1
00:00:01,022 --> 00:00:03,472
So robots.

2
00:00:03,472 --> 00:00:07,415
Robots can be programmed
to do the same task millions of times

3
00:00:07,415 --> 00:00:11,293
with minimal error, something
very difficult for us, right?

4
00:00:11,293 --> 00:00:14,477
And it can be very impressive
to watch them at work.

5
00:00:14,477 --> 00:00:15,756
Look at them.

6
00:00:15,756 --> 00:00:18,472
I could watch them for hours.

7
00:00:18,472 --> 00:00:19,622
No?

8
00:00:19,622 --> 00:00:23,559
But what is less impressive
that if you take this robot

9
00:00:23,559 --> 00:00:24,820
out of the factories

10
00:00:24,820 --> 00:00:29,223
where the environments are not
perfectly known and measured like here,

11
00:00:29,223 --> 00:00:33,588
to do even a simple task
which doesn't require much precision,

12
00:00:33,588 --> 00:00:35,222
and this is what can happen.

13
00:00:35,222 --> 00:00:37,512
I mean, opening a door,
you don't require much precision.

14
00:00:37,512 --> 00:00:38,901
(Laughter)

15
00:00:38,901 --> 00:00:41,706
Or, a small error in the measurements,

16
00:00:41,706 --> 00:00:43,472
you miss the ?? and that's it

17
00:00:43,472 --> 00:00:44,765
(Laughter)

18
00:00:44,765 --> 00:00:47,761
with no way of recovering
most of the time.

19
00:00:47,761 --> 00:00:49,331
So why is that?

20
00:00:49,331 --> 00:00:51,358
Well, for many years,

21
00:00:51,358 --> 00:00:54,856
robots have been designed
to emphasize speed and precision,

22
00:00:54,856 --> 00:00:57,828
and this translates
in a very specific architecture.

23
00:00:57,828 --> 00:00:58,874
If we take a robot term,

24
00:00:58,874 --> 00:01:01,626
it's a very well-defined
set of rigid links

25
00:01:01,626 --> 00:01:03,852
and mortars who are called actuators,

26
00:01:03,852 --> 00:01:05,920
they move the links above the joins.

27
00:01:05,920 --> 00:01:07,150
In this ?? structure,

28
00:01:07,150 --> 00:01:08,790
you have to perfectly
measure your environment,

29
00:01:08,790 --> 00:01:10,802
so what is around,

30
00:01:10,802 --> 00:01:13,649
and you have to perfectly
program every movement

31
00:01:13,649 --> 00:01:15,274
of the robot joints,

32
00:01:15,274 --> 00:01:19,158
because a small error
can generate a very large fault,

33
00:01:19,158 --> 00:01:22,194
so you can damage something
or you can get your robot damaged

34
00:01:22,194 --> 00:01:24,307
if something is harder.

35
00:01:24,307 --> 00:01:26,628
So let's talk about them a moment,

36
00:01:26,628 --> 00:01:29,397
and don't think about
the brains of these robots

37
00:01:29,397 --> 00:01:32,260
or how carefully we program them,

38
00:01:32,260 --> 00:01:34,607
but rather look at their bodies.

39
00:01:34,607 --> 00:01:37,557
There is obviously
something wrong with it,

40
00:01:37,557 --> 00:01:40,972
because what makes a robot
precise and strong

41
00:01:40,972 --> 00:01:45,504
also makes them ridiculously dangerous
and ineffective in the real world,

42
00:01:45,504 --> 00:01:47,572
because their body cannot deform

43
00:01:47,572 --> 00:01:51,506
or better adjust to the interaction
with the real world.

44
00:01:51,506 --> 00:01:54,647
So think about the opposite approach,

45
00:01:54,647 --> 00:01:58,178
being softer than
anything else around you.

46
00:01:58,178 --> 00:02:03,136
Well, maybe you think that you're not
really able to do anything if you're soft,

47
00:02:03,136 --> 00:02:04,150
probably.

48
00:02:04,150 --> 00:02:07,312
Well, nature teaches us the opposite.

49
00:02:07,312 --> 00:02:10,323
For example, at the bottom of the ocean
under thousands of pounds

50
00:02:10,323 --> 00:02:11,611
of ?? pressure,

51
00:02:11,611 --> 00:02:14,168
a completely soft animal

52
00:02:14,168 --> 00:02:18,078
can move and interact with a much
stiffer object than him.

53
00:02:18,078 --> 00:02:20,949
He works by carrying around
this coconut shell

54
00:02:20,949 --> 00:02:23,357
thanks to the flexibility
of his tentacles,

55
00:02:23,357 --> 00:02:26,465
which serve as both his feet and hands.

56
00:02:26,465 --> 00:02:30,820
And apparently, an octopus
can also open a jar.

57
00:02:30,820 --> 00:02:33,641
It's pretty impressive, right?

58
00:02:35,587 --> 00:02:40,354
But clearly, this is not enabled
just by the brain of this animal,

59
00:02:40,354 --> 00:02:42,614
but also by his body,

60
00:02:42,614 --> 00:02:46,895
and it's a clear example,
maybe the clearest example,

61
00:02:46,895 --> 00:02:48,608
of embodied intelligence,

62
00:02:48,608 --> 00:02:51,994
which is a kind of intelligence
that all living organisms have.

63
00:02:51,994 --> 00:02:53,309
We all have that.

64
00:02:53,309 --> 00:02:57,326
Our body, its shape,
material and structure,

65
00:02:57,326 --> 00:03:00,597
plays a fundamental role
during a physical task,

66
00:03:00,597 --> 00:03:05,702
because we can conform
to our environment

67
00:03:05,702 --> 00:03:08,660
so we can succeed in a large
variety of situations

68
00:03:08,660 --> 00:03:11,614
without much planning
or calculations ahead.

69
00:03:11,614 --> 00:03:15,445
So why don't we put
some of this embodied intelligence

70
00:03:15,445 --> 00:03:16,649
into our robotic machines

71
00:03:16,649 --> 00:03:18,559
to release them from relying
on excessive work

72
00:03:18,559 --> 00:03:20,868
on computation and sensing?

73
00:03:20,868 --> 00:03:23,918
Well, to do that we can follow
the strategy of nature,

74
00:03:23,918 --> 00:03:26,740
because with evolution,
she's done a pretty good job

75
00:03:26,740 --> 00:03:31,259
in designing machines
for environment interaction,

76
00:03:31,259 --> 00:03:35,776
and it's easy to notice that nature
uses soft material frequently

77
00:03:35,776 --> 00:03:38,094
and stiff material sparingly.

78
00:03:38,094 --> 00:03:41,909
And this is what is done
in this new field or robotics

79
00:03:41,909 --> 00:03:44,232
which is called soft robotics,

80
00:03:44,232 --> 00:03:47,991
in which the main objective
is not to make super-precise machines

81
00:03:47,991 --> 00:03:49,825
because we've already got them,

82
00:03:49,825 --> 00:03:54,117
but to make robots able to face
unexpected situations in the real world,

83
00:03:54,117 --> 00:03:56,562
so able to go out there.

84
00:03:56,562 --> 00:03:59,994
And what makes a robot soft
is first of all his compliant body,

85
00:03:59,994 --> 00:04:05,229
which is made of materials or structures
that can undergo very large deformations,

86
00:04:05,229 --> 00:04:07,308
so no more rigid links,

87
00:04:07,308 --> 00:04:11,366
and secondly to move them
we use what we call distributed actuation,

88
00:04:11,366 --> 00:04:15,973
so we have to control continuously
the shape of this very deformable body,

89
00:04:15,973 --> 00:04:19,294
which is the fact of having
a lot of links and joints,

90
00:04:19,294 --> 00:04:22,063
but we don't have
any stiff structure at all.

91
00:04:22,063 --> 00:04:25,305
So you can imagine that building
a soft robot is a very different process

92
00:04:25,305 --> 00:04:29,203
than stiff robotics, where
you have links, gears, screws

93
00:04:29,203 --> 00:04:31,458
that you must combine
in a very defined way.

94
00:04:31,458 --> 00:04:35,006
In soft robots, you just build
your actuator from scratch

95
00:04:35,006 --> 00:04:36,143
most of the time,

96
00:04:36,143 --> 00:04:38,394
but you shape your flexible material

97
00:04:38,394 --> 00:04:41,254
to the form that responds
to a certain input.

98
00:04:41,254 --> 00:04:44,538
For example here, you can just
deform a structure

99
00:04:44,538 --> 00:04:46,393
doing a fairly complex shape

100
00:04:46,393 --> 00:04:49,533
if you think about doing the same
with rigid links and joints,

101
00:04:49,533 --> 00:04:52,088
and here what you use is just one input,

102
00:04:52,088 --> 00:04:54,069
such as air pressure.

103
00:04:54,069 --> 00:04:57,582
Okay, but let's see
some cool examples of soft robots.

104
00:04:57,582 --> 00:05:00,921
Here is a little cute guy
developed by Harvard University,

105
00:05:00,921 --> 00:05:07,121
and he works thanks to waves
of pressure applied along its body,

106
00:05:07,121 --> 00:05:10,568
and thanks to the flexibility he can
also sneak under a low bridge,

107
00:05:10,568 --> 00:05:11,723
keep walking,

108
00:05:11,723 --> 00:05:15,434
and then keep walking
a little bit different afterwards.

109
00:05:15,434 --> 00:05:18,233
And it's a very preliminary prototype,

110
00:05:18,233 --> 00:05:21,076
but they also built a more robust version

111
00:05:21,076 --> 00:05:22,225
with power on board that can actually
be sent out in the world

112
00:05:22,225 --> 00:05:26,971
and face real-world interactions

113
00:05:26,971 --> 00:05:29,046
like a car passing it over it,

114
00:05:29,046 --> 00:05:30,764
and keep working.

115
00:05:30,764 --> 00:05:32,195
(Laughter)

116
00:05:32,195 --> 00:05:33,477
It's cute.

117
00:05:33,477 --> 00:05:34,983
(Laughter)

118
00:05:34,983 --> 00:05:38,625
Or a robotic fish which swims
like a real fish does

119
00:05:38,625 --> 00:05:42,486
in water simply because it has a soft tail
with distributed actuation

120
00:05:42,486 --> 00:05:44,154
using still air pressure.

121
00:05:44,154 --> 00:05:45,651
That was from MIT,

122
00:05:45,651 --> 00:05:48,365
and of course we have a robotic octopus.

123
00:05:48,365 --> 00:05:50,735
This was actually one of
the first projects developed

124
00:05:50,735 --> 00:05:52,618
in this new field of soft robots.

125
00:05:52,618 --> 00:05:54,733
Here you see the artificial tentacle,

126
00:05:54,733 --> 00:05:58,343
but they actually built and entire machine

127
00:05:58,343 --> 00:06:01,999
with several tentacles they could
just throw in the water,

128
00:06:01,999 --> 00:06:05,319
and you see that it can kind of go around
and do submarine exploration

129
00:06:05,319 --> 00:06:09,577
in a different way
than rigid robots would do.

130
00:06:09,577 --> 00:06:13,194
But this is very important for delicate
environments such as coral reefs.

131
00:06:13,194 --> 00:06:15,188
Let's go back to the ground.

132
00:06:15,188 --> 00:06:17,687
Here you see the view

133
00:06:17,687 --> 00:06:20,000
from a growing robot developed
by my colleagues in Stanford.

134
00:06:20,000 --> 00:06:21,970
You see the camera fixed on top.

135
00:06:21,970 --> 00:06:25,070
And this robot is particular
because using air pressure

136
00:06:25,070 --> 00:06:27,134
it grows from the tip
while the rest of the body

137
00:06:27,134 --> 00:06:29,627
stays in firm contact
with the environment.

138
00:06:29,627 --> 00:06:32,368
And this is inspired
by plants, not animals,

139
00:06:32,368 --> 00:06:35,597
which grows via the material
in a similar manner

140
00:06:35,597 --> 00:06:39,172
so it can face a pretty large
variety of situations.

141
00:06:39,172 --> 00:06:41,009
But I'm a biomedical engineer,

142
00:06:41,009 --> 00:06:43,459
and perhaps the application
I like the most

143
00:06:43,459 --> 00:06:45,016
it's in the medical field,

144
00:06:45,016 --> 00:06:49,929
and it's very difficult to imagine
a closer interaction with the human body

145
00:06:49,929 --> 00:06:51,744
than actually going inside the body,

146
00:06:51,744 --> 00:06:55,294
for example to inform
a minimally invasive procedure.

147
00:06:55,294 --> 00:06:58,680
And here, robots can be
very helpful with the surgeon

148
00:06:58,680 --> 00:07:00,600
because they must enter the body

149
00:07:00,600 --> 00:07:03,376
using small holes
and straight instruments,

150
00:07:03,376 --> 00:07:06,542
and these instruments must interact
with very delicate structures

151
00:07:06,542 --> 00:07:08,717
in a very uncertain environment,

152
00:07:08,717 --> 00:07:10,678
and this must be done safely.

153
00:07:10,678 --> 00:07:13,022
Also bringing the camera inside the body,

154
00:07:13,022 --> 00:07:17,126
so bringing the eyes of the surgeon
inside the subject, I feel,

155
00:07:17,126 --> 00:07:18,791
can be very challenging
if you use a rigid stick,

156
00:07:18,791 --> 00:07:20,422
like a classic endoscope.

157
00:07:20,422 --> 00:07:23,370
With my previous research group in Europe,

158
00:07:23,370 --> 00:07:26,175
we developed this
self-camera robot for surgery,

159
00:07:26,175 --> 00:07:29,500
which is very different
from a classic endoscope

160
00:07:29,500 --> 00:07:34,116
which can move thanks
to the flexibility of the module

161
00:07:34,116 --> 00:07:35,710
that can bend and elongate,

162
00:07:35,710 --> 00:07:37,955
bend in every direction
and also elongate.

163
00:07:37,955 --> 00:07:41,494
And this was actually used by surgeons
to see what they were doing

164
00:07:41,494 --> 00:07:43,443
with other instruments
from different points of view

165
00:07:43,443 --> 00:07:47,479
without caring that much
about what was touched around.

166
00:07:47,479 --> 00:07:51,245
And here you see the soft robot in action,

167
00:07:51,245 --> 00:07:54,576
and it just goes inside.

168
00:07:54,576 --> 00:07:57,461
This is a body simulator,
not a real human body.

169
00:07:57,461 --> 00:07:58,597
It goes around.

170
00:07:58,597 --> 00:08:01,076
You have a light, because usually
you don't have too many lights

171
00:08:01,076 --> 00:08:02,178
inside your body.

172
00:08:02,178 --> 00:08:03,423
(Laughter)

173
00:08:03,423 --> 00:08:04,588
We hope.

174
00:08:04,588 --> 00:08:07,970
(Laughter)

175
00:08:07,970 --> 00:08:12,222
But sometimes a surgical procedure
can even be done using a single needle,

176
00:08:12,222 --> 00:08:15,716
and in Stanford now, we are working
on a very flexible needle,

177
00:08:15,716 --> 00:08:19,124
kind of a very tiny soft robot

178
00:08:19,124 --> 00:08:23,709
which is mechanically designed
to use the interaction with the tissue

179
00:08:23,709 --> 00:08:24,837
and steer around inside a solid organ.

180
00:08:24,837 --> 00:08:29,068
This makes it possible to reach
many different targets such as tumors

181
00:08:29,068 --> 00:08:32,264
deep inside a solid organ

182
00:08:32,264 --> 00:08:32,987
by using one single insertion point,

183
00:08:32,987 --> 00:08:39,534
and you can even steer around
the structure that you want to avoid

184
00:08:39,534 --> 00:08:39,983
on the way to the target.

185
00:08:39,983 --> 00:08:42,946
So clearly, this is a pretty
exciting time for robotics.

186
00:08:42,946 --> 00:08:46,083
We have robots that have to deal
with soft structure,

187
00:08:46,083 --> 00:08:50,238
so this poses new
and very challenging questions

188
00:08:50,238 --> 00:08:51,204
for the robotics community,

189
00:08:51,204 --> 00:08:52,968
and indeed we are just starting
to learn how to control,

190
00:08:52,968 --> 00:08:55,935
how to put sensors
on these very flexible structures.

191
00:08:55,935 --> 00:08:58,619
But of course, we are not even close
to what nature figured out

192
00:08:58,619 --> 00:09:01,002
in millions of years of evolution.

193
00:09:01,002 --> 00:09:03,303
But one thing I know for sure:

194
00:09:03,303 --> 00:09:05,842
that robots will be softer and safer,

195
00:09:05,842 --> 00:09:08,375
and they will be out there helping people.

196
00:09:08,375 --> 00:09:09,191
Yeah.

197
00:09:09,191 --> 00:09:11,257
Thank you.

198
00:09:11,257 --> 00:09:14,516
(Applause)