Boston
Dynamics
has
shared
another
look
at
the
latest
version
of
its
humanoid
robot,
Atlas.
Earlier
videos
of
the
all-electric
robot
demonstrated
its
unique
range
of
motions,
but
this
time
Atlas
is
shown
using
machine
learning
and
its
upgraded
sensors
to
perform
sorting
tasks
in
a
simulated
factory
environment.
The
task
of
moving
engine
covers
“between
supplier
containers
and
a
mobile
sequencing
dolly”
isn’t
especially
exciting,
but
it
demonstrates
several
of
the
new
robot’s
capabilities
working
together.
After
being
provided
with
only
a
“list
of
bin
locations
to
move
parts
between,”
Atlas
uses
various
sensors
and
machine
learning
models
to
determine
the
location
of
bins
and
how
it
needs
to
manipulate
its
body,
arms,
and
three-fingered
hands
to
grasp
and
relocate
the
parts
inside
them.
The
new
Atlas
is
capable
of
completing
the
assigned
task
and
reacting
to
environmental
feedback
in
real
time.
At
one
point
in
the
video
a
part
is
positioned
a
little
too
high
to
easily
slide
into
one
one
of
the
sequencing
dolly’s
compartments.
When
the
robot
meets
resistance,
it
removes
the
part
and
then
re-evaluates
its
orientation
and
position
before
making
an
adjustment
and
trying
again
successfully.
As
is
also
evident
by
the
“Fully
Autonomous”
watermark
on
the
video
the
entire
time,
Boston
Dynamics
is
trying
to
emphasize
that
its
latest
humanoid
robot
can
perform
complex
tasks
with
minimal
human
intervention.
A
few
weeks
ago
it
was
revealed
that
Tesla’s
Optimus
humanoid
robots
that
were
mingling
with
guests
and
serving
drinks
during
the
company’s Cybercab
reveal were
mostly
being
controlled
remotely
by
human
operators.
As
with
its
other
robots
like
the
four-legged
Spot
and
the
one-armed
Stretch,
the
new
Atlas
may
one
day
be
more
than
just
a
platform
for
Boston
Dynamics
to
demonstrate
its
latest
and
greatest
technology.
But
that’s
assuming
its
advanced
capabilities
don’t
come
with
an
outrageous
price
tag.
Original author: Andrew Liszewski
Comments