i3 - May/June 2018 - 14
By Murray Slovick
A T EC H TO WATC H
Learning Robots Offer
a Glimpse of the Future
table tennisplaying robot.
But what if I then suggested that robots
also can learn on the fly? That they can
do things you want without having to be
told? Does that surprise you?
It shouldn't. Deep learning, the basis
for artificial intelligence (AI), is enabling
robots to learn much more quickly and
effectively, perform tasks autonomously
and find their way around unfamiliar
For example, by leveraging AI and
machine learning, the Aeolus Robot,
unveiled in prototype form at CES
2018, can recognize thousands of
objects and learn its owners' preferences, such as their favorite drink. It
can also pick up toys off the floor and
put them away in their storage areas
and use a vacuum to clean floors.
Aeolus can also determine things like a
person's change in posture, which
might occur if an elderly person is
falling or experiencing a seizure-and
then call for help. The robot is also
integrated with voice-command systems like Amazon Alexa and Google
Home. Named after the Greek god of
the wind, the Aeolus Robot is roughly
the height and weight of a 12-year old
and is expected to be available for purchase in the fourth quarter of 2018.
Robots have always been very effective
for precise, repetitive work, but for the
most part they were traditionally used
only in carefully controlled settings.
However, robots are now moving out of
the secure areas of industrial plants and
collaborating directly alongside human
co-workers. Known as "collaborative
robots," these machines work alongside
humans on light manufacturing lines,
and if a human might need to enter its
workspace, can be trusted not to injure
the interloper. Many of these smaller,
lighter robots can fit on a workbench or
counter. Their robotic arms have actuated joints that can be fitted with any
number of different tools.
An example of how robots can interact in real time with humans was demonstrated at CES 2018 by Omron, whose
AI-powered table tennis-playing robot
Forephus took on human opponents.
Forephus has a companion arm that can
serve up balls in the air and a five-axis
motor system lets it swing the paddle.
Omron says the machine can detect the
ball's speed and rotation up to 80 times
per second, which allows it to predict
the ball's trajectory. An AI-assisted
motion controller tells the machine how
and where to hit the ball. Omron says
the robot can adapt its playing style and
approach based on the skills and technique of its opponents.
What's Next in Robotics?
● Robots are becoming smaller, lighter,
cheaper and more flexible to use.
Thanks to the lower price of sensors
and the sheer volume of robots being
produced, affordability is a trend
expected to continue.
● Robots will be leveraging the cloud to
do processing and exchange information
with other robots, accelerating the
learning process, and allowing a robot to
benefit from the efforts of others.
● Advances in big data will help the
robotics community deal with the massive amount of information generated by
● New deep learning techniques and
algorithms will enable robots to learn
much more quickly and effectively than
in the past.
It isn't hard to see why some believe
the timing is ripe for robotic home companions and playmates.
I T I S I N N O VAT I O N
f I said that modern robots can detect obstacles,
navigate around them and sensitively handle objects
with their grippers, most people would agree.