The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study
Bortole et al. Journal of NeuroEngineering and Rehabilitation
The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study
Magdo Bortole 0 1
Anusha Venkatakrishnan 0 3
Fangshi Zhu 0
Juan C Moreno 1
Gerard E Francisco 2
Jose L Pons 1
Jose L Contreras-Vidal 0
0 Noninvasive Brain-Machine Interface Systems Laboratory, Department of Electrical and Computer Engineering, University of Houston , N308 Engineering Building I, 77204-4005 Houston , USA
1 Neural Rehabilitation Group, Cajal Institute, Spanish Research Council , Av. Doctor Arce 37, 28002 Madrid , Spain
2 TIRR Memorial Hermann and Department of PM&R, University of Texas Health Sciences Center , 1333 Moursund Street, 77030 Houston , USA
3 Currently at Palo Alto Research Center, a Xerox company , CA 94304, Palo Alto , USA
Background: Stroke significantly affects thousands of individuals annually, leading to considerable physical impairment and functional disability. Gait is one of the most important activities of daily living affected in stroke survivors. Recent technological developments in powered robotics exoskeletons can create powerful adjunctive tools for rehabilitation and potentially accelerate functional recovery. Here, we present the development and evaluation of a novel lower limb robotic exoskeleton, namely H2 (Technaid S.L., Spain), for gait rehabilitation in stroke survivors. Methods: H2 has six actuated joints and is designed to allow intensive overground gait training. An assistive gait control algorithm was developed to create a force field along a desired trajectory, only applying torque when patients deviate from the prescribed movement pattern. The device was evaluated in 3 hemiparetic stroke patients across 4 weeks of training per individual (approximately 12 sessions). The study was approved by the Institutional Review Board at the University of Houston. The main objective of this initial pre-clinical study was to evaluate the safety and usability of the exoskeleton. A Likert scale was used to measure patient's perception about the easy of use of the device. Results: Three stroke patients completed the study. The training was well tolerated and no adverse events occurred. Early findings demonstrate that H2 appears to be safe and easy to use in the participants of this study. The overground training environment employed as a means to enhance active patient engagement proved to be challenging and exciting for patients. These results are promising and encourage future rehabilitation training with a larger cohort of patients. Conclusions: The developed exoskeleton enables longitudinal overground training of walking in hemiparetic patients after stroke. The system is robust and safe when applied to assist a stroke patient performing an overground walking task. Such device opens the opportunity to study means to optimize a rehabilitation treatment that can be customized for individuals. Trial registration: This study was registered at ClinicalTrials.gov (NCT02114450).
Exoskeleton; Gait; Rehabilitation; Lower limb; Stroke
Stroke is the main cause of serious long-term disability
worldwide  and gait is one of the most important
activities affected in stroke survivors. In United States alone,
about 800.000 people have a new or recurrent stroke every
year  and most stroke victims experience significant
sensory-motor impairments and require rehabilitation to
achieve functional independence. In this context,
hemiparesis is a manifestation of stroke that affects the
contralesional side of the body, and commonly impacts gait
Task-oriented repetitive movements can improve
muscular strength and movement coordination in patients
with impairments due to neurological disorders such
hemiparesis. Gait training on a treadmill, sometimes
requiring manual assistance by therapists, is widely used
in clinical practice for stroke rehabilitation [5, 6].
Nevertheless, this technique has some limitations because the
manual assistance can lead to poor coordination and
synchronization of movements in both legs and fatigue of
therapist can limit the efficacy of the therapy.
Systematic reviews have considered the benefit of
traditional gait training paradigms in combination with
hightechnology, intensive rehabilitation approaches [7–10].
Compared with conventional therapy, robotic
rehabilitation can deliver highly controlled repetitive and intensive
training and reduce the burden of clinical staff, besides
providing a quantitative assessment of motion and forces.
The usage of robotic interventions in training tasks is
expected to augment plasticity and improve recovery.
Through these applications, patients will benefit as they
will be able to recover at a faster rate, thus enabling them
to resume their daily activities sooner and returning to
Body weight support with treadmill training machines,
such as Lokomat , ALEX  and LOPES , have
been used for gait training in stroke victims. For more
effective results in the rehabilitation process, it is known
that the patient’s involvement and participation in
voluntary movement of affected limbs is critical .
Stationary devices such as GaitTrainer  and HapticWalker
 have shown to facilitate the user’s cognitive
engagement and patient compliance during locomotor
training. However, a wearable robotic exoskeleton like the
H2 allows training functional movements in ecologically
valid ambulatory conditions. Therefore, cognitive
engagement can be further promoted while relevant sensory
inputs and central neuronal circuits may become activated
under physiological conditions (i.e., overground walking),
and lead to potentially important neural regeneration
Ambulatory exoskeletons with control strategies that
challenge the user to perform movements in these
reallife environments can be more effective to reinstate
neuroplasticity and improve motor functions . Hybrid
Assistive Limb (HAL)  is an ambulatory exoskeleton
that has been used for stroke rehabilitation. Two different
control strategies are used with HAL, depending on the
treatment purpose and user’s capabilities . The main
actuation mechanisms are based on surface
electromyography (sEMG) signals, which adjust the robot joint torques
for assistance depending on the measured muscle activity.
The second algorithm reproduces a stored movement
pattern based on acceleration and ground contact forces with
Safety and usability of HAL have been evaluated in
different studies [20–22]. In the study carried out by
Maeshima et al , which comprised 16 stroke patients
with severe hemiplegia, the authors conclude that sEMG
signals used to provide power assistance can make it
difficult for severely hemiplegic patients to perform activities
using their own muscles. This could lead to instability,
decreasing stride length and walking speed. Also, the
availability and quality of sEMG signals can vary from
patient to patient. Fragility and installation requirements
of electrodes can also be restrictive outside the laboratory
. The system based on sEMG signal requires a process
of adaptation and adjustment to a specific user that can
take up two months, depending of each person .
Other ambulatory exoskeletons that have been
evaluated for clinical use are ReWalk , Vanderbilt
exoskeleton , Ekso  and Kinesis . All these exoskeletons
have been evaluated with paraplegic users, except for the
Ekso, which is currently being tested in a clinical trial
with stroke patients . A recently developed
exoskeleton, Walk Assist, is also being tested in stroke patients
. However, there are no published reports on the safety
and usability of these systems in stroke patients. NASA
(National Aeronautics and Space Administration) in
partnership with IHMC (Institute for Human and Machine
Cognition) have also developed an exoskeleton called X1
 with powered hip and knee joints, intended for future
use in space as an compact exercise tool for astronauts.
This device was tested with 2 healthy subjects and 1
stroke patient in . It was noted that an actuation on
ankle joint in the X1 would be very clinically relevant
to counteract the foot drop problems in stroke stroke
The research presented here describes the
development and evaluation of a novel robotic exoskeleton for
gait rehabilitation in stroke survivors. The design
philosophy differs from treadmill devices, since motivation and
full engagement of patients are targeted with overground
walking in a real environment. The robotic exoskeleton,
named H2 (a totally improved version of the exoskeleton
described in ), has six joints actuated, including hip,
knee and ankle on both legs. To the best of our knowledge,
no ambulatory exoskeletons used for rehabilitation have
the ankle joint actuated. For gait rehabilitation in stroke
victims, actuation on ankle is important to target the foot
drop problem, which is common in most patients.
Furthermore, in the H2, an impedance controller
creates a force field around a desired joint trajectory to assist
patients, only applying the required joint torque to assist
completing the gait movements that patients are unable to
do so. Force field control uses the concept of
assistanceas-needed . Assisting only when patients need it can
lead to better results than fixed repetitive training as it is
more personalized to each individual ensuring consistent
patient engagement in the training .
Moreover, H2 presents an open architecture that allows
it to be integrated with other devices or systems. Both
wired and wireless communication interfaces are present
on the device for this purpose. This feature open means
for combined studies, allowing, for instance, integration of
H2 with neural interfaces. When associated with
exoskeletons, neural interfaces can be used for correlating the
aspects of learning during rehabilitation, as well as
creating brain-machine interfaces that further engage the
Here, we describe the H2 design and control algorithm
for gait assistance. The aim of this work is two-fold: to
present the development of the exoskeleton and secondly,
to perform a preliminary analysis, in a pre-clinical study,
of the safety and usability of the H2 robotic
exoskeleton when used for gait rehabilitation during 4 weeks
of intensive training in 3 participants with post-stroke
hemiparesis. Specifically, we wanted to explore the
applicability of the system in terms of safety and acceptance by
patients with impaired gait. Therefore, this study sought
to demonstrate safety and usability in post-stroke
hemiparetic patients who could at least walk short distances
but at low speeds and depending on external aids. While
this population does not represent target clinical users,
it helps test the applicability of the device in users with
gait deficits, and in whom successful overground walking
training with the H2 could provide added benefits (e.g.,
increased speed, improved intra- and inter-limb
coordination during gait, reduced dependence on assistive
H2 is a lower limb exoskeleton designed for
rehabilitation of adults between 1.50 and 1.95 m in height, with a
maximum body weight of 100 kg, such as stroke patients
following neurological insults. It also can be used for
Table 1 H2 range of motion
gait compensation in patients who have paralysis of the
lower limbs following spinal cord injuries. It is conceived
for overground gait training in a clinical environment
as a bilateral wearable device with six degrees of
freedom (DoF), in which hip, knee and ankle are powered
Various criteria informed the mechanical design. As
pointed out in previous work , an exoskeleton design
should be ergonomic, comfortable, lightweight, with a
strong structure, adaptable to different users and with
safety in mind. In H2, aluminum 7075 is primarily used in
the mechanical structure in consideration of mechanical
resistance and lightweight. The final device weights about
12 kg including its battery pack. The exoskeleton frame
has bilateral uprights for the thigh and the shank, hinged
hip, knee and ankles and articulated footplates (distally)
and a waist support (proximally). The mechanical
structure is designed to allow active and passive movements
in the sagittal plane. In the frontal plane, passive
movements of about 20 degrees are possible in the hip joint,
allowing for turns while walking. The range of motion
(ROM) in actuated joints is mechanically limited for safety
reasons. The maximum ROM possible across all joints is
shown in Table 1. For the ankle, plantarflexion is shown as
extension and dorsiflexion as flexion. These values were
chosen based on normal gait on healthy subjects , also
allowing users to perform sit-to-stand and stand-to-sit
The length of the thigh and the shank can be adjusted via
a mechanism of two telescopic bars that are pushed one
inside the other, and are fixed in different positions. The
same mechanism is used to change the position of the foot
relative to the exoskeleton’s ankle. The size and position of
adjustable rounded leg braces carriers with Velcro straps
allow for customization to individual requirements. Foam
pads are used to minimize pressure against the skin and
prevent damage. The exoskeleton supports its own weight
through the mechanical frame to the ground, so the users
do not feel any extra weight in their lower limbs.
Most types of actuators used in robotics cannot be used
in exoskeletons, since for this application high torques
are required while operating at higher speeds that most
actuators can provide . Main candidates available for
use as actuators in exoskeletons are electric, pneumatic,
hydraulic and series elastic actuators (SEA). The design
and selection of H2 actuators were based on average of
torque and power of each joint during normal gait (not
pathological) at normal speed . A study of different
possible candidates was evaluated. The most relevant
criteria to select the actuation technology to drive the human
joints were the specific power (ratio of actuator power to
actuator weight) and portability.
In this regard, linear hydraulic and pneumatic
actuators have high power density, but they usually are bulky
and present problems of internal leakage and friction
. SEAs have been used in some rehabilitation devices
, but they still face a common limitation about the
spring constant of the elastic element that is fixed. The
harmonious coordination of force and position between
patient and exoskeleton is difficult between different
subjects . The literature suggests that the use of electric
motors provide a reduction in power consumption
during gait . DC motors meet the criteria of necessary
power with a compact and portable solution for wearable
devices. Based on that, brushless DC motors coupled to a
type Harmonic Drive gearbox were selected.
A 100 W motor (Maxon, EC60 Flat Brushless) is used
in the hip, knee and ankle joints. This motor has a rated
voltage of 24 VDC and nominal torque of 220 mNm.
Furthermore, a gearbox (Harmonic Drive,
CSD-20-1602AGR) is coupled to the motor shaft in order to reduce
speed and increase torque. Harmonic Drive gearboxes
were selected to reduce the weight and size of the final
actuators. A gear ratio of 160:1 gives to each joint a
continuous net torque of 35 Nm and peak torques of 180
Nm. According to , an average torque of 35 Nm for the
hip actuator is presumed to be sufficient for most patients.
Power supply can be one of the most limiting factors for
an untethered exoskeleton embodiment. Although the H2
exoskeleton is designed to be used in a clinical setting,
a tethered device can lead to some drawbacks when
performing overground walking. Thus, the exoskeleton
was developed as an autonomous device. Different types
of energy sources have been used to power exoskeletons
. With improvements in battery technologies over the
years, a compact and higher capacity battery pack can
provide enough power for running an exoskeleton.
Autonomy also has to do with the performance of
the actuators. The developed exoskeleton was designed
with high efficiency motors and gearboxes, and
stateof-the-art electronic drives with very low dissipation.
Additionally, a compact lithium polymer battery pack
was specifically designed to power H2. The pack has
a nominal voltage of 22.5 VDC and a capacity of 12
Ah. The battery pack is integrated with the mechanical
frame and placed at the hip level, providing a
comfortable embodiment for the user and no extra weight on the
The interaction between user and exoskeleton is very
important for users’ comfort and safety in a wearable
robotic device . Also, when sensors have to be
physically placed on human limbs, several issues, specially
related to safety, comfort, reliability and donning/doffing
process need to be expected and appropriately dealt with.
In terms of physical interface with the human user, H2 is
designed in such a way that there are no sensors physically
attached to the human. All sensory information comes
from sensors placed on the exoskeleton: 6 potentiometers,
18 Hall Effect sensors, 24 strain gauges and 4 foot switches
are used to determine parameters such as angular position
and velocity, force and interaction torque between user’s
limb and exoskeleton.
Each joint is equipped with a precision industrial
potentiometer used as an angular position sensor. It
exhibits a high linearity and long rotational life. Its
stainless steel shaft is coupled to a toothed pulley, and a
toothed belt is used to transmit the joint’s motion, to
avoid slippage and therefore a loss of absolute reference
Strain gauges attached at each link are used as force
sensors. These sensors are designed to measure the torque
produced by the interaction between the user’s limb and
the exoskeleton. The strain gauges are connected in a full
Wheatstone bridge to enhance the measurement
sensitivity and accuracy. The bridge is excited with 5 VDC and
a custom-made electronic circuit balances the bridge for
null point measurement, also amplifying the output 500
times. Thus, the output signal is in a range that allows
torque measurements from –50 to +50 Nm. This range
was chosen based on the maximum torque of the
actuators with a safety factor for peak torques. A calibration
constant was obtained using a set of calibrated weights
and minimized with a least squares algorithm.
The footplate of the exoskeleton is equipped with two
foot switches based on resistive sensors, which binary
detect the contact between subject’s foot and the ground.
These sensors are located under the heel and the toe, and
their main goal is to detect the different phases during gait
segmentation. The exoskeleton, its actuators and sensors
are shown in Fig. 1.
The control hardware of the exoskeleton is shown in
Fig. 2. The main controller is based on a customized
electronic board (H2-ARM) designed specifically for
realtime control of H2. The small size of H2-ARM board
(56 x 44 mm) allows it to be placed on the exoskeleton
frame, reducing the bulk, as well as complexity and
difficulty of wiring and connections. Moreover, it eliminates
the need of a backpack being carried by the user, as most
lower limb exoskeletons have.
Fig. 1 H2 robotic exoskeleton. The six joints are powered by brushless DC motors coupled to Harmonic Drive gearboxes. All sensory information
comes from sensors placed on the exoskeleton: 6 potentiometers, 18 Hall Effect sensors, 24 strain gauges and 4 foot switches. A rechargeable
battery pack of lithium polymer powers the exoskeleton
H2-ARM computational power relies on an
STMicroelectronics ARM (Advanced Risk Machine)
microcontroller running at 168 MHz. The board has two
independent transceiver channels for real time
communication: one is used to connect to all six H2-Joint boards,
receiving sensory information and commanding the six
joint’s actuators; the other channel is intended to connect
to external devices.
The board also has two more communication ports,
both wireless: Bluetooth and Wi-Fi. Bluetooth
communication is intended to connect with a user interface
on a smartphone. The user interface is an
application that allows physical therapists to adjust certain
parameters as needed within the H2 during
rehabilitation. More details about the user interface are discussed
in the subsequent session. Wi-Fi link is used to send
Fig. 2 H2 overall control architecture. All sensors in both legs are connected to the H2-Joint1∼6 boards that communicate to H2-ARM board
through a deterministic real-time network. A Wi-Fi connection is used to capture the kinematic and kinetic data generated in the exoskeleton. A
Bluetooth link connects the exoskeleton to a user interface in a smartphone
data wirelessly via UDP protocol to a laptop, where
the data and information generated in the exoskeleton
can be visualized in real time and stored for offline
Each one of the six joints is equipped with an
H2Joint board (numbered from 1 to 6). Each board is in
charge of data acquisition of different joint’s sensors:
angular position, interaction torque, joint velocity and
foot-ground contact. H2-Joint1∼6 contain all the circuitry
of the analog filters for each joint sensor and also the
amplifiers for the strain gauges. The sensors’ analog
output are digitalized by a Digital Signal Processor (DSP)
microcontroller after the filtering and amplifying
process. A small data packet of six bytes aggregates the
sensor’s information on each joint and is sent to H2-ARM
controller every one millisecond.
The brushless DC motor’s drives are embedded directly
into the H2-Joint∼6 boards. The drive, developed
specifically for this application, is very compact and
lightweight, receives digital set points and it is small
enough to be mounted directly on the motor side in the
exoskeleton’s frame. This approach decreases the amount
of electromagnetic noise and the number of wires in the
A physical communication network that
guarantees strict determinism, data collision avoidance and
optimized data transfer for small data packets is used
to connect the H2-ARM and H2-Joint1∼6 boards.
The network structure has a deterministic real-time
communication based on Control Area Network (CAN)
technology running at 1 Mbps. The network allows an
unlimited number of nodes (limited only by the
electrical load on the bus) and does not require any alteration to
add or remove nodes. It is flexible in terms of
configuration, automatically avoids data collision and corrects data
packets errors in the transmission.
Each communication cycle in the network protocol
involves passing a message from the H2-ARM node to
all H2-Joint∼6 nodes in the network. As the message
travels on the bus, each H2-Joint∼6 reads its assigned
actuator command data byte (by looking for its own
ID number and message byte sequence). Then, each
H2-Joint∼6 returns one message back to H2-ARM node
with its locally collected sensor data.
Since the communication cycles occur at a fixed rate
(1 kHz) set by the control scheme on H2-ARM, this
protocol allows for deterministic control. Also, it provides
built-in network error detection as, for every message
received, each H2-Joint∼6 has to return data information
to H2-ARM. As a result, H2-ARM has a robust means
to determine the integrity of the network and the correct
operation of the joint’s actuators. If some failure occurs on
the network that cannot be corrected automatically (for
instance, a cable disconnection), H2-ARM instantly stops
the exoskeleton and shuts off the joint power for safety
Control algorithm for gait assistance
Position or trajectory control is a widely implemented
robotic strategy [11, 45–48]. In this control mode a
position controller guides the patient’s limb to a fixed
reference path, while receiving the joint angles as a feedback.
For lower limbs, the reference trajectory is a normal gait
pattern previously recorded from a healthy subject.
Because the gait pattern differs slightly between
individuals, there are some disadvantages to the implementation
of a trajectory control based on a pattern of another
individual. Recent research efforts are directed at
improving therapeutic robot transparency. It is critical that the
design of assistive controllers does not hinder the patients’
residual control and provides only the required amount
of torque. Thus, newer approaches to actively adjust joint
impedance during walking to alter the muscle torque
production for a functional purpose, such as
modulation of muscle activation, are preferred. When using fixed
position tracking controllers in overground exoskeletons,
undesired motions, such as those caused by spasms, may
cause large actuator forces that could led to unsafe
situations. Adding compliance operation to the robotic system
can naturally absorb large position errors, thereby
compensating for the consequences of these undesired
movements. In order to allow for a more compliant operation,
we developed an algorithm that takes into account the
interaction torque between subject and exoskeleton, in
order to produce an adaptive reference for gait assistance.
Figure 3 represents the algorithm scheme used in H2.
The adjusted reference trajectory θadj is given by (1) and
(2), where s is the Laplace operator.
θref is a vector containing the recorded angles based on
normal gait from a healthy subject and θint is the angle
related to the interaction torque Ti between
exoskeleton and subject’s limb. This angle is estimated using the
value of inertia J and damping B of the exoskeleton frame
alone, and it increases or decreases proportionally to the
interaction torque between the subject and exoskeleton. If
interaction increases, it means that the difference between
the trajectory of the subject’s limb and the trajectory of the
exoskeleton is higher. Thus, θint as given by (2) increases
and it is subtracted from θref , giving to the position
controller a corrected angle. The maximum value that the
adapted trajectory can deviate from the recorded
trajectory can be adjusted using Gint, which is a normalized gain
value between 0 and 1, where 0 allows no deviation from
reference trajectory. With the user interface, physical
therapists can change this gain value ad-hoc for each situation
based on the patient’s disability.
The second part of the algorithm is responsible for
assisting the patient’s gait based on their disability level.
To achieve this, the adapted trajectory feeds a
position controller whose output is converted to an input
to the torque controller. Consequently, H2 provides an
output torque for the actuators, which is proportional
to trajectory deviation. The output torque is estimated
based on the motor’s electrical current and the
gearbox reduction rate. Together, this algorithm creates a
force field control that guides patient’s limb in a
correct pattern, only assisting patient when he/she deviates
from the trajectory. Because all joints on the exoskeleton
have their own dedicated electronics and control
parameters, each actuator can be independently controlled.
This allows the algorithm to generate specific
assistance for each joint separately. Especially for hemiparetic
stroke patients, who have asymmetric functioning across
both lower limbs, this exoskeleton can adapt its
functionality in real time based on each individual patient’s
needs, without requiring a manual adjustment for each
Furthermore, during training, physical therapists can
also adjust the H2’s gait speed, across 10 different possible
speeds approximately between 0.5 to 1.8 km/h, to
personalize training for each patient. Since H2 adapts the
preprogrammed reference based on user’s gait, the absolute
final speed is, in some way, user-dependent.
In this pilot pre-clinical study, H2 functionality, safety and
usability were evaluated in 3 post-stroke hemiparetic users
during 4 weeks of gait training. The Institutional Review
Board (IRB) at the University of Houston approved all
Three eligible male participants with post-stroke left
hemiparesis participated in this study after providing
informed consent. Demographic data and lesion types for
participants are provided in Table 2.
Fig. 3 Control scheme for gait assistance. First, the algorithm generates an adaptative reference trajectory. Based on this reference, a force field
controller guides the patient limbs, applying the necessary torque to complete gait in each joint independently
In this pilot clinical investigation, the study design
consisted of an open-label assignment of participants to H2
robot-assisted gait training. During each training session,
subjects were asked to perform an overground walking
task guided by the H2 in assist-as-needed mode with
a pre-selected gait speed along a 50 m circular or 120
m linear path. After wearing the exoskeleton, patients
were instructed to walk as much as they were able and
encouraged to take rest breaks as necessary. Figure 4
illustrates a patient using H2 in the beginning of a
training session. The gait start and stop process was
controlled by the patient using two hand buttons placed on
a walker, which was used as a gait assistive device during
An experienced physical therapist followed patients
during the whole training period. Patients were allowed
to change the walking speed in real time during
continuous walking from level 1 (slowest) to 10 (fastest) based on
their comfort level. Based on patient feedback, the
therapist used the smartphone interface to adjust gait speed as
necessary. At least two more persons were present during
training sessions and followed patients to ensure patient
safety. Participants received approximately 12 sessions of
training over approximately a 4-week period. All 3
participants were chronic stroke patients and were not receiving
any additional gait training or physical therapy during this
period of experimental training.
Pre- and post-training (within 2 weeks after
training) standard clinical assessments were performed by an
independent rater (a second physical therapist that did
not participate in the robotic training). The assessment
included Berg’s Balance Score, Barthel Index, Functional
Gait Index, Fugl-Meyer’s assessment of motor recovery
(lower extremity), Timed Up and Go test and 6-minute
walk test. These assessments were included to help
document any clinically relevant behavioral changes that
may occur in response to training with the H2 powered
Table 2 Demographic data and lesion types for stroke participants
Fig. 4 Stroke patient using H2 exoskeleton at the beginning of one training session
exoskeleton. The study protocol is also registered and
available at ClinicalTrials.gov (NCT02114450).
Walking angular kinematics, interaction torques and
motor torques for left and right hip, knee and ankle
joints, together with toe and heel ground contact were
sampled at 100 Hz by the H2. All data were
transmitted wirelessly over a Wi-Fi link to a laptop via UDP
Robotic exoskeleton intervention
In this pilot clinical study, the usability and safety of using
the H2 for robot-assisted longitudinal gait training in
stroke patients has been established. The 3 participants
with stroke were able to finish 12 sessions of training over
a period of approximately 4 weeks (Subject 2 completed
10 sessions only as he had to miss 2 sessions due to a
personal schedule conflict). At the first session, all
participants started at the lowest walking speed (0.5 km/h)
and were able to increase the gait speed across sessions
as training progressed. The deviant gait pattern of the
three stroke patients was retrained into a more symmetric
pattern during the training time, as seen in the Fig. 5.
Further, the number of steps walked, a measure
indicative of walking distance, increased across sessions for all
participants (see Fig. 6). Additionally, the pre- and
posttraining clinical outcome scores for the 3 participants
are presented in Table 3. As shown, Subject 1 had no
changes in the clinical outcomes, while Subjects 2 and 3
had showed slight improvements in Six Minutes Walk test,
Time Up-and-Go test, as well as a minor improvement in
the Fugl-Meyer Lower Extremity Motor assessment score.
Subject 2 also had a slight improvement in Barthel Index
and Subject 3 showed a slight improvement in Functional
The time needed for donning and setup H2 was very
short, less than 10 minutes elapsed from the time
participants arrived before gait training was started. The doffing
process was even faster, less than 2 minutes. No adverse
effects were observed during training, including no skin
irritation or redness, no sore spots, any pain or discomfort
during or after training.
H2 also demonstrated significant autonomy in the
context of battery power. A totally charged battery pack could
run the exoskeleton for about 9 training sessions of an
average of 40 minutes each session. Considering that in
Fig. 5 Hip, knee and ankle trajectories performed by all subjects. Blue line is the reference trajectory that patients are guided through by means of a
force field. Red line represents the average of all steps performed by subject in the first training session. Black line is the average of all steps
performed at last session. Trajectories are represented based on stride length percentage, from heel strike to next heel strike
each session, a participant walked 30 minutes on average,
H2 could run for more than 4 hours of continuous walking
with a single battery charge. Also since the battery pack is
detachable from the mechanical frame, it is very easy to
replace an empty one with a fully charged battery pack.
The H2 tested was the first device built and thus, issues
might have been expected. Remarkably, the H2
robotassisted gait training was conducted without any major
problems. Only minor technical issues occurred without
impacting user’s safety and were easily fixed.
Total Number of Steps
Number of Steps per Minute
Fig. 6 Number of steps performed in all training sessions by all subjects. Although the number of steps and walk speed depends on patient’s
conditions and mood on the training day, the overall results clearly show an increase over time
Table 3 Pre- and pos-assessment data
Berg Balance Scale (0-56)
Functional Gait Index (0-30)
6 Min Walk Test (meters) Time Up-and-Go Test (sec) Fugl-Meyer LE (0-34) Barthel Index ADL (0-20)
Lastly, patients participating in the study were very
excited about training with the device. When asked to
evaluate the ease to use of device in each session on a
Likert scale, the average rating for the 3 patients in all 12
sessions were 7.2, where 0 indicates “extremely hard to
use" and 10 indicates “extremely ease to use". The main
positive feedback received from patients when training
with H2 was: “the device is lightweight"; “wearing it is fast
and simple"; “I can feel that it helps my knee flexion"; “it
is more exciting walking overground with this device than
my previous treadmill training with manual assistance"
and “I wish I had access to this device when I was in the
hospital for inpatient rehabilitation after my stroke". The
main negative feedback received from patients was: “it felt
weird at the first moment and took me some time to get
used to it in my first training session, since I have never
used a robotic device like this".
Here, we present the development and the first
evidence for safety and usability of the H2 wearable robotic
exoskeleton in the context of post-stroke gait
rehabilitation. The main finding of this work is that the developed
H2 exoskeleton provides a means for safe and intensive
gait training in hemiparetic stroke survivors. Across 4
weeks of training in 3 stroke subjects, the H2 exoskeleton
proved to be easy to use, with a fast donning and
doffing process and was very well accepted by patients as a
potential rehabilitation device.
Importantly, the results from this pilot clinical study
indicate that the H2 operated in “assist-as-needed"
control mode allows reshaping of the asymmetric, deviant
hemiparetic gait in stroke survivors through a relatively
short period of training. It is important to note that in
most stroke victims, the lack of knee flexion during swing
creates an abnormal compensatory movement in the hip,
commonly known as hip hiking . Also, most patients
do not rely on their paretic leg, hence, they do not shift
weight equally on both lower limbs during walking. This
behavior creates an asymmetric gait pattern where the
stance phase on the paretic leg is shorter than the
unaffected leg. The gait assistance force field implemented in
H2 guided patients in a correct gait pattern, creating a
stance phase that is equal across both lower limbs and this
could potentially prevent the compensatory hip hiking.
As a result, while using H2 patients are being trained to
the correct pattern of weight shifting between lower limbs
and knee flexion, which can further contribute to the H2’s
Actuation at the ankle was another important aspect of
H2 design. During training it helped avoid foot drop and
could help patients to work on dorsiflexion movements.
The H2’s control algorithm, therefore, helps these patients
relearn a symmetric gait pattern across both lower limbs
by providing assistance as needed at the appropriate
limb segments and joints. Importantly, the ability to
perform this training in a functional context such as over
ground walking is of major clinical significance.
Furthermore, it is very interesting to note that this training is
stimulating and challenging even for the participant with
chronic stroke (5 years ago). Coupled with the
motivational component of training provided by a novel robotic
gait training regimen, the H2 allows these participants to
experience kinesthetic feedback of near-normal gait
patterns in over ground walking. Since the 3 participants in
this study were able to increase walking speed and
distance across training sessions, it would appear that H2
robot-assisted training can potentially recruit extant
neuroplasticity and promote improved motor control in these
However, these findings must be considered with the
caveat that this study is limited to a small subgroup of
patients that is not representative of the entire stroke
population and therefore, conclusions cannot be drawn
regarding gait improvements after use of H2.
Furthermore, as seen from the patient demographic data and
functional outcome scores, the subgroup of stroke
participants included in this study is also heterogeneous in
terms of time at which H2-assisted training was
provided with regards to their stroke onset as well as their
individual functional impairments. This is an important
factor to be considered as this population is very diverse,
and therefore, no two patients are alike in terms of their
Therefore, it is critical that H2 robot-assisted training
be personalized to each individual based on his/her needs.
In this regard, further modifications can be implemented
in the control algorithm to provide variable resistance
once the user has reached a certain threshold in terms of
torque generation and/or joint angular position/velocity.
This will help ensure progressive, adaptive changes to the
training regime and is a clinically significant issue that
warrants further investigation.
Notably, the modular design of the H2 is particularly
relevant for stroke rehabilitation. Since various segments of
the device can be used independently, H2 offers promising
means of using unilateral Hip-Knee-Ankle, Knee-Ankle or
only Ankle versions of the device, customizing treatment
protocols to each patient’s specific needs. These questions
need to be addressed in future research, in order to help
develop optimal control algorithms to use these modular
components of the H2 for individualized rehabilitation.
Similarly, appropriate intervention durations, and
frequency of training i.e., “dosing schedules" are still not well
established for such wearable robotic rehabilitation
protocols, which also needs to be examined in careful detail
in further clinical investigations. Furthermore, in order to
fully utilize the functionality of the lightweight wearable
H2 device, future training protocols can also include other
functional tasks such as sit-to-stand, stand-to-sit and stair
As seen (Table 3), lack of major changes in
clinical outcomes precludes any conclusions about functional
improvements when training with H2 in this study. We
believe this is primarily because while the participants in
this study had qualitative gait asymmetries and
impairments, this is not captured by the granularity of the
standard clinical outcome measures. Further, in some
of the items such Berg’s Balance Scale and Functional
Gait Index, if participants achieve scores closer to the
ceiling, it is impossible to track any further qualitative
improvement using those items. This brings to light the
importance of developing novel metrics or outcome
measures that are sensitive and capable of tracking behavioral
changes quantitatively and qualitatively in robotic
rehabilitation paradigms. However, it is important to study
the relationship of these novel metrics to standard
clinical outcomes, in order to describe the functional domain
that is being assessed. Our research efforts are currently
directed towards this, as it is a very important factor in the
Finally, the factors discussed above such as inadequate
“dosing" in terms of frequency and duration of training
may have prevented sufficiently progressing treatment for
each participant based on their functional levels. These
questions need to be addressed in a clinical
investigation with a larger population, along with comparison
of H2 robot-assisted training to conventional physical
rehabilitation regimes. Our future work, therefore, is
focusing on a controlled clinical study in a larger sample
of participants with stroke.
In summary, this work presents the development and
evaluation of H2, a novel lower limb robotic
exoskeleton for rehabilitation of stroke survivors. This device
is lightweight and battery-powered, thereby allowing for
gait training in functional contexts such as overground
walking in comparison to more traditional tethered or
treadmill-based robotic rehabilitation devices. Further,
the control of H2 is based on a custom assist-as-needed
algorithm that creates a force field along a desired
trajectory, proportionally applying torque only when patient
deviates from the pre-programmed correct pattern. This
force field control, therefore, can help restore a symmetric
gait pattern in hemiparetic stroke survivors, by assisting
only the segments that need it and possibly preventing
undesired compensatory movement patterns, such as hip
Additionally, a customized mobile-based user interface
allows the therapist to personalize and adjust the
maximum allowed deviation from the reference based on a
specific patient’s condition. Finally, we also present early
findings from a clinical evaluation of the H2 for gait
rehabilitation in 3 participants with post-stroke
hemiparesis. Participants showed adaptive improvements in
their gait trajectories across the training sessions over
5 weeks. These results are encouraging and provide
the first evidence for safety and feasibility of using
the H2 for functional gait training in stroke patients.
Our future work aims to evaluate the therapeutic
benefits of active training with the exoskeleton in
restoring gait function in a larger population of stroke
In summary, the developed H2 device opens up future
research avenues to study methods to optimize
rehabilitation protocols that can be customized for individuals
with gait impairments following neurological injuries and
with the capacity to deliver high dosage and high
intensity therapies. Taken together, these advances can have a
huge clinical impact by helping accelerate recovery and
improve functional independence and quality of life in
The authors declare that they have no competing interests.
MB, JCM and JLP designed the exoskeleton. MB, AV, GEF, JLP and JLC designed
the experiments. MB, AV and FZ ran the experiments, collected and analyzed
data. Data interpretation was done by MB, AV, JLC and JLP. MB drafted and
wrote the manuscript. AV contributed to writing the clinical results and
discussion sections of the manuscript. JCM, GEF, JLP and JLC reviewed the
draft and made substantial comments. JLP and JLC were responsible for
funding. All authors have read and approved the final manuscript.
The authors gratefully acknowledge the voluntary participation of the
patients, as well as the assistance of Jonathan Kung, Victor Issa and Zachary
Hernandez in data collection. This work has been partially supported by the
Noninvasive Brain-Machine Interface Lab at the University of Houston and the
HYPER Project (Hybrid Neuroprosthetic and Neurorobotic Devices for
Functional Compensation and Rehabilitation of Motor Disorders). Ministerio
de Ciencia y Innovación, Spain (CSD2009 - 00067 CONSOLIDER INGENIO 2010).
1. Feigin VL , Lawes CM , Bennett DA , Barker-Collo SL , Parag V. Worldwide stroke incidence and early case fatality reported in 56 population-based studies: a systematic review . Lancet Neurol . 2009 ; 8 : 355 - 69 .
2. Roger VL , Go AS , Lloyd-Jones DM , Adams RJ , Berry JD , Brown TM , et al. Executive summary: Heart disease and stroke statistics - 2011 update. Circulation . 2011 ; 123 : 459 - 63 .
3. Duncan PW , Zorowitz R , Bates B , Choi JY , Glasberg JJ , Graham GD , et al. Management of adult stroke rehabilitation care: a clinical practice guideline . Stroke . 2005 ; 36 : 100 - 43 .
4. Jorgensen HS , Nakayama H , Raaschou HO , Olsen TS . Recovery of walking function in stroke patients: the copenhagen stroke study . Arch Phys Med Rehabil . 1995 ; 76 : 27 - 32 .
5. Lau KW , Mak MK . Speed-dependent treadmill training is effective to improve gait and balance performance in patients with sub-acute stroke . J Rehabil Med . 2011 ; 43 : 709 - 13 .
6. Franceschini M , Carda S , Agosti M , Antenucci r , Malgrati D , Cisari C. Walking after stroke: What does treadmill training with body weight support add to overground gait training in patients early after stroke?: A single-blind , randomized, controlled trial. Stroke . 2009 ; 40 : 3079 - 085 .
7. Moseley AM , Stark A , Cameron ID , Pollock A. Treadmill training and body weight support for walking after stroke . Cochrane Database Syst Rev . 2005 . 4.
8. Díaz I , Gil JJ , Sánchez E. Lower-limb robotic rehabilitation: literature review and challenges . J Robot . 2011 . 2001 : Article ID 759764.
9. Dollar AM , Herr H. Lower extremity exoskeletons and active orthoses: Challenges and state-of-the-art . IEEE Trans Robot . 2008 ; 24 : 144 - 58 .
10. Herr H. Exoskeletons and orthoses: classification, design, challenges and future directions . J NeuroEngineering Rehabil . 2009 ; 6 : 21 .
11. Colombo G , Joerg M , Schreier R , Dietz V. Treadmill training of paraplegic patients using a robotic orthosis . J Rehabil Res Dev . 2000 ; 37 ( 6 ): 693 - 700 .
12. Banala SK , Agrawal SK , Kim SH , Scholz JP . Novel gait adaptation and neuromotor training results using an active leg exoskeleton . IEEE/ASME Trans Mechatron . 2010 ; 15 ( 2 ): 216 - 25 .
13. Veneman JF , Kruidhof R , Hekman EEG , Ekkelenkamp R , Van Asseldonk EHF , van der Kooij H. Design and evaluation of the lopes exoskeleton robot for interactive gait rehabilitation . IEEE Trans Neural Syst Rehabil Eng . 2007 ; 15 : 379 - 86 .
14. Kaelin-Lang A , Sawaki L , Cohen LG . Role of voluntary drive in encoding an elementary motor memory . J Neurophysiol . 2004 ; 93 ( 2 ): 1099 - 1103 .
15. Hesse S , Uhlenbrock D. A mechanized gait trainer for restoration of gait . J Rehabil Res Develop . 2000 ; 37 : 701 - 8 .
16. Schmidt H , Werner C , Bernhardt R , Hesse S , Kruger J. Gait rehabilitation machines based on programable footplates . J NeuroEngineering Rehabil . 2007 ; 4 : 2 .
17. Belda-Lois JM , del Horno SM , Bermejo-Bosch I , Moreno JC , Pons JL , Farina D , et al. Rehabilitation of gait after stroke: a review towards a top-down approach . J NeuroEngineering Rehabil . 2011 ; 8 ( 66 ):.
18. Kawamoto H , Sankai Y. Power assist method based on phase sequence and muscle force condition for hal . Adv Robot . 2005 ; 19 ( 7 ): 717 - 34 .
19. Tsukahara A , Kawanishi R , Hasegawa Y , Y S. Sit-to-stand and stand-to-sit transfer support for complete paraplegic patients with robot suit hal . Adv Robot . 2010 ; 24 ( 11 ): 1615 - 1638 .
20. Kawamoto H , Hayashi T , Sakurai T , Eguchi K , Y S. Development of single leg version of hal for hemiplegia . In: Proceedings of the 31st Annual International Conference of the IEEE Engineering in Medicine and Biology Society: 3-6 September 2009 . Minneapolis; 2009 . p. 5038 - 043 .
21. Maeshima S , Osawa A , Nishio D , Hirano Y , Takeda K , Kigawa H , et al. Efficacy of a hybrid assistive limb in post-stroke hemiplegic patients: a preliminary report . BMC Neurology . 2011 ; 11 : 116 .
22. Nilsson A , Vreede KS , Haglund V , Kawamoto H , Sankai Y , Borg J. Gait training early after stroke with a new exoskeleton - the hybrid assistive limb: a study of safety and feasibility . J NeuroEngineering Rehabil . 2014 ; 11 : 92 .
23. Jimenez-Fabian R , Verlinden O. Review of control algorithms for robotic ankle systems in lower-limb orthoses, prostheses, and exoskeletons . Med Eng Phys . 2012 ; 34 ( 4 ): 397 - 408 .
24. Guizzo E , Goldstein H. The rise of the body bots . IEEE Spectrum . 2005 ; 42 ( 10 ): 50 - 6 .
25. Esquenazi A , Talaty M , Packel A , Saulino M. The rewalk powered exoskeleton to restore ambulatory function to individuals with thoracic-level motor-complete spinal cord injury . Am J Phys Med Rehabil . 2012 ; 91 ( 11 ): 911 - 21 .
26. Farris RJ , Quintero HA , Goldfarb M. Preliminary evaluation of a powered lower limb orthosis to aid walking in paraplegic individuals . IEEE Trans Neural Syst Rehabil Eng . 2011 ; 19 ( 6 ): 652 - 9 .
27. Strausser KA , Swift TA , Zoss AB , Kazerooni H. Prototype medical exoskeleton for paraplegic mobility: first experimental results . In: In Proceedings of ASME 2010 Dynamic Systems and Control Conference: 12-15 September 2010 . Massachusetts; 2010 . p. 453 - 8 .
28. del-Ama AJ , Gil-Agudo A , Pons JL , Moreno JC . Hybrid fes-robot cooperative control of ambulatory gait rehabilitation exoskeleton . J NeuroEngineering Rehabil . 2014 ; 11 ( 27 ):.
29. Mobility Training Using Exoskeletons for Functional Recovery After Stroke. https://clinicaltrials.gov/ct2/show/NCT02128152?term=ekso&rank=3.
30. Development of Walk Assist Device to Improve Community Ambulation . https://clinicaltrials.gov/ct2/show/NCT01994395?term= honda+assist& rank=1.
31. Rea R , Beck C , Rovekamp RN , Diftler MA , Neuhaus P. X1: a Robotic Exoskeleton for In-space Countermeasures and Dynamometry . In: AIAA Space Conference ; 2013 .
32. He Y , Nathan K , Venkatakrishnan A , Rovekamp R , Beck C , Ozdemir R , et al. An integrated neuro-robotic interface for stroke rehabilitation using the nasa x1 powered lower limb exoskeleton . In: Proceedings of 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society : 26 -30 August 2014 . Chicago; 2014 .
33. Bortole M , del Ama A , Rocon E , Moreno JC , Brunetti F , Pons JL . A robotic exoskeleton for overground gait rehabilitation . In: In Proceedings of 2013 IEEE International Conference on Robotics and Automation : 6 - 10 May 2013 . Karlsruhe; 2013 . p. 3356 - 361 .
34. Banala SK , Kim SH , Agrawal SK , Scholz JP . Robot assisted gait training with active leg exoskeleton (alex) . In: In Proceedings of 2nd IEEE RAS & EMBS International Conference on Biomedical Robotics and Biomechatronics : 19 -22 October 2008 . Scottsdale; 2008 . p. 653 - 8 .
35. Lotze M , Braun C , Birbaumer N , Anders S , Cohen LG. Motor learning elicited by voluntary drive . Brain . 2003 ; 126 ( 4 ): 866 - 72 .
36. Venkatakrishnan A , Francisco GE , Contreras-Vidal JL . Applications of brain-machine interface systems in stroke recovery and rehabilitation . Curr Phys Med Rehabil Rep . 2014 ; 2 : 93 - 105 .
37. Onen U , Botsali FM , Kalyoncu M , Tinkir M , Yilmaz N , Sahin Y. Design and actuator selection of a lower extremity exoskeleton . IEEE/ASME Trans Mechatron . 2013 ; 99 : 1 - 10 .
38. Winter DA . Biomechanics and Motor Control of Human Movement , 4th edn.: John Wiley & Sons; 2009 .
39. Hong YW , King Y , Yeo W , Ting C , Chuah Y , Lee J , et al. Lower extremity exoskeleton: review and challenges surrounding the technology and its role in rehabilitation of lower limbs . Australian J Basic Appl Sci . 2013 ; 7 ( 7 ): 520 - 4 .
40. Zoss AB , Kazerooni H , Chu A. Biomechanical design of the berkeley lower extremity mechatronics . IEEE/ASME Trans Mechatron . 2006 ; 11 ( 2 ): 128 - 38 .
41. Vallery H , Veneman J , van Asseldonk E , Ekkelenkamp R , Buss M , van Der Kooij H. Compliant actuation of rehabilitation robots . IEEE Robot Automation Mag . 2008 ; 15 : 60 - 9 .
42. Fan Y , Yin Y. Active and progressive exoskeleton rehabilitation using multisource information fusion from emg and force-position epp . IEEE Trans Biomed Eng . 2013 ; 60 ( 12 ): 3314 - 321 .
43. Zoss AB . Design of an electrically actuated lower extremity exoskeleton . Adv Robot . 2006 ; 20 : 967 - 88 .
44. Gams A , Petric T , Debevec T , Babic J. Effects of robotic knee exoskeleton on human energy expenditure . IEEE Trans Biomed Eng . 2013 ; 60 ( 6 ): 1636 - 44 .
45. Hussain S , Xie SQ , Liu G . Robot assisted treadmill training: mechanisms and training strategies . Med Eng Phys . 2011 ; 33 ( 5 ): 527 - 33 .
46. Aoyagi D , Ichinose WE , Harkema SJ , Reinkensmeyer DJ , Bobrow JE. A robot and control algorithm that can synchronously assist in naturalistic motion during body-weight-supported gait training following neurologic injury . IEEE Trans Neural Syst Rehabil Eng . 2007 ; 15 ( 3 ): 387 - 400 .
47. Wisneski KJ , Johnson MJ . Quantifying kinematics of purposeful movements to real, imagined, or absent functional objects: implications for modelling trajectories for robot-assisted adl tasks . J NeuroEngineering Rehabil . 2007 ; 4 : 7 .
48. Montagner A , Frisoli A , Borelli L , Procopio C , Bergamasco M , Carboncini MC , et al. A pilot clinical study on robotic assisted rehabilitation in vr with an arm exoskeleton device . In: Proceedings of Virtual Rehabilitation : 27 -29 September 2007 . Venice; 2007 . p. 57 - 64 .
49. Pietrusinski M , Cajigas I , Severini G , Bonato P , Mavroidis C. Robotic gait rehabilitation trainer . IEEE/ASME Trans Mechatron . 2014 ; 19 ( 2 ): 490 - 9 .