636 16 AUGUST 2019 • VOL 365 ISSUE 6454 sciencemag.org SCIENCEILLUSTRATION: V. ALTOUNIAN/SCIENCE; (MOTION CAPTURE DATA) HARVARD BIODESIGN LABINSIGHTS
ROBOTICSWitnessing a wearables transition
Assistive robots must mimic human dynamics and move toward neural-interface control
By José L. Pons1, 2, 3W
earable robots, such as exoskel-
etons and soft exosuits, can aug-
ment normal function or serve
as prostheses for missing limbs.
In both cases, they extend, com-
plement, substitute, or enhance
human functions and capability and can
empower or replace human limbs. Cognitive
and physical interactions between human
and robot are key for these robots to seam-
lessly deliver assistance when required. The
physical interaction between a robot and
its wearer generates forces to overcome
the wearer’s physical limits, and cognitive
interactions allow the wearer to guide and
control the robot at all times. On page 668
of this issue, Kim et al. ( 1 ) report on a soft
exosuit that switches assistance profiles for
different physical interaction strategies—in
this case, walking versus running—through
a versatile cognitive interaction in whichrobots and exosuits can compromise these
dynamics. Anatomical and artificial joints
can become misaligned, the robot or exosuit
exerts forces that oppose rather than assist
the movements intended by the wearer, or
the inertial characteristics of body segments
can be modified in unfamiliar ways by the
added mass (for example, the movements can
feel sluggish or top-heavy).
In spite of these altered neuromechan-
ics, the effect of wearing a wearable robot
on human motor coordination is small ( 3 ),
which supports the use of these technolo-
gies to assist human movement. Moreno et
al. ( 3 ) observed that the dimensionality of
motor control when using a wearable ro-
bot—how many independent components
in the muscle space of wearers are required
to explain how muscles are activated—did
not change for different walking speeds or
amount of assistive force. They also showed
that the timing impulsive motor structure—
that is, the gait cycle phase and timing at
which muscle groups are recruited during
locomotion—was maintained along a broad
range of walking conditions (flat surfaces as
well as uneven terrain). These findings sup-
port the value of robotic-based strategies,
both with rigid wearable robots and withPERSPECTIVES
(^1) Legs + Walking AbilityLab, Shirley Ryan AbilityLab, Chicago,
IL, USA.^2 Department of Physical Medicine and Rehabilitation,
Department of Biomedical Engineering, and Department of
Mechanical Engineering, Northwestern University, Chicago,
IL, USA.^3 Neural Rehabilitation Group, Cajal Institute, Spanish
National Research Council, Madrid, Spain. Email: [email protected]
algorithms accurately determine and detect
the wearer’s gait.
Early wearable robots and exosuits were
used to assist healthy wearers by reducing
the metabolic cost of walking or running
and, in the health care arena, to assist physi-
cal therapy for individuals with neurological
disorders. In health care, they promote motor
recovery or assist locomotion through assis-
tive force to the patients, but how can assis-
tive forces be efficiently delivered? Kim et al.
showed that by assisting hip extension at key
walking and running phases, metabolic cost
is reduced (see the figure). The soft exosuit
ke pt added mass as close to the wearer as
possible, which reduced the number of joint
movements restricted by the robot. Assist-
ing hip extension can aid both walking and
running, and a robust online classification al-
gorithm helped the suit switch between walk-
ing and running assistive modes.
Previous studies have highlighted the im-
pact of neuromechanics and dynamic prop-
erties of human limbs in achieving efficient
locomotion and to adapt locomotion modes
over a wide speed range ( 2 ). To a large extent,
efficient locomotion in humans is based on
exploiting our passive lower-extremity dy-
namics, but several problems with wearable