IEEE Robotics & Automation Magazine - March 2016 - 90

Table 2 shows the seven remaining participants' features
and trials ordering. For the assistive order nomenclature, see
the "Experimental Protocol" section.
Experimental Setup
During the experiments, assistance was provided to the hips
through the APO device worn by the subjects. A person
wearing the full experimental setup is shown in
The DLMP-based approach
Figure 4. In the figure, the
following elements are
provided larger flexion
visible: APO, WSA, and
Oxycon Mobile.
torque than the
To evaluate the optimal assistance level and
NLMP-based approach.
the difference between
both assistive methods
with respect to the transparent mode (TM), the following data were collected during
the experiments.
● The position, velocity, and acceleration of the body segments from the inertial measurement units (IMUs) of
the WSA (Figure 4). The IMUs are wireless, batterypowered, and enclosed into a small casing appropriate
for attachment on body segments. The IMUs were

Oxycon Mobile
APO

IMUs

Shoes Instrumented
with Sensitive Insoles
Figure 4. The frontal view of a person wearing the experimental
setup. The following components are visible: Oxycon Mobile,
APO, the sensory system with IMUs, and the shoes instrumented
with custom pressure-sensitive insoles. For more details on the
latter two, see [30]. (Photos courtesy of Virginia Ruiz Garate.)

IEEE ROBOTICS & AUTOMATION MAGAZINE

march 2016

placed on both shoes, shanks, and tights, and one was
placed on the back.
● Ground reaction forces from sensorized insoles of the
WSA (Figure 4).
● Oxygen uptake rate O2 from a portable Oxycon Mobile
(Carefusion, St. Albans, the United Kingdom) with a sampling time of 5 s (Figure 4).
Moreover, the following data were computed in real time
and recorded for offline analyses: gait phase and cadence of
each leg from the oscillators, reference joint torques generated
by the controller, and actual hip torque and power provided
by the APO.
Experimental Protocol
Several subcases were evaluated depending on the type and
level of provided assistance. Assistance was provided with
two different methods: through direct torques using DLMP
or through NLMP and the use of the musculoskeletal model.
Before starting the experiment, each subject underwent a
phase of familiarization that lasted around 10 min. During
this phase, the subject walked on a treadmill at a self-selected
speed, while assistive torques were provided through the pelvis module using both methods. This phase aimed at establishing an estimate of the percentage of assistance to be
provided to have similar torque magnitudes for both methods
and across subjects. In any case, a saturation of 15 Nm was
applied for safety reasons.
Afterward, all subjects underwent a series of 6-min walking tests (WTs) with trials under the following conditions:
● No APO (NO)-ground-level walking without wearing the
orthosis. It served as a basis for the kinematics and dynamics comparison and to evaluate the natural cadence and oxygen cost during normal walking.
● APO in TM-ground-level walking wearing the APO. The
APO was controlled in TM, i.e., rendering zero impedance.
This condition served to evaluate the influence of wearing
the pelvis module on the natural kinematics, dynamics, cadence, and oxygen cost.
● Low assistance (LA)-ground-level walking wearing the
APO while providing assistance. This condition was applied twice: a walking trial while providing assistance directly by outputting torques [low torque-based
assistance (LTA)] and a walking trial while providing assistance through the musculoskeletal model [low stimulation-based assistance (LSA)]. During LA, assistive peak
torques were in the range between 2 and 4 Nm.
● High assistance (HA)-ground-level walking wearing the
APO while providing assistance. This condition was applied
twice: a walking trial while providing assistance directly by
outputting torques [high torque-based assistance (HTA)]
and a walking trial while providing assistance through the
musculoskeletal model [high stimulation-based assistance (HSA)]. During HA, assistive torques were in the
range between 4 and 6 Nm.
Importantly, subjects performed these trials by walking in
a free environment, i.e., not on a treadmill. Specifically, they

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2016