1
00:00:01,531 --> 00:00:03,368
So, robots.

2
00:00:03,392 --> 00:00:04,806
Robots can be programmed

3
00:00:04,830 --> 00:00:08,521
to do the same task millions of times
with minimal error,

4
00:00:08,545 --> 00:00:11,059
something very difficult for us, right?

5
00:00:11,083 --> 00:00:14,244
And it can be very impressive
to watch them at work.

6
00:00:14,268 --> 00:00:15,524
Look at them.

7
00:00:15,548 --> 00:00:17,456
I could watch them for hours.

8
00:00:18,108 --> 00:00:19,407
No?

9
00:00:19,431 --> 00:00:21,638
What is less impressive

10
00:00:21,662 --> 00:00:24,595
is that if you take this robot
out of the factories,

11
00:00:24,619 --> 00:00:28,999
where the environments are not
perfectly known and measured like here,

12
00:00:29,023 --> 00:00:33,301
to do even a simple task
which doesn't require much precision,

13
00:00:33,325 --> 00:00:34,936
this is what can happen.

14
00:00:34,960 --> 00:00:37,689
I mean, opening a door,
you don't require much precision.

15
00:00:37,713 --> 00:00:38,743
(Laughter)

16
00:00:38,767 --> 00:00:41,221
Or a small error in the measurements,

17
00:00:41,245 --> 00:00:43,071
you miss the valve, and that's it --

18
00:00:43,095 --> 00:00:44,365
(Laughter)

19
00:00:44,389 --> 00:00:46,833
with no way of recovering,
most of the time.

20
00:00:47,561 --> 00:00:49,236
So why is that?

21
00:00:49,260 --> 00:00:51,134
Well, for many years,

22
00:00:51,158 --> 00:00:54,458
robots have been designed
to emphasize speed and precision,

23
00:00:54,482 --> 00:00:57,444
and this translates
into a very specific architecture.

24
00:00:57,468 --> 00:00:58,619
If you take a robot arm,

25
00:00:58,643 --> 00:01:01,402
it's a very well-defined
set of rigid links

26
00:01:01,426 --> 00:01:03,485
and motors, what we call actuators,

27
00:01:03,509 --> 00:01:05,279
they move the links about the joints.

28
00:01:05,303 --> 00:01:06,610
In this robotic structure,

29
00:01:06,624 --> 00:01:08,851
you have to perfectly
measure your environment,

30
00:01:08,865 --> 00:01:10,762
so what is around,

31
00:01:10,786 --> 00:01:13,425
and you have to perfectly
program every movement

32
00:01:13,449 --> 00:01:15,584
of the robot joints,

33
00:01:15,608 --> 00:01:18,870
because a small error
can generate a very large fault,

34
00:01:18,894 --> 00:01:21,907
so you can damage something
or you can get your robot damaged

35
00:01:21,931 --> 00:01:23,462
if something is harder.

36
00:01:24,107 --> 00:01:26,312
So let's talk about them a moment.

37
00:01:26,336 --> 00:01:29,559
And don't think
about the brains of these robots

38
00:01:29,583 --> 00:01:32,328
or how carefully we program them,

39
00:01:32,352 --> 00:01:34,170
but rather look at their bodies.

40
00:01:34,606 --> 00:01:37,485
There is obviously
something wrong with it,

41
00:01:37,509 --> 00:01:40,636
because what makes a robot
precise and strong

42
00:01:40,660 --> 00:01:45,049
also makes them ridiculously dangerous
and ineffective in the real world,

43
00:01:45,073 --> 00:01:47,058
because their body cannot deform

44
00:01:47,082 --> 00:01:50,311
or better adjust to the interaction
with the real world.

45
00:01:51,226 --> 00:01:54,344
So think about the opposite approach,

46
00:01:54,368 --> 00:01:57,186
being softer than
anything else around you.

47
00:01:57,827 --> 00:02:02,912
Well, maybe you think that you're not
really able to do anything if you're soft,

48
00:02:02,936 --> 00:02:04,103
probably.

49
00:02:04,127 --> 00:02:06,977
Well, nature teaches us the opposite.

50
00:02:07,001 --> 00:02:09,032
For example, at the bottom of the ocean,

51
00:02:09,056 --> 00:02:11,492
under thousands of pounds
of hydrostatic pressure,

52
00:02:11,516 --> 00:02:13,944
a completely soft animal

53
00:02:13,968 --> 00:02:17,245
can move and interact
with a much stiffer object than him.

54
00:02:17,878 --> 00:02:20,725
He works by carrying around
this coconut shell

55
00:02:20,749 --> 00:02:23,133
thanks to the flexibility
of his tentacles,

56
00:02:23,157 --> 00:02:25,661
which serve as both his feet and hands.

57
00:02:26,241 --> 00:02:30,066
And apparently,
an octopus can also open a jar.

58
00:02:31,883 --> 00:02:33,637
It's pretty impressive, right?

59
00:02:35,918 --> 00:02:40,418
But clearly, this is not enabled
just by the brain of this animal,

60
00:02:40,442 --> 00:02:42,456
but also by his body,

61
00:02:42,480 --> 00:02:46,512
and it's a clear example,
maybe the clearest example,

62
00:02:46,536 --> 00:02:48,336
of embodied intelligence,

63
00:02:48,360 --> 00:02:51,646
which is a kind of intelligence
that all living organisms have.

64
00:02:51,670 --> 00:02:53,236
We all have that.

65
00:02:53,260 --> 00:02:57,102
Our body, its shape,
material and structure,

66
00:02:57,126 --> 00:03:00,308
plays a fundamental role
during a physical task,

67
00:03:00,332 --> 00:03:05,945
because we can conform to our environment

68
00:03:05,969 --> 00:03:08,373
so we can succeed in a large
variety of situations

69
00:03:08,397 --> 00:03:11,390
without much planning
or calculations ahead.

70
00:03:11,414 --> 00:03:14,129
So why don't we put
some of this embodied intelligence

71
00:03:14,153 --> 00:03:15,708
into our robotic machines,

72
00:03:15,732 --> 00:03:18,081
to release them from relying
on excessive work

73
00:03:18,105 --> 00:03:20,122
on computation and sensing?

74
00:03:21,097 --> 00:03:23,747
Well, to do that, we can follow
the strategy of nature,

75
00:03:23,771 --> 00:03:26,383
because with evolution,
she's done a pretty good job

76
00:03:26,407 --> 00:03:30,903
in designing machines
for environment interaction.

77
00:03:30,927 --> 00:03:35,421
And it's easy to notice that nature
uses soft material frequently

78
00:03:35,445 --> 00:03:37,740
and stiff material sparingly.

79
00:03:37,764 --> 00:03:41,556
And this is what is done
in this new field or robotics,

80
00:03:41,580 --> 00:03:43,880
which is called "soft robotics,"

81
00:03:43,904 --> 00:03:47,640
in which the main objective
is not to make super-precise machines,

82
00:03:47,664 --> 00:03:49,601
because we've already got them,

83
00:03:49,625 --> 00:03:54,545
but to make robots able to face
unexpected situations in the real world,

84
00:03:54,569 --> 00:03:56,126
so able to go out there.

85
00:03:56,150 --> 00:03:59,674
And what makes a robot soft
is first of all its compliant body,

86
00:03:59,698 --> 00:04:05,229
which is made of materials or structures
that can undergo very large deformations,

87
00:04:05,253 --> 00:04:07,084
so no more rigid links,

88
00:04:07,108 --> 00:04:10,656
and secondly, to move them,
we use what we call distributed actuation,

89
00:04:10,680 --> 00:04:15,712
so we have to control continuously
the shape of this very deformable body,

90
00:04:15,736 --> 00:04:19,034
which has the effect
of having a lot of links and joints,

91
00:04:19,058 --> 00:04:21,681
but we don't have
any stiff structure at all.

92
00:04:21,705 --> 00:04:25,135
So you can imagine that building
a soft robot is a very different process

93
00:04:25,159 --> 00:04:28,039
than stiff robotics,
where you have links, gears, screws

94
00:04:28,063 --> 00:04:30,294
that you must combine
in a very defined way.

95
00:04:30,948 --> 00:04:34,473
In soft robots, you just build
your actuator from scratch

96
00:04:34,497 --> 00:04:35,648
most of the time,

97
00:04:35,672 --> 00:04:38,054
but you shape your flexible material

98
00:04:38,078 --> 00:04:40,481
to the form that responds
to a certain input.

99
00:04:41,054 --> 00:04:43,512
For example, here,
you can just deform a structure

100
00:04:43,536 --> 00:04:46,007
doing a fairly complex shape

101
00:04:46,031 --> 00:04:49,309
if you think about doing the same
with rigid links and joints,

102
00:04:49,333 --> 00:04:51,666
and here, what you use is just one input,

103
00:04:51,690 --> 00:04:53,054
such as air pressure.

104
00:04:53,869 --> 00:04:57,358
OK, but let's see
some cool examples of soft robots.

105
00:04:57,765 --> 00:05:02,312
Here is a little cute guy
developed at Harvard University,

106
00:05:02,336 --> 00:05:06,829
and he works thanks to waves
of pressure applied along its body,

107
00:05:06,853 --> 00:05:10,139
and thanks to the flexibility,
he can also sneak under a low bridge,

108
00:05:10,163 --> 00:05:11,314
keep walking,

109
00:05:11,338 --> 00:05:14,535
and then keep walking
a little bit different afterwards.

110
00:05:15,345 --> 00:05:17,576
And it's a very preliminary prototype,

111
00:05:17,600 --> 00:05:21,276
but they also built a more robust version
with power on board

112
00:05:21,300 --> 00:05:26,747
that can actually be sent out in the world
and face real-world interactions

113
00:05:26,771 --> 00:05:28,477
like a car passing it over it ...

114
00:05:30,090 --> 00:05:31,240
and keep working.

115
00:05:32,056 --> 00:05:33,207
It's cute.

116
00:05:33,231 --> 00:05:34,652
(Laughter)

117
00:05:34,676 --> 00:05:38,540
Or a robotic fish, which swims
like a real fish does in water

118
00:05:38,564 --> 00:05:41,748
simply because it has a soft tail
with distributed actuation

119
00:05:41,772 --> 00:05:43,416
using still air pressure.

120
00:05:43,954 --> 00:05:45,312
That was from MIT,

121
00:05:45,336 --> 00:05:48,141
and of course, we have a robotic octopus.

122
00:05:48,165 --> 00:05:50,244
This was actually one
of the first projects

123
00:05:50,268 --> 00:05:52,394
developed in this new field
of soft robots.

124
00:05:52,418 --> 00:05:54,304
Here, you see the artificial tentacle,

125
00:05:54,328 --> 00:05:59,007
but they actually built an entire machine
with several tentacles

126
00:05:59,031 --> 00:06:01,642
they could just throw in the water,

127
00:06:01,666 --> 00:06:05,959
and you see that it can kind of go around
and do submarine exploration

128
00:06:05,983 --> 00:06:09,286
in a different way
than rigid robots would do.

129
00:06:09,310 --> 00:06:12,970
But this is very important for delicate
environments, such as coral reefs.

130
00:06:12,994 --> 00:06:14,390
Let's go back to the ground.

131
00:06:14,414 --> 00:06:15,604
Here, you see the view

132
00:06:15,628 --> 00:06:19,776
from a growing robot developed
by my colleagues in Stanford.

133
00:06:19,800 --> 00:06:21,650
You see the camera fixed on top.

134
00:06:21,674 --> 00:06:23,112
And this robot is particular,

135
00:06:23,136 --> 00:06:25,552
because using air pressure,
it grows from the tip,

136
00:06:25,576 --> 00:06:28,922
while the rest of the body stays
in firm contact with the environment.

137
00:06:29,316 --> 00:06:32,034
And this is inspired
by plants, not animals,

138
00:06:32,058 --> 00:06:35,373
which grows via the material
in a similar manner

139
00:06:35,397 --> 00:06:38,357
so it can face a pretty large
variety of situations.

140
00:06:39,043 --> 00:06:40,711
But I'm a biomedical engineer,

141
00:06:40,735 --> 00:06:43,004
and perhaps the application
I like the most

142
00:06:43,028 --> 00:06:44,481
is in the medical field,

143
00:06:44,505 --> 00:06:49,346
and it's very difficult to imagine
a closer interaction with the human body

144
00:06:49,370 --> 00:06:51,289
than actually going inside the body,

145
00:06:51,313 --> 00:06:54,084
for example, to perform
a minimally invasive procedure.

146
00:06:54,958 --> 00:06:58,360
And here, robots can be
very helpful with the surgeon,

147
00:06:58,384 --> 00:07:00,133
because they must enter the body

148
00:07:00,157 --> 00:07:02,784
using small holes
and straight instruments,

149
00:07:02,808 --> 00:07:06,318
and these instruments must interact
with very delicate structures

150
00:07:06,342 --> 00:07:08,390
in a very uncertain environment,

151
00:07:08,414 --> 00:07:10,089
and this must be done safely.

152
00:07:10,113 --> 00:07:12,225
Also bringing the camera inside the body,

153
00:07:12,249 --> 00:07:15,867
so bringing the eyes of the surgeon
inside the subject, I feel,

154
00:07:15,891 --> 00:07:18,242
can be very challenging
if you use a rigid stick,

155
00:07:18,266 --> 00:07:19,873
like a classic endoscope.

156
00:07:20,517 --> 00:07:23,106
With my previous research group in Europe,

157
00:07:23,130 --> 00:07:25,726
we developed this
self-camera robot for surgery,

158
00:07:25,750 --> 00:07:29,518
which is very different
from a classic endoscope,

159
00:07:29,542 --> 00:07:32,646
which can move thanks
to the flexibility of the module

160
00:07:32,670 --> 00:07:37,558
that can bend in every direction
and also elongate.

161
00:07:37,582 --> 00:07:40,692
And this was actually used by surgeons
to see what they were doing

162
00:07:40,716 --> 00:07:43,454
with other instruments
from different points of view,

163
00:07:43,478 --> 00:07:46,684
without caring that much
about what was touched around.

164
00:07:47,247 --> 00:07:50,990
And here you see the soft robot in action,

165
00:07:51,014 --> 00:07:53,832
and it just goes inside.

166
00:07:53,856 --> 00:07:57,125
This is a body simulator,
not a real human body.

167
00:07:57,149 --> 00:07:58,300
It goes around.

168
00:07:58,324 --> 00:07:59,998
You have a light, because usually,

169
00:08:00,022 --> 00:08:03,143
you don't have too many lights
inside your body.

170
00:08:03,167 --> 00:08:04,340
We hope.

171
00:08:04,364 --> 00:08:07,366
(Laughter)

172
00:08:07,390 --> 00:08:12,088
But sometimes, a surgical procedure
can even be done using a single needle,

173
00:08:12,112 --> 00:08:16,159
and in Stanford now, we are working
on a very flexible needle,

174
00:08:16,183 --> 00:08:18,835
kind of a very tiny soft robot

175
00:08:18,859 --> 00:08:22,153
which is mechanically designed
to use the interaction with the tissues

176
00:08:22,177 --> 00:08:24,407
and steer around inside a solid organ.

177
00:08:24,431 --> 00:08:28,511
This makes it possible to reach
many different targets, such as tumors,

178
00:08:28,535 --> 00:08:30,233
deep inside a solid organ

179
00:08:30,257 --> 00:08:32,582
by using one single insertion point.

180
00:08:32,606 --> 00:08:36,645
And you can even steer around
the structure that you want to avoid

181
00:08:36,669 --> 00:08:38,033
on the way to the target.

182
00:08:39,377 --> 00:08:42,682
So clearly, this is a pretty
exciting time for robotics.

183
00:08:42,706 --> 00:08:45,859
We have robots that have to deal
with soft structures,

184
00:08:45,883 --> 00:08:48,468
so this poses new
and very challenging questions

185
00:08:48,492 --> 00:08:49,849
for the robotics community,

186
00:08:49,873 --> 00:08:52,548
and indeed, we are just starting
to learn how to control,

187
00:08:52,572 --> 00:08:55,576
how to put sensors
on these very flexible structures.

188
00:08:55,600 --> 00:08:58,560
But of course, we are not even close
to what nature figured out

189
00:08:58,584 --> 00:09:00,778
in millions of years of evolution.

190
00:09:00,802 --> 00:09:02,906
But one thing I know for sure:

191
00:09:02,930 --> 00:09:05,446
robots will be softer and safer,

192
00:09:05,470 --> 00:09:08,452
and they will be out there helping people.

193
00:09:08,809 --> 00:09:09,960
Thank you.

194
00:09:09,984 --> 00:09:14,396
(Applause)