ﬁxed to the load cell at the top
of the housing unit at the same point. The pressure is varied
from 0 kPa to 100 kPa in 10 kPa increments, while the angle
between actuators is held constant and the contraction of the
actuator is again ﬁxed at 0%, 10%, 20%, and 30%.
D. Actuator Performance Validation
The SR-AFO exosuit needs sufﬁcient cycle time to inﬂate
and deﬂate within a short window. To determine the inﬂation
and deﬂation speeds of the soft actuators, a pressure sensor
(ASDXAVX 100PGAA5, Honeywell Sensing and Produc-
tivity Solutions, Charlotte, USA) is used to monitor the
instantaneous pressure provided to the actuator. The pressure
is held constant at the source and a solenoid valve (320
12VDC, Humphrey, USA) is opened to create inﬂation. The
pressure is monitored at a baud rate of 9600 Hz and the
output is evaluated for the rise time. Fig. 6 shows inﬂation
and deﬂation time of both single and double actuators under
no-load conditions. Observing the dynamic force response
it can provide is critical in understanding how the actuator
performs when used during walking. A universal tensile
testing machine (UTM) (Instron 5565, Instron Corp., High
Wycombe, United Kingdom) is used to measure the dynamic
response of the actuators. Fig. 7a-d shows the actuator force
output with various actuation times pre-deﬁned. For each
instance, the actuator pressurizes for 0.1, 0.2, 0.3, and 0.4
sec for a ﬁxed period of 1 sec, averaged across 30 trials. A
manual valve is used to ensure the actuators reach the desired
pressure of 100 kPa set for each trial. Under 0% contraction
and loaded conditions, the actuators are able to reliably reach
the peak pressure and corresponding force output of 118.2
±3.1 Nat an interval of 0.3 sec. Fig. 7d shows that the
resultant force from the actuators becomes more accurate in
both rise time and the deviation of the peak force at higher
pressures when observed between 100 kPa and 40 kPa. The
force returns to 0 Nat the end of each trial, suggesting that
that inﬂation and deﬂation time of the actuators under load
will be sufﬁcient for walking.
III. CON TROL O F THE SR-AFO EXOSUIT AND
PRELIMINARY HUMAN EXPERIMENTS
The SR-AFO exosuit is controlled by a control box, which
houses all of the electro-pneumatic components required to
provide dynamic control to the SR-AFO exosuit, and allows
the SR-AFO exosuit to function autonomously without con-
tinuous monitoring or input from the user. The timing of the
controlled perturbation for the soft actuators is critical for
ensuring the user is receiving assistance in the right location
at the proper interval. When this is achieved, muscle exertion
of the GA and SOL muscles will be reduced in the forward
swinging motion of the leg. This reduces the effort exerted
during walking for the user, creating a useful solution to gait
rehabilitation. The actuator will inﬂate during mid-stance,
when the GA and SOL isometrically contract to activate the
Achilles tendon to provide pre-tensioning for plantarﬂexion.
A. Control of the SR-AFO Exosuit
The control box houses the logic controller (Arduino
Mega 2560 Rev3), ﬁtted with a ProtoShield to connect
Fig. 8. Logic control box for the SR-AFO exosuit, which controls
the timing of actuation, monitors pressure levels, and collects kinematic
behavioral data of the user through FSR sensors.
and read analog signals from force-sensitive resistor (FSR)
sensors (Interlink 406, Adafruit, New York, USA). The FSR
sensors are embedded in the user’s shoe at the anterior
and posterior ends of the shoe insoles to provide kinematic
data information that is critical to time-events to detect heel
strike and the weight shift preceding pre-swing. A 12V,
5000 mAh LiPo Battery is used to power the system. A
custom voltage regulator board is used to step the voltage
down to 5V to power both the logic board and the pressure
sensors (ASDXAVX 100PGAA5, Honeywell Sensing and
Productivity Solutions, Charlotte, USA), which monitor the
resulting pressure throughout operation. A portable air com-
pressor (Model 8010A, California Air Tools, USA) is used
to provide a pneumatic actuation source and is controlled
through the use of 3-way, 2-channeled solenoid values (320
12VDC, Humphrey, USA), which activate during speciﬁed
times in the human gait cycle to provide instantaneous
pressure to the actuators through the tether (Fig. 8). The
valves are powered through MOSFETs (IRF520 MOSFET
Driver Module) which connect the signal from the logic
controller to the 12V power rail. The housing of the control
box is made from a custom 3D printed PLA design to house
all of the electro-pneumatic components. LEDs are used as
indicators once the lid of the control black box is closed to
provide information to the user regarding the current state or
status of the system. A power switch is provided on the side
of the box to shut down the system and safely disconnect
from battery power when not in use.
B. Human Experiments
A total of three healthy participants are recruited for this
preliminary study [ages: 21 - 26 years, weight: 60.5 - 65.8
kg, height: 1.67 - 1.72 m]. This study has been approved by
the Institutional Review Board of Arizona State University
(STUDY00004351). In preparation for the walking trials,
surface EMG sensors (Bagnoli Desktop System, Delsys,
Natick, MA) are used to monitor the tibialis anterior (TA),
Fig. 9. Controller logic based on bang-bang control, based on input from
the FSR, producing inﬂation in the actuator when FSR passes a threshold.
Fig. 10. State of the system during a single gait cycle going from heel
strike to heel strike showing the threshold being reached prior to the active
range of the actuator which spans from mid-stance to pre-swing.
GA, and SOL throughout the duration of the trials after
the maximum voluntary contraction (MVC) and the muscle
activity during relaxation is collected as per standard Interna-
tional Society of Electrophysiology and Kinesiology (ISEK)
protocols . Ankle angle in the sagittal plane for ankle
Dorsiﬂexion-Plantarﬂexion (DP) is tracked throughout the
trials using a wireless dual-axis goniometer (SG110, Biomet-
rics Ltd, UK). All data is streamed through a motion capture
system (Bonita 10, VICON Inc., Los Angeles, CA), and the
participant dons a custom shoe retroﬁtted with FSR sensors
as described in section II.E. The participant is asked to walk
on a split-belt treadmill (Bertec Treadmill, Columbus, OH,
USA) for two minutes prior to data collection to allow for
acclimation to the repetitive and controlled treadmill walking
and speed. This time is used to let the participant select a
comfortable walking speed which ranged between 1.0 m/s
and 1.2 m/sfor the three participants.
The participant is asked to stand naturally with even
weight distribution on each foot. Bodyweight and applied
force on the FSR sensor in the shoe is recorded, and the
threshold is set at 10% above this calibrated value for the
FSR in the toe to detect the weight shift and transition out of
mid-stance during walking (Fig. 10). Three walking trials are
performed for each participant with 1) no SR-AFO exosuit,
2) the donned passive SR-AFO exosuit, and 3) the SR-AFO
exosuit active. During each walking trial, the participant is
asked to walk on the treadmill at their selected walking speed
Fig. 11. Average EMG data for individual steps collected and averaged
over 2 minute walking trials, shown here for a representative subject. The
EMG signal for the GA muscle is shown in (a), for the SOL in (b) and the
TA in (c).
Fig. 12. Average ankle angle for DP across each two minute walking trial,
with SR-AFO exosuit passive, active, and with no SR-AFO exosuit for all
three healthy participants.
for the aforementioned experimental setups with a 1 min
resting period between trials.
For each participant, the trials are averaged across the
duration of each step, synchronizing heel strike as the start
of the gait cycle for the right foot. Heel strike is detected
via the force plates in the split-belt treadmill. The difference
between muscle activity recorded with no SR-AFO exosuit
and the passive SR-AFO exosuit was negligible. During
the trials where the SR-AFO exosuit was active, muscle
activity reduction is seen in the GA, and SOL while the
TA increases slightly during swing most likely due to the
hysteresis in the system deﬂating during that window. The
reduction is calculated by taking the area under the curve
between the average of the trials with and without SR-AFO
exosuit assistance between 40% and 60% of the gait cycle.
Average GA reduction observed for the three participants was
WALKI NG SP EE DS A ND M US CL E RE DU CT IO N FO R EAC H PART IC IPA NT
13.4%, and the average reduction for SOL was measured at
16.6% (Fig. 11). The reduction of each participant is shown
in Table II. The reduction observed during this window of
gait indicates that the SR-AFO exosuit is able to provide
assistance to the user during late stance by taking effort away
from the primary two muscles responsible for powering this
critical stage of the gait cycle. The increase in the TA muscle
is anticipated due to its antagonistic behavior in relation to
the actuation for plantarﬂexion assistance provided.
The angle of the ankle is monitored for DP and syn-
chronized across the gait cycle starting with heel strike,
and averaged across each 2 minute trial. With no SR-AFO
exosuit, the average DP range of motion is observed at
31.27◦, while the passive SR-AFO exosuit trials showed an
ankle DP ROM of 29.83◦. This falls within 0.57% difference
from the 30◦required ROM for ankle DP angle during gait.
When the SR-AFO exosuit is providing active assistance, the
average ROM in ankle DP is observed at 30.54◦, which meets
the minimum aforementioned requirement and falls within
2.3% difference from the original ROM with no SR-AFO
exosuit. An increase in maximum dorsiﬂexion angle was
recorded with the passive SR-AFO exosuit and active SR-
AFO exosuit, while a decrease in maximum plantarﬂexion
is observed. The increased dorsiﬂexion angle during active
stance is likely a compensatory motion by the user due to
the presence of the SR-AFO exosuit, which is supported by
the increased TA muscle activity during this window (40% -
60%). Decreased plantarﬂexion during late stance can likely
be attributed to an increased pretension effect caused by the
actuators, where the user does not need to push off with their
toe as far as they previously needed to without the SR-AFO
exosuit. The negligible change in muscle effort for GA and
SOL during this period can help support this hypothesis and
suggests the SR-AFO exosuit was able to provide assistance
to the healthy subjects for plantarﬂexion.
This paper presents the design of the SR-AFO exosuit
for plantarﬂexion assistance, characterization of actuator
properties and behaviors, and preliminary results for walking
trials of healthy participants. The design is lightweight and
form-ﬁtting for the user and can be easily worn over a
variety of athletic shoes of various sizes. The actuators are
characterized for payload capacity for a single actuator, and
two actuators in parallel. When an additional actuator is used,
the payload capacity increases 45.3% to produce a 13.51 ±
0.31 kg payload when inﬂated to 100 kPa. Three healthy
participants are recruited to participate in a walking study,
with no SR-AFO exosuit, the SR-AFO exosuit passive, and
the SR-AFO exosuit active and providing assistance. EMG
sensors and a wireless goniometer were used to monitor
the muscle activity and kinematics of the ankle joint. The
actuators are inﬂated during the window of late stance (40%
- 60% of the gait cycle). During this window, a reduction of
13.4% is seen for the GA, and 16.6% for the SOL. There is
a slight increase in the TA muscle during this stage as well
as through the swing phase, however, this is anticipated as
there may be some residual energy still stored in the system.
The ankle angle is able to verify that the minimum ROM
for ankle DP for normal walking (30◦) is still maintained
throughout gait, with a slightly offset proﬁle that does not
affect muscle reduction. These preliminary results suggest
that the SR-AFO exosuit would be an effective assistive
device for plantarﬂexion for impaired user testing in the
Future work will consist of the inclusion of further sensory
information about the system, such as embedding wireless
Inertial Measurement Unit (IMU) sensors for continuous
ankle angle monitoring, as well as developing algorithms to
provide pressure sensor feedback control based on the ankle’s
current position and stage of the gait cycle. Future studies
will also investigate the integration of these algorithms with
actuators proposed in future works to support additional
degrees of freedom and stages of gait. Finally, this work
will aim to begin testing with impaired users in entrainment
studies to evaluate the usefulness of controlled perturbations
from a completely soft, fabric-based actuation system for gait
training and regulating stride length.
ACK NO WL EDG EME NT S
C. M. Thalman is funded by the National Science Foun-
dation, Graduate Research Fellowship Program (NSF-GRFP)
award #1841051. This work is funded by the Global Sport
Institute of the adidas and Arizona State University (ASU)
Global Sport Alliance. The authors would like to thank
Kayleigh Gavin and Omik Save for contribution.
 D J Farris and G S Sawicki. The mechanics and energetics of human
walking and running: a joint level perspective. Journal of The Royal
Society Interface, 9(66):110–118, 2011.
 D A Winter. Overall principle of lower limb support during stance
phase of gait. Journal of biomechanics, 13(11):923–927, 1980.
 D A Winter. Energy generation and absorption at the ankle and knee
during fast, natural, and slow cadences. Clinical orthopaedics and
related research, (175):147–154, 1983.
 G. S Sawicki and D. P Ferris. Powered ankle exoskeletons reveal
the metabolic cost of plantar ﬂexor mechanical work during walking
with longer steps at constant step frequency. Journal of Experimental
Biology, 212(1):21–31, 2009.
 R R Neptune, SA Kautz, and FE Zajac. Contributions of the individual
ankle plantar ﬂexors to support, forward progression and swing
initiation during walking. Journal of biomechanics, 34(11):1387–1398,
 J Perry. Kinesiology of lower extremity bracing. Clinical Orthopaedics
and Related Research (1976-2007), 102:18–31, 1974.
 M J Mueller, S D Minor, J A Schaaf, M J Strube, and S A Sahrmann.
Relationship of plantar-ﬂexor peak torque and dorsiﬂexion range
of motion to kinetic variables during walking. Physical therapy,
 I. D´
ıaz, J. J. Gil, and E. S´
anchez. Lower-limb robotic rehabilitation:
literature review and challenges. Journal of Robotics, 2011, 2011.
 J F Geboers, M R Drost, F Spaans, H Kuipers, and H A Seelen.
Immediate and long-term effects of ankle-foot orthosis on muscle
activity during walking: a randomized study of patients with unilateral
foot drop. Archives of physical medicine and rehabilitation, 83(2):240–
 B Shi, X Chen, Z Yue, S Yin, Q Weng, X Zhang, J Wang, and
W Wen. Wearable ankle robots in post-stroke rehabilitation of gait:
A systematic review. Frontiers in neurorobotics, 13:63, 2019.
 J. Kwon, J. Park, S. Ku, Y. Jeong, N. Paik, and Y.L. Park. A Soft
Wearable Robotic Ankle-Foot-Orthosis for Post-Stroke Patients. IEEE
RA-L, International Conference on Soft Robotics (RoboSoft), 2019.
 J. Chung, R. Heimgartner, C. T. O’Neill, N. S. Phipps, and C. J. Walsh.
Exoboot, a soft inﬂatable robotic boot to assist ankle during walking:
Design, characterization and preliminary tests. In 2018 7th IEEE In-
ternational Conference on Biomedical Robotics and Biomechatronics
(Biorob), pages 509–516. IEEE, 2018.
 Y.L. Park, B.R. Chen, N.O. P´
erez-Arancibia, D. Young, L. Stirling,
R.J. Wood, E.C. Goldﬁeld, and R. Nagpal. Design and control of a
bio-inspired soft wearable robotic device for ankle-foot rehabilitation.
Bioinspiration & biomimetics, 9(1):016007, 2014.
 S. Lee, S. Crea, P. Malcolm, I. Galiana, A. Asbeck, and C. Walsh.
Controlling negative and positive power at the ankle with a soft exo-
suit. In IEEE International Conference on Robotics and Automation
(ICRA), pages 3509–3515. IEEE, 2016.
 Y. Ren, Y. Wu, C. Yang, T. Xu, R. L. Harvey, and L. Zhang.
Developing a Wearable Ankle Rehabilitation Robotic Device for in-
Bed Acute Stroke Rehabilitation. IEEE Transactions on Neural
Systems and Rehabilitation Engineering, 2017.
 J Ahn, T Patterson, H Lee, D Klenk, A Lo, H I Krebs, and N Hogan.
Feasibility of entrainment with ankle mechanical perturbation to treat
locomotor deﬁcit of neurologically impaired patients. In 2011 Annual
international conference of the IEEE engineering in medicine and
biology society, pages 7474–7477. IEEE, 2011.
 R C Browning, J R Modica, R Kram, and A Goswami. The effects
of adding mass to the legs on the energetics and biomechanics of
walking. Medicine & Science in Sports & Exercise, 39(3):515–525,
 A.T. Asbeck, R.J. Dyer, A.F. Larusson, and C.J. Walsh. Biologically-
inspired soft exosuit. In 2013 IEEE 13th International Conference on
Rehabilitation Robotics (ICORR), pages 1–8, 2013.
 C. M. Thalman, J. Hsu, L. Snyder, and P. Polygerinos. Design of
a soft ankle-foot orthosis exosuit for foot drop assistance. In 2019
International Conference on Robotics and Automation (ICRA), pages
8436–8442. IEEE, 2019.
 CM Thalman and H Lee. Design and validation of a soft robotic ankle-
foot orthosis (sr-afo) exosuit for inversion and eversion ankle support.
In 2019 International Conference on Robotics and Automation (ICRA).
IEEE, 2020. [Accepted].
 R Niiyama, D Rus, and S Kim. Pouch motors: Printable/inﬂatable
soft actuators for robotics. In Robotics and Automation (ICRA), 2014
IEEE International Conference on, pages 6332–6337. IEEE, 2014.
 A T Asbeck, R J Dyer, A F Larusson, and C J Walsh. Biologically-
inspired soft exosuit. In Rehabilitation robotics (ICORR), 2013 IEEE
international conference on, pages 1–8. IEEE, 2013.
 J. J Eng and D A Winter. Kinetic analysis of the lower limbs during
walking: what information can be gained from a three-dimensional
model? Journal of Biomechanics, 28(6):753–758, 1995.
 C. L Brockett and G. J Chapman. Biomechanics of the ankle.
Orthopaedics and trauma, 30(3):232–238, 2016.
 L M Mooney and H M Herr. Biomechanical walking mechanisms
underlying the metabolic reduction caused by an autonomous ex-
oskeleton. Journal of neuroengineering and rehabilitation, 13(1):4,
 C. Chou and B. Hannaford. Measurement and modeling of mckibben
pneumatic artiﬁcial muscles. IEEE Transactions on robotics and
automation, 12(1):90–102, 1996.
 M. Doumit, A. Fahim, and M. Munro. Analytical modeling and
experimental validation of the braided pneumatic muscle. IEEE
transactions on robotics, 25(6):1282–1291, 2009.
 R. Merletti and P Di Torino. Standards for reporting emg data. J
Electromyogr Kinesiol, 9(1):3–4, 1999.