Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
Exoskeletons for Gait Assistance and Training of the MotorImpaired
Sunil K. Agrawal, Sai K. Banala, Kalyan Mankala, Vivek Sangwan, John P. Scholz, Vijaya
Krishnamoorthy, Wei-Li Hsu
Abstract—Robotics is emerging as a promising tool for
training of human functional movement. The current research
in this area is focused primarily on upper extremity
movements. This paper describes novel designs of three lower
extremity exoskeletons, intended for gait assistance and
training of motor-impaired patients. The design of each of
these exoskeletons is novel and different. Force and position
sensors on the exoskeleton provide feedback to the user during
training. The exoskeletons have undergone limited tests on
healthy and stroke survivors to assess their potential for
treadmill walking. GBO is a Gravity Balancing un-motorized
Orthosis which can alter the gravity acting at the hip and knee
joints during swing. ALEX is an Actively driven Leg
Exoskeleton which can modulate the foot trajectory using
motors at the joints. SUE is a bilateral Swing-assist Unmotorized Exoskeleton to propel the leg during gait.
I
I. INTRODUCTION
n the past decade, robotics has been used to evaluate and
treat upper extremity functions in individuals with severe
motor impairments. The MANUS, MIME, and ARM [1,2,3]
represent early advances in upper extremity devices which
demonstrated the usefulness of robotics in functional
training by exploiting the built-in neural plasticity in
humans. In recent years, a variety of powered arms [4,5,6]
and un-powered arms have been developed and tested
[7,8,9].
Machines for functional training of the lower extremity
are far more challenging to design compared to those for the
upper extremity due to issues such as posture and balance
during walking. From an engineering perspective, the
designs must be flexible to allow both upper and lower body
motions, once a subject is in the exoskeleton, since walking
involves synergy between upper and lower body motions.
From a neuro-motor perspective, an exoskeleton must be
adjustable to anatomical parameters of a subject and should
Sunil K. Agrawal* a Professor, Kalyan K. Mankala a post-doctoral
researcher, Sai K. Banala and Vivek Sangwan Ph.D. students - all with the
Department of Mechanical Engineering,.University of Delaware, Newark,
DE 19716, USA Website: http://mechsys4.me.udel.edu E-mail of the
corresponding author: [email protected],
John P. Scholz a Professor, Vijaya Krishnamoorthy a post-doctoral
researcher, Wei-Li Hsu a PhD student, all with the Department of Physical
Therapy, University of Delaware, Newark, DE 19716, USA.
The work was supported by NIH R01 HD 38582 and NIDRR
subcontract through Rehabilitation Institute of Chicago.
Videos of training with the exoskeletons are available on our website:
http://mechsys4.me.udel.edu/research/medical_robotics/
1-4244-1320-6/07/$25.00 (c)2007 IEEE
1108
facilitate human learning under subject’s control. Lokomat
is a motorized lower extremity exoskeleton, now
commercially available, for spinal cord injury patients on a
treadmill [10]. Mechanized Gait Trainer (MGT) is a one
degree-of-freedom powered machine that drives the foot
[11]. However, due to limited performance and excessive
costs, these machines are not common in rehabilitation
clinics. These machines drive patients on a specified
trajectory but do not provide flexibility to subjects to move
under their own neuro-motor control. Training with fixed
trajectories has been known to result in “learned
helplessness”, i.e., habituation to given sensory inputs and
poor motor response to variations in these inputs [12]. Other
current trends in motorized exoskeletons include human
strength amplification [13, 14] or targeting assistance at
specific joints [15, 16, 17, 18].
In the last five years, our group at the University of
Delaware has fabricated and tested three lower extremity unmotorized and motorized exoskeletons for human motor
training [19, 20, 21, 22]. Each of these exoskeletons has a
unique design principle and addresses the issue of providing
a flexible motor-learning environment. The un-motorized
exoskeletons can be fabricated at a very reasonable cost,
thereby, making these accessible to users from different
socio-economic backgrounds. We believe that these
exoskeletons have the potential to make an impact on the
emerging field of lower extremity gait rehabilitation of
motor-impaired patients.
The rest of this paper is organized as follows: Section II
summarizes the design principles and features of these
exoskeletons. Section III presents a brief summary of
preliminary results obtained with these exoskeletons on
healthy persons and a person with stroke. These are followed
by concluding remarks.
II. EXOSKELETON DESIGNS FOR THE LOWER EXTREMITY
A. Gravity Balancing Orthosis (GBO)
Gravity balancing Orthosis (GBO) is an un-motorized
exoskeleton for right hemiparetic patients that can alter the
level of gravity acting at the hip and knee joints during
motion. Gravity plays an important role in human movement.
It assists (or resists) the motion of a joint, such as the hip and
knee, over different parts of the swing. With an orthosis that
alters the gravity at the joints, i.e., provides partial or full
gravity assistance during motion, the hip and knee joints can
swing through a larger range of motion for the same
nominally applied human joint torques (See Fig. 1). As a
1
Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
result, a person with a weak musculature or poor motorcontrol, when wearing such an exoskeleton, may be able to
attain a larger swing and adapt to it as the gravity is altered
during training (see data in Fig.??). This provides the
guiding principle behind the design of the gravity balancing
orhosis.
B. Active Leg Exoskelton (ALEX)
ALEX is a motorized exoskeleton designed to assist right
hemiparetic patients. The significant aspect of ALEX is the
nature of assistance it provides to the leg during gait,
designed to maximize human-motor learning. The control
architecture can prescribe a tunnel around the desired foot
trajectory, see Fig.3 [20]. Depending on the current position
of the foot with respect to the tunnel, the controller provides
normal and a small tangential force on the foot to help
achieve this desired foot trajectory. If the foot is within the
tunnel, the normal force is close to zero. As the foot goes
outside the tunnel, the normal force acts as a restoring force
to bring the foot closer to the center line of the tunnel. The
controller allows the user to select the tunnel width and the
nature of the force field outside the tunnel. Unlike Lokomat,
it does not move the leg on a fixed trajectory but allows the
user to modulate it according to their current capability.
The exoskeleton design of ALEX is structurally similar to
Figure 1. Effect of reduced gravity on motion of a
swing leg, modeled as a two degree-of-freedom
pendulum. For nominal joint torques corresponding to a
human swing and anthropomorphic mass and geometry
data , range of motion of the joints increases as the
gravity assistance is increased from 0% to 100% [22].
Figure 2. (i) A schematic showing the conceptual
design of the GBO, (ii) close-up view of the mechanical
design, (iii) a stroke patient wearing it on a treadmill.
During training, the GBO is attached to a walker and is
strapped on to the subject at the trunk and segments of the
leg. The trunk has four degrees-of-freedom with respect to
the walker to allow natural motion of the upper body during
walking. The hip segment can abduct with respect to the
1109
Vertical axis (cm)
-65
The GBO is a simple device composed of rigid links,
joints and springs, which are adjustable to the geometry and
inertia of the leg of a human wearing it. The GBO does not
use any motors or controllers, yet can still unload the joints
of the leg joints from gravity over the full range of motion of
the leg [19]. The mechanical design aims to make the
potential energy of the combined system invariant with
configuration of the leg. Additionally, parameters of the
GBO can be changed to achieve a prescribed level of partial
balancing, between 0-gravity and 1-gravity (See Fig. 2).
1-4244-1320-6/07/$25.00 (c)2007 IEEE
trunk. The thigh and shank segments of the machine are
telescopic to accommodate a range of subjects. The orthosis
is fitted with encoders that record the movement of the trunk
and the leg. In addition, the GBO has two 6-axis forcetorque sensors that record the interaction forces between the
GBO and corresponding segments on the human. The
trajectory of the foot and joint angles, recorded by the joint
encoders, are provided to the subject as visual feedback.
Aspects of gait rehabilitation of a stroke subject are
presented in Section IIIA.
-70
Width of the tunnel
-75
-80
-85
-90
-50
0
Horizontal axis (cm)
50
Figure 3. (i) A healthy subject wearing ALEX on a treadmill,
(ii) A template of foot motion with a bounding tunnel motors help to keep the foot within the tunnel.
the GBO (Fig. 3). The hip and the knee joints are actuated
by servo motors. The human-machine interface of ALEX is
similar to GBO. It is attached to a walker and is adjustable to
a range of subjects. ALEX has sensors to record kinematics
and kinetics of the gait similar to the GBO. A major design
and control challenge in the development of ALEX was to
achieve back drivability of motors while keeping the
required torques small. When we amplified the peak torque
by a large gear reduction, this introduced significant friction
in the gear train that prevented back drivability. This was
resolved by characterizing the friction empirically and
compensating for it using a load cell with a PI control law.
Details of gait training of healthy subjects using this
motorized exoskeleton are presented in Section IIIB.
2
Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
C. Swing-Assist Un-motorized Exoskelton (SUE)
SUE is a bilateral un-motorized exoskeleton (SUE)
optimized for propulsion of the leg, during treadmill walking
[21]. The design is based on a mathematical model of a
swinging leg strapped to SUE. The dynamic model provides
a framework for optimization of the torsion springs at the hip
and knee joints. The optimixed design was aimed at
achieving a given final configuration with a positive ground
clearance during motion. Figure 4(i) shows a subject wearing
force due to springs in SUE. The training exploits the ability
of a human, under assistance, to extend the range of motion.
GBO and ALEX are well suited to improve the motion of
the hemiparetic leg following a stroke, while patients with
spinal cord injury would typically benefit from SUE or
future bilateral designs of ALEX. A given gait template, in
general, can be achieved by multiple combinations of torque
inputs applied by the human and is selected by the human
motor system.
A. Gravity Balancing Orthosis (GBO)
The first data, presented here, was collected from four
healthy young adults and three subjects with right
hemiparesis, following a stroke. Stroke patients walked at
their preferred speeds, while the healthy subjects walked at
30% and 60% of their preferred speeds to make the speeds
comparable to those of the stroke subjects. Five trials of
walking were collected and the time duration of each trial
was about 30 seconds. Walking task was conducted within
the device with the following settings: (i) leg and device
fully gravity balanced, referred to as `full-balanced
condition, (ii) device only is gravity balanced, referred to as
`device balanced' condition.
Device-balanced only
Flex 100
80
Knee Angle (deg)
60
100% Fully-balanced
Healthy subject
Healthy subject
Toe off
40
Heel strike
20
0
-20
100
80
Person with stroke
Person with stroke
60
40
20
0
Ext -20
-60
60 -40
Ext
Figure 4. (i) A healthy subject on a treadmill wearing
SUE, (ii) Motor training paradigm with GBO, ALEX,
SUE, where the assistance is respectively gravity,
motor torque, propulsive spring force.
SUE.
In contrast to GBO and ALEX, it is not attached to a
walker. It is un-motorized, light-weight, and portable.
Kinematic and kinetic data are collected using joint encoders
and interface force-torque sensors similar to the GBO and
ALEX. Details of subject data collected with SUE are
described in Section IIIC.
III. RESULTS AND PERFORMANCE EVALUATION
A schematic of the training paradigm, used with GBO
and ALEX, is shown in Figure 4 (ii), where a subject is
presented with a gait template of the foot, which is made
progressively closer to the normal trajectory over sessions.
The assistance is in the form of gravity balancing in the
GBO, hip and knee joint torque in ALEX, or propulsive
1-4244-1320-6/07/$25.00 (c)2007 IEEE
1110
-20
-
0
20
40
60
-60
-40 -20
Hip Angle (deg)
0
20
40
60
Flex
Figure 5. Plot of knee versus hip joint angles during
treadmill walking with the GBO during “device-balanced”
(left) and “full-balanced” (right) conditions for a
representative patient (bottom) and subject without
neurological involvement (top).
Fig. 5 shows the plots of the hip joint angle versus the
knee joint angle during an entire gait cycle for a
representative healthy subject and a patient performing
walking [19]. It is clear from these plots from healthy and
stroke subjects that for the `full balanced' condition, the
range of motion at both hip and knee joints is larger than
with `device balanced' condition. For the stroke patient, this
increase in range of motion is nearly 50%. This increased
range of motion was expected according to our simulation
results in Fig. 1.
A chronic stroke survivor, 56 year old male, with right
hemiparesis volunteered for training with the GBO to
determine long-term training effects. Formal evaluations of
kinematics and kinetics were performed during both over-
3
Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
ground and treadmill walking, mid-way through the 15
training sessions, and after the final training session. The
over-ground walking evaluation was also repeated four
weeks following the last training session.
Gravity assistance began at 100% and was gradually
reduced over sessions to 0%. Each training session consisted
of four blocks, 10 minutes each, of treadmill walking, each
followed by five minutes of rest, or longer if requested. The
subject began training at his preferred treadmill walking
speed of 1.5 mph, increased to 1.7 mph by mid-training and
1.9 mph by the final evaluation.
B. Active Leg Exoskeleton (ALEX)
Tests with the active leg exoskeleton were performed
with six healthy adult subjects. This was done to evaluate
whether the device could be used to induce short-term
adaptations of the walking pattern, even of healthy subjects.
Device-balanced only
Heel Strike
Toe off
60
40
40
20
20
0
0
A. Pre-test
C. Post-test
0.12
-20
-40
-20
0
20
40
-20
-40
-20
0
20
40
100% Fully-balanced
60
60
40
40
20
Ext
0.14
Vertical Foot Position (m)
Knee Angle (deg)
Flex 60
decrease of applied hip torque in the initial period of the
swing. Following the 15 training sessions with the GBO,
improvements measured during over ground gait were as
follows: Knee flexion excursion during swing improved
from the initial evaluation significantly as did the hip
flexion. In addition, the hip was slightly more extended at
terminal stance during over ground walking compared to the
first evaluation. The ankle was less plantar flexed at heel
strike, indicating improvement in ankle dorsiflexion. These
were encouraging findings that indicated continued
improvement after training.
20
0
0
-20
-40
-20
-40
-20
0
20
40
-20
0
20
Ext
40
0.08
0.06
0.04
0.02
0.00
-0.4 -0.3 -0.2 -0.1
0.14
0.12
0.10
0
0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1
B. Mid-test
0 0.1 0.2 0.3 0.4
D. Retention
Prescribed
Actual
0.08
0.06
0.04
Flex
0.02
Hip Angle (deg)
0.00
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -0.4 -0.3 -0.2 -0.1 0
Figure 6. Hip-Knee angle plots in GBO without
gravity balancing (top) for sessions 1 (left) and 15
(right), respectively, and 100% gravity balancing
(Bottom) for sessions 1 (left) and 2 (right).
Posterior
The top panel in Figure 6 provides an example of the
patient’s knee-hip trajectories obtained from the initial
walking evaluation (left) and from the last training session
(session 15; right) using the GBO with device-only
balancing. The patient’s knee excursion during the swing
phase showed impressive increases over the training
sessions although he still maintained the double-loop
pattern, indicating an early knee extension followed by
additional knee flexion prior to heel strike. The bottom panel
in Fig. 6 illustrates changes of hip-knee coordination in the
fully-balanced condition (100% assistance) between
sessions 1 and 3. Note that, in contrast to device-only
balancing (top panel), the patient was able to perform a
single loop pattern of hip-knee coordination when the leg
was fully gravity balanced that was more like the normal
pattern. Moreover, within three sessions of practice, the
pattern became more normal looking during the first half of
the cycle following toe-off.
The swing kinematics data over the 15 days showed that
the range of motion gradually increased, accompanied by a
1-4244-1320-6/07/$25.00 (c)2007 IEEE
0.10
1111
0.1 0.2 0.3 0.4
Horizontal Foot Position (m)
Anterior
Figure 7. Plots of sagittal plane foot path of a healthy
subject from four evaluation points after training with the
ALEX. Although it is shown in the plot for reference, the
prescribed trajectory was not shown to the subject during
the retention test (bottom, right panel).
Three subjects were assigned to an experimental group
who received active assistance from ALEX to produce the
sagittal plane foot trajectory during training by application
of force fields along a desired foot path. In addition, virtual
tunnels of variable widths and compliance were defined,
prescribing allowable limits of sagittal plane foot movement,
as described in Section IIB. The tunnel width could be set to
allow more or less deviation of the current foot path from
the prescribed foot path. Three subjects in the control group
tried to match a prescribed template without receiving
assistance from ALEX. A prescribed foot trajectory was
devised by scaling down each subjects’ typical foot path,
particularly in the vertical dimension, enough to make
replication of the prescribed path challenging.
Subjects in both groups received four 15 minute blocks
of training during which the prescribed template and the
subject’s current foot path were displayed to the subject. All
four training blocks were identical for the control subjects.
4
Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
For the experimental subjects, both the compliance of the
virtual walls and assisting force were kept constant. What
changed was the width of the virtual walls, i.e. the degree of
constraint on the foot path. The constraint was maximal for
the first block of training trials. The foot path was less
constrained for blocks 2 and 3. The constraint was further
relaxed for training block 4. Evaluation of the subjects’
walking performance on the treadmill was obtained with the
motors of ALEX turned off at three time points: (1) prior to
the initial block of training (pre-test), (2) prior to block 3 of
training (mid-test), and (3) after block 4 of training (posttest). All of these data collections were performed with the
template and current foot trajectory displayed to the subject.
In addition, following the post-test, video feedback was
turned off and the subject’s performance was recorded to
test for evidence of foot path adaptation.
Results of one experimental subject’s performance during
the four evaluation sessions are illustrated in Figure 7, where
the thin solid line is the subject’s average foot path and the
heavy dashed line is the prescribed foot path, The actual foot
path deviated substantially from the required trajectory at
the initial evaluation but was shaped to approximate the
required trajectory even by the mid-training evaluation. That
the trained pattern was maintained reasonably well in this
subject following training, after the feedback was turned off
completely, is remarkable.
Figure 8. Torque at the hip and knee joints for a swing
motion at treadmill speed of 2 mph: mathematical model
(blue line), force-torque sensors (red line)
The control subjects’ performance remained relatively
constant throughout, while the experimental group showed a
consistent reduction in the deviation from the required
trajectory during training that was partially maintained even
after removal of visual feedback, similar to the subject in
Fig. 7. This result is striking given that walking patterns are
highly ingrained through many years of practice. Of course,
interpretation of this result requires caution given that
treadmill walking differs significantly from over ground
locomotion and is not well-practiced.
1-4244-1320-6/07/$25.00 (c)2007 IEEE
1112
C. Swing-Assist Un-motorized Exoskeleton (SUE)
SUE was used to collect data on a single healthy subject,
who walked on a treadmill at different speeds. In order to
evaluate the effectiveness of the device, we computed the
human applied joint torque from the swing data in two ways:
(i) via a mathematical model, (ii) via force-torque sensors
on the exoskeleton. If the device was working as intended,
one would expect to see that the torques required in case (ii)
are less than that those required in case (i) during swing.
Representative torque plots at treadmill speeds of 2 mph, for
which the exoskeleton was designed, are shown in Fig.8.
From these plots, we observe that a smaller hip joint
torque was used by the subject with SUE, while the knee
joint torque was smaller in magnitude during the first half of
the cycle. These results are remarkable considering that only
four design parameters were optimized in the design with
simplistic model of walking on a treadmill [21].
IV. CONCLUSIONS
This paper described the design principles, training
paradigms, and preliminary human subject results for three
novel lower extremity exoskeletons intended for gait
assistance and training of motor-impaired patients, such as
after stroke or spinal cord injury. Gravity Balancing
Orthosis (GBO) is an un-motorized exoskeleton that can
alter the gravity at the joints of a swinging leg during
walking. Experiments on healthy and stroke subjects
confirmed that GBO extends the range of motion of a
swinging leg, thereby, proposing reduced gravity to be an
important learning paradigm for gait training of stroke
patients. A 6-week gait training of a chronic stroke patient
with GBO, while the gravity assistance was progressively
reduced from 100% to 0%, showed
significant
improvements in gait and walking speed of the patient.
ALEX is a motorized exoskeleton that provides a flexible
environment to maximize human-motor learning. Its control
architecture prescribes a tunnel around a desired foot
trajectory and laws that bring the foot close to the template
trajectory. Tests on healthy subjects showed that humans
could be trained to walk in a significantly altered gait on a
treadmill, within 45 minutes. We believe that this result has
deep implications in motor training of impaired patients.
SUE is an exoskeleton that provides propulsive forces to a
user during swing through clever designs of the exoskeleton
spring parameters.
We believe that devices like GBO and ALEX can make a
substantial difference in future gait training of patients
suffering from stroke, while exoskeletons such as SUE and
ALEX can impact training of spinal cord injury patients.
REFERENCES
[1]
[2]
H. I. Krebs, N. Hogan N., et al. ``Robot-aided Neuro-rehabilitation."
IEEE Trans Rehabil Eng}, 6(1), p. 75-87, 1998.
P.S. Lum, C. G., Burgar, C.G., P. C. Shor, M. Majmundar, and M. Van
Der Loos, M. ``Robot-assisted movement training compared with
conventional therapy techniques for the rehabilitation of upper limb
motor function following stroke'', Arch. Phys. Med. Rehabil., Vol. 83,
5
Proceedings of the 2007 IEEE 10th International Conference on Rehabilitation Robotics, June 12-15, Noordwijk, The Netherlands
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
D. J.Reinkensmeyer, L. E. Kahn, M. Averbuch, A. N. McKenna-Cole,
B. D. Schmit, and W. Z. and Rymer, ``Understanding and treating arm
movement impairment after chronic brain injury: Progress with the
arm guide'', J. Rehabil. Res. Develop., Vol. 37, p. 653, 2000.
K. Kiguchi, T. Tanaka, "Exoskeleton for the human upper-limb
motion support." Proc. of the IEEE International Conference on
Robotics and Automation, 2003. p. 2206-2211.
R.J. Sanchez,
J.Liu,
S.Rao, P. Shah, R.Smith, T.Rahman,
S.C.Cramer, J.E.Bobrow, D.J.Reinkensmeyer, ``Automating Arm
Movement Training Following Severe Stroke: Functional Exercises
With Quantitative Feedback in a Gravity-Reduced Environment,''
IEEE
Transactions on Neural systems and Rehabilitation
Engineering, vol. 14, no.3, pp.378-389, 2006.
R. Colombo, F. Pisano, S. Micera, A. Mazzone, C. Delconte, M. C.
Carrozza, P. Dario, G. Minuco, “Robotic techniques for upper limb
evaluation and rehabilitation of stroke patients.” IEEE Trans. on
Rehabilitation Engineering, 2005. 13(3), p. 311-324.
T.Rahman, R.Ramanathan, S.Stroud, W.Sample, R.Seliktar,
W.Harwin, M.Alexander, and M.Scavina, ``Towards the control of a
powered orthosis
for people with muscular dystrophy,''
\emph{Proceedings of the Institution of Mechanical Engineers, Part
H: Journal of Engineering in Medicine, vol. 215, no.3, pp. 267--274,
2001.
S. K. Agrawal, G. Gardener, S. Pledgie, ``Design and Fabrication of a
Gravity Balanced Planar Mechanism Using Auxiliary Parallelograms'',
Journal of Mechanical Design, Transactions of the ASME, Vol. 123,
No. 4, p. 525-528, 2001.
L. F. Cardoso, S. Tomazio S., et al. Conceptual design of a passive
arm orthosis." Proceedings, ASME Design Engineering Technical
Conferences, 2002.
G.Colombo, M.Joerg, R.Schreier, and V.Dietz, ``Treadmill training of
paraplegic patients using a robotic orthosis,'' Journal of Rehabilitation
Research and Development, vol.37, no.6, 2000.
S. Hesse, D. Uhlenbrock, C. Werner, A. Bardeleben, ``A mechanized
gait trainer for restoring gait in nonambulatory subjects'', Arch Phys
Med Rehabil., 81, p. 1158-1161, 2000.
V.R. Edgerton, N.J. Tilakaratne, A.J. Bigbee, R.D. deLeon, and R.R.
Roy, ``Plasticity of the spinal circuitry after injury,'' Ann. Rev.
Neurosci., vol.27, pp. 145--167, 2004.
A. B. Zoss, H. Kazerooni, and A. Chu, ``Biomechanical design of the
Berkeley Lower Extermity Exoskeleton (BLEEX)'', IEEE/ASME
Trans. Mechatronics, 11(2), p.128-138, 2006.
H. Kawamoto and Y. Sankai, ``Power assist system hal-3 for gait
disorder person'', International Conference on Computers for
Handicapped Persons, p. 196-203, 2002.
D. Aoyagi, W. E. Ichinose, S. J. Harkema, D. J. Reinkensmeyer, and J.
Bobrow, ``An assistive robotic device that can synchronize to the
pelvic motion during human gait training,'' IEEE, International
Conference on Rehabilitation Robotics, 2003, p. 565-568.
E. Rocon,, A. Ruiz, J. Pons, J.Belda-Lois, and J. S. Anchez-Lacuesta,
``Rehabilitation robotics: a wearable exo-skeleton for tremor
assessment and suppression.'' IEEE, International Conference on
Robotics and Automation, 2005. p. 2283-2288.
J. B. W. Nikitczuk and C. Mavroidis, ``Rehabilitative knee orthosis
driven by electro-rheological fluid based actuators,'' IEEE,
International Conference on Robotics and Automation, 2005. p. 22942300.
G. S. Sawicki, K. E. Gordon, and D P. Ferris., ``Powered lower limb
orthoses: Applications in motor adaptation and rehabilitation,'' IEEE,
International Conference on Rehabilitation Robotics, 2005., p. 206211.
S. Banala, S. K. Agrawal, A. Fattah, V. Krishnamoorthy, W. L. Hsu, J.
P. Scholz, K. Rudolph, ``Gravity-Balancing Leg Orthosis and Its
Performance Evaluation'', IEEE Transcation on Robotics, 22(6),
p.1228-1237, 2006.
S. Banala, A. Kulpe, and S. K. Agrawal, ``A Powered Leg Orthosis
for Gait Rehabilitation of Motor-Impaired Patients'', In Proceedings
IEEE International Conference on Robotics and Automation, Rome,
April 2007.
Mankala, K. and Agrawal, S. K., “Passive Swing Assistive
Exoskeletons for Motor-Incomplete Spinal Cord Injury Patients”, In
1-4244-1320-6/07/$25.00 (c)2007 IEEE
1113
Proceedings IEEE International Conference on Robotics and
Automation, Rome, April 2007.
[22] S. K. Agrawal, S. Banala, A. Fattah, V. Sangwan, V. Krishnamoorthy,
W. L. Hsu, J. P. Scholz,, ``Assessment of Motion of a Swing Leg and
Gait Rehabilitation with a Gravity Balancing Exoskeleton'', to appear
in IEEE Transactions on Neural systems and Rehabilitation
Engineering, 2007.
6