0:00:01.531,0:00:03.368
So, robots.

0:00:03.392,0:00:04.806
Robots can be programmed

0:00:04.830,0:00:08.521
to do the same task millions of times[br]with minimal error,

0:00:08.545,0:00:11.059
something very difficult for us, right?

0:00:11.083,0:00:14.244
And it can be very impressive[br]to watch them at work.

0:00:14.268,0:00:15.524
Look at them.

0:00:15.548,0:00:17.456
I could watch them for hours.

0:00:18.108,0:00:19.407
No?

0:00:19.431,0:00:21.638
What is less impressive

0:00:21.662,0:00:24.595
is that if you take this robot[br]out of the factories,

0:00:24.619,0:00:28.999
where the environments are not[br]perfectly known and measured like here,

0:00:29.023,0:00:33.301
to do even a simple task[br]which doesn't require much precision,

0:00:33.325,0:00:34.936
this is what can happen.

0:00:34.960,0:00:37.689
I mean, opening a door,[br]you don't require much precision.

0:00:37.713,0:00:38.743
(Laughter)

0:00:38.767,0:00:41.221
Or a small error in the measurements,

0:00:41.245,0:00:43.071
you miss the valve, and that's it --

0:00:43.095,0:00:44.365
(Laughter)

0:00:44.389,0:00:46.833
with no way of recovering,[br]most of the time.

0:00:47.561,0:00:49.236
So why is that?

0:00:49.260,0:00:51.134
Well, for many years,

0:00:51.158,0:00:54.458
robots have been designed[br]to emphasize speed and precision,

0:00:54.482,0:00:57.444
and this translates[br]into a very specific architecture.

0:00:57.468,0:00:58.619
If you take a robot arm,

0:00:58.643,0:01:01.402
it's a very well-defined[br]set of rigid links

0:01:01.426,0:01:03.485
and motors, what we call actuators,

0:01:03.509,0:01:05.279
they move the links about the joints.

0:01:05.303,0:01:06.610
In this robotic structure,

0:01:06.624,0:01:08.851
you have to perfectly[br]measure your environment,

0:01:08.865,0:01:10.762
so what is around,

0:01:10.786,0:01:13.425
and you have to perfectly[br]program every movement

0:01:13.449,0:01:15.584
of the robot joints,

0:01:15.608,0:01:18.870
because a small error[br]can generate a very large fault,

0:01:18.894,0:01:21.907
so you can damage something[br]or you can get your robot damaged

0:01:21.931,0:01:23.462
if something is harder.

0:01:24.107,0:01:26.312
So let's talk about them a moment.

0:01:26.336,0:01:29.559
And don't think[br]about the brains of these robots

0:01:29.583,0:01:32.328
or how carefully we program them,

0:01:32.352,0:01:34.170
but rather look at their bodies.

0:01:34.606,0:01:37.485
There is obviously[br]something wrong with it,

0:01:37.509,0:01:40.636
because what makes a robot[br]precise and strong

0:01:40.660,0:01:45.049
also makes them ridiculously dangerous[br]and ineffective in the real world,

0:01:45.073,0:01:47.058
because their body cannot deform

0:01:47.082,0:01:50.311
or better adjust to the interaction[br]with the real world.

0:01:51.226,0:01:54.344
So think about the opposite approach,

0:01:54.368,0:01:57.186
being softer than[br]anything else around you.

0:01:57.827,0:02:02.912
Well, maybe you think that you're not[br]really able to do anything if you're soft,

0:02:02.936,0:02:04.103
probably.

0:02:04.127,0:02:06.977
Well, nature teaches us the opposite.

0:02:07.001,0:02:09.032
For example, at the bottom of the ocean,

0:02:09.056,0:02:11.492
under thousands of pounds[br]of hydrostatic pressure,

0:02:11.516,0:02:13.944
a completely soft animal

0:02:13.968,0:02:17.245
can move and interact[br]with a much stiffer object than him.

0:02:17.878,0:02:20.725
He works by carrying around[br]this coconut shell

0:02:20.749,0:02:23.133
thanks to the flexibility[br]of his tentacles,

0:02:23.157,0:02:25.661
which serve as both his feet and hands.

0:02:26.241,0:02:30.066
And apparently,[br]an octopus can also open a jar.

0:02:31.883,0:02:33.637
It's pretty impressive, right?

0:02:35.918,0:02:40.418
But clearly, this is not enabled[br]just by the brain of this animal,

0:02:40.442,0:02:42.456
but also by his body,

0:02:42.480,0:02:46.512
and it's a clear example,[br]maybe the clearest example,

0:02:46.536,0:02:48.336
of embodied intelligence,

0:02:48.360,0:02:51.646
which is a kind of intelligence[br]that all living organisms have.

0:02:51.670,0:02:53.236
We all have that.

0:02:53.260,0:02:57.102
Our body, its shape,[br]material and structure,

0:02:57.126,0:03:00.308
plays a fundamental role[br]during a physical task,

0:03:00.332,0:03:05.945
because we can conform to our environment

0:03:05.969,0:03:08.373
so we can succeed in a large[br]variety of situations

0:03:08.397,0:03:11.390
without much planning[br]or calculations ahead.

0:03:11.414,0:03:14.129
So why don't we put[br]some of this embodied intelligence

0:03:14.153,0:03:15.708
into our robotic machines,

0:03:15.732,0:03:18.081
to release them from relying[br]on excessive work

0:03:18.105,0:03:20.122
on computation and sensing?

0:03:21.097,0:03:23.747
Well, to do that, we can follow[br]the strategy of nature,

0:03:23.771,0:03:26.383
because with evolution,[br]she's done a pretty good job

0:03:26.407,0:03:30.903
in designing machines[br]for environment interaction.

0:03:30.927,0:03:35.421
And it's easy to notice that nature[br]uses soft material frequently

0:03:35.445,0:03:37.740
and stiff material sparingly.

0:03:37.764,0:03:41.556
And this is what is done[br]in this new field or robotics,

0:03:41.580,0:03:43.880
which is called "soft robotics,"

0:03:43.904,0:03:47.640
in which the main objective[br]is not to make super-precise machines,

0:03:47.664,0:03:49.601
because we've already got them,

0:03:49.625,0:03:54.545
but to make robots able to face[br]unexpected situations in the real world,

0:03:54.569,0:03:56.126
so able to go out there.

0:03:56.150,0:03:59.674
And what makes a robot soft[br]is first of all its compliant body,

0:03:59.698,0:04:05.229
which is made of materials or structures[br]that can undergo very large deformations,

0:04:05.253,0:04:07.084
so no more rigid links,

0:04:07.108,0:04:10.656
and secondly, to move them,[br]we use what we call distributed actuation,

0:04:10.680,0:04:15.712
so we have to control continuously[br]the shape of this very deformable body,

0:04:15.736,0:04:19.034
which has the effect[br]of having a lot of links and joints,

0:04:19.058,0:04:21.681
but we don't have[br]any stiff structure at all.

0:04:21.705,0:04:25.135
So you can imagine that building[br]a soft robot is a very different process

0:04:25.159,0:04:28.039
than stiff robotics,[br]where you have links, gears, screws

0:04:28.063,0:04:30.294
that you must combine[br]in a very defined way.

0:04:30.948,0:04:34.473
In soft robots, you just build[br]your actuator from scratch

0:04:34.497,0:04:35.648
most of the time,

0:04:35.672,0:04:38.054
but you shape your flexible material

0:04:38.078,0:04:40.481
to the form that responds[br]to a certain input.

0:04:41.054,0:04:43.512
For example, here,[br]you can just deform a structure

0:04:43.536,0:04:46.007
doing a fairly complex shape

0:04:46.031,0:04:49.309
if you think about doing the same[br]with rigid links and joints,

0:04:49.333,0:04:51.666
and here, what you use is just one input,

0:04:51.690,0:04:53.054
such as air pressure.

0:04:53.869,0:04:57.358
OK, but let's see[br]some cool examples of soft robots.

0:04:57.765,0:05:02.312
Here is a little cute guy[br]developed at Harvard University,

0:05:02.336,0:05:06.829
and he works thanks to waves[br]of pressure applied along its body,

0:05:06.853,0:05:10.139
and thanks to the flexibility,[br]he can also sneak under a low bridge,

0:05:10.163,0:05:11.314
keep walking,

0:05:11.338,0:05:14.535
and then keep walking[br]a little bit different afterwards.

0:05:15.345,0:05:17.576
And it's a very preliminary prototype,

0:05:17.600,0:05:21.276
but they also built a more robust version[br]with power on board

0:05:21.300,0:05:26.747
that can actually be sent out in the world[br]and face real-world interactions

0:05:26.771,0:05:28.477
like a car passing it over it ...

0:05:30.090,0:05:31.240
and keep working.

0:05:32.056,0:05:33.207
It's cute.

0:05:33.231,0:05:34.652
(Laughter)

0:05:34.676,0:05:38.540
Or a robotic fish, which swims[br]like a real fish does in water

0:05:38.564,0:05:41.748
simply because it has a soft tail[br]with distributed actuation

0:05:41.772,0:05:43.416
using still air pressure.

0:05:43.954,0:05:45.312
That was from MIT,

0:05:45.336,0:05:48.141
and of course, we have a robotic octopus.

0:05:48.165,0:05:50.244
This was actually one[br]of the first projects

0:05:50.268,0:05:52.394
developed in this new field[br]of soft robots.

0:05:52.418,0:05:54.304
Here, you see the artificial tentacle,

0:05:54.328,0:05:59.007
but they actually built an entire machine[br]with several tentacles

0:05:59.031,0:06:01.642
they could just throw in the water,

0:06:01.666,0:06:05.959
and you see that it can kind of go around[br]and do submarine exploration

0:06:05.983,0:06:09.286
in a different way[br]than rigid robots would do.

0:06:09.310,0:06:12.970
But this is very important for delicate[br]environments, such as coral reefs.

0:06:12.994,0:06:14.390
Let's go back to the ground.

0:06:14.414,0:06:15.604
Here, you see the view

0:06:15.628,0:06:19.776
from a growing robot developed[br]by my colleagues in Stanford.

0:06:19.800,0:06:21.650
You see the camera fixed on top.

0:06:21.674,0:06:23.112
And this robot is particular,

0:06:23.136,0:06:25.552
because using air pressure,[br]it grows from the tip,

0:06:25.576,0:06:28.922
while the rest of the body stays[br]in firm contact with the environment.

0:06:29.316,0:06:32.034
And this is inspired[br]by plants, not animals,

0:06:32.058,0:06:35.373
which grows via the material[br]in a similar manner

0:06:35.397,0:06:38.357
so it can face a pretty large[br]variety of situations.

0:06:39.043,0:06:40.711
But I'm a biomedical engineer,

0:06:40.735,0:06:43.004
and perhaps the application[br]I like the most

0:06:43.028,0:06:44.481
is in the medical field,

0:06:44.505,0:06:49.346
and it's very difficult to imagine[br]a closer interaction with the human body

0:06:49.370,0:06:51.289
than actually going inside the body,

0:06:51.313,0:06:54.084
for example, to perform[br]a minimally invasive procedure.

0:06:54.958,0:06:58.360
And here, robots can be[br]very helpful with the surgeon,

0:06:58.384,0:07:00.133
because they must enter the body

0:07:00.157,0:07:02.784
using small holes[br]and straight instruments,

0:07:02.808,0:07:06.318
and these instruments must interact[br]with very delicate structures

0:07:06.342,0:07:08.390
in a very uncertain environment,

0:07:08.414,0:07:10.089
and this must be done safely.

0:07:10.113,0:07:12.225
Also bringing the camera inside the body,

0:07:12.249,0:07:15.867
so bringing the eyes of the surgeon[br]inside the subject, I feel,

0:07:15.891,0:07:18.242
can be very challenging[br]if you use a rigid stick,

0:07:18.266,0:07:19.873
like a classic endoscope.

0:07:20.517,0:07:23.106
With my previous research group in Europe,

0:07:23.130,0:07:25.726
we developed this[br]self-camera robot for surgery,

0:07:25.750,0:07:29.518
which is very different[br]from a classic endoscope,

0:07:29.542,0:07:32.646
which can move thanks[br]to the flexibility of the module

0:07:32.670,0:07:37.558
that can bend in every direction[br]and also elongate.

0:07:37.582,0:07:40.692
And this was actually used by surgeons[br]to see what they were doing

0:07:40.716,0:07:43.454
with other instruments[br]from different points of view,

0:07:43.478,0:07:46.684
without caring that much[br]about what was touched around.

0:07:47.247,0:07:50.990
And here you see the soft robot in action,

0:07:51.014,0:07:53.832
and it just goes inside.

0:07:53.856,0:07:57.125
This is a body simulator,[br]not a real human body.

0:07:57.149,0:07:58.300
It goes around.

0:07:58.324,0:07:59.998
You have a light, because usually,

0:08:00.022,0:08:03.143
you don't have too many lights[br]inside your body.

0:08:03.167,0:08:04.340
We hope.

0:08:04.364,0:08:07.366
(Laughter)

0:08:07.390,0:08:12.088
But sometimes, a surgical procedure[br]can even be done using a single needle,

0:08:12.112,0:08:16.159
and in Stanford now, we are working[br]on a very flexible needle,

0:08:16.183,0:08:18.835
kind of a very tiny soft robot

0:08:18.859,0:08:22.153
which is mechanically designed[br]to use the interaction with the tissues

0:08:22.177,0:08:24.407
and steer around inside a solid organ.

0:08:24.431,0:08:28.511
This makes it possible to reach[br]many different targets, such as tumors,

0:08:28.535,0:08:30.233
deep inside a solid organ

0:08:30.257,0:08:32.582
by using one single insertion point.

0:08:32.606,0:08:36.645
And you can even steer around[br]the structure that you want to avoid

0:08:36.669,0:08:38.033
on the way to the target.

0:08:39.377,0:08:42.682
So clearly, this is a pretty[br]exciting time for robotics.

0:08:42.706,0:08:45.859
We have robots that have to deal[br]with soft structures,

0:08:45.883,0:08:48.468
so this poses new[br]and very challenging questions

0:08:48.492,0:08:49.849
for the robotics community,

0:08:49.873,0:08:52.548
and indeed, we are just starting[br]to learn how to control,

0:08:52.572,0:08:55.576
how to put sensors[br]on these very flexible structures.

0:08:55.600,0:08:58.560
But of course, we are not even close[br]to what nature figured out

0:08:58.584,0:09:00.778
in millions of years of evolution.

0:09:00.802,0:09:02.906
But one thing I know for sure:

0:09:02.930,0:09:05.446
robots will be softer and safer,

0:09:05.470,0:09:08.452
and they will be out there helping people.

0:09:08.809,0:09:09.960
Thank you.

0:09:09.984,0:09:14.396
(Applause)