Meet Spot, the robot dog that can run, hop and open doors | Marc Raibert

3,212,884 views ・ 2017-08-14

TED


Please double-click on the English subtitles below to play the video.

00:19
(Laughter)
0
19085
1800
00:24
(Laughter)
1
24323
1150
00:36
That's SpotMini.
2
36710
1150
00:37
He'll be back in a little while.
3
37884
1618
00:39
I --
4
39526
1164
00:40
(Applause)
5
40714
3838
00:45
I love building robots.
6
45472
1737
00:48
And my long-term goal is to build robots
7
48419
2452
00:50
that can do what people and animals do.
8
50895
2082
00:53
And there's three things in particular
9
53964
3099
00:57
that we're interested in.
10
57541
1703
01:00
One is balance and dynamic mobility,
11
60008
3302
01:03
the second one is mobile manipulation,
12
63334
2630
01:05
and the third one is mobile perception.
13
65988
2365
01:08
So, dynamic mobility and balance --
14
68944
2911
01:11
I'm going to do a demo for you.
15
71879
1637
01:15
I'm standing here, balancing.
16
75099
1547
01:18
I can see you're not very impressed. OK, how about now?
17
78121
2714
01:20
(Laughter)
18
80859
1160
01:22
How about now?
19
82043
1193
01:23
(Applause)
20
83260
2043
01:26
Those simple capabilities mean that people can go almost anywhere on earth,
21
86049
4373
01:30
on any kind of terrain.
22
90446
1652
01:32
We want to capture that for robots.
23
92122
2655
01:35
What about manipulation?
24
95772
1392
01:37
I'm holding this clicker in my hand;
25
97614
1716
01:39
I'm not even looking at it,
26
99354
1292
01:40
and I can manipulate it without any problem.
27
100670
2730
01:43
But even more important,
28
103424
2640
01:46
I can move my body while I hold the manipulator, the clicker,
29
106088
5160
01:52
and stabilize and coordinate my body,
30
112115
2174
01:54
and I can even walk around.
31
114313
1522
01:56
And that means I can move around in the world
32
116367
2898
01:59
and expand the range of my arms and my hands
33
119289
3302
02:02
and really be able to handle almost anything.
34
122615
2169
02:04
So that's mobile manipulation.
35
124808
1592
02:07
And all of you can do this.
36
127454
1775
02:09
Third is perception.
37
129253
1917
02:11
I'm looking at a room with over 1,000 people in it,
38
131194
3265
02:14
and my amazing visual system can see every one of you --
39
134483
4775
02:19
you're all stable in space,
40
139282
1961
02:21
even when I move my head,
41
141267
1319
02:22
even when I move around.
42
142610
1546
02:24
That kind of mobile perception is really important for robots
43
144180
3779
02:27
that are going to move and act
44
147983
1644
02:29
out in the world.
45
149651
1213
02:31
I'm going to give you a little status report
46
151865
2278
02:34
on where we are in developing robots toward these ends.
47
154167
3846
02:40
The first three robots are all dynamically stabilized robots.
48
160338
4540
02:44
This one goes back a little over 10 years ago --
49
164902
2648
02:47
"BigDog."
50
167574
1168
02:48
It's got a gyroscope that helps stabilize it.
51
168766
3310
02:52
It's got sensors and a control computer.
52
172100
3336
02:55
Here's a Cheetah robot that's running with a galloping gait,
53
175460
3167
02:58
where it recycles its energy,
54
178651
1696
03:00
it bounces on the ground,
55
180371
1567
03:01
and it's computing all the time
56
181962
1554
03:03
in order to keep itself stabilized and propelled.
57
183540
2887
03:07
And here's a bigger robot
58
187934
1828
03:09
that's got such good locomotion using its legs,
59
189786
2705
03:12
that it can go in deep snow.
60
192515
1404
03:13
This is about 10 inches deep,
61
193943
2411
03:16
and it doesn't really have any trouble.
62
196378
2111
03:20
This is Spot, a new generation of robot --
63
200166
2629
03:22
just slightly older than the one that came out onstage.
64
202819
2808
03:26
And we've been asking the question --
65
206258
1832
03:28
you've all heard about drone delivery:
66
208114
1974
03:30
Can we deliver packages to your houses with drones?
67
210112
2623
03:32
Well, what about plain old legged-robot delivery?
68
212759
2787
03:35
(Laughter)
69
215570
1132
03:36
So we've been taking our robot to our employees' homes
70
216726
3397
03:40
to see whether we could get in --
71
220147
1583
03:41
(Laughter)
72
221754
1024
03:42
the various access ways.
73
222802
1228
03:44
And believe me, in the Boston area,
74
224054
1884
03:45
there's every manner of stairway twists and turns.
75
225962
3174
03:49
So it's a real challenge.
76
229160
1237
03:50
But we're doing very well, about 70 percent of the way.
77
230421
2817
03:54
And here's mobile manipulation,
78
234479
1565
03:56
where we've put an arm on the robot,
79
236068
2591
03:58
and it's finding its way through the door.
80
238683
2334
04:01
Now, one of the important things about making autonomous robots
81
241742
3159
04:04
is to make them not do just exactly what you say,
82
244925
3220
04:08
but make them deal with the uncertainty of what happens in the real world.
83
248169
5141
04:14
So we have Steve there, one of the engineers,
84
254054
3011
04:17
giving the robot a hard time.
85
257089
1608
04:18
(Laughter)
86
258721
1032
04:19
And the fact that the programming still tolerates all that disturbance --
87
259777
4190
04:23
it does what it's supposed to.
88
263991
1497
04:25
Here's another example, where Eric is tugging on the robot
89
265512
2791
04:28
as it goes up the stairs.
90
268327
1298
04:29
And believe me,
91
269649
1152
04:30
getting it to do what it's supposed to do in those circumstances
92
270825
3272
04:34
is a real challenge,
93
274121
1341
04:35
but the result is something that's going to generalize
94
275486
2740
04:38
and make robots much more autonomous than they would be otherwise.
95
278250
3871
04:43
This is Atlas, a humanoid robot.
96
283317
2431
04:46
It's a third-generation humanoid that we've been building.
97
286427
3905
04:50
I'll tell you a little bit about the hardware design later.
98
290671
2818
04:53
And we've been saying:
99
293513
1176
04:54
How close to human levels of performance and speed could we get
100
294713
4423
04:59
in an ordinary task,
101
299160
1575
05:00
like moving boxes around on a conveyor?
102
300759
2469
05:03
We're getting up to about two-thirds of the speed that a human operates
103
303931
5306
05:09
on average.
104
309261
1162
05:10
And this robot is using both hands, it's using its body,
105
310758
2929
05:13
it's stepping,
106
313711
1169
05:14
so it's really an example of dynamic stability,
107
314904
2727
05:17
mobile manipulation
108
317655
1379
05:19
and mobile perception.
109
319058
1611
05:22
Here --
110
322192
1173
05:23
(Laughter)
111
323905
1907
05:26
We actually have two Atlases.
112
326389
1610
05:28
(Laughter)
113
328483
1185
05:30
Now, everything doesn't go exactly the way it's supposed to.
114
330198
3396
05:33
(Laughter)
115
333618
1702
05:38
(Laughter)
116
338318
1528
05:40
(Laughter)
117
340536
1871
05:45
And here's our latest robot, called "Handle."
118
345292
2544
05:48
Handle is interesting, because it's sort of half like an animal,
119
348457
4118
05:52
and it's half something else
120
352599
2314
05:54
with these leg-like things and wheels.
121
354937
2756
05:58
It's got its arms on in kind of a funny way,
122
358141
3026
06:01
but it really does some remarkable things.
123
361191
2096
06:03
It can carry 100 pounds.
124
363311
3331
06:06
It's probably going to lift more than that,
125
366666
2059
06:08
but so far we've done 100.
126
368749
1778
06:10
It's got some pretty good rough-terrain capability,
127
370551
2436
06:13
even though it has wheels.
128
373011
1428
06:17
And Handle loves to put on a show.
129
377881
2366
06:20
(Laughter)
130
380686
1358
06:24
(Applause)
131
384644
5129
06:30
I'm going to give you a little bit of robot religion.
132
390740
2998
06:34
A lot of people think that a robot is a machine where there's a computer
133
394356
4329
06:38
that's telling it what to do,
134
398709
1634
06:40
and the computer is listening through its sensors.
135
400900
2816
06:44
But that's really only half of the story.
136
404349
2404
06:46
The real story is that the computer is on one side,
137
406777
3109
06:49
making suggestions to the robot,
138
409910
1944
06:51
and on the other side are the physics of the world.
139
411878
2492
06:54
And that physics involves gravity, friction, bouncing into things.
140
414778
4922
07:00
In order to have a successful robot,
141
420238
1835
07:02
my religion is that you have to do a holistic design,
142
422097
4591
07:06
where you're designing the software, the hardware and the behavior
143
426712
3553
07:10
all at one time,
144
430289
1236
07:11
and all these parts really intermesh and cooperate with each other.
145
431549
3702
07:15
And when you get the perfect design, you get a real harmony
146
435275
3107
07:18
between all those parts interacting with each other.
147
438406
2879
07:22
So it's half software and half hardware,
148
442269
2260
07:24
plus the behavior.
149
444553
1291
07:26
We've done some work lately on the hardware, where we tried to go --
150
446698
3542
07:30
the picture on the left is a conventional design,
151
450264
2477
07:32
where you have parts that are all bolted together,
152
452765
2900
07:35
conductors, tubes, connectors.
153
455689
2740
07:38
And on the right is a more integrated thing;
154
458453
2049
07:40
it's supposed to look like an anatomy drawing.
155
460526
2401
07:43
Using the miracle of 3-D printing,
156
463330
2510
07:45
we're starting to build parts of robots
157
465864
2637
07:48
that look a lot more like the anatomy of an animal.
158
468525
2897
07:51
So that's an upper-leg part that has hydraulic pathways --
159
471446
3344
07:54
actuators, filters --
160
474814
1930
07:56
all embedded, all printed as one piece,
161
476768
2380
07:59
and the whole structure is developed
162
479172
3202
08:02
with a knowledge of what the loads and behavior are going to be,
163
482398
3036
08:05
which is available from data recorded from robots
164
485458
2903
08:08
and simulations and things like that.
165
488385
1809
08:10
So it's a data-driven hardware design.
166
490218
2937
08:13
And using processes like that,
167
493547
1726
08:15
not only the upper leg but some other things,
168
495297
2244
08:17
we've gotten our robots to go from big, behemoth, bulky, slow, bad robots --
169
497565
5140
08:22
that one on the right, weighing almost 400 pounds --
170
502729
3619
08:26
down to the one in the middle which was just in the video,
171
506372
3109
08:29
weighs about 190 pounds,
172
509505
1563
08:31
just a little bit more than me,
173
511092
1697
08:32
and we have a new one,
174
512813
1486
08:34
which is working but I'm not going to show it to you yet,
175
514323
2742
08:37
on the left,
176
517089
1162
08:38
which weighs just 165 pounds,
177
518275
1644
08:39
with all the same strength and capabilities.
178
519943
2266
08:42
So these things are really getting better very quickly.
179
522233
2730
08:46
So it's time for Spot to come back out,
180
526280
3262
08:49
and we're going to demonstrate a little bit of mobility,
181
529566
3748
08:53
dexterity and perception.
182
533338
1600
08:55
This is Seth Davis, who's my robot wrangler today,
183
535681
3764
08:59
and he's giving Spot some general direction
184
539469
3040
09:02
by steering it around,
185
542533
1826
09:04
but all the coordination of the legs and the sensors
186
544383
3071
09:07
is done by the robot's computers on board.
187
547478
2509
09:10
The robot can walk with a number of different gaits;
188
550543
3446
09:14
it's got a gyro,
189
554013
2159
09:16
or a solid-state gyro,
190
556196
1337
09:17
an IMU on board.
191
557557
1420
09:19
Obviously, it's got a battery, and things like that.
192
559442
3027
09:23
One of the cool things about a legged robot is,
193
563352
2506
09:25
it's omnidirectional.
194
565882
1449
09:27
In addition to going forward, it can go sideways,
195
567355
2773
09:31
it can turn in place.
196
571413
1440
09:36
And this robot is a little bit of a show-off.
197
576482
2268
09:39
It loves to use its dynamic gaits,
198
579675
1960
09:41
like running --
199
581659
1158
09:42
(Laughter)
200
582841
1044
09:43
And it's got one more.
201
583909
1516
09:47
(Laughter)
202
587104
1830
09:49
Now if it were really a show-off, it would be hopping on one foot,
203
589858
3236
09:53
but, you know.
204
593118
1168
09:54
Now, Spot has a set of cameras here, stereo cameras,
205
594310
4043
09:58
and we have a feed up in the center.
206
598377
1830
10:00
It's kind of dark out in the audience,
207
600591
1824
10:02
but it's going to use those cameras in order to look at the terrain
208
602439
3220
10:05
right in front of it,
209
605683
1173
10:06
while it goes over these obstacles back here.
210
606880
2761
10:09
For this demo, Seth is steering,
211
609665
3313
10:13
but the robot's doing all its own terrain planning.
212
613002
2505
10:15
This is a terrain map,
213
615531
1502
10:17
where the data from the cameras is being developed in real time,
214
617057
5170
10:22
showing the red spots, which are where it doesn't want to step,
215
622251
3237
10:25
and the green spots are the good places.
216
625512
1991
10:27
And here it's treating them like stepping-stones.
217
627527
2449
10:30
So it's trying to stay up on the blocks,
218
630000
2755
10:32
and it adjusts its stride,
219
632779
1273
10:34
and there's a ton of planning
220
634076
1461
10:35
that has to go into an operation like that,
221
635561
2241
10:37
and it does all that planning in real time,
222
637826
2262
10:40
where it adjusts the steps a little bit longer
223
640112
2430
10:42
or a little bit shorter.
224
642566
1276
10:45
Now we're going to change it into a different mode,
225
645216
2430
10:47
where it's just going to treat the blocks like terrain
226
647670
3423
10:51
and decide whether to step up or down
227
651117
3719
10:54
as it goes.
228
654860
1285
10:57
So this is using dynamic balance
229
657085
2969
11:00
and mobile perception,
230
660078
1858
11:01
because it has to coordinate what it sees along with how it's moving.
231
661960
5387
11:08
The other thing Spot has is a robot arm.
232
668988
4318
11:14
Some of you may see that as a head and a neck,
233
674337
2490
11:16
but believe me, it's an arm.
234
676851
1585
11:18
Seth is driving it around.
235
678460
1765
11:20
He's actually driving the hand and the body is following.
236
680249
3768
11:24
So the two are coordinated in the way I was talking about before --
237
684041
4133
11:28
in the way people can do that.
238
688198
1864
11:30
In fact, one of the cool things Spot can do we call, "chicken-head mode,"
239
690086
4643
11:34
and it keeps its head in one place in space,
240
694753
3184
11:37
and it moves its body all around.
241
697961
1845
11:40
There's a variation of this that's called "twerking" --
242
700553
2602
11:43
(Laughter)
243
703179
1016
11:44
but we're not going to use that today.
244
704219
1823
11:46
(Laughter)
245
706066
1071
11:47
So, Spot: I'm feeling a little thirsty. Could you get me a soda?
246
707161
3586
11:51
For this demo, Seth is not doing any driving.
247
711293
3773
11:55
We have a LIDAR on the back of the robot,
248
715090
2113
11:57
and it's using these props we've put on the stage
249
717227
2537
11:59
to localize itself.
250
719788
1393
12:01
It's gone over to that location.
251
721205
2281
12:03
Now it's using a camera that's in its hand
252
723510
2882
12:06
to find the cup,
253
726416
1786
12:08
picks it up --
254
728853
1255
12:10
and again, Seth's not driving.
255
730132
1951
12:13
We've planned out a path for it to go --
256
733372
3241
12:16
it looked like it was going off the path --
257
736637
2107
12:18
and now Seth's going to take over control again,
258
738768
2358
12:21
because I'm a little bit chicken about having it do this by itself.
259
741150
3449
12:24
Thank you, Spot.
260
744623
1376
12:28
(Applause)
261
748231
5283
12:34
So, Spot:
262
754889
1590
12:36
How do you feel about having just finished your TED performance?
263
756503
3567
12:41
(Laughter)
264
761014
2689
12:44
Me, too!
265
764068
1152
12:45
(Laughter)
266
765244
1032
12:46
Thank you all,
267
766300
1816
12:48
and thanks to the team at Boston Dynamics,
268
768140
2395
12:50
who did all the hard work behind this.
269
770559
2114
12:52
(Applause)
270
772697
2123
13:03
Helen Walters: Marc, come back in the middle.
271
783059
2163
13:05
Thank you so much.
272
785246
1154
13:06
Come over here, I have questions.
273
786424
2095
13:08
So, you mentioned the UPS and the package delivery.
274
788543
3415
13:11
What are the other applications that you see for your robots?
275
791982
3937
13:15
Marc Raibert: You know, I think that robots
276
795943
2027
13:17
that have the capabilities I've been talking about
277
797994
2406
13:20
are going to be incredibly useful.
278
800424
1660
13:22
About a year ago, I went to Fukushima
279
802108
2497
13:24
to see what the situation was there,
280
804629
2031
13:26
and there's just a huge need
281
806684
1877
13:28
for machines that can go into some of the dirty places
282
808585
3338
13:31
and help remediate that.
283
811947
1968
13:34
I think it won't be too long until we have robots like this in our homes,
284
814528
4599
13:39
and one of the big needs is to take care of the aging
285
819151
5321
13:44
and invalids.
286
824496
1366
13:45
I think that it won't be too long till we're using robots
287
825886
3989
13:49
to help take care of our parents,
288
829899
2462
13:52
or probably more likely, have our children help take care of us.
289
832385
4415
13:57
And there's a bunch of other things.
290
837721
1750
13:59
I think the sky's the limit.
291
839495
1366
14:00
Many of the ideas we haven't thought of yet,
292
840885
2258
14:03
and people like you will help us think of new applications.
293
843167
3595
14:06
HW: So what about the dark side?
294
846786
1587
14:08
What about the military?
295
848397
1833
14:10
Are they interested?
296
850254
1503
14:12
MR: Sure, the military has been a big funder of robotics.
297
852441
3634
14:16
I don't think the military is the dark side myself,
298
856099
4202
14:20
but I think, as with all advanced technology,
299
860325
3871
14:24
it can be used for all kinds of things.
300
864220
2047
14:26
HW: Awesome. Thank you so much.
301
866291
1641
14:27
MR: OK, you're welcome.
302
867956
1405
14:29
Thank you.
303
869385
1152
14:30
(Applause)
304
870561
1586
About this website

This site will introduce you to YouTube videos that are useful for learning English. You will see English lessons taught by top-notch teachers from around the world. Double-click on the English subtitles displayed on each video page to play the video from there. The subtitles scroll in sync with the video playback. If you have any comments or requests, please contact us using this contact form.

https://forms.gle/WvT1wiN1qDtmnspy7