IEEE Robotics & Automation Magazine - March 2020 - 58

Protocol
After the electrodes and markers were placed, subjects were
asked to perform functional and nonfunctional tasks. These
tasks were designed to simulate gestures typical of automotive
underbody scenarios (such as cladding, repairing, and bolt
tightening) and to represent generic upper-limb movements
that may happen in ordinary working conditions (such as
reaching for tools or maintaining awkward arm postures for a
relatively long time).
Functional Tests
During the functional tests, participants were asked to stand
still under a horizontal panel and perform overhead tasks.
The panel depicted a cross inside a circle, and its position was
adjusted based on the user's height to allow subjects to perform the tasks with the dominant arm elevated (shoulder
flexed at ~90° and elbow flexed at ~90°).
Four trials were carried out, in which each subject was
asked to retrace the AP part of the cross, the internal medial-
lateral (i-ML) part of the cross, the external medial-lateral
(e-ML) part of the cross, and the circumference (Figure 3).
Each trial consisted of 20 repetitions of the gesture.
Nonfunctional Tests
Nonfunctional tests included three sets of tasks: repetitive
reaching movements, static tasks, and quasistatic tasks that
are determinants of the complex functional work tasks
analyzed.
1) Repetitive reaching tests. Each participant was requested to
perform reaching movements, starting from a relaxed arm
position to a forward-reaching arm position (Figure 3).

External
Frontal
Internal
e-ML
AP
i-ML

Repetitive Reaching

Functional

Static

Quasistatic

Figure 3. A schematic of the tasks. The image on the left shows
functional movements while nonfunctional movements are
shown on the right. C: circumference.

IEEE ROBOTICS & AUTOMATION MAGAZINE

MARCH 2020

The arm was considered straight when the shoulder
reached 90° in the flexion direction and the elbow was
fully extended (0°). A wooden bar was placed vertically in
front of the user, with a sign at the height of the user's right
acromion. The sign on the bar was used as a target point
that participants had to approach more closely each time
they repeated the reaching movement. The distance
between the vertical bar and the subject was adapted to be
equal to the user's arm length. Subjects were asked to perform the task from a sitting position and keep their backs
straight. Reaching trials were executed three times: the
first with the bar positioned in the external workspace, i.e.,
45° medial angle of the arm (external reaching), the second with the bar positioned in front of the right acromion
of the user (frontal reaching), and the third with the bar
positioned in the internal workspace, i.e., −45° medial
angle of the arm (internal reaching). In each condition,
subjects were requested to repeat the movements repetitively 12 times at a self-selected pace.
2) Static test. Each subject was asked to stand still and keep
the right arm in a flexed position (90°), with the elbow fully
extended, for 60 s (Figure 3).
3) Quasistatic test. Each subject was asked to retrace four
arrays of 20 sinewaves on a vertical transparent panel.
The panel was vertically regulated to enable alignment
between the user's shoulder and the sinusoidal trace.
Movements were executed with the arms straight (i.e.,
shoulder flexed at ~90° and elbow fully extended) and the
person standing still (Figure 3). EMG and kinematics
data related to the tracking movements of the first and
last arrays were recorded, to compare the performance at
the beginning and end of the task.
Experimental Conditions
Each task was executed under two conditions: without and
with the exoskeleton (these conditions are referred in the
text as FREE and EXO, respectively). The order of FREE
and EXO conditions was randomized across participants.
Before starting the EXO trials, the level of assistance was
selected as the minimum that allowed compensation for
about 50% of the estimated gravitational torque of the user's
upper-limb [13] (Table 1). The assistive torque is tuned to
compensate for about 50% of the estimated gravitational
torque weighing on the upper limb (assessed based on
weight and height).
Notably, an additional reflective marker was positioned on
the proto-MATE back frame in EXO trials. Its purpose was to
assess any relative translational movement between the trunk
and the frame of the proto-MATE.
Data Analysis
EMG data were sampled at 1,000 Hz. Raw EMG signals were
bandpass filtered with a zero-lag, second-order Butterworth
filter (frequency range 20-450 Hz) to remove movement artifacts and high-frequency noise. A zero-lag, second-order infinite impulse response notch filter at 50 Hz was applied to

IEEE Robotics & Automation Magazine - March 2020

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2020