Research and Challenges in Commercializing Robots

Transcription

Research and Challenges in Commercializing Robots
Research
and
Challenges
in
Commercializing
Robots
Eduardo
Torres‐Jara
Assistant
Professor
Robo<cs
Engineering
Worcester
Polytechnic
Ins<tute
Outline
•  State
of
the
art
•  Sensi<ve
Manipula<on
•  Commercializa<on
Robots
•  What
are
our
expecta<ons
of
a
robot?
–  Usually
a
humanoid
•  with
capabili<es
beyond
humans’
•  capable
of
handling
any
task
–  We
could
hire
or
buy
a
robo<c
maid.
–  They
could
take
over
the
world.
Where
are
we?
• 
Humanoids
‐ Asimo
‐ Advanced
control.
‐ Precision
mechanics.
•  Autonomous
Naviga<on
‐
Advanced
sensors.
•  Teleoperated
‐
UAVs,
Soujouner,
Packbot
Where
are
we?
•  Autonomous
Manipula<on
•  Other
PlaRorms
‐  PR2.
Willow
Garage
‐  Big
Dog.
Boston
Dynamics
Ar<ficial
Intelligence
•  Deep
blue
–  Chess
computer
•  Watson
–  Plays
Jeopardy
Ready
to
take
over
the
world?
What
is
happening?
•  Ar<ficial
Intelligence
–  We
are
not
working
in
the
right
aspects.
•  Robots
that
play
basketball
?
–  We
do
not
have
the
right
“soYware.”
•  It
is
believed
that
the
existent
hardware
with
more
complex
so2ware
can
solve
the
current
problems.
The
case
of
manipula<on
•  Remote
sensing
(for
example
vision)
or
a
CAD
provides
a
model
of
the
environment.
•  SoYware
assumes
everything
is
known
a
priori.
–  The
robot
only
needs
to
know
where
to
move
the
arm.
–  Focus
on
planning
the
posi<on
of
the
end
effector.
•  Rigid
robo<c
arms
–  The
posi<on
of
the
arm
is
controlled
with
high
accuracy.
–  The
control
gets
rid
of
the
dynamical
effects.
In
prac<ce
•  Automa<c
mapping
of
the
environment
does
not
work
with
the
required
precision.
–  Even
with
current
technologies.
•  Engineering
of
the
environment
is
required
–  Expensive.
About
10
<mes
the
cost
of
the
robot.
•  It
is
dangerous
and
no
flexible.
–  Any
change
requires
reprogramming.
Example:
Picking
up
an
object
Problems
• 
The
robot
is
dangerous
because
of
its
necessary
s<ffness.
• 
Coming
in
contact
with
an
object
is
not
gentle.
• 
A
different
model
for
each
object
is
needed.
• 
In
structure
environments,
ok.
But
what
about
your
kitchen?
•  We
need
a
different
approach
What
could
we
do?
•  Animals
and
humans
cannot
posi<on
their
limbs
as
precisely
as
a
robot
but
are
much
more
dexterous.
•  Have
a
lot
of
sensors.
–  1000
sensing
points
in
your
finger
<p.
•  A
possible
solu<on:
–  Robots
need
to
sense
more
their
environments.
Let’s
consider
naviga<on
•  Worked
because
they
have
a
beeer
sensor
(LIDAR)
•  New
algorithms
and
models
to
deal
with
rich
sensing
capabili<es.
My
approach
to
manipula<on
•  Sensi<ve
Manipula<on
•  The
basic
idea:
feel
the
environment.
•  It
requires:
–  gentle
interac<on
with
environment
–  feeling
the
contact
•  Changes
at
the
conceptual
and
hardware
level.
The
plaRorm
OBRERO
Tac<le
Sensors
• 
Existent
technology
is
not
adequate
• 
Designed
biological
inspired
sensors
–  Ridges
• 
Dome
shaped
–  Deformable
• 
Sensors
favor
compliance
over
spa<al
resolu<on
Sensi<ve
Manipula<on
•  The
robot
feels
its
way
around
the
object
•  No
models
are
needed
•  It
is
not
dangerous
Precision
Sensi<ve
Manipula<on
•  GoBot
–  General
purpose
robot.
–  It
can
manipulate
GO
stones.
–  Uses
sensi<vity,
no
precise
posi<oning
Is
it
ready
for
the
market?
Possible
applica<ons
•  Sensi<ve
Manipula<on
– 
– 
– 
– 
Medical/Surgery
robots
Collec<ng
samples
in
space.
Helping
the
astronauts
Domes<c
chores
Any
applica<on
that
requires
grabbing
objects
•  Tac<le
Sensors
–  Sensing
in
seats
•  Massage
chair
•  Car
seats
–  Human
prostheses
–  Car
Wheels
–  New
space
suits
Feasibility
•  Sensi<ve
Manipula<on
and
Tac<le
Sensors
–  Brainstormed
about
the
applica<ons
with
3
different
teams.
•  Lots
at
the
beginning.
None
aYer
the
analysis.
–  The
typical
outcome:
•  There
is
not
market
for
it.
•  Ques<oning
the
technology.
–  It
is
not
ready.
–  It
is
too
complex.
–  It
is
too
simple.
Should
we
give
up?
Innova<on
gap
Science
University
Start
up
Technology
(10%)
Commercial
Applica<ons
Company
Let’s
consider
the
LASER
Optoelectronics
CD
,
CD‐ROM,
fiber
op<c
comm.
Medicine
Ophthalmology,
Dermatology
Measurement
&
instrumenta7on
Mining
and
tunnel
surveying,
Measurement
of
workpiece
surfaces
Analysis
Manufacturing
technology
Culng,
Welding,
Surface
treatment
etc.
Research
Laser
fusion,
Diagnos<cs,
Measurement
of
earth‐moon
distance,
robots.
?
Scien<st
hep://www.ilt.fraunhofer.de/eng/100049.html
Market
We
would
have
missed
the
benefits
of
the
laser
because
the
industries
did
not
exist.
Considera<ons
to
bring
robots
to
market
•  Most
likely
the
most
sophis<cated
humanoid
robot
will
not
be
the
first
robo<c
product.
–  Too
much
is
expected
from
a
humanoid
robot.
•  There
are
markets
for
robo<cs
technologies.
–  Manipulators
that
dexterously
interact
with
the
environment
are
needed:
Bomb
disposal,
Medical
robots,
Rehabilita<on.
–  Tac<le
sensors
in
car
seats.
–  Aided
Naviga<on.
•  Fully
autonomous
machines
–  No
well
received
by
the
users
because
they
want
control.
–  Semi‐autonomy
is
accepted.
Car
cruise
control.
–  The
expecta<on
of
autonomy
in
a
robot
is
high.
Considera<ons
to
bring
manipulators
to
market
•  Industrial
manipulators
–  Complemented
by
new
types
of
manipulators.
•  High
precision/s<ff
robots
need
feeders.
•  New
type
of
robots
can
mount
the
parts.
–  Explore
markets
beyond
the
car
industry
model.
•  The
current
expecta<on
of
the
robot
makes
it
difficult
to
innovate.
•  Bring
the
cost
down
–  Complex
technology
can
be
introduced
to
the
consumer
market
at
the
right
price.
For
ex.
:Kinect.
–  Exchange
payload
capacity
for
cost.
–  Different
fabrica<on
technologies.
What
about
the
research
contribu<on?
•  An
example
•  In
manipulator
joints
•  Cables,
actuators,
sensors
that
stretch
are
useful.
•  “Dangerous”
university
research
•  Stretchable
circuits
•  SoY
robo<cs
Considera<ons
to
bring
manipulators
to
market
•  Programming
interfaces
–  More
flexible
interfaces
than
a
joys<ck
are
needed.
Conclusions
•  The
innova<on
gap
exists
in
robo<cs
–  Robo<c
entrepreneurs
need
to
be
aware
of
the
work
required
to
bring
a
product
to
market.
•  Robo<cs
technologies
are
easier
to
introduce
to
market
than
complete
robots.
–  Less
expecta<ons
•  Making
robots
more
sensi<ve
to
their
environment
enable
different
types
of
applica<ons.