Вы находитесь на странице: 1из 32

1

AbstractThis paper presents a novel reconfigurable ankle rehabilitation robot to cover


various rehabilitation exercise modes. The designed robot can allow desired ankle and foot
motions including toe and heel raising as well as traditional ankle rotations since the
mechanism can generate relative rotation between the fore and rear platforms as well as pitch
and roll motions. In addition, the robotic device can be reconfigured from a range of motion
(ROM)/strengthening exercise device to a balance/proprioception exercise device by simply
incorporating additional plate. Further, the action of the device is two folded in the sense that
while a patients foot is fastened firmly to the ROM/strengthening device for task specific
training, s/he can also stand on the balance/proproception device. To perform each mode of
ROM, strengthening, and proproception exercises, a unified position-based impedance control
is systematically developed taking into account of desired position and velocity.

Index TermsAnkle rehabilitation, exercise modes, metatarsophalangeal joint, parallel
mechanism, reconfigurable robot, and impedance control.



A Novel Reconfigurable Ankle Rehabilitation
Robot for Various Exercises
Jungwon Yoon
1
, Jeha Ryu*
2
, and Kil-Byung Lim
3

1
School of Mechanical and Aerospace Engineering and ReCAPT
Gyeongsang National University, Jinju, Gyeongnam 660-701, Korea.
email:jwyoon@gsnu.ac.kr

2
Department of Mechatronics, Gwangju Institute of Science and Technology
Bukgu, Gwangju 500-712, Korea
e-mail: ryu@gist.ac.kr

3
Department of Rehabilitation Medicine, Ilsan Paik Hospital,
Inje University, Ilsangu, Goyang, Korea
2
1. INTRODUCTION
Rehabilitation aims to bring back the patient's physical, sensory, and mental capabilities that
were lost due to injury, illness, and disease and to support the patient to compensate for deficits
that cannot be treated medically.
1
Primary types of rehabilitations are physical, occupational,
and speech therapy. Physical therapy aids the patient to restore the muscles, bones, and nervous
system through the use of thermal treatments and exercises. Among physical therapy, exercises
are the most widely used method.
2
Exercises are used to restore the original range of motion
(ROM) and mobility and to strengthen the affected limbs or parts.
Among lower limb injuries, the ankle sprain that occurs with ligaments stretched or torn is
the most common in sports and daily life in general. Full functional activities such as ROM,
progressive muscle strengthening, and proprioceptive exercise after acute cure plays a pivotal
role in speeding up return to preinjury level.
3
Achilles tendon stretching should be performed
within 48 to 72 hours of injury, regardless of weight-bearing capacity and ROM must be
regained. Once ROM is achieved, strengthening of weakened muscles is essential to rapid
recovery and is a preventive measure against reinjury. Manual therapy such as toe raises, heel
walks, and toe walks, toe curls, and marble pick-ups as well as resistance exercises to ankle
joints are attempted to regain strength and coordination.

As the patient achieves full weight
bearing capability without pain, proprioceptive exercise is initiated for the recovery of balance
and postural control. Finally, advanced exercises should be performed to regain functions
specific to normal activities.
3
Modal exercise procedures are summarized in Fig. 1.
In order to cover some of the exercises satisfying the above procedures, several devices for
ankle rehabilitation have been developed.

Traditionally primitive passive devices
4
such as
elastic bands and the active reflex treatment unit for strength exercises, wobbles board and
foam rollers for balance exercises have been in use. These devices, however, can allow very
simple rehabilitation exercise by only patients efforts and cannot store the past information of
the exercises. Therefore, task specific training by programmable robotic systems will be very
helpful not only to provide patient with more diverse and useful exercises in a suitable manner
3
but also to reduce the physical efforts due to repetitive nature of exercise and furthermore to
store the patients exercise histories. To meet the need, the design and development of
advanced exercise tools are in progress. Representatives of some commercial developments
are biodexs multi joint system3 and balance system which can facilitate to achieve
ROM/strengthening, and balance exercises, respectively. Girone et. al suggested the Rutgers
Ankle
5
that can provide active 6-dof motions at ankle with a Stewart platform structure to
cover various exercise modes. Using this, the patients foot is fixed firmly to the platform of the
Rutgers Ankle and is synchronized to virtual reality scenarios.
6,7
Recently, the Rutgers
Ankle is extended to Dual Stewart platform
8
with 12-dof motions for gait rehabilitation.
Even though presently most of the exercise modes are possible by the use of two Stewart
platforms
8
, balance/proprioception exercises can also be achieved with a single device. Since
ankle movement in most exercises needs less than 4-dof motions, Dai and Zhao proposed a
sprained ankle device
9
using a 3-or 4- dof parallel mechanisms with a central strut and
analyzed the orientation and stiffness of the mechanism with considerations of the central strut.
Commercial developments and the existing robotic devices
5-9
as mentioned above with a
single rigid upper platform, however, cannot generate the relative rotation i.e.,
metatarsophalangeal (MTP) joint motion (shown in Fig. 2) between fore and rear foot for
natural foot motions. The relative rotation around MTP joint entails more natural exercises
such as toe raise, heel raise, etc. Therefore, more diverse exercise modes should be available
with the robotic devices in order to be able to facilitate more natural MTP joint motions.
Further, in order to accommodate the desired exercise modes, a reconfigurable platform is
inevitable. Eventually a versatile single platform-based reconfigurable robot mechanism with
satisfactory MTP joint motions and less than 6-dof motions should be developed for allowing
more natural foot and ankle motions to cover various exercise modes as shown in Fig. 1.
Girone et. al
6
commented that adequate programs should be developed to perform desired
rehabilitative motions satisfying exercise parameters and constraints, and suitable control
methods should be implemented for certain exercise modes. As regards to transduction
4
scheme and methods, impedance-based combined position and force control is preferred for
such interactive type applications. A short review on impedance control especially applied to
rehabilitation devices follows. For upper limbs, Noritsugu and Tanake
10
and Richardson et. al
11
applied a position-based impedance control using pneumatic actuators. Krebs et. al
12

implemented a force-based impedance control on a physiotherapy robot using electric motors.
For lower limbs, Yoon et. al
13
presented position-based impedance control for the Rutgers
Ankle rehabilitation device that is driven by six double-acting pneumatic actuators. Jezernik et.
al
14
suggested a novel gait-pattern adaptation algorithm based on the impedance control for
adapting the reference walking trajectories according to the physical interaction of the patient
with the robotic orthosis. Even though these controllers satisfied their specific purposes, design
and development of unified stable controller that can realize various exercise modes including
flexibility, strength, and proprioception, however, draws attention for effective control of
versatile reconfigurable rehabilitation robots.
In this paper, we present design and control of a novel ankle rehabilitation robot to cover
various exercises modes with natural foot and ankle motions using a parallel mechanism that
can generate pitch and roll motions as well as relative rotation between fore and rear platforms.
In addition, we propose a simple reconfiguration in the sense that it can be used as a
ROM/strengthing exercise device as well as a balance/proprioception device simply by adding
an extra large plate. For systematic implementation of the exercise, a unified position-based
impedance controller is suggested that can cover desired exercise modes. The following
section describes requirements specification based on ankle anatomy. It follows with the
mechanism design. Section 3 presents a position-based impedance control. Section 4 presents
impedance parameter design for each exercise mode. Section 5 presents experimental results
for each exercise mode. Finally, conclusions and future research scenarios are summarized in
Section 6.
5
2. MECHANISM DESIGN FOR ANKLE REHABILITATION EXERCISE

2.1. Ankle Anatomy and Design Requirements

Based on ankle anatomy, this section identifies the design requirements for ankle
rehabilitation device. Normally the ROM (flexibility) exercise of ankle joint is performed
before other functional rehabilitation exercises are initiated. ROM exercise is followed by
strengthening exercise to increase muscle strength and endurance. The device should generate
pitch motions: dorsiflexion (pull foot upward toward the head) and plantar flexion (push foot
downward away from the head), as well as roll motions: inversion (push foot inward toward the
midline of the body) and eversion (push foot outward away from the midline of the body) about
the ankle joint. In addition, since a patients foot will be fastened to the device platform during
flexibility and strength exercises, the platform should generate MTP joint motion between fore
and rear foot to facilitate natural foot motions (see Fig. 2).
Even though the ankle joint can generate three rotations in all three pitch, roll, and yaw axes,
two rotations about ankle axes are generally utilized for the sprained ankle rehabilitation.
3
Research in biomechanics areas showed the importance of inter-segments foot motions based
on the extent of forefoot motion during stance phase
15
, and foot could be simply modeled as
two segments with one revolute joint to represent more closely natural foot motion during
gait.
16
For ankle rehabilitation, this motion will allow more diverse ankle exercises such as toe
raise, heel raise, etc, which cannot be achieved with the single rigid platform.
Further, lack of sufficient ankle proprioception is responsible for subsequent sprains in the
ankle.
17
Therefore, a proprioceptive exercise is desirable. This reduces the risk of re-injury and
increases the quality of the training. A hemispheric base is a suitable choice to increase
proprioception. Moreover, the device should have enough orientation workspace for wobble
bode motions
17
and should support full human weight. The designed angle orientation should
be more than 20 and the device should generate at least continuous 1000N to allow patient to
6
stand. Furthermore, the heave motion can be utilized to the advanced dynamic balancing that is
performed by moving platform after static balancing exercise. In essence, such rehabilitation
device must have functional platforms; to generate pitch and roll motions as well as relative
rotations at a foot. Table I summarizes ROM and force/torque-bearing capability requirements.
5,9, 18-19

2.2. Mechanism Description
Parallel mechanism robots have recently been proposed for ankle
4-9
and hand
20
rehabilitations because they have the desirable characteristics of low inertia, high rigidity,
compactness, greater portability, and precise resolution compared with serial robots. Currently,
most parallel robots consist of a single rigid platform. In clinical fields, however, robots with
more than two platforms are necessary in order to support multi-body objects. For this purpose,
a 4-dof parallel mechanism
21
integrated with two platforms have been suggested. The
mechanism with two platforms can generate pitch motion (
f
and
r
) at each platform, and roll
() and heave motions (Z
f
and Z
r
) of both platforms as shown in Fig. 3 (a). Fig. 3(b) shows the
manufactured robot, which has x-y stages with two serial LM guides inserted between
spherical joints and plates. This device is composed of front and rear platforms, and three limbs
as shown in Fig. 3, in which the letter P represents a prismatic joint, while the letter R
represents a revolute joint, the letter S a spherical joint, and an underlined letter represents an
actuated joint. Two outer limbs with 6-dof serial joints (P-S-P-P) are attached to both platforms
and are perpendicular to the base plate, while the middle limb is attached to a revolute joint that
connects the front and rear platforms. The middle limb is composed of 3-dof serial joints
(P
e
-R
e
-R), where P
e
-R
e
is a 2-dof driving mechanism that is equivalent to active serial prismatic
and revolute joints with two base-fixed prismatic actuators.
Even when the human foot is fastened firmly to the device, the structure of the suggested
robot allows relative rotation
m
between the fore and rear foot. Moreover, since the 2-dof
driving mechanism has very simple kinematic solutions, the computational load of the
7
proposed 4-dof mechanism is much smaller than that of general parallel mechanisms.
Furthermore, the platform of the mechanism is equipped with only two serial limbs instead of
four as seen in other parallel mechanisms, providing parallel mechanisms increased orientation
workspace as well as high stiffness to support human weight. The mechanism also inherits
characteristics that are intermediate between those of parallel and serial types, and provides
adequate solution for MTP motions.
Moreover, reconfiguration can be utilized to implement various exercise modes with single
device. Modular design has already been pursued for both serial and parallel mechanism robots.
22-25
Yoon and Ryu
26
proposed a kind of top-down approach, in which a base multiple
degree-of-freedom mechanism is invented first and then reconfigurability is sought by more
specifically changing extra part and/or by changing the direction of a revolute joint with adding
an extra part. Similar to Yoon and Ryu
26
, it is suggested that if a large plate is attached to both
fore and rear platforms with one actuator located at front platform having detachable feature,
the robot as shown in Fig. 4(a) with two platforms for ankle and foot motion control can act as
3-dof robot as shown in Fig. 4(b) with single platform, on which a patient can stand for both
proprioception and balance exercises. This facilitates reconfigurablity.
2.3. Kinematic Performance
Based on the design requirements computer simulations have been performed and the design
values are: L
a
= 5 cm, L
b
= 15 cm, L
base
= 10 cm, and actuator stroke = 20 cm, where L
a
and L
b
are the distances from the middle limb to the point of the fore and rear prismatic actuators, and
L
base
is the distance between the two active prismatic joints of the 2-dof mechanism (Fig. 2(a)).
For ROM exercises, necessary orientation workspaces are obtained with the designed
parameters and Fig. 5 shows the resultant workspace envelope. The workspace analysis uses
the inverse kinematics of the mechanism and actuator length constraints. Fig. 5(a) shows the
orientation workspace about pitch angle
f
and roll angle versus the front platform heave
motion Z
f
, while Fig. 5(b) shows the orientation workspace of pitch angle
r
and roll angle
8
versus the rear platform heave motion Z
r
. The maximum heave motion is 20 cm, and the
maximum pitch angles of the front and rear plates exceed 50 and 70, respectively. The
maximum roll angle of the platforms exceeds 55. Since the relative rotation angle
m

(
r f m
= ) can be obtained by relative rotation of the front and rear platform, the maximum
allowable MTP joint angle can be generated up to 70 when
r
=0 and
f
= maximum or
minimum. Fig. 6 shows the geometric representation at roll angle of 55 and MTP joint angle
of 70.
After design of desirable workspaces, it is necessary to satisfy adequate force-bearing
capability requirement throughout range of motions for strengthening exercises. In addition,
when a patient stands on the device platform during balance exercises, the device need to
support the human weight. In the content of analysis and design for load-bearing capability,
given maximum force of actuators, the maximum force/torque at the device platforms can be
computed by using force/torque relationship between actuators and platform as

i
T
i
F J = ( r or f i = ) (1)

where
T
] , , , [
2 1 m m r f
= is the vector of the linear forces of the actuators,
i
F
T
z
r
T F T T
i f
] , , , [

= is the vector of the generalized forces of each platform, and J is the
Jacobian matrix
21
of the designed mechanism, subscript f and r represent front and rear
platform, respectively. For bounded actuators force, maximal force and torques applied on the
upper platform of the mechanism can be determined by using numerical discretization.
27
The
maximum force and torques at each axis of the platform are measured by checking whether or
not the maximum force of each single actuator is over with respect to platform rotations. For a
pneumatic cylinder (max force: 1000N at 100psi) with a 50mm diameter and actuator force
bounded by 1000N, force-bearing capability requirements for both strengthening and
balancing exercises can be satisfied. For strengthening exercises, the maximum allowable
torques that can be applied on the upper platform are shown in Fig. 7. As two rations vary
within prescribed workspace, the maximum torque

T about the roll axis in Fig. 7(a) exceeds


9
104Nm, while the maximum torque
r
T

about the pitch axis in Fig. 7(b) is above 248Nm. It


should be noted that the maximum torque
r
T

about the pitch axis are independent on the roll


motions since the roll motions generated by the 2-dof driving mechanism are independent on
the pitch motions. Fig. 8 shows that the maximum force
r
z
F at z-axis exceeds at least 1100N at
the center of the rear platform with respect to the two rotations for balancing exercises. It
should be noted that the maximum force
r
z
F at z-axis is 3000N at zero angles of two rotations,
which is the sum of the bounded force of the three actuators. Finally, the torque at MTP joint
can be obtained by relative torque of the front and rear platform until 300Nm. Based on the
specification requirements, the design (Table I) can satisfy motion range and force-bearing
capabilities for ROM/strengthening device and balance/proprioception device. Fig. 9 shows
each device that is designed based on these requirements.

3. IMPEDANCE CONTROL FOR VARIOUS EXERCISE MODES
For physical therapy, impedance control of pneumatically-driven robotic devices can be
integrated depending on specific rehabilitation situations. Pneumatic actuators have been in
use because of their high power-to-weight ratio, good compliance, ease of maintenance,
cleanliness, and the ability to maintain high forces without overheating.
28
Since the
force-based scheme generally provides low impedance compared to position-based scheme
29
and lower limbs generate higher forces/ torques than upper limbs, the rehabilitation device
entails high impedance suggesting position-based adoption. Additionally, position-based
impedance may be more appropriate for pneumatic systems because the valve dynamics to
generate desirable torque as actuators have large lag compared with electrical motors and
unlike torque based control, the controller requires no dynamic model of the robot or the
environments, simplifying the control.
30
For patients who want to exercise from small to large
torque to increase their muscular strength, a wide range of achievable impedance must be
10
provided without causing any stability problem. The suggested impedance controller is
therefore very convenient and systematic way of implementing various exercise modes
requiring pure position, pure force, and hybrid position and force control of the system at given
specific rehabilitations without causing any instability during rehabilitation trial by setting the
impedance parameters within a stable range.
3.1. Position-based Impedance Control
3.1.1. Outer loop (position compensation loop with a force feedback)
The position-based impedance controller (Fig. 10) is composed of a position controller (inner
loop) with a force feedback loop (outer loop) that modifies position control command in order
to satisfy target impedance. The desired force F
d
and desired foot position X
d
in the Cartesian
space are related as
29
;
) ( ) (
m
d
m
d
d
X X B X X K F
& &
+ = (2)
where K=diag {
i
z
K ,
i
K

K } is the desired stiffness matrix, B=diag {


i
z
B ,
i
B

B } is the
desired damping matrix, and X
m
is the measured platform position. The desired stiffness,
damping matrices, and foot position X
d
should be selected (explained later). The position
adjustment vector X
imp
can be computed by following relations

imp
imp m
X B KX F
&
+ = (3)
where F
m
is the measured interaction force that is estimated by pressure sensors. The desired
set point position command X
c
and the position adjustment vector X
imp
are related by

imp d c
X X X = (4)
Thus, the position command X
c
is modified by the position adjustment vector X
imp
calculated
from the target impedance relationship in equation (3) with the measured force F
m
. If the
position control of the robot is perfectly operating (X
c
= X
m
), the desirable force F
d
in equation
11
(2) become ideally equal to the measured force F
m
in equation (3), which means that the
desired impedance have been achieved ideally. Note that if there are no interaction forces, the
desired foot position X
d
becomes the desired position command X
c
and the impedance control
is operating as only a position control. Also, the larger impedance implies the harder movement
and the smaller impedance implies the easier movement of the platform through the foot.
3.1.2. Inner loop (position control loop)
The position control loop should be designed to operate as ideally as possible to achieve
desirable impedance more closely. The position control is implemented with feedback control
of PD type with feedforward control input.
10
If the desired command input X
c
is given, the
resultant measured platform coordinates X
m
can be obtained from the measured actuator
lengths by forward kinematics (FK). The position control input will be transformed into the
desired actuator input L
c
after being mapped through inverse kinematics (IK).
Then, the independent joint control will be applied to actuator, which is controlled by the
four-way proportional valve (Enfield Technology). The control input of the double acting
pneumatic cylinder with special low-friction type (Bimba Corp.) is proportional to differential
pressure. The generated differential pressure is expressed as
) ( ) )( / (
u l m m c i d p
P P K L L s K s K K P + + = (5)
where L
c
and L
m
are the commanded and measured cylinder displacements, K
p
, K
d
, and K
i
are
proportional, derivative, integral gains, P
l
and P
u
are lower and upper pressures of the cylinder
chamber, respectively, and K
m
is the pressure feedback gain. The pressure feedback is
proportional to the external force applied to the piston. The system stability can be improved
by augmenting differential pressure feedback to a position controller.
31
We define instability
as the situation in which vibration occurs while maintaining a fixed pose (i.e. at constant X
d
)
during foot interaction with the robot. The pressure sensor can directly detect the forces such
as friction or external disturbances, and can efficiently reject force disturbances that are
12
generated by patients legs during trials. The PID scheme with pressure feedback gives
comparable results with other robust controllers.
32

3.2. Achievable Impedance
For the suggested impedance controller, the maximum stiffness
33-34
, the ability to create
force in response to a disturbance, can be obtained by measuring the changed positions from
home position and the resultant forces of the device with the largest possible position gain
while not inducing control instabilities. Similarly, the minimum DC stiffness can be achieved
by measuring the changed positions and the resultant forces. In the case of position control
with pressure differential feedback, the achievable stiffness range was increased about 20%,
which shows the importance of the position controller performance for wider impedance range.
When the minimum achievable impedance experiments are performed at the position
controller with and without pressured feedback, Fig. 11 shows that the platform driven by
position controller without differential pressure feedback becomes unstable at a stiffness gain
of
r z
K =10kN/m, whereas the platform driven by position controller with differential pressure
feedback is stable at the same stiffness gain. The minimum achievable stiffness with pressured
feedback was stable until 7kN/m without causing instability. Since damping coefficient is not
utilized for implementation of each exercise of ankle rehabilitation stages in the later section,
this coefficient is set to zero. When the position controller is set to its maximum and minimum
stable gains, the achievable maximum and minimum stiffness were measured (Table II).

4. IMPEDANCE DESIGN FOR EACH EXERCISE MODE
Parameters of the position-based impedance control such as the desirable foot position X
d
,
the desirable foot velocity X
d
, and the desired impedance are designed for each exercise mode.
13
4.1. Range of Motion Exercise
There are three types of ROM exercises: passive, active-assist, and active. Passive ROM
(PROM) is the movement that is applied to a joint only by efforts such as another persons or a
machine while the patient does not resist against the motion. To realize PROM, a fine position
control should be guaranteed to keep the range of motion within acceptable range without
inducing any pain even though some external force is exerted. This demands implication of
infinite impedance Z
inf
. By setting the desired impedance infinite, the position can be
controlled through the impedance controller. This reduces X
imp
to become zero. Since the
resultant impedance generated by setting of the desirable impedance of the robot cannot exceed
the maximum achievable impedance, the desirable impedance was set to the maximum
achievable impedance. Active assist range of motion (AAROM) represents a joint movement
that needs partial assistance from an external force. Active range of motion (AROM) is the
movement of a joint that is provided totally by the patient without external assistance. Thus, the
amount of selected impedance determines the effect of assistance suggesting minimum and
maximum achievable range of the specific device. In PROM and AAROM mode, a desired
staircase position command X
d
can be designated by a physical therapist who can directly
diagnose the patients conditions during testing as well as operations, keeping in view of static
movements. The staircase command input may be more convenient to generate necessary
change of the patients current ROM with respect to the patients direct response.
4.2. Strengthening Exercise
Strengthening exercises enhance muscle strength by generating more strain on a muscle than
it is in normal condition. Strengthening exercises can be classified into the form of isometric,
isotonic and isokinetic strengthening.
4.2.1. Isometric exercise
Strengthening can be applied to the proper joint angle with exact isolation about the region of
14
weak muscles. Isometric training is efficient to develop overall strength of particular muscles.
Even though muscles contract, there is no movement in the affected joints during isometric
exercises. To satisfy this mode, the infinite impedance Z
inf
should be applied keeping position
command.
4.2.2. Isotonic exercise
Isotonic exercise is a dynamic form of exercise that is carried out against a constant or
variable load through the available ROM. Isotonic exercise is different from isometric exercise
in that there is movement of a joint during the exercise. Weight training with dumbbells and
barbells, or heel raise are standard examples of an isotonic exercise. The muscle can be
shortened or lengthened with a constant weight throughout the range of motion. The isotonic
controller is implemented when the human moves the foot, while the desired force/torque is
maintained constant. By using (5) with B=0, the desired foot position X
d
can be represented by


m d d
X K F X + = / (6)
Then, the desired position control command
c
X can be represented by

m m d m m d imp d c
X K F F K F X K F X X X + = + = = / ) ( / / (7)
Eq. (7) implies that X
c
will be changed to keep the measured force F
m
to desired force F
d
.
4.2.3. Isokinetic exercise
Isokinetic exercise allows constant preset speed with variable resistance as the muscle moves
through full ROM. This exercise can provide maximal loading throughout. Despite how hard
and fast the patient works, the isokinetic properties of the robot should permit patients to move
only as fast as preset speed. To satisfy isokinetic motions, the desired velocity command
d
X
&
should be preset. In addition, the infinite impedance Z
inf
should be applied to generate
velocity control, which should prevent the robot from deviating from the desired speed by
external disturbances like patients muscle force. Since the impedance controller with the
infinite impedance Z
inf
behaves a position controller, the desirable foot position X
d
can be
15
represented by the desirable foot velocity
d
X
&
and time t
X
d
= X
0
+
d
X
&
t (8)
Where X
0
is the initial foot position. The desirable foot velocity
d
X
&
should be constant in
piecewise periods to provide isokinetic mode.
4.3. Proprioception Exercise
The loss of proprioception will cause a patient not to recognize his joint motions, while people
with a strong proprioceptive sense can have good balance and coordination. A common
progression when performing balance exercise is to move from a position of non-weight
bearing to weight bearing, bilateral stance to unilateral stance, eyes open to eyes closed, firm
surface to soft, uneven or moving surface.
3
For static balancing exercise, motions similar to
wobble board are required to be generated by impedance control. The wobble board allows
only rotational motions of pitch, roll, and yaw about a hemispherical pivot at the center of
lower side. In order to mimic the wobble board motions, the impedance of translational
motions should be maximized to fix the platform at a constant position, while the impedance of
the orientation motions should be minimized to give free orientation motions at platform center.
Then, the dynamic balancing exercises are performed to improve effectiveness. Dynamic
balancing can be implemented by varying the desired positions and changing the desired
impedance of the platform. Table III summarizes the design parameters of impedance Z,
desired position and velocity command, and the desirable exercise examples with respect to
each exercise mode.

5. EXPERIMENTS OF EXERCISE MODES
For experiments, the total ankle rehabilitation system is suggested for patient to perform each
exercise mode with visual feedback, which composed of the robotic device, the controller with
pneumatic actuators, the host PC, and a patient. The host PC displays virtual environments and
16
connects the controller through RS232C serial communication. The rotation and torque
information of the robotic platform will be transferred to the virtual environments and will be
stored for the patients histories at DB. The suggested system is shown in Fig. 12. In this
section, the experiment results of each exercise mode that is designed based on impedance
parameters will be shown.
5.1. ROM Exercises
Based on the patients condition, initial commands utilizing stair type exercise scenario are set.
Fig. 13 shows rotation and torque of the inversion (negative roll motion of the platform) in the
AAROM and PROM modes. For AAROM mode, desired stair input was displayed at
computer monitor for visual feedback of a patient. Fig. 13(a) shows the desired and measured
stair response for each ROM mode. PROM mode with infinite impedance setting follows the
desired trajectory well, while AAROM mode with

K =150Nm/rad and

B =0 has bigger
trajectory errors due to the lower impedance setting. Fig. 13(b) shows the measured torque for
each case. The relatively small measured torque value at AAROM means that the subject
applied his own torque in order to follow the desired trajectory. Conversely, the PROM mode
needs larger torque because of inactive efforts from the subject.
5.2. Strengthening Exercise
5.2.1. Isometric exercise
For isometric exercise, virtual environments have been developed in the previous research
for ankle and hand rehabilitation.
35-36
Fig. 14 shows the image viewers that are used to give
motivation for isometric exercise. There are two images for which the front image will be
widened proportional to the torque exerted by a patient. If the device driven by human foot
exceeds beyond threshold (torque) selected by a therapist, the hidden image will appear at the
front. Then, another mage will be positioned on the back. The torques of the inversion and
17
eversion directions will divide the front image into horizontal direction, while the torques of
the dorsiflexion and plantar flexion directions will divide the hidden image into horizontal
direction. Fig. 15 shows that even though the position variations are relatively small (within 2
degrees), which is required to be kept constant for isometric exercises, the exerted maximum
torques about inversion and eversion increases or decreases according to a patients efforts at
the ankle joint. If the exerted torque is higher than the present threshold (e.g. 50Nm), the next
scene will appear in the front. The torque threshold is determined by physical therapists
considering the patients conditions.
5.2.2. Heel raise for isotonic excise
Heel raise motion is a standard isotonic exercise since there is constant torque to ankle axis
due to a human weight. This exercise can be generated by setting the desired position control
command X
c
at equation (7) with desired force and stiffness. In order to generate heel raise
motion, the only pitch motion
r
of the rear platform should be moved, while the pitch motion

f
of the front platform and roll motion should be zero since toe attached to the front platform is
steady. Fig. 16 shows that even though the pitch motion
r
(platar flexion or dorsiflexion) is
changing, the device can keep the measured torque constant; 25Nm with small oscillations,
where
r
K

=500Nm/rad,
r
B

=0, and X
d
(
r
)=0.5 rad (28.65).
5.2.3. Gait trajectories training for isokinetic exercise
The ankle and foot motions during walking at the ground level is generated based on normal
gait.
37
The task specific and repetitive gait training of a patient at foot and ankle can increase
the strength of the lower limbs with an isokinetic exercise. Since the mechanisms for the
unrolling foot during stance phase could generate the development of ankle sprains, Willems et.
al
38
suggested that effective prevention and rehabilitation of inversion sprains should include
attention to gait patterns and adjustments of foot biomechanics in subjects at risk of a sprain.
Therefore, the test of ankle and foot motions about the repetitive normal gait trajectories was
18
executed while the subject was sitting on the chair.
When the heel-striking angle, heel-off angle, and toe-off angle are 30, respectively, the foot
trajectory configurations at planar surface with respect to gait cycle are illustrated in Fig. 17.
Fig.18 (a) shows the desired and measured rotation value of pitch angles
r
of the real platform
in order to generate desirable trajectories with respect to gait cycle. Fig. 18(b) shows the
desired and measured angular velocity
r

&
of the rear platform. At piecewise periods, the
angular velocities of the rear platform generated constant values, a desirable characteristic for
isokinetic exercises. Through experiments, the ioskinetic motions were well generated
regardless of subjects intentions. These kinds of advanced functional training including MTP
joint motions will not be possible by using Cybex and Biodex, which may be helpful to
increase the coordination of the lower limbs during real walking.
5.3. Proprioception/Balance Exercise
A ball stabilization game may be used for these exercises (See Fig. 19). By standing on the
large plate and shifting his/her weight from left to right or from fore to rear, the patient can
make the board tilt about the two ankle axes. The balance by maintaining center of gravity
within bodys base of support and the coordination by moving patients two limbs can be
improved simultaneously. Fig. 20 shows experimental results for the condition that a subject
was standing on the platform with double legs and shifted his center of gravity from fore to rear
while keeping his bodys positions stable. The impedance values are set to
r
z
K =80kN/m for
hard surface,
r
K

=0.2kNm/rad, and
r
K

=0.1kNm/rad for free rotations. The desirable position


commands are set to the X
d
(z)=150mm, X
d
(
r
)=0 and X
d
()=0 for keeping the plate stationary.
Fig. 20 shows that the rear platform z-axis reaction force of about 800N is kept almost constant
(Fig. 20(a)) with small displacements changes (Fig. 20(b)). On the other hand, there are large
pitch and roll motions with relatively small torques as shown in Fig. 20 (c) and Fig. 20(d).
Decreasing rotational motions in Fig. 20(d) showed that the subject standing on the center of
19
the platform was changing his weight to keep the balance of his body. Patients will try to
maintain constant height and zero rotations to improve proprioception/balance. The recorded
motion histories should be useful for progress reports of patients. Comparing to conventional
wobble board, the suggested proprioception/balance type robot can adjust the level of motions
stability by changing the impedance of the platforms. Also, the more difficult dynamic balance
training can be achievable by moving the platform with the desired foot position X
d
for
generations of moving surfaces.

6. CONCLUSIONS AND DISCUSSIONS
This paper presents a novel reconfigurable ankle rehabilitation robot for allowing various
exercise modes. The proposed 4-dof robot is composed of two upper platforms and three limbs
driven by four pneumatic actuators. This new design can generate MTP motion between the
fore and rear foot and two rotations about the ankle joint including the heave motion. These
motions are adequate for natural foot and ankle motions. The mechanism has very simple
kinematics and allows wide workspace enough to cover the required ROM. Additional
exercises such as heel and toe raise that were not possible with the previous rehabilitation robot
systems can be implemented. Furthermore, simple reconfiguration of the upper platforms
allows covering more exercises such as proprioception and balance training. Position control
method utilizing impedance parameter of each exercise mode is adopted.
The developed ankle rehabilitation robot can substitute not only the traditional therapy in
various exercise modes but also supply the advanced functional exercises such as desirable gait
training during stance phase by generating natural foot trajectories with the MTP and ankle
joints motions and dynamic balancing by controlling levels of unstable surfaces with the
variable impedance and varied movement of the platform. Some standard exercise examples
are experimented and presented.
As regards to further work, comprehensive clinical trials with real patients should be
20
performed to tune the impedance parameters and to evaluate the usefulness of the proposed
system. Moreover, more entertaining virtual environments for each exercise mode should be
developed for motivating patients in order to shorten recovery periods. Clinical test for the
advanced functional exercises with the developed ankle rehabilitation robot and extension to
gait rehabilitation should be considered too.
ACKNOWLEDGMENT
Research reported here was supported by grant (No. R01-2002-000-00254-0) from the Basic
Research Program of the Korea Science & engineering Foundation. The authors would like to
thank T.S. Song and J. W. Choi for their discussions, especially on experiments.

REFERENCES
1. http://www.ehendrick.org/healthy/
2. Bookhout, Mark R., and Grenman, Philip. Principles of exercise prescription. Woburn,
MA: Butterworth-Heinemann, 2001.
3. C. G. Mattacola and M. K. Dwyer, Rehabilitaion of the ankle after acute sprain or
chronic instability, Journal of Athletic Training, 37: (4)(2002), 413-429.
4. M. Girone, G. Burdea, and M. Bouzit, Ankle rehabilitation system, U.S. patent,
6162189, Dec 2000.
5. M. Girone, G. Burdea, and M. Bouzit, Rutgers Ankle orthopedic rehabilitation
interface, Proc. of the ASME, Dynamic Systems and Control Division, DSC Vol.67,
1999, pp. 305-312.
6. M. Girone, G. Burdea, M. Bouzit, V. Popescu, and J. E. Deutsch, A Stewart
platform-based system for ankle telerehabilitation, Autonomous Robots, 10:(2)(2001),
203212.
21
7. Deutsch, J., J. Latonio, G. Burdea and R. Boian, Post-stroke rehabilitation with the
Rutgers Ankle system - A case study, Presence, MIT Press, 10:(4)(2001), 416-430.
8. R. F. Boian, H. Kourtev, K. Ericson, J.E. Deutsch, J.A. Lewis, and G.C. Burdea, Dual
Stewart-platform gait rehabilitation system for individuals post-stroke, Proc. Second
Int. Workshop on Virtual Rehabilitation, 2003, pp. 92.
9. J. S. Dai and T. Zhao, Sprained ankle physiotherapy based mechanism synthesis and
stiffness analysis of a robotic rehabilitation device, Autonomous Robot, 16: (2)(2004),
207-218.
10. T. Noritsugu and T. Tanaka, Application of rubber artificial muscle manipulator as a
rehabilitation robot, IEEE/ASME Tran. on Mecha., 2: (4)(1997), 259-267.
11. R.Richardson, M. Brown, B. Bhakta, and M. C. Levesley, Design and control of a
three degree of freedom pneumatic physiotherapy robot, Robotica, 21:(6)(2003),
589-604.
12. H. I. Krebs, N. Hogan, M. Aisen, and B. Volper, Robot-aided neurorehabilitation,
IEEE Trans. Rehab.Eng., 6: (1)(1998), pp. 7587.
13. J. Yoon, R. Boian, J. Ryu, and G. Burdea, Control of the Rutgers dual-platform
interface for ankle rehabilitation, ASME Robot Dynamics and Control Conference,
IMECE2002- DSC- 32655, New Orleans, USA, 2002.
14. S.Jezernik, K.Jezernik, and M.Morari, Impedance control based gait-pattern
adaptation for a robotic rehabilitation device, 2nd IFAC Conference on Mechatronic
Systems, December 9-11, Berkeley, California, USA, 2002.
15. A. E. Hunt, R. M. Smith, M. Torode, A. M. Keenan, Inter-segment foot motion and
ground reaction forces over the stance phase of walking, Clinical biomechanics, 16: (7)
(2001), 592-600.
16. L. A. Gilchrist and D. A. Winter, A two-part, viscoelastic foot model for use in gait
simulation, Journal of Biomechanical Engineering, 29: (6)(1996), 795-798.
22
17. Tropp, H. and H. Alaranta, Proprioception and coordination training in injury
prevention, Sports Injuries: Basic Principles of Prevention and Care, Oxford, 1993.
18. C. S. Parenteau, D. C. Viano, and P. Petit, Biomechanical properties of human
cadaveric ankle-subtalar joints in quasi-static loading, Journal of Biomechanical
Engineering, 120:(1)(1998), 105-111.
19. Donatelli, R. A., The biomechanics of the foot and ankle (Philadelphia, F.A. Davis,
1996).
20. H. Choi and J. Ryu, Development of a tele-rehabilitation system for upper limb using
virtual environment and haptics, 5
th
International Workshop on Human-Friendly
Welfare Robotic Systems, KAIST, Daejeon, 2004, pp. 68-75.
21. J. Yoon and J. Ryu, A new family of hybrid 4-dof parallel mechanisms with two
platforms and its application to a footpad device, J. Robot. Syst., 22: (5)(2005),
287-298.
22. R. Cohen, M. G. Lipton, M. Q. Dai, and B. Benhabib, Conceptual design and a
modular robot, ASME J. Mech. Des., 114: (1)(1992), 117-125.
23. Z. Ji, and P. Song, Design of a reconfigurable platform manipulator, J. Robot. Syst.,
15: (6)(1998), 341-346.
24. Yang G., Chen, I-M., and Yeo, S. H., Design consideration and kinematic modeling for
modular reconfigurable parallel robots, 10
th
World Congress on the Theory of
Machines and Mechanisms, Oulu, June 20-24, 1999 , pp. 1079-1084.
25. Lee, W.H. and Sanderson, A.C., Dynamics and distributed control of modular robotic
systems, Proc. of the 26th Annual Conf. of the IEEE Electronics society, 2000, pp.
2479-2484.
26. J. Yoon and J. Ryu, Reconfigurability of a parallel manipulator: A case study,
Proceedings of the WORKSHOP on Fundamental Issues and Future Research
Directions for Parallel Mechanisms and Manipulators, Quebec, Canada, 2002, pp.
94-97.
23
27. J. P. Merlet, Parallel Robots, Dordrecht, The Netherlands: Kluwer, 2000, pp. 254-255.
28. S. R. Pandian, F. Takemura, Y. Hayakawa, and S. Kawamura, Control performance of
an air motor, IEEE Int. Conf. on Rob. and Auto., 1999, pp. 518-524.
29. D.A. Lawrence, Impedance control stability properties in common implementations,
IEEE Int. Conf. on Rob. and Auto., 1988, pp. 1185-1190.
30. B. Heinrichs, N. Sepehri and A. B. Thorton-Trump, Position based impedance control
of an industrial hydraulic manipulator, IEEE Control Systems Magazine, 17:(1)(1997),
4652.
31. J. Wang, J. Pu, and P. Moore, A Practical strategy for servo-pneumatic actuator
systems, Control Engineering Practice, 7: (12)(1999), 1483-1488.
32. S. Chillari, S.Guccione and G. Muscato, An Experimental comparison between
several pneumatic position control methods, Proc. of the 40
th
IEEE Conference on
Decision and Control, 2001, pp. 1168-1173.
33. G. Ellis and R.D. Lorenz, Comparison of motion control loops for industrial
applications, Proc. of IEEE, IAS Annual Tech. Conf., Oct. 3-7, 1999, pp. 2599-2605.
34. R. E. Ellis, O. M. Ismaeil, and M. G. Lipsett, Design and evaluation of a
high-performance haptic interface, Robotica, 14: (3)(1996), pp. 321327.
35. J. Ryu, H. Song, J. Yoon, J. Choi, J. Cha, G. Yim, Y. Nah, K. Lee, and K. Kim, Virtual
environments and haptics for ankle rehabilitation, Journal of Control, Automation, and
Systems Engineering, 9: (8)(2003), 577-583 (Korean).
36. W. Son, K. Kim, B. Jang, and B. Choi, Interactive Dynamic Simulation Schemes for
Articulated Bodies through Haptic Interface, ETRI Journal, 25:(1)(2003), (25-33).
37. J. Perry, Gait analysis: normal and pathological function (Thorofare, N. J. :Slack Inc.,
1992).
38. T. Willems, E. Witvrouw, K. Delbaere, A. De Cockb, and D. De Clercq, Relationship
between gait biomechanics and inversion sprains: a prospective study of risk factors,
Gait Posture, 21:(4)(2005), 379-87.
24


Figure 1. Functional Ankle Rehabilitation

Figure 2. Ankle and foot motions

(a) CAD model (b) fabricated robot
Figure 3. New 4-dof mechanism with two platforms
25

(a) ROM/strengthening type (b) balance/proprioception type
Figure 4. Novel reconfigurable ankle robot

(a) Front plate (b) Rear plate
Figure 5. Workspaces at ankle joint for ROM exercises


(a) roll ( ) (b) MTP joint

Figure 6. Geometric representation of maximum allowable rotations
26


(a)

T (roll axis) (b)


r
T

(pitch axis)

Figure 7. Maximum torques of the two axes for strengthening exercises

Figure 8. The maximum z-axis forces at center of the rear platform for balancing exercises

(a) ROM/ strengthening type (b) Balance/proprioception type
Figure 9. Ankle rehabilitation device

27

Figure 10. Impedance control block diagram for rehabilitation exercises

(a) position (b) force
Figure 11. Minimum achievable stiffness at
r
z
K =10kN/m


Figure 12. Total Ankle Rehabilitation System
28

(a) X
m
( axis) (b) F
m
( axis)
Figure 13. AAROM and PROM for inversion



Figure 14. Image viewer window
35



a) X
m
( axis) b) F
m
( axis)
Figure 15. Isometric exercise for inversion and eversion
29


(a) X
m
(
r
axis) (b) F
m
(
r
axis)
Figure 16. Isotonic exercise at heel raise




(a) Heel-strike (0%) (b) Foot-flat (10%) (c) Heel-off (50%)

(d) Toe-off (60%) (e) Initial swing (70%) (f) Terminal swing (90%)
Figure 17. The mechanism configuration according to gait cycle at ground level
30

(a) measured angles
r
(b) measured angular velocity (
r

&
)
Fig. 18. The walking motions for isokinetic exercise


Figure 19. Ball Stabilization Virtual Environment


(a) F
m
(Z
r
axis) (b) X
m
(Z
r
axis)
31

(c) F
m
(
r
and axes) (d) X
m
(
r
and axes)
Figure 20. Balance /coordination exercise results




TABLE I
REQUIREMENTS OF ANKLE JOINT AND DESIGNED RESULTS OF THE DEVICE.

Motion
DF
(ROM)
PF
(ROM)
Inv
(ROM)
Eve
(ROM)
MTP joint
(ROM)
DF
(Torq)
PF
(Torq)
Inv
(Torq)
Eve
(Torq)
Spec 20 40 35 25
Flex (45)
Ext (70)
40.7-
97.6Nm
20.3-36.6
Nm
Max
48Nm
Max 34Nm
Designed 50 55 70 104Nm 248Nm




TABLE II
MINIMUM AND MAXIMUM ACHIEVABLE STIFFNESS

f
z
K
f
K

K
r
z
K
r
K


Minimum 4.5kN/m 0.25kNm/rad 0.05kNm/rad 4.5kN/m 0.1kN/rad
Maximum 200kN/m 6.5kNm/rad 2kNm/rad 200kN/m 2.5kNm/rad

32
TABLE III
IMPEDANCE PARAMETERS FOR EACH EXERCISE MODE


ROM (stretching)

Strengthening

Proprioception


Impedance
parameter
PROM AAROM AROM Isometric Isotonic Isokinetic AROM + Isometric

Z

Z
inf


Z
zero
- Z
inf



Z
zero

Z
inf
Z
const
Z
inf
Z
zero
at orientations
Z
inf
at position
X
d
Vari Vari Const Vari Vari Const (static)
or Vari (dynamic)
d
X
&

Zero Zero

N/A
Zero Vari Const Zero (static)
or Const (dynamic)
Control
type

Position
Variable
Impedance
N/A
Position
Position
based force

Velocity

Wobble board motions

Exercise
Example


Stair input

Stair input
with visual
feedback



N/A

Image
viewer

Heel
raise

Gait
generation


Ball stabilization

Вам также может понравиться