1025

Study of an upper arm exoskeleton for gravity

balancing and minimization of transmitted forces
V N Dubey1* and S K Agrawal2
1
School of Design, Engineering and Computing, Bournemouth University, Poole, UK
2
Mechanical Systems Laboratory, Department of Mechanical Engineering, University of Delaware, Newark, USA

The manuscript was received on 10 April 2010 and was accepted after revision for publication on 28 July 2011.

DOI: 10.1177/0954411911420664

Abstract: An upper-arm wearable exoskeleton has been designed for the assistance and func-
tional training of humans. One of the goals of this design is to provide passive assistance to a
user by gravity balancing, while keeping the transmitted forces to the shoulder joints at a min-
imum. Consistent with this goal, this paper discusses: analytical gravity balancing design con-
ditions for the structure of the exoskeleton; a possible implementation of these conditions
into practical designs; the minimization of transmitted joint forces to the shoulder while satis-
fying the gravity balancing conditions; the numerical optimization of the system for gravity
balancing and minimization of transmitted forces; and the effect of parameter variation on
joint moments and joint forces via numerical optimization.
An implementation of the design was undertaken using zero-free-length springs. The design
idea presented in this paper may be useful in relieving the actuators effort of exoskeletons to
support the weight of the arm and therefore the possibility of using small actuators and mak-
ing the system light and portable or even a stand-alone passive support device can be devel-
oped based on these gravity balancing conditions.

Keywords: arm exoskeleton, gravity balancing, assistive device, portable systems,

rehabilitative device

1 INTRODUCTION which has a need for exoskeletons consists of peo-

A large number of people are affected by arm condi-
tions whose symptoms include profound muscle
weakness: these conditions include spinal muscular
atrophy, Duchenne muscular dystrophy, and limb
girdle [1]. As a result sufferers have problems lifting
their arm against gravity. Consequently, they are
unable to independently perform various activities
of daily living (ADL) [2, 3]. The functional capabil-
ity of this group can be enhanced by augmenting
the arm with external mechanical supports in the
form of exoskeletons. Such devices can provide
assistance to movement of the arm. Another group
*Corresponding author: School of Design, Engineering and
Computing, Bournemouth University, Fern Barrow, Poole, BH12
5BB, UK.
email: vdubey@bmth.ac.uk
ple who suffer from impaired motor skills such as
those affected by strokes. The exoskeletons can be
used as a rehabilitative tool to improve functional
movement as recently demonstrated in lower
extremity studies with the exoskeletons GBO and
ALEX [4, 5].
To allow these devices to be worn continuously
they should be light in weight and provide anatomi-
cally consistent motions. This precludes the use of
many existing upper limb exoskeletons as they are
not designed for continuous use and are bulky.
Some well known designs for post-stroke upper
extremity functional rehabilitation are ARM [6],
MIME [7], and MIT-MANUS [8]. There have been
many improvements to these designs in recent years
[9–14]. Some newer designs that feature cable
actuation are promising and could be further devel-
oped but for now they still remain bulky [15, 16].

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
1026 V N Dubey and S K Agrawal

A challenging problem in exoskeleton design is

the need for a portable power source. Whilst such
power sources will be developed in future, an alter-
native solution to this problem would be to design a
gravity support system which could partially or fully
eliminate the effects of gravity on the arm over its
full range of motion, thus mitigating the need for a
power source. Such a design would be dependent
on parameters of the human and the exoskeleton
[17, 18]. Recent studies in gravity balancing have
focused primarily on eliminating the torques at the
joints to support the arm in any configuration.
However, another issue that is equally important Fig. 1 Model of the exoskeleton on a skeleton arm
and also needs to be addressed is the nature of the
forces transmitted at the joints. Ideally, the design
should be such that it does not exceed the force- have four degrees of freedom; three at the shoulder
bearing capacity of the joints in the arm. In addi- and one at the elbow. As seen in Fig. 2, body 1 is
tion, sensitivity of the design should be studied with the cross-bar connected to the shoulder, body 2 is
respect to variation in the parameters of the human the U-link connecting the cross-bar to the arm,
and the exoskeleton design. This paper addresses body 3 and body 4 are the upper and lower arms,
some of these issues from a design implementation respectively. The shown angles (u1, u2, u3) are joint
perspective. rotations about the Z, X, and Y-axis, respectively
The paper is organized into the following sec- that define the plane of elevation, angle of eleva-
tions. Section 2 discusses the design and a simple tion, and internal/external rotation; and the angle
kinematic model of the upper arm, section 3 devel- u4 is the elbow rotation about the X-axis with
ops gravity balancing conditions for the arm, section respect to the shown coordinate frame. The skele-
4 evaluates design parameters and their effects on ton emulates the motion of a human arm with
the transmitted forces at the shoulder joint, section
three rotational joints at the shoulder and a revo-
5 presents a numerical optimization for minimiza-
lute joint at the elbow [20]. The details of the arm
tion of joint torques and forces, and section 6 dis-
parameters including that for gravity balancing are
cusses design implementation, finally, conclusions
given in the appendix. The kinematic parameters
are drawn in section 7.
of the arm are shown in Fig. 3 and the correspond-
ing Denavit–Hartenberg (DH) parameters are
2 KINEMATIC MODEL OF THE UPPER ARM listed in Table 1. The axes of rotation of the joints
in the initial configuration are shown in Fig. 4. The
An upper-arm exoskeleton has been designed to transformation matrices that can be obtained
assist in movement training of the human arm. The based on these DH parameters are well known
solid model of the exoskeleton over a skeleton arm [21], for brevity these are not included in the
with comparable degrees-of-freedom to the human paper.
arm is shown in Fig. 1. The complete design details
of the cable-driven exoskeleton using cuffs at the
shoulder and arm are described in reference [19].
The exoskeleton has cuffs on the upper and forearm
to which cables are connected. An inverted U-
shaped shoulder cuff holds the motors. The arm
cuffs have channel sections which hold circular
tubes that can be inflated or deflated for sufficient
grip on the arm.
The design using cuffs and cables allows a flex-
ible structure for the exoskeleton that makes it
light in weight, portable, adjustable to different
sizes, and able to accommodate any auxiliary units
that may be required for gravity balancing. The Fig. 2 A solid model of an arm having the same num-
arm supported by the exoskeleton is considered to ber of degrees of freedom as the human arm

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
 Study of an upper arm exoskeleton 1027

obtain the gravity balancing condition for the arm,

the identification of various arm parameters includ-
ing axes of rotation and locations of the local frames
is important. The location of the centre of mass of
the upper arm as seen in frames 3 and 0 is given (Fig.
3); note S and C are sine and cosine trigonometric
functions which represent the arm configuration
2 3
2 3 0
Fig. 3 Kinematic chain of the arm 0 6 d3 7
F3 = 4 d3 5 F0 = 0 T 3 64 0 5
7
0
1
2 3
Table 1 DH parameters of the skeleton arm d3 C1 S2 + d3 C1 S2
6 d3 S1 S2 + d3 S1 S2 7
Link/parameters ai ai di ui =64 d3 C2  d3 C2 5
7
1 0 90 0 u1 1
2 0 90 0 u2
3 0 90 d3 u3
4 a4 0 0 u4 The vertical height of the centre of mass of body 3
determines its potential energy, which can be writ-
ten as

Vg3 = m3 g(d3  d3 ) C2 (1)

Similarly, the location of the centre of mass of the

forearm in frame 4 is
2 3
a4
F4 = 4 0 5
0

On expressing this in frame 0 and taking its vertical

component, the potential energy of body 4 is

Vg4 = m4 g ½(S2 C3 C4  C2 S4 )(a4  a4 )  d3 C2  (2)

Fig. 4 Zero configuration and axes of rotation of the
arm
3 GRAVITY BALANCING CONDITIONS OF
THE ARM
A device is said to be gravity balanced if its total
potential energy remains invariant with configuration
[22]. This can be achieved if the centre of mass of the
device remains inertially fixed [23, 24] or if elastic
elements are added which compensate for variations
in the potential energy due to motion of the system
[25]. A number of designs have been reported for bal-
ancing mechanical manipulators using springs
directly connected to links [26, 27] or springs
through cable-pulley arrangements [28, 29]. In the
implementation proposed in this paper a cable-
pulley arrangement is used to achieve a zero-free-
length spring between the attachment points. To
The initial approach adopted in this paper is to use
two springs to gravity balance the system with the
parameters described and as shown in Fig. 5. As will
be shown later, a minimum of two springs is neces-
sary to achieve this feature.
The attachment points of the spring in the local
frame are P10 (h2Z ^ 0 ) and P14 (r2 X
^ 4 ), writing the
expression for spring length in frame 0
kP14  P10 k2 = (a4  r2 )2 + 2d3 (a4  r2 )S4
+ d32 + 2d3 h2 C2 + h22  2(a4  r2 ) (3)
h2 (S2 C3 C4  C2 S4 )
Therefore, the potential energy stored in the spring is

Vs1 = 12k1 (a4  r2 )2 + 2d3 (a4  r2 )S4 + d32
(4)
+ 2d3 h2 C2 + h22  2(a4  r2 )h2 (S2 C3 C4  C2 S4 )

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
1028 V N Dubey and S K Agrawal
Fig. 5 Spring attachment for gravity balancing of the
arm using two springs (refer to the notation in
the appendix for the shown parameters)
Combining all of the above energy terms, equations
(1), (2), and (4), the total energy of the system is
given by
V = Vg3 + Vg4 + Vs1
V =  ½m3 g(d3  d3 ) + m4 gd3 C2 + m4 g(a4  a4 )
 
(S2 C3 C4  C2 S4 ) + 12k1 (a4  r2 )2 + d32 + h22
+ 12k1 (2d3 h2 )C2  k1 (a4  r2 )h2 (S2 C3 C4  C2 S4 )
+ 12k1 (2d3 )(a4  r2 )S4
The gravity balancing conditions are obtained by
grouping the configuration-dependent terms together
and equating them to zero
V = fk1 d3 h2  ½m3 g(d3  d3 ) + m4 gd3 gC2
+ ½m4 g(a4  a4 )  k1 (a4  r2 )h2 (S2 C3 C4  C2 S4 ) (6)
 
+ k1 d3 (a4  r2 )S4 + 12k1 (a4  r2 )2 + d32 + h22
Two of the terms that arise from equation (6) are
k1 (a4  r2 )h2 = m4 g(a4  a4 )
k1 d3 h2 = m3 g(d3  d3 ) + m4 g d3
The coefficient of S4 in equation (6) is still not zero,
thus in order to cancel this term another spring is
required between P23(h3 Y ^ 3 ) and P24(  r3 X
^ 4 ). This
has been achieved by introducing spring (k2) as
shown in Fig. 5. The location of the attachment
points in frame 3 are
2 3 2 3
0 r3
P23 in F3 = 4 h3 5, and P24 in F3 = 3 T 4 4 0 5
0 0
2 3
(a4  r3 )C4
= 4 (a4  r3 )S4 5
0
The square of spring length for this attachment is
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
kP23  P24 k2 = (a4  r2 )2 C42 + ½(a4  r3 )S4  h3 2
(9)
= (a4  r3 )2 + h23  2h3 (a4  r3 )S4
The associated spring energy is
 
Vs2 = 12k2 (a4  r3 )2 + h23  2h3 (a4  r3 )S4 (10)
Combining equation (10) with equation (6) and
equating the coefficient of the trigonometric func-
tion S4 to zero provides the third design condition
k1 d3 (a4  r2 ) = k2 h3 (a4  r3 ) (11)
Equations (7), (8), and (11) give necessary and suffi-
cient conditions for gravity balancing of the skeleton
arm and parameters satisfying these conditions will
(5) eliminate the torque from each joint over the whole
range of arm motion.
4 DESIGN PARAMETER EVALUATION UNDER
GRAVITY BALANCING CONDITIONS
As seen from equations (7), (8), and (11) several
combinations of the design variables will satisfy the
gravity balancing conditions. r2 can be directly
obtained from equations (7) and (8) and this vari-
able only depends on the arm parameters
m4 g(a4  a4 )d3
r 2 = a4  (12)
m3 g(d3  d3 ) + m4 g d3
Equation (12) can be substituted in equation (7) to
obtain the stiffness of the spring k1 as given by
equation (13). Depending on the choice of the
(7)
attachment point h2, the stiffness of the spring is
given by (the rest being the fixed arm parameters)
(8)
m4 g(a4  a4 )
k1 = (13)
(a4  r2 ) h2
In the same way, the stiffness of the spring k2 can
be obtained by rearranging equation (11), which
results in equation (14). In this equation two para-
meters h3 and r3, are unknown. Therefore, one para-
meter has to be fixed in order to get the other. Due
to the practicality of implementation of the attach-
ment points, r3 can be taken to be at the same loca-
tion as r2. This may, however, be taken differently
from a parameter optimization point of view (see
section 5)
k1 d3 (a4  r2 )
k2 = (14)
(a4  r3 ) h3
 Study of an upper arm exoskeleton
Table 2 Parameters based on anthropomorphic data
for gravity balancing and estimation of joint
forces
Parameter
m3
m4
d3
a4
r2 (calculated)
r3 (assumed)
d3* (assumed half of arm length)
a4* (assumed half of arm length)
The initial design parameters for the human arm
based on anthropomorphic data of an average male
of height 170 cm and weight 75 kg are listed in
Table 2 [30]. These values are used for a gravity bal-
ancing design and for joint force evaluation.
These values give k1 = 147.15 N/m and
k2 = 163.5 N/m for h2 = 0.2 m and h3 = 0.15 m to
achieve the full gravity balancing condition of the
arm. However, different values of (h2, h3) can be
used to be compatible with standard values of
spring stiffness. Figure 6 shows a plot of spring
stiffness versus attachment point for the para-
meters listed in Table 2. These graphs are non-line-
ar and can serve as a reference for parameter
selection.
Prior to gravity balancing, moments at the
shoulder joint are purely due to the gravity forces
acting on the arm, which are given by equations
(15) and (16). Note that there is no moment about
the Z-axis (variation about u1 ) since it is parallel to
the gravity direction
Fig. 6 Spring stiffness variation with attachment
points for gravity balancing conditions
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
1029
Mx =  m4 g ½(a4  r2 )(S1 C2 C3 C4  C1 S3 C4 (15)
+ S1 S2 S4 ) + d3 S1 S2   m3 g(d3  d3 ) S1 S2
My =  m4 g ½(a4  r2 )(C1 C2 C3 C4 + S1 S3 C4
Value (16)
+ C1 S2 S4 ) + d3 C1 S2   m3 g(d3  d3 ) C1 S2
2 kg
2 kg
0.3 m On selecting the parameters using gravity balancing
0.25 m conditions, the arm becomes fully gravity balanced
0.1667 m
0.1 m for the full range of arm motion. Figures 7(a) and
0.15 m (b) show the joint moments prior to gravity balan-
0.125 m cing; the plots show variation of the joint moments
at the shoulder over the range 10°<u2 <170° and
60°<u3 <60°. The effect of variation of the elbow
joint (u4 ) is shown as separate plots for comparison.
The last plot (Fig. 7(c)) shows the resultant
moments along the two shoulder joints after gravity
balancing. It clearly shows that the arm is fully
balanced with zero moments in the plot for the full
range of the arm motion.
The gravity balancing, however, changes the joint
forces due to the addition of the spring. Prior to
gravity balancing, the only joint force was in the Z-
direction due to the weight of the arm. However, the
spring creates additional forces in the X and Y direc-
tions. These spring forces are given by equations
(17) to (19)
Fx = k1 ½(a4  r2 )(C1 C2 C3 C4 + S1 S3 C4 (17)
+ C1 S2 S4 ) + d3 C1 S2 
Fy = k1 ½(a4  r2 )(S1 C2 C3 C4  C1 S3 C4 + S1 S2 S4 ) (18)
+ d3 S1 S2 
Fz =  (m3 + m4 )g + k1 ½(h2  a4 + r2 )
(19)
(S2 C3 C4  C2 S4 )  d3 C2 
The initial joint forces (Fz) and the resultant force at
the shoulder after gravity balancing are shown in
Fig. 8. As can be seen the maximum resultant force
has changed from approximately 40 to 60 N in
Figs 8(a) and (b). Even though the joint moments
have been fully balanced, the increase in joint forces
may have a detrimental effect on the wearer, there-
fore these must be optimized for acceptable range
of moments and forces at the shoulder joint.
5 OPTIMIZATION FOR MINIMIZATION OF
JOINT TORQUES AND FORCES
In order to avoid excessive joint forces, the exoske-
leton parameters may be adjusted by optimization
with the goal to have a partially gravity balanced
system with minimum joint forces or have a fully
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
1030 V N Dubey and S K Agrawal

Fig. 7 Shoulder joint moments (a) and (b) before gravity balancing and (c) after gravity balancing

Fig. 8 Initial joint forces (a) and the resultant joint force (b) after gravity balancing, the three sur-
faces in each plot show the force variation due to elbow orientation at (0°, 60°, –45°)

gravity balanced system with minimal joint forces.
The objective function used for parameter optimi-
zation is minimization of the resultant force given
by equation (20) which involves the arm para-
meters. This may be used to validate the para-
meters for gravity balancing of the system (to that
described in section 4). The optimization of the
design parameters was achieved by applying the
fmincon function of MATLABä with lower and
upper bounds on the design parameters. This is a
gradient-based method designed to work with con-
tinuous functions such as the current formulation
for the resultant joint force at the shoulder
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
F= Fx2 + Fy2 + Fz2
In the following section initially a generalized opti-
mization is considered where all variables are free
to be chosen from their specified range and later a
constrained optimization is carried out keeping the
moments about the joint fully balanced i.e. a fully
gravity balanced arm.
5.1 Generalized parameter optimization
The three parameters listed in Table 3 were consid-
ered for the generalized optimization; no other con-
straints were applied in this case. The objective
function in this optimization was to minimize the
resultant joint forces (equation (20)). The para-
meters after optimization are also presented in
Table 3.
It can be seen that the optimized values of the
parameters are different from those obtained for
the fully balanced system and do not match with
the pair-value of k1 = 100 N/m and h2 = 0.3 m as
seen in Fig. 6. This means that the arm is only par-
tially balanced. This, however, results in the mini-
mization of the joint forces and it has been found
(20) that the maximum joint force (Fz) is almost back
to the original level when the arm was not
balanced.
The resulting unbalance in moments at the
shoulder joint is shown in Fig. 9, clearly this is the
condition for a partially gravity balanced arm
(compare this with Fig. 7). The plots show that
approximately 80 per cent of joint moments are

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
 Study of an upper arm exoskeleton
Table 3 Parameters and constraints for joint force
minimization and optimized parameters
Parameters Constraints Optimized parameters
k1 (N/m) 100\k1 \400 100
h2 (m) 0.1\h2 \0.4 0.1954
r2 (m) 0.05\r2 \0.15 0.15
balanced (estimated by comparing the maximum
value of the unbalanced moment in Figs 7 and 9).
However, if the moments are constrained to be
fully balanced the results of the optimization are
found to be different as discussed in the following
section.
5.2 Constrained optimization for a fully balanced
system with a minimum of joint forces
For this optimization the objective function still
remains the minimization of the resultant joint
forces (equation (20)), however, this is constrained
in such a way that full gravity balancing conditions
are satisfied. This was achieved by fixing r2 to
0.1667 m which satisfies the full gravity balancing
conditions (equation (12)). The upper and lower
bounds for the parameters used in the optimization
are listed in Table 4. This results in a different set of
optimal parameters, which is compatible with
Fig. 6, that satisfies the full gravity balancing condi-
tions. It has been found that the forces are the least
from all the previous cases.
The resulting moments at the shoulder joint are
found to be close to zero. Thus, the analysis pro-
vides a tool for the evaluation of design parameters
subject to different levels of moment and force bal-
ance via numerical optimization. The same tech-
nique can also be used to study the effect of
parameter variation on joint moments and joint
forces.
Fig. 9 Unbalanced moments due to the minimization of joint forces resulting in partial gravity
balancing, the three surfaces in each plot show the force variation due to elbow orientation
at (0°, 60°, –45°)
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
1031
Table 4 Parameters for force minimization of fully
balanced arm and optimized parameters
Parameters Constraints Optimized parameters
k1 (N/m) 100\k1 \400 100.2248
h2 (m) 0.1\h2 \0.4 0.2929
r2 (m) 0.1667 (fixed)
6 INITIAL DESIGN VALIDATION
In order to validate the proposed techniques a
simple test set-up was developed. It uses a cuff on
the forearm (Fig. 10(a) and a shoulder-mount
(Fig. 10(b)) for the spring attachment. Note that a
single spring balances the shoulder joint as evident
from equations (7) and (8), the third condition in
equation (11) simply balances the elbow joint. The
implementation uses a zero-free-length spring i.e. a
linear extension spring in which the force is propor-
tional to the length of the spring rather than to its
elongation, connected from the waist-belt passing
over a small pulley on the shoulder-mount through
the cable connected to the cuff, depicted in Fig. 11.
This allows the spring to be located remotely and
provides space for elongation. The cuff was pro-
duced using the rapid prototyping technique which
provides a special feature of self-orienting of the
attachment point depending on the location of the
arm. The cuff is also adjustable to fit different sizes
of arms.
Tests were conducted (after obtaining informed
consent) on a healthy subject with approximate
arm parameters as listed in Table 5. The subject
used in this experiment was asked to provide quali-
tative feedback based on ease of moving his arm
when wearing the device, thus precise measure-
ment of involuntary muscle contraction was
not considered in this experiment. For this user, it
was recorded that the person felt weightlessness
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
1032
Fig. 10 Adjustable forearm cuff (a) and shoulder-mount with straps (b) for gravity balancing test
Fig. 11 Experimental setup and verification of gravity
balancing of a natural arm
of his arm over a limited range (20°<u2 <70°,
50°<u3 <0°, u4 = 620°). The experiment used the
closest match springs available in the laboratory
(spring parameters are given in Table 5). To verify
the extent of gravity balancing of the arm under
these conditions these parameters were fed back to
the theoretical model, it can be seen in Fig. 12 that
the shoulder moments were relatively small for the
above specified joint angle ranges. It was also
found that the maximum moment of the unba-
lanced system was originally significantly higher
(6.5 Nm) and the maximum unbalanced moment
for the specified range of arm motion was 0.9 Nm,
whereas the maximum unbalanced moment for the
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
V N Dubey and S K Agrawal
smaller range was 0.75 Nm. This supports the idea
that the joint torque required to move the arm is
relatively small for a smaller range and as a result
the user felt that it was easy to move his arm when
he was wearing the device. This also supports the
idea that for a smaller range partial gravity balan-
cing could be easily implemented for the human
arm with careful selection of the design
parameters.
7 DISCUSSION AND CONCLUSIONS
One of the goals of this research was to present the
design criterion for cancelling the effect of gravity
and therefore eliminating or minimizing the need
for a power source in the upper arm exoskeleton
while minimizing the force transferred to the
joints. As evident from the design and implementa-
tion considerations it is clear that it may be very
difficult to accurately model the human arm for
gravity balancing and verify it experimentally.
There are various factors involved in this design
such as: the kinematic model of the mechanical
arm does not fully match with the human arm (as
this is based on a four-degrees-of-freedom system),
the joints are not accurately aligned with the
model, masses are spread throughout as opposed
to the idealized model-masses, and spring attach-
ments as per the model is not possible due to arm’s
girth. However, precise control of these parameters
is only required for a fully gravity balancing system;
a small deviation from the idealized condition may
still render a partially balanced system. It may be
useful to implement this idea for relieving the
actuators effort to lift the weight of the arm or even
for the possibility of a stand-alone passive support
device. It is also important to note that in this
implementation it has been assumed that the per-
son using this device is in an upright position,
 Study of an upper arm exoskeleton
Table 5 Parameters used for implementation of grav-
ity balancing on the natural arm
Spring/arm parameters
Wire diameter
Coil diameter
Free length
Maximum extension
Maximum load
Stiffness (k1)
Approximate upper arm mass (m1)
Approximate forearm mass (m2)
Approximate hand mass (m3)
Cuff mass (m4)
Cuff diameter
r2
h2
d3
a4
d3*
a4*
which is mostly the case in ADL; any change in the
base orientation may not provide the same gravity
balancing effect as expected from the theoretical
model. This has been the basis for all gravity balan-
cing systems but the advantage of using this idea
on a powered exoskeleton is that the actuators may
still be able to support the arm when the gravity
balancing is not effective. Therefore, the contribu-
tion of this research is to help select various design
parameters to achieve tunable gravity balancing
conditions for the upper arm.
In this paper gravity balancing conditions for the
upper arm were derived and demonstrated with the
help of a simple experimental set-up for the human
arm. Theoretically, only two springs are required to
achieve full gravity balancing conditions on
shoulder and elbow joints. However, a single spring
is sufficient to achieve balancing of the arm at the
shoulder joint. The locations of spring attachment
Fig. 12 Moment plots at the shoulder joint: (a) unbalanced system: (b) balanced system for the
full range; and (c) balanced system for a smaller range of the arm motion
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
1033
points were selected in such a way that the design
could be practically implemented on a shoulder cuff
[19]. Various conditions of moment and force unba-
Value
lance were examined and an optimization solution
0.001 143 m was presented to minimize the joint forces, which
0.0127 m
0.127 m may have a detrimental effect on the wearer. The
0.38 m proposed technique offers an evaluation tool for
33 N designing a partially or fully balanced system with
100 N/m
1.4 kg appropriate parameter selection resulting in mini-
1 kg mization of joint forces. The aim of this research
0.3 kg
0.32 kg was to explore this feature and conditions of gravity
0.15 m balancing for a human arm so that a lightweight
0.1 m
0.2 m
wearable upper arm support system could be devel-
0.3 m oped. With partial balancing, one can use small and
0.25 m low-power motors on an exoskeleton device for the
0.15 m
0.1 m arm movement. This has been experimentally veri-
fied in a simple set-up. This idea could be extended
for development of a stand-alone system for arm
support. The design shows the potential for devel-
oping a portable arm exoskeleton for training and
assistance.
FUNDING
The financial support for the first author was
partly provided by the Department for Business,
Innovation and Skills (UK) and a Global Research
Award from the Royal Academy of Engineering
(UK) which is gratefully acknowledged. This
allowed him to visit the Mechanical Systems
Laboratory of the University of Delaware (USA).
ACKNOWLEDGEMENTS
The support received from Xavier Velay, Martin
Mathias, Prashant Prashun, and Kevin Smith in
devising the experiment is thankfully acknowledged.
Ó Authors 2011
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
1034
REFERENCES
1 Pohlschmidt, M. and Meadowcroft, R. Muscle dis-
ease: the impact – incidence and prevalence of
neuromuscular conditions in the UK. Report by the
Muscular Dystrophy Campaign, London, UK, 2010.
2 Magermans, D. J., Chadwick, E. K., Veeger, H. E.,
and van der Helm, F. C. Requirements for upper
extremity motions during activities of daily living.
Clin. Biomech., 2005, 20(6), 591–599.
3 van Andel, C. J., Wolterbeek, N., Doorenbosch, C.
A., Veeger, D. H., and Harlaar, J. Complete 3D
kinematics of upper extremity functional tasks.
Gait Post., 2008, 27(1), 120–127.
4 Banala, S. K., Agrawal, S. K., and Scholz, J. Active
leg exoskeleton (ALEX) for gait rehabilitation of
motor-impaired patients. In Proceedings of the
Tenth IEEE International Conference on Rehabili-
tation robotics, Noordwijk, the Netherlands, 13–15
June 2007, pp. 401–407 (IEEE, Piscataway, New
Jersey).
5 Banala, S. K., Agrawal, S. K., Fattah, A.,
Krishnamoorthy, V., Hsu, W. L., Scholz, J., and
Rudolph, K. Gravity balancing leg orthosis and its
performance evaluation. IEEE Trans. Robot., 2006,
22(6), 1228–1239.
6 Reinkensmeyer, D. J., Hogan, N., Krebs, H. I.,
Lehman, S. L., and Lum, P. S. Rehabilitators,
robots, and guides: new tools for neurological reha-
bilitation. In Biomechanics and neural control of
movement (Eds J. Winters and P. E. Crago), 2000
(Springer, Berlin, Germany) pp. 516–533.
7 Lum, P. S., Reikensmeyer, D. J., and Lehman, S.
L. Robotic assist devices for bimanual physical
therapy: preliminary experiments. IEEE Trans.
Rehabil. Engng, 1993, 1, 185–191.
8 Aisen, M. L. and Krebs, H. I. The effect of robot-
assisted therapy and rehabilitative training on
motor recovery following stroke. Arch. Neurol.,
1997, 54(4), 443–446.
9 Perry, J. C. and Rosen, J. Design of a 7 degree-of-
freedom upper-limb powered exoskeleton. In Pro-
ceedings of the First IEEE /RAS EMBS International
Conference on Biomedical robotics and biomecha-
tronics, Pisa, Italy, 2006.
10 Sanchez, R. J., Liu, J., Rao, S., Shah, P., Smith, R.,
Rahman, T., Cramer, S. C., Bobrow, J. E., and
Reinkensmeyer, D. J. Automating arm movement
training following severe stroke: functional exer-
cises with quantitative feedback in a gravity-
reduced environment. IEEE Trans. Neural Syst.
Rehabil. Engng, 2006, 14(3), 378–389.
11 Rahman, T., Sample, W., Jayakumar, S., King, M.
M., Wee, J. Y., Seliktar, R., Alexander, M.,
Scavina, M., and Clark, A. Passive exoskeletons for
assisting limb movement. J. Rehabil. Res. Develop.,
2006, 43(5), 583–590.
12 Herder, J. L., Vrijlandt, N., Antonides, T.,
Cloosterman, M., and Mastenbroek, P. L. Principle
and design of a mobile arm support for people with
muscular weakness. J. Rehabil. Res. Develop., 2006,
43(5), 591–604.
Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine
Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015
V N Dubey and S K Agrawal
13 Carignan, C., Liszka, M., and Roderick, S. Design
of an exoskeleton with scapula motion for shoulder
rehabilitation. In Proceedings of the IEEE Interna-
tional Conference on Advanced robotics, Seattle,
Washington, 18–20 July 2005, pp. 524–531 (IEEE,
Piscataway, New Jersey).
14 Riener, R., Nef, T., and Colombo, G. Robot-aided
neurorehabilitation of the upper extremities. Med.
Biol. Engng Comput., 2005, 43, 2–10.
15 Chen, W., Chen, Q., Zhang, J., and Yu, S. Kine-
matics control for a 7-DOF cable-driven anthropo-
morphic arm. In Proceedings of the IEEE/RSJ
International Conference on Intelligent Robots and
Systems, Beijing, China, 2006.
16 Mustafa, S. K., Yang, G., Yeo, S. H., and Lin, W.
Optimal design of a bio-inspired anthropocentric
shoulder rehabilitator. Appl. Bionics Biomech.,
2006, 3(3), 199–208.
17 Agrawal, S. K. and Fattah, A. Design of an orthotic
device for full or partial gravity-balancing of a human
upper arm during motion. In Proceedings of the
IEEE/RSJ International Conference on Intelligent
Robots and Systems, Las Vegas, Nevada 2003 pp.
2841–2846.
18 Agrawal, S. K. and Fattah, A. Theory and design of
an orthotic device for full or partial gravity-
balancing of a human leg during motion. IEEE Trans.
Neural Syst. Rehabil. Engng, 2004, 12(2), 157–165.
19 Agrawal, S. K., Dubey, V. N., Gangloff, Jr, J. J.,
Brackbill, E., Mao, Y., and Sangwan, V. Design and
optimization of a cable-driven upper arm exoskele-
ton. J. Med. Devices, 2009, 3(3), 031004 (8 pages).
20 Mustafa, S. K., Yang, G., Yeo, S. H., and Lin, W.
Self-identification of the joint centre of a cable-
driven shoulder rehabilitator. In Proceedings of the
IEEE International Conference on Robotics and
automation, Roma, Italy, 10–14 April 2007, pp.
3767–3772 (IEEE, Piscataway, New Jersey).
21 Spong, M. W., Hutchinson, S., and Vidyasagar, M.
Robot modeling and control, 2006 (John Wiley &
Sons, USA).
22 Agrawal, S. K. and Fattah, A. Gravity-balancing of
special robotic manipulators. J. Mech. Mach. The-
ory, 2004, 39, 1331–1334.
23 Gokce, A. and Agrawal, S. K. Mass center of planar
mechanisms using auxiliary parallelograms. Trans.
ASME, J. Mech. Des., 1999, 121(1), 166–168.
24 Agrawal, S. K., Gardner, G., and Pledgie, S. Design
and fabrication of a gravity balanced planar
mechanism using auxiliary parallelograms. Trans.
ASME, J. Mech. Des., 2001, 123(4), 525–528.
25 Fattah, A. and Agrawal, S. K. Gravity-balancing
of classes of industrial robots. In Proceedings of
the IEEE Conference on Robotics and automation,
Orlando, Florida, 15–19 May 2006, pp. 2872–2877
(IEEE, Piscataway, New Jersey).
26 Nathan, U. and Kumar, V. Passive mechanical
gravity compensation for robot manipulators, In
Proceedings of the IEEE Conference on Robotics
and automation, Sacramento, California, 9–11
April 1991, 1536-1541 (IEEE, Piscataway, New
Jersey).
 Study of an upper arm exoskeleton 1035

27 Simionescu, I. and Ciupitu, L. The static balancing d3* centre of mass of the upper arm
of the industrial arms. Part I: continuous balancing. h2 height of attachment point of
J. Mech. Mach. Theory, 2000, 35, 1299–1311. spring k1 above shoulder
28 Streit, D. A. and Gilmore, B. J. Perfect spring equi- h3 attachment point of spring k2 on
librators for rotatable bodies. Trans. ASME, J. Mech. upper arm
Transm. Autom. Des., 1989, 111, 451–458.
k1 stiffness of the spring connected
29 Nathan, R. H. A constant force generation
mechanism. Trans. ASME, J. Mech. Transm. Autom. between the shoulder and forearm
Des., 1985, 107, 508–512. k2 stiffness of the spring connected
30 Anthropometric source book vol. 1: anthropometry between upper arm and forearm
for designers, NASA reference publication 1024, 1978 m3 mass of the upper arm
(Webb Associates, Yellow Springs, Ohio, USA). m4 mass of the forearm
P10, P14 attachment points for spring k1
APPENDIX P23, P24 attachment points for spring k2
r2 attachment point of spring k1
Notation from forearm end
r3 attachment point of spring k2
a4 length of the forearm on body 4
from forearm end
a4* centre of mass of the forearm
d3 length of the upper arm on body 3

Proc. IMechE Vol. 225 Part H: J. Engineering in Medicine

Downloaded from pih.sagepub.com at UNIVERSITE DE MONTREAL on June 24, 2015