WEBVTT

99:59:59.999 --> 99:59:59.999
So robots.

99:59:59.999 --> 99:59:59.999
Robots can be programmed
to do the same task millions of times

99:59:59.999 --> 99:59:59.999
with minimal error, something
very difficult for us, right?

99:59:59.999 --> 99:59:59.999
And it can be very impressive
to watch them at work.

99:59:59.999 --> 99:59:59.999
Look at them.

99:59:59.999 --> 99:59:59.999
I could watch them for hours.

99:59:59.999 --> 99:59:59.999
No?

99:59:59.999 --> 99:59:59.999
But what is less impressive
that if you take this robot

99:59:59.999 --> 99:59:59.999
out of the factories

99:59:59.999 --> 99:59:59.999
where the environments are not
perfectly known and measured like here,

99:59:59.999 --> 99:59:59.999
to do even a simple task
which doesn't require much precision,

99:59:59.999 --> 99:59:59.999
and this is what can happen.

99:59:59.999 --> 99:59:59.999
I mean, opening a door,
you don't require much precision.

99:59:59.999 --> 99:59:59.999
(Laughter)

99:59:59.999 --> 99:59:59.999
Or, a small error in the measurements,

99:59:59.999 --> 99:59:59.999
you miss the ?? and that's it

99:59:59.999 --> 99:59:59.999
(Laughter)

99:59:59.999 --> 99:59:59.999
with no way of recovering
most of the time.

99:59:59.999 --> 99:59:59.999
So why is that?

99:59:59.999 --> 99:59:59.999
Well, for many years,

99:59:59.999 --> 99:59:59.999
robots have been designed
to emphasize speed and precision,

99:59:59.999 --> 99:59:59.999
and this translates
in a very specific architecture.

99:59:59.999 --> 99:59:59.999
If we take a robot term,

99:59:59.999 --> 99:59:59.999
it's a very well-defined
set of rigid links

99:59:59.999 --> 99:59:59.999
and mortars who are called actuators,

99:59:59.999 --> 99:59:59.999
they move the links above the joins.

99:59:59.999 --> 99:59:59.999
In this ?? structure,

99:59:59.999 --> 99:59:59.999
you have to perfectly
measure your environment,

99:59:59.999 --> 99:59:59.999
so what is around,

99:59:59.999 --> 99:59:59.999
and you have to perfectly
program every movement

99:59:59.999 --> 99:59:59.999
of the robot joints,

99:59:59.999 --> 99:59:59.999
because a small error
can generate a very large fault,

99:59:59.999 --> 99:59:59.999
so you can damage something
or you can get your robot damaged

99:59:59.999 --> 99:59:59.999
if something is harder.

99:59:59.999 --> 99:59:59.999
So let's talk about them a moment,

99:59:59.999 --> 99:59:59.999
and don't think about
the brains of these robots

99:59:59.999 --> 99:59:59.999
or how carefully we program them,

99:59:59.999 --> 99:59:59.999
but rather look at their bodies.

99:59:59.999 --> 99:59:59.999
There is obviously
something wrong with it,

99:59:59.999 --> 99:59:59.999
because what makes a robot
precise and strong

99:59:59.999 --> 99:59:59.999
also makes them ridiculously dangerous
and ineffective in the real world,

99:59:59.999 --> 99:59:59.999
because their body cannot deform

99:59:59.999 --> 99:59:59.999
or better adjust to the interaction
with the real world.

99:59:59.999 --> 99:59:59.999
So think about the opposite approach,

99:59:59.999 --> 99:59:59.999
being softer than
anything else around you.

99:59:59.999 --> 99:59:59.999
Well, maybe you think that you're not
really able to do anything if you're soft,

99:59:59.999 --> 99:59:59.999
probably.

99:59:59.999 --> 99:59:59.999
Well, nature teaches us the opposite.

99:59:59.999 --> 99:59:59.999
For example, at the bottom of the ocean
under thousands of pounds

99:59:59.999 --> 99:59:59.999
of ?? pressure,

99:59:59.999 --> 99:59:59.999
a completely soft animal

99:59:59.999 --> 99:59:59.999
can move and interact with a much
stiffer object than him.

99:59:59.999 --> 99:59:59.999
He works by carrying around
this coconut shell

99:59:59.999 --> 99:59:59.999
thanks to the flexibility
of his tentacles,

99:59:59.999 --> 99:59:59.999
which serve as both his feet and hands.

99:59:59.999 --> 99:59:59.999
And apparently, an octopus
can also open a jar.

99:59:59.999 --> 99:59:59.999
It's pretty impressive, right?

99:59:59.999 --> 99:59:59.999
But clearly, this is not enabled
just by the brain of this animal,

99:59:59.999 --> 99:59:59.999
but also by his body,

99:59:59.999 --> 99:59:59.999
and it's a clear example,
maybe the clearest example,

99:59:59.999 --> 99:59:59.999
of embodied intelligence,

99:59:59.999 --> 99:59:59.999
which is a kind of intelligence
that all living organisms have.

99:59:59.999 --> 99:59:59.999
We all have that.

99:59:59.999 --> 99:59:59.999
Our body, its shape,
material and structure,

99:59:59.999 --> 99:59:59.999
plays a fundamental role
during a physical task,

99:59:59.999 --> 99:59:59.999
because we can conform
to our environment

99:59:59.999 --> 99:59:59.999
so we can succeed in a large
variety of situations

99:59:59.999 --> 99:59:59.999
without much planning
or calculations ahead.

99:59:59.999 --> 99:59:59.999
So why don't we put
some of this embodied intelligence

99:59:59.999 --> 99:59:59.999
into our robotic machines

99:59:59.999 --> 99:59:59.999
to release them from relying
on excessive work

99:59:59.999 --> 99:59:59.999
on computation and sensing?

99:59:59.999 --> 99:59:59.999
Well, to do that we can follow
the strategy of nature,

99:59:59.999 --> 99:59:59.999
because with evolution,
she's done a pretty good job

99:59:59.999 --> 99:59:59.999
in designing machines
for environment interaction,

99:59:59.999 --> 99:59:59.999
and it's easy to notice that nature
uses soft material frequently

99:59:59.999 --> 99:59:59.999
and stiff material sparingly.

99:59:59.999 --> 99:59:59.999
And this is what is done
in this new field or robotics

99:59:59.999 --> 99:59:59.999
which is called soft robotics,

99:59:59.999 --> 99:59:59.999
in which the main objective
is not to make super-precise machines

99:59:59.999 --> 99:59:59.999
because we've already got them,

99:59:59.999 --> 99:59:59.999
but to make robots able to face
unexpected situations in the real world,

99:59:59.999 --> 99:59:59.999
so able to go out there.

99:59:59.999 --> 99:59:59.999
And what makes a robot soft
is first of all his compliant body,

99:59:59.999 --> 99:59:59.999
which is made of materials or structures
that can undergo very large deformations,

99:59:59.999 --> 99:59:59.999
so no more rigid links,

99:59:59.999 --> 99:59:59.999
and secondly to move them
we use what we call distributed actuation,

99:59:59.999 --> 99:59:59.999
so we have to control continuously
the shape of this very deformable body,

99:59:59.999 --> 99:59:59.999
which is the fact of having
a lot of links and joints,

99:59:59.999 --> 99:59:59.999
but we don't have
any stiff structure at all.

99:59:59.999 --> 99:59:59.999
So you can imagine that building
a soft robot is a very different process

99:59:59.999 --> 99:59:59.999
than stiff robotics, where
you have links, gears, screws

99:59:59.999 --> 99:59:59.999
that you must combine
in a very defined way.

99:59:59.999 --> 99:59:59.999
In soft robots, you just build
your actuator from scratch

99:59:59.999 --> 99:59:59.999
most of the time,

99:59:59.999 --> 99:59:59.999
but you shape your flexible material

99:59:59.999 --> 99:59:59.999
to the form that responds
to a certain input.

99:59:59.999 --> 99:59:59.999
For example here, you can just
deform a structure

99:59:59.999 --> 99:59:59.999
doing a fairly complex shape

99:59:59.999 --> 99:59:59.999
if you think about doing the same
with rigid links and joints,

99:59:59.999 --> 99:59:59.999
and here what you use is just one input,

99:59:59.999 --> 99:59:59.999
such as air pressure.

99:59:59.999 --> 99:59:59.999
Okay, but let's see
some cool examples of soft robots.

99:59:59.999 --> 99:59:59.999
Here is a little cute guy
developed by Harvard University,

99:59:59.999 --> 99:59:59.999
and he works thanks to waves
of pressure applied along its body,

99:59:59.999 --> 99:59:59.999
and thanks to the flexibility he can
also sneak under a low bridge,

99:59:59.999 --> 99:59:59.999
keep walking,

99:59:59.999 --> 99:59:59.999
and then keep walking
a little bit different afterwards.

99:59:59.999 --> 99:59:59.999
And it's a very preliminary prototype,

99:59:59.999 --> 99:59:59.999
but they also built a more robust version

99:59:59.999 --> 99:59:59.999
with power on board that can actually
be sent out in the world

99:59:59.999 --> 99:59:59.999
and face real-world interactions

99:59:59.999 --> 99:59:59.999
like a car passing it over it,

99:59:59.999 --> 99:59:59.999
and keep working.

99:59:59.999 --> 99:59:59.999
(Laughter)

99:59:59.999 --> 99:59:59.999
It's cute.

99:59:59.999 --> 99:59:59.999
(Laughter)

99:59:59.999 --> 99:59:59.999
Or a robotic fish which swims
like a real fish does

99:59:59.999 --> 99:59:59.999
in water simply because it has a soft tail
with distributed actuation

99:59:59.999 --> 99:59:59.999
using still air pressure.

99:59:59.999 --> 99:59:59.999
That was from MIT,

99:59:59.999 --> 99:59:59.999
and of course we have a robotic octopus.

99:59:59.999 --> 99:59:59.999
This was actually one of
the first projects developed

99:59:59.999 --> 99:59:59.999
in this new field of soft robots.

99:59:59.999 --> 99:59:59.999
Here you see the artificial tentacle,

99:59:59.999 --> 99:59:59.999
but they actually built and entire machine

99:59:59.999 --> 99:59:59.999
with several tentacles they could
just throw in the water,

99:59:59.999 --> 99:59:59.999
and you see that it can kind of go around
and do submarine exploration

99:59:59.999 --> 99:59:59.999
in a different way
than rigid robots would do.

99:59:59.999 --> 99:59:59.999
But this is very important for delicate
environments such as coral reefs.

99:59:59.999 --> 99:59:59.999
Let's go back to the ground.

99:59:59.999 --> 99:59:59.999
Here you see the view

99:59:59.999 --> 99:59:59.999
from a growing robot developed
by my colleagues in Stanford.

99:59:59.999 --> 99:59:59.999
You see the camera fixed on top.

99:59:59.999 --> 99:59:59.999
And this robot is particular
because using air pressure

99:59:59.999 --> 99:59:59.999
it grows from the tip
while the rest of the body

99:59:59.999 --> 99:59:59.999
stays in firm contact
with the environment.

99:59:59.999 --> 99:59:59.999
And this is inspired
by plants, not animals,

99:59:59.999 --> 99:59:59.999
which grows via the material
in a similar manner

99:59:59.999 --> 99:59:59.999
so it can face a pretty large
variety of situations.

99:59:59.999 --> 99:59:59.999
But I'm a biomedical engineer,

99:59:59.999 --> 99:59:59.999
and perhaps the application
I like the most

99:59:59.999 --> 99:59:59.999
it's in the medical field,

99:59:59.999 --> 99:59:59.999
and it's very difficult to imagine
a closer interaction with the human body

99:59:59.999 --> 99:59:59.999
than actually going inside the body,

99:59:59.999 --> 99:59:59.999
for example to inform
a minimally invasive procedure.

99:59:59.999 --> 99:59:59.999
And here, robots can be
very helpful with the surgeon

99:59:59.999 --> 99:59:59.999
because they must enter the body

99:59:59.999 --> 99:59:59.999
using small holes
and straight instruments,

99:59:59.999 --> 99:59:59.999
and these instruments must interact
with very delicate structures

99:59:59.999 --> 99:59:59.999
in a very uncertain environment,

99:59:59.999 --> 99:59:59.999
and this must be done safely.

99:59:59.999 --> 99:59:59.999
Also bringing the camera inside the body,

99:59:59.999 --> 99:59:59.999
so bringing the eyes of the surgeon
inside the subject, I feel,

99:59:59.999 --> 99:59:59.999
can be very challenging
if you use a rigid stick,

99:59:59.999 --> 99:59:59.999
like a classic endoscope.

99:59:59.999 --> 99:59:59.999
With my previous research group in Europe,

99:59:59.999 --> 99:59:59.999
we developed this
self-camera robot for surgery,

99:59:59.999 --> 99:59:59.999
which is very different
from a classic endoscope