IEEE Robotics & Automation Magazine - March 2016 - 20

from the guest editors

Bioinspired Motor Control
for Articulated Robots
By Nicola Vitiello, Auke J. Ijspeert, and Stefan Schaal

he pervasive deployment of
robot companions that can actually help humans to improve
our quality of life is becoming a
progressively more urgent need for a
number of domains of our lives, including personal, urban, environmental, and
social aspects.
While the robotics community is
working toward this ambitious goal, it is
clear that robots are still most deployable when they operate in structured
environments rather than in unpredictable, real-life scenarios.
Evolution has developed many of the
required solutions to cope with unstructured environments: living organisms
can perform many tasks that robots cannot perform yet, like the simultaneous
execution and coordination of multiple
movements, the effective exploration of
the surrounding
physical environment, and
Evolution has
the adaptation
developed many
to novel situations. Living orof the required
ganisms are able
solutions to cope
to accomplish
these tasks dewith unstructured
spite fairly long
environments.
delays in nervous information processing
(as compared with silicon-based computational systems), stringent physical
constraints, and energy rationing. In
other words, biological systems have
Digital Object Identifier 10.1109/MRA.2015.2511680
Date of publication: 8 March 2016

IEEE ROBOTICS & AUTOMATION MAGAZINE

overcome major challenges that remain
key obstacles in current robots.
Unveiling the secrets of living beings
and translating them into effective new
design principles for more effective robotic devices and controllers is the strategy at the core of the emerging field of
bioinspired robotics, with the hope to
overcome the current robotics bottleneck. In addition, robots can also serve
as scientific tools to test hypotheses
about animal motor control. That is,
there can be interesting bidirectional interactions between biology and robotics.
This special issue of IEEE Robotics
and Automation Magazine (edited by
Nicola Vitiello, Auke J. Ijspeert, and Stefan Schaal) collects recent works on the
development and experimental validation of bioinspired control strategies for
articulated robots. Selected articles cover
a large range of robotics applications
from humanoid robotics to animaloids
and wearable assistive technologies, and
all have a strong focus on the experimentation with real robotic systems.
Overall, this special issue comprises
seven articles, reporting on different
control systems: three systems are for
humanoid robots, one is for a snake-like
robot, two are for legged robots, and one
is for a wearable assistive robot.
The first article, "Good Posture,
Good Balance," approaches the control
of posture of humanoid bipedal robots,
with emphasis on postural balance
against unforeseen external disturbances, definitely one of the most challenging features to address for humanoid
robots. The article carries out a comparison between two strategies for postural

march 2016

balance: one strategy is engineering inspired and model based, and the second
strategy is bioinspired and relies on the
disturbance estimation and compensation concept. The comparison is made
both from a qualitative and quantitative
viewpoint with respect to the different
values of sampling rate and time delays
in the feedback loop. This article can
also be considered a relevant contribution in the field of neurorobotics: the
humanoid torque-controlled platform
adopted in this article for experimental
activities has been used as a test bed for
providing insights into the principles
underpinning human motor control.
The second article, "Practice Makes
Perfect," presents an innovative bioinspired control methodology for quadrupedal robots. The proposed architecture
is based on a parameterized, modelbased and state-feedback controller,
which integrates a generic set of motion
primitives encoded through cost functions. Optimization of the cost functions is addressed by exploiting motor
learning principles observed in nature:
by means of a trial-and-error process,
the optimal choice of parameters is selected by executing a slightly different
sequence of the same motion skill. The
optimization process was carried out in
simulation thanks to a reliable model of
the quadruped robot. The use of a dynamic model was necessary to speed up
the optimization process and prevent
the robot from possible damages. The
overall approach was successfully validated with StarlETH: a quadruped
robot equipped with a compliant actuation system. Results showed that the

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