IEEE Robotics & Automation Magazine - March 2015 - 101

and roller pulls the tendon from the spool. A passive brake
mechanism, also shown in Figure 6, was added to the actuation units [17] to increase the energy efficiency of the entire
system. The detailed design and performance of the actuation module, including the passive brake mechanism and
slack prevention mechanism, can be found in previous publications [17], [18].
Underactuation Mechanism for
Soft Tendon Routing System
To achieve stable grasping using fewer actuators than the
number of fingers, an underactuation mechanism for adaptation is required. Stable grasping requires the index and middle
fingers to touch the object, and the contact forces of the two
fingers should be well distributed. If only one of the two fingers touches the object or one with much more force than the
other, the object will rotate and the grasp will fail.
The Exo-Glove uses a mechanism developed for adaptive
grasping in underactuated soft wearable robots [13]. The
basic concept of the mechanism is the same as that of a conventional and differential mechanism, i.e., distribution of the
pulled length of the tendon to multiple fingers. To implement
the mechanism, we installed U-shaped tubes at the tips of the
index and middle fingers and between the fingers. The tendons used for finger flexion were then passed through the
tubes, as shown in Figure 7. To flex the fingers, the motor
winds the spool, and the total length of the tendon from the
spool to the fingers is reduced. The shortened length and the
tension of the tendon are then distributed to both fingers
because the path of the tendon, which is formed by the tubes,
does not restrict the tendon. The mechanism, therefore, also
enables adaptation when the glove is in contact with an
uneven surface. The U-shaped tubes were used because a
mechanical pulley is unsuitable for a soft wearable robot. The
tubes act as a pulley, and the tendons can freely move through
the tubes, possibly changing its direction.
Figure 7 shows the movement of the tendon in our proposed mechanism. Figure 7(a) shows the index and middle
fingers moving identically. Figure 7(b) shows the index finger
fixed by the environment while the middle finger is flexed.
The mechanism can be extended to actuate more fingers at
the same time.
Control and Command
Considering that the expected users of the Exo-Glove are
people with some type of disability, the input command for
the control has been made as simple as possible. In particular,
admittance control is used. This control makes the tendon act
as a virtual compliant spring, with the single input command
being used to determine the pulled length of the virtual
spring. Moreover, the compliance characteristic of the admittance control guarantees the safety of the device.
Regarding the control input, various types of human-
robot interfaces, such as electromyography (EMG), electroencephalography, and analog switches for detecting body
motion could be used. Here, wrist motion has been consid-

(a)

(b)

Figure 7. The tendon movement in the Exo-Glove adaptive
mechanism. (a) The two fingers move identically. (b) The middle
finger is fixed while the index finger moves.

ered for the following two reasons. First, wrist motion is easy
to detect and is reliable. For example, a simple bend sensor
can detect wrist motion with high reliability. Second, people
with a paralyzed hand are generally familiar with performing
wrist motion to induce finger motion to grasp an object.
Wrist motion is connected to finger motion by a mechanism
referred to as the tenodesis effect, which is the flexing of the
finger when the wrist is
extended. It should be
noted that the use of wrist
After the experiment, the
motion as control input
eliminates one degree of
subject was very satisfied
freedom from the movement of the hand, thereby
that he could grasp various
restricting the position in
which the object can be
objects by himself.
grabbed. This issue can,
however, be solved by
adding another combinable control input. Nevertheless, the
simplicity of wrist motion makes it a good candidate for the
control input for assisting the disabled. Here, a resistor bend
sensor installed at the back of the wrist is used to measure the
angle of the wrist.
Performance of the Exo-Glove
We experimentally investigated the grasp performance of the
Exo-Glove by testing the device with two male subjects, one a
healthy 30-year-old and the other a disabled 41-year-old who
had sustained traumatic injury to the sixth cervical spine six
months earlier, resulting in tetraplegia. The disabled subject
could neither flex nor extend his fingers but could move his
elbow and shoulder. After measuring the Exo-Glove kinematics, pinching force, and wrap grasping force with the healthy
subject, we observed both subjects as they attempted to grasp
variously shaped objects while wearing the device. All procedures were approved by the Institutional Review Board of
Seoul National University (IRB No. 1401/001-019).
march 2015

IEEE ROBOTICS & AUTOMATION MAGAZINE

101

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2015