Robots that fly ... and cooperate | Vijay Kumar

2,184,417 views ใƒป 2012-03-01

TED


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืžืชืจื’ื: Yubal Masalker ืžื‘ืงืจ: Ido Dekkers
00:20
Good morning.
0
20566
1159
ื‘ื•ืงืจ ื˜ื•ื‘.
00:22
I'm here today to talk about autonomous flying beach balls.
1
22853
4153
ืื ื™ ื›ืืŸ ื›ื“ื™ ืœื“ื‘ืจ
ืขืœ ื›ื“ื•ืจื™-ืžื™ื ื”ื˜ืกื™ื ืขืฆืžืื™ืช.
00:27
(Laughter)
2
27030
1008
ืœื, ืืœื ืขืœ ืจื•ื‘ื•ื˜ื™ื ืงืœื™-ืชื ื•ืขื” ืฉื˜ืกื™ื ื›ืžื• ื–ื” ื›ืืŸ.
00:28
No, agile aerial robots like this one.
3
28062
2865
00:31
I'd like to tell you a little bit about the challenges in building these,
4
31624
3460
ืืกืคืจ ืœื›ื ืžืขื˜ ืขืœ ื”ืืชื’ืจื™ื ื‘ื‘ื ื™ื™ืชื
ื•ืขืœ ื›ืžื” ืžื”ืืคืฉืจื•ื™ื•ืช ื”ื ืคืœืื•ืช
00:35
and some of the terrific opportunities for applying this technology.
5
35108
3392
ืœืฉื™ืžื•ืฉ ื‘ื˜ื›ื ื•ืœื•ื’ื™ื” ื–ื•.
00:38
So these robots are related to unmanned aerial vehicles.
6
38957
4608
ืจื•ื‘ื•ื˜ื™ื ื”ืœืœื•
ืงืฉื•ืจื™ื ื‘ื›ืœื™-ื˜ื™ืก ื‘ืœืชื™ ืžืื•ื™ื™ืฉื™ื.
ืื‘ืœ ื”ื›ืœื™ื ืฉืจื•ืื™ื ื›ืืŸ ื”ื ื’ื“ื•ืœื™ื.
00:44
However, the vehicles you see here are big.
7
44199
2651
ืžืฉืงืœื ืืœืคื™ ืงื™ืœื•ื’ืจืžื™ื,
00:47
They weigh thousands of pounds, are not by any means agile.
8
47318
3269
ื•ื”ื ื›ืœืœ ืื™ื ื ืงืœื™-ืชื ื•ืขื”.
00:50
They're not even autonomous.
9
50611
1625
ื”ื ืืคื™ืœื• ืœื ืขืฆืžืื™ื™ื.
00:52
In fact, many of these vehicles are operated by flight crews
10
52886
4182
ืœืžืขืฉื”, ืจื‘ื™ื ืžื”ื
ืžื•ืคืขืœื™ื ืขืœ-ื™ื“ื™ ืฆื•ื•ืชื™ ื”ื˜ืกื”
ื”ืขืฉื•ื™ื™ื ืœื›ืœื•ืœ ืžืกืคืจ ื˜ื™ื™ืกื™ื,
00:57
that can include multiple pilots,
11
57092
2560
00:59
operators of sensors,
12
59676
2206
ืžืคืขื™ืœื™ ื—ื™ื™ืฉื ื™ื
01:01
and mission coordinators.
13
61906
1330
ื•ืžื ื”ืœื™ ืžืฉื™ืžื•ืช.
01:03
What we're interested in is developing robots like this --
14
63922
2743
ืื ื• ืžืขื•ื ื™ื™ื ื™ื ืœืคืชื— ืจื•ื‘ื•ื˜ื™ื ื›ืžื• ืืœื” --
ืืœื• ืฉืชื™ ืชืžื•ื ื•ืช ื ื•ืกืคื•ืช --
01:06
and here are two other pictures --
15
66689
1690
ืฉืœ ืจื•ื‘ื•ื˜ื™ื ืฉืืคืฉืจ ืœืจื›ื•ืฉ ืžื”ืžื“ืฃ.
01:08
of robots that you can buy off the shelf.
16
68403
2683
ื•ื‘ื›ืŸ, ืืœื” ื”ื ืžืกื•ืงื™ื ื‘ืขืœื™ 4 ืจื•ื˜ื•ืจื™ื
01:11
So these are helicopters with four rotors,
17
71110
3284
ื’ื•ื“ืœื ื›ืžื˜ืจ
01:14
and they're roughly a meter or so in scale,
18
74418
3690
ื•ืžืฉืงืœื ืงื™ืœื•ื’ืจืžื™ื ืื—ื“ื™ื.
01:18
and weigh several pounds.
19
78132
1460
ืžืชืงื™ื ื™ื ืขืœื™ื”ื ื—ื™ื™ืฉื ื™ื ื•ืžืขื‘ื“ื™ื,
01:20
And so we retrofit these with sensors and processors,
20
80076
3355
ื•ืจื•ื‘ื•ื˜ื™ื ื”ืœืœื• ื™ื›ื•ืœื™ื ืœื˜ื•ืก
01:23
and these robots can fly indoors.
21
83455
2206
ื‘ืžื‘ื ื” ืกื’ื•ืจ ืœืœื GPS (ื ื™ื•ื•ื˜ ืœื•ื•ื™ื™ื ื™).
01:25
Without GPS.
22
85685
1319
ื”ืจื•ื‘ื•ื˜ ืฉืื ื™ ืžื—ื–ื™ืง ื‘ื™ื“ื™
01:27
The robot I'm holding in my hand
23
87457
1888
ื”ื•ื ื–ื”,
01:29
is this one,
24
89369
1650
ื•ื”ื•ื ื ื‘ื ื” ืขืœ-ื™ื“ื™ 2 ืกื˜ื•ื“ื ื˜ื™ื,
01:31
and it's been created by two students,
25
91043
3143
ืืœื›ืก ื•ื“ื ื™ืืœ.
01:34
Alex and Daniel.
26
94210
1648
ื”ื•ื ืฉื•ืงืœ
01:36
So this weighs a little more than a tenth of a pound.
27
96493
3115
ื›-50 ื’ืจื.
01:39
It consumes about 15 watts of power.
28
99632
2579
ื”ื•ื ืฆื•ืจืš ื›-15 ื•ืื˜ ืื ืจื’ื™ื”.
ื•ื›ืคื™ ืฉื ื™ืชืŸ ืœืจืื•ืช,
01:42
And as you can see, it's about eight inches in diameter.
29
102235
2833
ืงื•ื˜ืจื• ื›-20 ืก"ืž.
ืื– ืชื ื• ืœื™ ืœืชืืจ ื‘ืงืฆืจื” ื›ื™ืฆื“
01:46
So let me give you just a very quick tutorial
30
106330
2595
01:48
on how these robots work.
31
108949
1666
ื”ืจื•ื‘ื•ื˜ื™ื ื”ืœืœื• ืคื•ืขืœื™ื.
ื›ืืžื•ืจ ื™ืฉ ืœื• 4 ืจื•ื˜ื•ืจื™ื.
01:51
So it has four rotors.
32
111043
1193
01:52
If you spin these rotors at the same speed,
33
112260
2048
ืื ืžืกื•ื‘ื‘ื™ื ืื•ืชื ื‘ืžื”ื™ืจื•ืช ื–ื”ื”,
01:54
the robot hovers.
34
114332
1199
ื”ืจื•ื‘ื•ื˜ ืžืจื—ืฃ.
01:56
If you increase the speed of each of these rotors,
35
116420
3636
ืื ืžื’ื‘ื™ืจื™ื ืžื”ื™ืจื•ืช ืฉืœ ื›ืœ ืื—ื“ ืžื”ืจื•ื˜ื•ืจื™ื,
ื”ืจื•ื‘ื•ื˜ ื˜ืก ืœืžืขืœื”, ื”ื•ื ืžืื™ืฅ ื›ืœืคื™ ืžืขืœื”.
02:00
then the robot flies up, it accelerates up.
36
120080
2539
02:02
Of course, if the robot were tilted,
37
122643
2432
ื‘ืจื•ืจ ืฉืื ืžื˜ื™ื ืื•ืชื•,
ืœืžืฆื‘ ืื•ืคืงื™,
02:05
inclined to the horizontal,
38
125099
1310
02:06
then it would accelerate in this direction.
39
126433
2803
ื”ื•ื ื™ื˜ื•ืก ืœื›ื™ื•ื•ืŸ ื”ื–ื”.
02:09
So to get it to tilt,
40
129686
1413
ื›ื“ื™ ืœื”ื˜ื•ืช ืื•ืชื•, ื™ืฉ ืฉืชื™ ื“ืจื›ื™ื.
02:11
there's one of two ways of doing it.
41
131123
2093
ื‘ืชืžื•ื ื” ื–ื•
02:13
So in this picture, you see that rotor four is spinning faster
42
133240
3636
ืจื•ืื™ื ืฉืจื•ื˜ื•ืจ 4 ืžืกืชื•ื‘ื‘ ื™ื•ืชืจ ืžื”ืจ
02:16
and rotor two is spinning slower.
43
136900
1968
ื•ืจื•ื˜ื•ืจ 2 ืžืกืชื•ื‘ื‘ ื™ื•ืชืจ ืœืื˜.
02:18
And when that happens,
44
138892
1560
ื•ื›ืืฉืจ ื–ื” ืงื•ืจื”
02:20
there's a moment that causes this robot to roll.
45
140476
3051
ื ื•ืฆืจ ืžื•ืžื ื˜ ื”ื’ื•ืจื ืœืจื•ื‘ื•ื˜ ืœื”ืชื’ืœื’ืœ.
ื•ืœืฆื“ ื”ืฉื ื™,
02:24
And the other way around,
46
144495
1255
02:25
if you increase the speed of rotor three and decrease the speed of rotor one,
47
145774
5579
ืื ืžื’ื‘ื™ืจื™ื ืืช ื”ืžื”ื™ืจื•ืช ืฉืœ ืจื•ื˜ื•ืจ 3
ื•ืžื•ืจื™ื“ื™ื ืืช ื”ืžื”ื™ืจื•ืช ืฉืœ ืจื•ื˜ื•ืจ 1,
ืื– ื”ืจื•ื‘ื•ื˜ ืžืชืงื“ื.
02:31
then the robot pitches forward.
48
151377
1859
02:33
And then finally,
49
153820
1168
ื•ืœื‘ืกื•ืฃ,
02:35
if you spin opposite pairs of rotors
50
155012
2894
ืื ืžืกื•ื‘ื‘ื™ื ื–ื•ื’ ืจื•ื˜ื•ืจื™ื ืžื ื•ื’ื“ื™ื
02:37
faster than the other pair,
51
157930
1692
ื™ื•ืชืจ ืžื”ืจ ืžื”ื–ื•ื’ ื”ืื—ืจ,
02:39
then the robot yaws about the vertical axis.
52
159646
2721
ื”ืจื•ื‘ื•ื˜ ื—ื’ ืกื‘ื™ื‘ ื”ืฆื™ืจ ื”ืื ื›ื™.
ืžืขื‘ื“ ื‘ื›ืœื™-ื”ื˜ื™ืก ืขื•ืงื‘
02:42
So an on-board processor
53
162391
1460
02:43
essentially looks at what motions need to be executed
54
163875
3307
ืื—ืจ ื”ืชื ื•ืขื•ืช ืฉื“ืจื•ืฉ ืœื”ื•ืฆื™ืืŸ ืœืคื•ืขืœ
ื•ืžื—ืฉื‘ ืืœื• ืคืงื•ื“ื•ืช
02:47
and combines these motions,
55
167206
1970
ืœืฉื’ืจ ืœืžื ื•ืขื™ื ื›ื“ื™ ืœืฉืœื‘ ืืช ื”ืชื ื•ืขื•ืช,
02:49
and figures out what commands to send to the motors --
56
169200
3110
600 ืคืขื ื‘ืฉื ื™ื”.
02:52
600 times a second.
57
172334
1412
02:53
That's basically how this thing operates.
58
173770
1969
ื›ืš ื‘ืขื™ืงืจื•ืŸ ื“ื‘ืจ ื–ื” ืขื•ื‘ื“.
ืื—ื“ ื”ื™ืชืจื•ื ื•ืช ืฉืœ ื”ืฆื•ืจื” ื”ื–ื•
02:56
So one of the advantages of this design
59
176247
2142
ื”ื•ื ืฉื›ืืฉืจ ืžืงื˜ื™ื ื™ื ืืช ื”ื’ื•ื“ืœ,
02:58
is when you scale things down,
60
178413
1899
ื”ืจื•ื‘ื•ื˜ ื”ื•ืคืš ืœื”ื™ื•ืช ืงืœ-ืชื ื•ืขื”.
03:00
the robot naturally becomes agile.
61
180336
2318
R ื”ื•ื ืื•ืจืš
03:03
So here, R is the characteristic length of the robot.
62
183194
4063
ืื•ืคื™ื™ื ื™ ืฉืœ ื”ืจื•ื‘ื•ื˜.
ื‘ืขืฆื ื–ื” ืžื—ืฆื™ืช ืžื”ืงื•ื˜ืจ.
03:07
It's actually half the diameter.
63
187281
1674
03:09
And there are lots of physical parameters that change as you reduce R.
64
189595
4641
ื™ืฉื ื ื”ืžื•ืŸ ืคืจืžื˜ืจื™ื ืคื™ื–ื™ืงืœื™ื™ื ืฉืžืฉืชื ื™ื
ื›ื›ืœ ืฉืžื•ืจื™ื“ื™ื ืืช R.
03:14
The one that's most important is the inertia,
65
194968
2631
ืื—ื“ ืฉื”ื•ื ื”ื—ืฉื•ื‘ ื‘ื™ื•ืชืจ
ื”ื•ื ื”ืื™ื ืจืฆื™ื” ืื• ื”ื”ืชื ื’ื“ื•ืช ืœืชื ื•ืขื”.
03:17
or the resistance to motion.
66
197623
1921
ืžืชื‘ืจืจ ืฉื”ืื™ื ืจืฆื™ื”,
03:19
So it turns out the inertia, which governs angular motion,
67
199568
4476
ืืฉืจ ืฉื•ืœื˜ืช ื‘ืชื ื•ืขื” ื–ื•ื™ืชื™ืช,
ืžื•ืฉืคืขืช ืž-R ื‘ืžืขืœื” ื”ื—ืžื™ืฉื™ืช.
03:24
scales as a fifth power of R.
68
204068
2600
ืœื›ืŸ ื›ื›ืœ ืฉืžืงื˜ื™ื ื™ื ืืช R,
03:27
So the smaller you make R,
69
207046
1696
03:28
the more dramatically the inertia reduces.
70
208766
2317
ื”ืื™ื ืจืฆื™ื” ื™ื•ืจื“ืช ื‘ืื•ืคืŸ ื“ืจืžื˜ื™.
03:31
So as a result, the angular acceleration,
71
211956
2935
ื›ืชื•ืฆืื” ืžื–ื”, ื”ืชืื•ืฆื” ื”ื–ื•ื™ืชื™ืช,
03:34
denoted by the Greek letter alpha here,
72
214915
2039
ื”ืžืฆื•ื™ื™ื ืช ื›ืืŸ ื‘ืืžืฆืขื•ืช ื”ืื•ืช ืืœืคื,
03:36
goes as 1 over R.
73
216978
1619
ื”ื™ื ืื—ื“ ื—ืœืงื™ R.
03:38
It's inversely proportional to R.
74
218621
1729
ื”ื™ื ื‘ื™ื—ืก ื”ืคื•ืš ืœ-R.
03:40
The smaller you make it, the more quickly you can turn.
75
220374
2770
ื›ื›ืœ ืฉืžืงื˜ื™ื ื™ื ืืช R, ื›ืš ื ื™ืชืŸ ืœื”ืกืชื•ื‘ื‘ ื™ื•ืชืจ ืžื”ืจ.
ื–ื” ื‘ืจื•ืจ ืžืชื•ืš ื”ืกืจื˜ื•ื ื™ื ื”ืืœื”.
03:44
So this should be clear in these videos.
76
224219
1961
ื‘ืชื—ืชื™ืช ืžื™ืžื™ืŸ ืจื•ืื™ื ืจื•ื‘ื•ื˜ ื”ืžื‘ืฆืข
03:46
On the bottom right, you see a robot performing a 360-degree flip
77
226204
4459
ืกืœื˜ื” ืฉืœ 360 ืžืขืœื•ืช
03:50
in less than half a second.
78
230687
1548
ืชื•ืš ืคื—ื•ืช ืžื—ืฆื™ ืฉื ื™ื”.
03:52
Multiple flips, a little more time.
79
232656
2580
ื•ืกืœื˜ื•ืช ืžืจื•ื‘ื•ืช ื‘ืงืฆืช ื™ื•ืชืจ ื–ืžืŸ.
ื›ืืŸ ื”ืžืขื‘ื“ื™ื ืฉื‘ืจื•ื‘ื•ื˜ ื”ื˜ืก
03:56
So here the processes on board
80
236259
1881
ืžืงื‘ืœื™ื ืžืฉื•ื‘ ืžืžื“ื™-ื”ืชืื•ืฆื”
03:58
are getting feedback from accelerometers and gyros on board,
81
238164
3184
ื•ืžื’'ืื™ื™ืจื•ืื™ื ืฉื‘ื›ืœื™
04:01
and calculating, like I said before,
82
241372
2118
ื•ืžื‘ืฆืขื™ื ื—ื™ืฉื•ื‘ื™ื ืฉืœ ืคืงื•ื“ื•ืช,
04:03
commands at 600 times a second,
83
243514
1809
ื›ืืžื•ืจ, 600 ืคืขื ื‘ืฉื ื™ื”
04:05
to stabilize this robot.
84
245347
1889
ื›ื“ื™ ืœื™ื™ืฆื‘ ืืช ื”ืจื•ื‘ื•ื˜.
04:07
So on the left, you see Daniel throwing this robot up into the air,
85
247607
3350
ืžืฉืžืืœ ืจื•ืื™ื ืืช ื“ื ื™ืืœ ื–ื•ืจืง ืืช ื”ืจื•ื‘ื•ื˜ ืœืื•ื™ืจ.
04:10
and it shows you how robust the control is.
86
250981
2027
ื•ื–ื” ืžืจืื” ืœื›ื ืืช ืขื•ืฆืžืช ื”ืฉืœื™ื˜ื”.
ืœื ืžืฉื ื” ื›ื™ืฆื“ ื–ื•ืจืงื™ื ืื•ืชื•,
04:13
No matter how you throw it,
87
253032
1310
04:14
the robot recovers and comes back to him.
88
254366
2912
ื”ืจื•ื‘ื•ื˜ ืžืชืื•ืฉืฉ ื•ื—ื•ื–ืจ ืืœื™ื•.
04:18
So why build robots like this?
89
258881
1777
ืื– ืžื“ื•ืข ืœื‘ื ื•ืช ืจื•ื‘ื•ื˜ื™ื ื›ืืœื”?
ืœืจื•ื‘ื•ื˜ื™ื ื›ืืœื” ื™ืฉ ื”ืจื‘ื” ืฉื™ืžื•ืฉื™ื.
04:21
Well, robots like this have many applications.
90
261079
2571
ื ื™ืชืŸ ืœืฉื’ืจื ืœืชื•ืš ื‘ื ื™ื™ื ื™ื ื›ืžื• ื–ื”
04:24
You can send them inside buildings like this,
91
264198
2471
04:26
as first responders to look for intruders,
92
266693
3587
ื›ืžืขื ื” ืจืืฉื•ืŸ ื›ื“ื™ ืœื—ืคืฉ ืคื•ืจืฆื™ื,
ืื•ืœื™ ื›ื“ื™ ืœื—ืคืฉ ื“ืœื™ืคื•ืช ื‘ื™ื•ื›ื™ืžื™ื•ืช,
04:30
maybe look for biochemical leaks,
93
270304
3301
ื“ืœื™ืคื•ืช ื’ื–.
04:33
gaseous leaks.
94
273629
1151
ื ื™ืชืŸ ื’ื ืœื”ืฉืชืžืฉ ื‘ื”ื
04:35
You can also use them for applications like construction.
95
275102
3714
ืœืžื˜ืจื•ืช ื›ืžื• ื‘ื ื™ื”.
04:38
So here are robots carrying beams, columns
96
278840
4524
ื”ื ื” ืจื•ื‘ื•ื˜ื™ื ื ื•ืฉืื™ื ืงื•ืจื•ืช, ืขืžื•ื“ื™ื
ื•ืžืจื›ื™ื‘ื™ื ืžื‘ื ื™ื ื“ืžื•ื™ื™-ืงื•ื‘ื™ื•ืช.
04:43
and assembling cube-like structures.
97
283388
1848
04:45
I'll tell you a little bit more about this.
98
285260
2056
ืืกืคืจ ืœื›ื ืขืœ ื›ืš ืงืฆืช ื™ื•ืชืจ.
04:48
The robots can be used for transporting cargo.
99
288695
2541
ื ื™ืชืŸ ืœื”ืฉืชืžืฉ ื‘ืจื•ื‘ื•ื˜ื™ื ืœื”ืขื‘ืจืช ืžื˜ืขืŸ.
04:51
So one of the problems with these small robots
100
291583
3113
ืœื›ืŸ ืื—ื“ ื”ืงืฉื™ื™ื ืขื ืจื•ื‘ื•ื˜ื™ื ืงื˜ื ื™ื ื›ืืœื”
04:54
is their payload-carrying capacity.
101
294720
2001
ื”ื™ื ื™ื›ื•ืœืชื ืœืฉืืช ืžื˜ืขืŸ.
04:56
So you might want to have multiple robots carry payloads.
102
296745
3017
ืœื›ืŸ ื™ื™ืชื›ืŸ ื•ื ืจืฆื” ืจื•ื‘ื•ื˜ื™ื ืžืจื•ื‘ื™ื
ื›ื“ื™ ืœืฉืืช ืžื˜ืขืŸ.
05:00
This is a picture of a recent experiment we did --
103
300746
2355
ื–ื• ืชืžื•ื ื” ืฉืœ ื ื™ืกื•ื™ ืฉื‘ื™ืฆืขื ื• ืœืื—ืจื•ื ื” --
ื‘ืขืฆื ืœื ื›ืœ-ื›ืš ืœืื—ืจื•ื ื” --
05:03
actually not so recent anymore --
104
303125
1595
05:04
in Sendai, shortly after the earthquake.
105
304744
2492
ื‘ืกื ื“ืื™, ื–ืžืŸ ืงืฆืจ ืœืื—ืจ ืจืขื™ื“ืช ื”ืื“ืžื”.
05:07
So robots like this could be sent into collapsed buildings,
106
307938
3191
ืจื•ื‘ื•ื˜ื™ื ื›ืืœื” ื ืฉืœื—ื• ืœื‘ื ื™ื™ื ื™ื ืฉื”ืชืžื•ื˜ื˜ื•
ื›ื“ื™ ืœืืžื•ื“ ืืช ื”ื ื–ืง ืœืื—ืจ ืืกื•ื ื•ืช ื˜ื‘ืข,
05:11
to assess the damage after natural disasters,
107
311153
2891
ืื• ืœื‘ื ื™ื™ื ื™ื ืจื“ื™ื•ืืงื˜ื™ื‘ื™ื™ื
05:14
or sent into reactor buildings,
108
314068
1582
05:15
to map radiation levels.
109
315674
1619
ื›ื“ื™ ืœืžืคื•ืช ืจืžื•ืช ืงืจื™ื ื”.
05:19
So one fundamental problem that the robots have to solve
110
319583
3521
ืงื•ืฉื™ ืžื”ื•ืชื™ ืื—ื“
ืฉืขืœ ื”ืจื•ื‘ื•ื˜ื™ื ืœื”ืชื’ื‘ืจ ืขืœื™ื• ื›ื“ื™ ืœืคืขื•ืœ ืขืฆืžืื™ืช
05:23
if they are to be autonomous,
111
323128
1579
05:24
is essentially figuring out how to get from point A to point B.
112
324731
3505
ื”ื•ื ืœืžืฆื•ื
ื›ื™ืฆื“ ืœื”ื’ื™ืข ืžื ืงื•ื“ื” A ืœื ืงื•ื“ื” B.
05:28
So this gets a little challenging,
113
328937
1667
ื–ื” ื”ื•ืคืš ืืช ื”ืขื ื™ื™ืŸ ืœืžืืชื’ืจ ื‘ืžื™ื“ืช ืžื”
05:30
because the dynamics of this robot are quite complicated.
114
330628
3007
ืžื›ื™ื•ื•ืŸ ืฉืขื™ืงืจื•ืŸ ื”ืชื ื•ืขื” ืฉืœ ื”ืจื•ื‘ื•ื˜ ื”ื•ื ื“ื™ ืžื•ืจื›ื‘.
05:33
In fact, they live in a 12-dimensional space.
115
333659
2492
ืœืžืขืฉื”, ื”ื ืžืชื ื”ืœื™ื ื‘ืžืจื—ื‘ 12-ืžื™ืžื“ื™.
ืœื›ืŸ ืื ื• ืขื•ืฉื™ื ืชื›ืกื™ืก ืงื˜ืŸ.
05:36
So we use a little trick.
116
336524
1444
05:37
We take this curved 12-dimensional space,
117
337992
3413
ืื ื• ื ื•ื˜ืœื™ื ืืช ื”ืžืจื—ื‘ ื”-12-ืžื™ืžื“ื™
ื•ื”ื•ืคื›ื™ื ืื•ืชื•
05:41
and transform it into a flat, four-dimensional space.
118
341429
3894
ืœืžืจื—ื‘ ืฉื˜ื•ื— 4-ืžื™ืžื“ื™.
ื•ืื•ืชื• ืžืจื—ื‘ 4-ืžื™ืžื“ื™
05:45
And that four-dimensional space consists of X, Y, Z,
119
345347
3071
ืžื•ืจื›ื‘ ืž-X, Y, Z ื•ืžื–ื•ื™ืช ื”ืกื˜ื™ื™ื”.
05:48
and then the yaw angle.
120
348442
1412
05:49
And so what the robot does,
121
349878
1358
ื•ื›ืš ืžื” ืฉื”ืจื•ื‘ื•ื˜ ืขื•ืฉื”
05:51
is it plans what we call a minimum-snap trajectory.
122
351260
3976
ื–ื” ืœืชื›ื ืŸ ืืช ืžื” ืฉื ืงืจื ืžืกืœื•ืœ ืขื ืžื™ื ื™ืžื•ื ืคื ื™ื•ืช.
ืื–ื›ื™ืจ ืฉื‘ืคื™ื–ื™ืงื”,
05:56
So to remind you of physics:
123
356209
1397
05:57
You have position, derivative, velocity;
124
357630
2054
ื™ืฉ ืžื™ืงื•ื, ื”ื ื’ื–ืจืช (ืฉืœื•) ืžื”ื™ืจื•ืช,
05:59
then acceleration;
125
359708
1856
ืื—ืจ-ื›ืš ืชืื•ืฆื”,
06:01
and then comes jerk,
126
361588
2101
ื•ืื– ื‘ื ื”ื˜ื™ืœื˜ื•ืœ
06:03
and then comes snap.
127
363713
1420
ื•ื‘ืกื•ืฃ ืžื’ื™ืขื” ื”ืคื ื™ื”.
06:05
So this robot minimizes snap.
128
365578
2452
ืจื•ื‘ื•ื˜ ื–ื” ืžืžื–ืขืจ ืคื ื™ื•ืช.
06:08
So what that effectively does,
129
368800
1436
ืžื” ืฉื–ื” ื‘ืกื•ืฃ ื™ื•ืฆืจ
06:10
is produce a smooth and graceful motion.
130
370260
2624
ื–ื• ืชื ื•ืขื” ื—ืœืงื” ื•ืžืœืืช ื—ืŸ.
06:12
And it does that avoiding obstacles.
131
372908
2842
ื•ื”ื•ื ืžื‘ืฆืข ื–ืืช ืชื•ืš ื”ืชื—ืžืงื•ืช ืžืžื›ืฉื•ืœื™ื.
ืžืกืœื•ืœื™ื ื”ืœืœื• ืฉืœ ืคื ื™ื•ืช ืžื™ื ื™ืžืœื™ื•ืช ื‘ืžืจื—ื‘ ืฉื˜ื•ื— ื–ื”
06:16
So these minimum-snap trajectories in this flat space are then transformed
132
376313
3563
ืžื•ืขื‘ืจื™ื ื‘ื—ื–ืจื” ืœืฆื•ืจื”
06:19
back into this complicated 12-dimensional space,
133
379900
3325
ื”ืžื•ืจื›ื‘ืช ื”ื–ื• ืฉืœ 12-ืžื™ืžื“ื™ื,
ืฉื”ืจื•ื‘ื•ื˜ื™ื ื—ื™ื™ื‘ื™ื ืœื‘ืฆืข
06:23
which the robot must do for control and then execution.
134
383249
2986
ืœืฉื ืฉืœื™ื˜ื” ื•ื‘ื™ืฆื•ืข ืคืขื•ืœื•ืช.
06:26
So let me show you some examples
135
386749
1540
ืืจืื” ืœื›ื ื›ืžื” ื“ื•ื’ืžืื•ืช
06:28
of what these minimum-snap trajectories look like.
136
388313
2777
ื›ื™ืฆื“ ื ืจืื™ื ืžืกืœื•ืœื™ ื”ืžื™ื ื™ืžื•ื ืคื ื™ื•ืช.
ื‘ืกืจื˜ื•ืŸ ื”ืจืืฉื•ืŸ, ืชืจืื• ืืช ื”ืจื•ื‘ื•ื˜
06:31
And in the first video,
137
391114
1151
06:32
you'll see the robot going from point A to point B,
138
392289
2729
ืขื•ื‘ืจ ืžื ืงื•ื“ื” A ืœื ืงื•ื“ื” B
ื“ืจืš ื ืงื•ื“ืช ื‘ื™ื ื™ื™ื.
06:35
through an intermediate point.
139
395042
1623
06:36
(Whirring noise)
140
396990
2730
ื”ืจื•ื‘ื•ื˜ ืžืกื•ื’ืœ ื‘ื‘ื™ืจื•ืจ
06:43
So the robot is obviously capable of executing any curve trajectory.
141
403377
3621
ืœื‘ืฆืข ื›ืœ ืžืกืœื•ืœ ืžืขื•ืงืœ.
ืืœื” ืžืกืœื•ืœื™ื ืžืขื’ืœื™ื™ื
06:47
So these are circular trajectories,
142
407022
1714
06:48
where the robot pulls about two G's.
143
408760
2500
ื‘ื”ื ื”ืจื•ื‘ื•ื˜ ืžื•ืฉืš ืขื“ 2 ื’'ื™.
06:52
Here you have overhead motion capture cameras on the top
144
412844
3702
ื›ืืŸ ื™ืฉ ืžืžืขืœ ืžืฆืœืžื•ืช ืชื ื•ืขื” ื”ืžืขื“ื›ื ื•ืช ืืช ื”ืจื•ื‘ื•ื˜
06:56
that tell the robot where it is 100 times a second.
145
416570
3206
ื‘ื ื•ื’ืข ืœืžื™ืงื•ืžื• 100 ืคืขื ื‘ืฉื ื™ื”.
06:59
It also tells the robot where these obstacles are.
146
419800
2436
ื”ืŸ ื’ื ืžืขื“ื›ื ื•ืช ืื•ืชื• ื”ื™ื›ืŸ ื ืžืฆืื™ื ื”ืžื›ืฉื•ืœื™ื.
ื”ืžื›ืฉื•ืœื™ื ื™ื›ื•ืœื™ื ื’ื ืœื ื•ืข.
07:03
And the obstacles can be moving.
147
423140
1556
07:04
And here, you'll see Daniel throw this hoop into the air,
148
424720
3127
ื›ืืŸ ืจื•ืื™ื ืืช ื“ื ื™ืืœ ื–ื•ืจืง ื—ื™ืฉื•ืง ืœืื•ื™ืจ,
07:07
while the robot is calculating the position of the hoop,
149
427871
2659
ื‘ืขื•ื“ ื”ืจื•ื‘ื•ื˜ ืžื—ืฉื‘ ืืช ืžื™ืงื•ื ื”ื—ื™ืฉื•ืง
ื•ืžื ืกื” ืœืžืฆื•ื ื›ื™ืฆื“ ืœืขื‘ื•ืจ ื“ืจื›ื• ื‘ืื•ืคืŸ ื”ืžื™ื˜ื‘ื™.
07:10
and trying to figure out how to best go through the hoop.
150
430554
2880
ื”ืจื™ ื‘ืชื•ืจ ืืงื“ืžืื™ื,
07:14
So as an academic,
151
434002
1234
07:15
we're always trained to be able to jump through hoops
152
435260
2528
ืื ื• ื›ื‘ืจ ืžื•ืจื’ืœื™ื ืœืขืฉื•ืช ืฉืžื™ื ื™ื•ืช ื‘ืื•ื™ืจ ื›ื“ื™ ืœื–ื›ื•ืช ื‘ืžื™ืžื•ืŸ
07:17
to raise funding for our labs,
153
437812
1485
ืœื ื™ืกื•ื™ื™ื ื•ืขื›ืฉื™ื• ืื ื• ืžืœืžื“ื™ื ืืช ื”ืจื•ื‘ื•ื˜ื™ื ืœืขืฉื•ืช ื–ืืช.
07:19
and we get our robots to do that.
154
439321
1915
07:21
(Applause)
155
441260
6365
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
ื“ื‘ืจ ื ื•ืกืฃ ืฉื”ืจื•ื‘ื•ื˜ ื™ื›ื•ืœ ืœื‘ืฆืข
07:28
So another thing the robot can do
156
448524
1611
ื”ื•ื ืœื–ื›ื•ืจ ืงื˜ืขื™ ืžืกืœื•ืœื™ื ืฉื”ื•ื ืœื•ืžื“
07:30
is it remembers pieces of trajectory
157
450159
2077
07:32
that it learns or is pre-programmed.
158
452260
2904
ืื• ืฉื”ื•ื ืžืชื•ื›ื ืช ืžืจืืฉ ืœื‘ืฆืข.
ื›ืืŸ ืจื•ืื™ื ืืช ื”ืจื•ื‘ื•ื˜
07:35
So here, you see the robot combining a motion that builds up momentum,
159
455466
4770
ืžืฉืœื‘ ืชื ื•ืขื”
ื”ืฆื•ื‘ืจืช ืžื•ืžื ื˜ื•ื,
07:40
and then changes its orientation and then recovers.
160
460260
2976
ืžืฉื ื” ืืช ืชื ื•ื—ืชื• ื•ืื– ื—ื•ื–ืจ ืœืžืฆื‘ื• ื”ื”ืชื—ืœืชื™.
ืขืœื™ื• ืœื‘ืฆืข ื–ืืช ื‘ื’ืœืœ ืฉื”ืžืจื•ื•ื— ื‘ื—ืœื•ืŸ
07:44
So it has to do this because this gap in the window
161
464004
2960
07:46
is only slightly larger than the width of the robot.
162
466988
3458
ืจื—ื‘ ืจืง ืžืขื˜ ื™ื•ืชืจ ืžืจื•ื—ื‘ ื”ืจื•ื‘ื•ื˜.
ื‘ื“ื™ื•ืง ื›ืžื• ืฆื•ืœืœืŸ ืฉืขื•ืžื“ ืขืœ ืžืงืคืฆื”
07:51
So just like a diver stands on a springboard
163
471050
2753
07:53
and then jumps off it to gain momentum,
164
473827
2453
ื•ืงื•ืคืฅ ืขืœื™ื” ื›ื“ื™ ืœืฆื‘ื•ืจ ืžื•ืžื ื˜ื•ื,
ื•ืื– ืขื•ืฉื” ืกืœื˜ื”, ืกืœื˜ื” ืฉืœ ืฉื ื™ื™ื ื•ื—ืฆื™ ืกื™ื‘ื•ื‘ื™ื
07:56
and then does this pirouette, this two and a half somersault through
165
476304
3229
ื•ืื– ื—ื•ื–ืจ ื‘ื ื•ื ืฉืœื ื˜ื™ื•ืช ืœืžืฆื‘ื• ื”ื”ืชื—ืœืชื™,
07:59
and then gracefully recovers,
166
479557
1402
08:00
this robot is basically doing that.
167
480983
1698
ื”ืจื•ื‘ื•ื˜ ื”ื–ื” ื‘ืขื™ืงืจื•ืŸ ืขื•ืฉื” ื–ืืช.
08:02
So it knows how to combine little bits and pieces of trajectories
168
482705
3080
ื”ื•ื ื™ื•ื“ืข ื›ื™ืฆื“ ืœืฉืœื‘ ื‘ื™ื—ื“ ืคื™ืกื•ืช ืงื˜ื ื•ืช
08:05
to do these fairly difficult tasks.
169
485809
2999
ืฉืœ ืžืกืœื•ืœื™ื ื›ื“ื™ ืœื‘ืฆืข ืžืฉื™ืžื•ืช ื“ื™ ืžื•ืจื›ื‘ื•ืช ื›ืืœื•.
ืื ื™ ืจื•ืฆื” ืœืขื‘ื•ืจ ืฉืœื‘.
08:10
So I want change gears.
170
490018
1218
08:11
So one of the disadvantages of these small robots is its size.
171
491260
3839
ืื—ื“ ื”ื—ืกืจื•ื ื•ืช ืฉืœ ื”ืจื•ื‘ื•ื˜ื™ื ื”ืœืœื• ื”ื•ื ื’ื•ื“ืœื.
ืกื™ืคืจืชื™ ืœื›ื ืงื•ื“ื
08:15
And I told you earlier
172
495654
1152
08:16
that we may want to employ lots and lots of robots
173
496830
2349
ืฉืื•ืœื™ ื ืฉืชืžืฉ ื‘ืžืกืคืจื™ื ืžืื•ื“
ื’ื“ื•ืœื™ื ืฉืœ ืจื•ื‘ื•ื˜ื™ื ื›ื“ื™ ืœื”ืชื’ื‘ืจ ืขืœ ืžื™ื’ื‘ืœื•ืช ื”ื’ื•ื“ืœ.
08:19
to overcome the limitations of size.
174
499203
2033
ืงื•ืฉื™ ืื—ื“ ื›ื–ื” ื”ื•ื
08:22
So one difficulty is:
175
502010
1574
08:23
How do you coordinate lots of these robots?
176
503608
2793
ื›ื™ืฆื“ ืœืชืื ื‘ื™ืŸ ืจื•ื‘ื•ื˜ื™ื ืจื‘ื™ื ื›ืœ-ื›ืš?
08:26
And so here, we looked to nature.
177
506425
1811
ื‘ืฉืœื‘ ื–ื” ืคื ื™ื ื• ืœื˜ื‘ืข.
08:28
So I want to show you a clip of Aphaenogaster desert ants,
178
508544
4608
ืืจืื” ืœื›ื ืงื˜ืข ื•ื™ื“ืื•
ืฉืœ ื ืžืœื™ ืžื“ื‘ืจ,
ื‘ืžืขื‘ื“ืชื• ืฉืœ ืคืจื•ืค' ืกื˜ืคืŸ ืคืจืื˜, ื”ื ื•ืฉืื•ืช ืขืฆื.
08:33
in Professor Stephen Pratt's lab, carrying an object.
179
513176
3388
ื–ื•ื”ื™ ืœืžืขืฉื” ื—ืชื™ื›ืช ืชืื ื”.
08:36
So this is actually a piece of fig.
180
516588
1715
ืœืžืขืฉื” ื ื•ื˜ืœื™ื ืขืฆื ื›ืœืฉื”ื•
08:38
Actually you take any object coated with fig juice,
181
518327
2400
ื”ืžืฆื•ืคื” ื‘ืžื™ืฅ ืชืื ื” ื›ื“ื™ ืฉื”ื ืžืœื™ื ื™ืฉืื• ืื•ืชื• ืœืงืŸ ืฉืœื”ืŸ.
08:40
and the ants will carry it back to the nest.
182
520751
2091
08:42
So these ants don't have any central coordinator.
183
522866
3169
ืœื ืžืœื™ื ืืœื• ืื™ืŸ ืื™ื–ื” ื’ื•ืจื ืžืจื›ื–ื™ ืฉืžืชืื ื‘ื™ื ื™ื”ืŸ.
ื”ืŸ ืงื•ืœื˜ื•ืช ืืช ืฉื›ื ื™ื”ืŸ.
08:46
They sense their neighbors.
184
526487
1547
ืื™ืŸ ืชืงืฉื•ืจืช ื—ื“ื” ื•ื‘ืจื•ืจื”.
08:48
There's no explicit communication.
185
528058
2285
ืื‘ืœ ืžืื—ืจ ื•ื”ืŸ ืงื•ืœื˜ื•ืช ืืช ื”ืฉื›ื ื™ื
08:50
But because they sense the neighbors
186
530367
1739
ื•ื”ืŸ ื—ืฉื•ืช ืืช ื”ืขืฆื,
08:52
and because they sense the object,
187
532130
1776
08:53
they have implicit coordination across the group.
188
533930
2830
ื ื•ืฆืจ ืชื™ืื•ื ืกืคื•ื ื˜ื ื™ ื‘ื™ืŸ ื—ืœืงื™ ื”ืงื‘ื•ืฆื”.
ื•ืžื™ืŸ ืกื•ื’ ื›ื–ื” ืฉืœ ืชื™ืื•ื
08:57
So this is the kind of coordination we want our robots to have.
189
537474
3420
ืื ื• ืจื•ืฆื™ื ืฉื™ื”ื™ื” ื‘ื™ืŸ ื”ืจื•ื‘ื•ื˜ื™ื.
09:01
So when we have a robot which is surrounded by neighbors --
190
541569
4778
ืื– ื›ืืฉืจ ื™ืฉ ืœื ื• ืจื•ื‘ื•ื˜
ื”ืžื•ืงืฃ ื‘ืฉื›ื ื™ื --
ื”ื‘ื” ื ื‘ื™ื˜ ื‘ืจื•ื‘ื•ื˜ I ื•ืจื•ื‘ื•ื˜ J --
09:06
and let's look at robot I and robot J --
191
546371
2229
ืžื” ืฉืื ื• ืจื•ืฆื™ื ืฉื”ืจื•ื‘ื•ื˜ื™ื ื”ืœืœื• ื™ืขืฉื•
09:08
what we want the robots to do,
192
548624
1548
ื–ื” ืœื ื˜ืจ ืืช ื”ืžืจื•ื•ื— ื‘ื™ื ื™ื”ื
09:10
is to monitor the separation between them,
193
550196
2111
09:12
as they fly in formation.
194
552331
1905
ื‘ืขื•ื“ื ื˜ืกื™ื ื‘ืžื‘ื ื”.
09:14
And then you want to make sure
195
554537
1481
ื•ืื– ื‘ืจืฆื•ื ื ื• ืœื•ื•ื“ื
09:16
that this separation is within acceptable levels.
196
556042
2904
ืฉื”ืžืจื•ื•ื— ื”ื–ื” ื”ื•ื ื‘ืžื™ื’ื‘ืœื•ืช ื”ืžื•ืชืจ.
ื”ืจื•ื‘ื•ื˜ื™ื ืขื•ืงื‘ื™ื ืื—ืจ ื”ืฉื’ื™ืื”
09:19
So again, the robots monitor this error
197
559303
2468
09:21
and calculate the control commands 100 times a second,
198
561795
4149
ื•ืžื—ืฉื‘ื™ื ืืช ืคืงื•ื“ื•ืช ื”ื‘ืงืจื”
100 ืคืขื ื‘ืฉื ื™ื”,
09:25
which then translates into motor commands,
199
565968
2023
ื”ืžืชื•ืจื’ืžื•ืช ืœืคืงื•ื“ื•ืช ืœืžื ื•ืข 600 ืคืขื ื‘ืฉื ื™ื”.
09:28
600 times a second.
200
568015
1215
ื›ืœ ื–ื” ืฆืจื™ืš ืœื”ืชื‘ืฆืข
09:29
So this also has to be done in a decentralized way.
201
569254
3222
ื‘ืื•ืคืŸ ืžื‘ื•ื–ืจ.
09:32
Again, if you have lots and lots of robots,
202
572785
2334
ื›ืืžื•ืจ, ืื ื™ืฉ ืžืกืคืจ ื’ื“ื•ืœ ืฉืœ ืจื•ื‘ื•ื˜ื™ื,
ื–ื” ื‘ืœืชื™ ืืคืฉืจื™ ืœืกื ื›ืจืŸ ืืช ื›ืœ ื”ืžื™ื“ืข ืžื”ืจ ืžืกืคื™ืง ืžืžืงื•ื ืื—ื“ ืžืจื›ื–ื™
09:35
it's impossible to coordinate all this information centrally
203
575143
3277
09:38
fast enough in order for the robots to accomplish the task.
204
578444
3177
ื‘ืื•ืคืŸ ื›ื–ื” ืฉื”ืจื•ื‘ื•ื˜ื™ื ื™ื•ื›ืœื• ืœื‘ืฆืข ืืช ืžืฉื™ืžืชื.
09:41
Plus, the robots have to base their actions only on local information --
205
581645
4348
ื‘ื ื•ืกืฃ ืขืœ ื”ืจื•ื‘ื•ื˜ื™ื ืœื‘ืกืก ืืช ืคืขื™ืœื•ืชื
ืืš ื•ืจืง ืขืœ ืžื™ื“ืข ืžืงื•ืžื™,
ืขืœ ืžื” ืฉื”ื ืงื•ืœื˜ื™ื ืžืฉื›ื ื™ื”ื.
09:46
what they sense from their neighbors.
206
586017
2119
ื•ืœื‘ืกื•ืฃ,
09:48
And then finally,
207
588160
1386
09:49
we insist that the robots be agnostic to who their neighbors are.
208
589570
4100
ืื ื• ืžืงืคื™ื“ื™ื ืฉื”ืจื•ื‘ื•ื˜ื™ื
ื™ื”ื™ื• ืื“ื™ืฉื™ื ืœื–ื”ื•ืช ืฉื›ื ื™ื”ื.
09:53
So this is what we call anonymity.
209
593694
2117
ืœื–ื” ืื ื• ืงื•ืจืื™ื ืืœืžื•ื ื™ื•ืช.
ืžื” ืฉืื ื™ ืจื•ืฆื” ืœื”ืจืื•ืช ืœื›ื ื›ืขืช
09:57
So what I want to show you next is a video of 20 of these little robots,
210
597258
6417
ื–ื” ืกืจื˜ื•ืŸ
ื”ืžืฆื™ื’ 20 ืจื•ื‘ื•ื˜ื™ื ืงื˜ื ื™ื ื›ืืœื”
10:03
flying in formation.
211
603699
1537
ื”ื˜ืกื™ื ื‘ืžื‘ื ื”.
ื”ื ืžื ื˜ืจื™ื ืืช ืžื™ืงื•ื ืฉื›ื ื™ื”ื.
10:06
They're monitoring their neighbors' positions.
212
606144
2896
ื”ื ืฉื•ืžืจื™ื ืขืœ ื”ืžื‘ื ื”.
10:09
They're maintaining formation.
213
609064
1902
10:10
The formations can change.
214
610990
1780
ื”ืžื‘ื ื™ื ื™ื›ื•ืœื™ื ืœื”ืฉืชื ื•ืช.
10:12
They can be planar formations,
215
612794
1679
ื”ืžื‘ื ื™ื ื™ื›ื•ืœื™ื ืœื”ื™ื•ืช ืžื™ืฉื•ืจื™ื™ื,
10:14
they can be three-dimensional formations.
216
614497
2103
ื™ื›ื•ืœื™ื ืœื”ื™ื•ืช ืชืœืช-ืžื™ืžื“ื™ื™ื.
ื›ืคื™ ืฉื ื™ืชืŸ ืœืจืื•ืช ื›ืืŸ,
10:17
As you can see here,
217
617331
1341
10:18
they collapse from a three-dimensional formation into planar formation.
218
618696
3687
ื”ื ืขื•ื‘ืจื™ื ืžืžื‘ื ื” ืชืœืช-ืžื™ืžื“ื™ ืœืžื‘ื ื” ืžื™ืฉื•ืจื™.
ื•ื›ื“ื™ ืœื˜ื•ืก ื“ืจืš ืžื›ืฉื•ืœื™ื,
10:22
And to fly through obstacles,
219
622407
1531
10:23
they can adapt the formations on the fly.
220
623962
2952
ื”ื ื™ื›ื•ืœื™ื ืœื˜ื•ื•ืช ืืช ื”ืžื‘ื ื™ื ืชื•ืš ื›ื“ื™ ื˜ื™ืกื”.
ื›ืืžื•ืจ, ื”ืจื•ื‘ื•ื˜ื™ื ื”ืืœื” ื™ื›ื•ืœื™ื ืžืžืฉ ืœื”ืชืงืจื‘ ื–ื” ืœื–ื”.
10:28
So again, these robots come really close together.
221
628026
2463
10:30
As you can see in this figure-eight flight,
222
630513
2388
ื›ืคื™ ืฉืจื•ืื™ื ื‘ื˜ื™ืกืช ืกื™ืคืจื”-8 ื–ื•,
10:32
they come within inches of each other.
223
632925
1984
ื”ื ืžืชืงืจื‘ื™ื ืขื“ ื›ื“ื™ ืกื ื˜ื™ืžื˜ืจื™ื ื–ื” ืœื–ื”.
ื•ืขืœ ืืฃ ื”ืฉืคืขื•ืช ืื•ื•ื™ืจื•ื“ื™ื ืžื™ื•ืช ื”ื“ื“ื™ื•ืช
10:35
And despite the aerodynamic interactions with these propeller blades,
224
635377
4317
ืฉืœ ืœื”ื‘ื™ ื”ืžื“ื—ืคื™ื,
10:39
they're able to maintain stable flight.
225
639718
2148
ื”ื ืžืฆืœื™ื—ื™ื ืœืฉืžื•ืจ ืขืœ ื˜ื™ืกื” ื™ืฆื™ื‘ื”.
10:41
(Applause)
226
641890
6896
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
ื›ืš ืฉื‘ืจื’ืข ืฉื™ื•ื“ืขื™ื ื›ื™ืฆื“ ืœื˜ื•ืก ื‘ืžื‘ื ื”,
10:49
So once you know how to fly in formation,
227
649246
1999
ื ื™ืชืŸ ืœื”ืจื™ื ื—ืคืฆื™ื ื‘ืžืฉื•ืชืฃ.
10:51
you can actually pick up objects cooperatively.
228
651269
2216
ื–ื” ืจืง ืžืจืื”
10:53
So this just shows that we can double, triple, quadruple
229
653509
4577
ืฉืื ื• ื™ื›ื•ืœื™ื ืœื”ื›ืคื™ืœ, ืœื”ืฉืœื™ืฉ, ืœื”ื›ืคื™ืœ ืคื™-4
ืืช ื™ื›ื•ืœืช ื”ืจื•ื‘ื•ื˜ ืคืฉื•ื˜ ืขืœ-ื™ื“ื™
10:58
the robots' strength,
230
658110
1158
10:59
by just getting them to team with neighbors, as you can see here.
231
659292
3077
ื”ืฆื™ื•ื•ืช ืฉืœื”ื ืขื ืฉื›ื ื™ื”ื, ื›ืคื™ ืฉืจื•ืื™ื.
ืื—ื“ ื”ื—ืกืจื•ื ื•ืช ืฉืœ ื–ื”
11:02
One of the disadvantages of doing that is, as you scale things up --
232
662393
4142
ื”ื•ื ืฉื›ื›ืœ ืฉื”ืžืกืคืจื™ื ืขื•ืœื™ื --
11:06
so if you have lots of robots carrying the same thing,
233
666559
2564
ื›ื›ืœ ืฉื™ืฉ ื™ื•ืชืจ ืจื•ื‘ื•ื˜ื™ื ื”ื ื•ืฉืื™ื
ื“ื‘ืจ ืื—ื“ ืžืกื•ื™ื™ื, ื‘ื”ื›ืจื— ื’ื ืžืขืœื™ื ืืช ื”ืื™ื ืจืฆื™ื”,
11:09
you're essentially increasing the inertia,
234
669147
2720
11:11
and therefore you pay a price; they're not as agile.
235
671891
2619
ื•ืœื›ืŸ ืžืฉืœืžื™ื ืžื—ื™ืจ; ื”ื ื›ื‘ืจ ืœื ืงืœื™-ืชื ื•ืขื”.
11:14
But you do gain in terms of payload-carrying capacity.
236
674978
3096
ืื‘ืœ ืžืจื•ื™ื—ื™ื ื™ื›ื•ืœืช ื ืฉื™ืืช ืžื˜ืขืŸ.
ืฉื™ืžื•ืฉ ื ื•ืกืฃ ืฉื‘ืจืฆื•ื ื™ ืœื”ืจืื•ืช --
11:18
Another application I want to show you -- again, this is in our lab.
237
678098
3215
ืฉื•ื‘, ื–ื” ื‘ืžืขื‘ื“ื” ืฉืœื ื•.
11:21
This is work done by Quentin Lindsey, who's a graduate student.
238
681337
3165
ื–ื• ืขื‘ื•ื“ื” ืฉื ืขืฉืชื” ืขืœ-ื™ื“ื™ ืงื•ื•ื™ื ื˜ื™ืŸ ืœื™ื ื“ืกื™ื™ ืฉื”ื•ื ืกื˜ื•ื“ื ื˜ ืœืžื—ืงืจ.
ื”ืืœื’ื•ืจื™ืชื ืฉืœื• ืื•ืžืจ ื‘ืขื™ืงืจื•ืŸ ืœืจื•ื‘ื•ื˜ื™ื
11:24
So his algorithm essentially tells these robots
239
684526
2620
ื›ื™ืฆื“ ืœื‘ื ื•ืช ืขืฆืžืื™ืช
11:27
how to autonomously build cubic structures
240
687170
4072
ืžื‘ื ื™ื ืงื•ื‘ื™ื™ืชื™ื™ื
ืžืขืฆืžื™ื ื›ืžื• ืงื•ืจื•ืช, ืกืžื•ื›ื•ืช ื•ื›ื“ื•ืžื”.
11:31
from truss-like elements.
241
691266
1970
ื”ืืœื’ื•ืจื™ืชื ืฉืœื• ืื•ืžืจ ืœืจื•ื‘ื•ื˜ื™ื
11:34
So his algorithm tells the robot what part to pick up,
242
694441
3760
ืื™ื–ื” ื—ืœืง ืœื”ืจื™ื,
ืžืชื™ ื•ื”ื™ื›ืŸ ืœื”ื ื™ื—ื•.
11:38
when, and where to place it.
243
698225
1825
ื›ืš ืฉื‘ืกืจื˜ื•ืŸ ื–ื” ืจื•ืื™ื --
11:40
So in this video you see --
244
700471
1483
11:41
and it's sped up 10, 14 times --
245
701978
2024
ื•ื”ื•ื ืžื•ืืฅ ืคื™ 10, 14 --
ืจื•ืื™ื ืฉืœื•ืฉื” ืžื‘ื ื™ื ืฉื•ื ื™ื ื”ืžื•ืงืžื™ื ืขืœ-ื™ื“ื™ ื”ืจื•ื‘ื•ื˜ื™ื.
11:44
you see three different structures being built by these robots.
246
704026
3089
ื•ืฉื•ื‘ ื›ืืžื•ืจ, ื”ื›ืœ ื‘ืื•ืคืŸ ืขืฆืžืื™,
11:47
And again, everything is autonomous,
247
707139
2026
ื•ื›ืœ ืžื” ืฉืงื•ื•ื™ื ื˜ื™ืŸ ืฆืจื™ืš ืœืขืฉื•ืช
11:49
and all Quentin has to do
248
709189
1283
11:50
is to give them a blueprint of the design that he wants to build.
249
710496
3935
ื–ื” ืœืกืคืง ืœื”ื ืฉืจื˜ื•ื˜
ืฉืœ ื”ืžื‘ื ื” ืฉืจื•ืฆื™ื ืœื‘ื ื•ืช.
11:56
So all these experiments you've seen thus far,
250
716925
2811
ื›ืœ ื”ื ื™ืกื•ื™ื™ื ืฉืจืื™ืชื ืขื“ ืขื›ืฉื™ื•,
11:59
all these demonstrations,
251
719760
1350
ื›ืœ ื”ืชืฆื•ื’ื•ืช,
12:01
have been done with the help of motion-capture systems.
252
721134
3102
ื ืขืฉื• ื‘ืขื–ืจืช ืžืขืจื›ื•ืช ืœืœื›ื™ื“ืช ืชื ื•ืขื”.
ืื‘ืœ ืžื” ืงื•ืจื” ื›ืืฉืจ
12:05
So what happens when you leave your lab,
253
725029
2507
ื™ื•ืฆืื™ื ืžื”ืžืขื‘ื“ื” ืœืขื•ืœื ื”ืืžื™ืชื™ ื‘ื—ื•ืฅ?
12:07
and you go outside into the real world?
254
727560
2037
12:09
And what if there's no GPS?
255
729922
1706
ื•ืžื” ืื ืื™ืŸ GPS?
12:12
So this robot is actually equipped with a camera,
256
732755
4337
ืœื›ืŸ ืจื•ื‘ื•ื˜ ื–ื”
ืžืฆื•ื™ื™ื“ ื‘ืžืฆืœืžื”
ื•ื‘ืœื™ื™ื–ืจ ื”ืžื•ื“ื“ ื˜ื•ื•ื—, ืกื•ืจืง ืœื™ื™ื–ืจ.
12:17
and a laser rangefinder, laser scanner.
257
737116
2373
ื”ื•ื ืžืฉืชืžืฉ ื‘ื—ื™ื™ืฉื ื™ื ืืœื”
12:20
And it uses these sensors to build a map of the environment.
258
740140
3920
ืœื™ืฆื™ืจืช ืžืคื” ืฉืœ ื”ืกื‘ื™ื‘ื”.
ืžื” ืฉื”ืžืคื” ืžื›ื™ืœื” ื–ื” ื”ืžืืคื™ื™ื ื™ื --
12:24
What that map consists of are features --
259
744084
3286
ื›ื’ื•ืŸ ืคืชื—ื™ื, ื—ืœื•ื ื•ืช,
12:27
like doorways, windows, people, furniture --
260
747394
3871
ืื ืฉื™ื, ืจื”ื™ื˜ื™ื --
ื•ืื– ื”ื•ื ืžื—ืฉื‘ ืืช ืžื™ืงื•ืžื•
12:31
and it then figures out where its position is,
261
751289
2143
ื‘ื™ื—ืก ืœืžืืคื™ื™ื ื™ื ืืœื”.
12:33
with respect to the features.
262
753456
1401
12:34
So there is no global coordinate system.
263
754881
2210
ื›ืœื•ืžืจ, ืื™ืŸ ืžืขืจื›ืช ืชื™ืื•ื ืžืจื›ื–ื™ืช ืื—ืช.
ืžืขืจื›ืช ื”ืงื•ืื•ืจื“ื™ื ื˜ื•ืช ืžื•ื’ื“ืจืช ื‘ื”ืชื‘ืกืก ืขืœ ื”ืจื•ื‘ื•ื˜ ืขืฆืžื•,
12:37
The coordinate system is defined based on the robot,
264
757115
2506
12:39
where it is and what it's looking at.
265
759645
2057
ืขืœ ืžื™ืงื•ืžื• ื•ืขืœ ืžื” ืฉื”ื•ื ืžื‘ื™ื˜ ืขืœื™ื•.
12:42
And it navigates with respect to those features.
266
762575
2755
ื•ื”ื•ื ืžื ื•ื•ื˜ ื‘ื™ื—ืก ืœืžืืคื™ื™ื ื™ื ืืœื”.
ื›ืขืช ืืจืื” ืœื›ื ืงื˜ืข ื•ื™ื“ืื•
12:46
So I want to show you a clip
267
766316
1420
12:47
of algorithms developed by Frank Shen and Professor Nathan Michael,
268
767760
3952
ืฉืœ ืืœื’ื•ืจื™ืชืžื™ื ืฉืคื•ืชื—ื• ืขืœ-ื™ื“ื™ ืคืจื ืง ืฉืŸ
ื•ืคืจื•ืค' ื ืชืŸ ืžื™ื™ืงืœ,
12:51
that shows this robot entering a building for the very first time,
269
771736
3769
ื”ืžืจืื” ืืช ื”ืจื•ื‘ื•ื˜ ื”ื–ื” ื ื›ื ืก ืœื‘ื ื™ื™ืŸ ืžืžืฉ ื‘ืคืขื ื”ืจืืฉื•ื ื”
12:55
and creating this map on the fly.
270
775529
2422
ื•ื™ื•ืฆืจ ืžืคื” ื–ื• ืชื•ืš ื›ื“ื™ ื˜ื™ืกื”.
12:58
So the robot then figures out what the features are,
271
778642
3158
ืื—ืจ-ื›ืš ื”ืจื•ื‘ื•ื˜ ืžื—ืฉื‘ ืžื”ื ื”ืžืืคื™ื™ื ื™ื ื”ืืœื”.
13:01
it builds the map,
272
781824
1151
ื”ื•ื ื‘ื•ื ื” ืืช ื”ืžืคื”.
13:02
it figures out where it is with respect to the features,
273
782999
2967
ื”ื•ื ืžื—ืฉื‘ ื”ื™ื›ืŸ ื”ื•ื ื ืžืฆื ื‘ื™ื—ืก ืœืžืืคื™ื™ื ื™ื ื”ืœืœื•
13:05
and then estimates its position 100 times a second,
274
785990
3948
ื•ืื– ืื•ืžื“ ืืช ืžื™ืงื•ืžื•
100 ืคืขื ื‘ืฉื ื™ื” ื•ื›ืš ืžืืคืฉืจ ืœื ื•
13:09
allowing us to use the control algorithms that I described to you earlier.
275
789962
3818
ืœื”ืฉืชืžืฉ ื‘ืืœื’ื•ืจื™ืชืžื™ ื”ืฉืœื™ื˜ื”
ืฉื“ื™ื‘ืจืชื™ ืขืœื™ื”ื ืงื•ื“ื.
13:13
So this robot is actually being commanded remotely by Frank,
276
793804
4433
ืจื•ื‘ื•ื˜ ื–ื” ื ืฉืœื˜ ืžืจื—ื•ืง
ืขืœ-ื™ื“ื™ ืคืจื ืง.
ืื‘ืœ ื”ืจื•ื‘ื•ื˜ ื™ื›ื•ืœ ื’ื ืœื—ืฉื‘
13:18
but the robot can also figure out where to go on its own.
277
798261
3539
ื‘ืื•ืคืŸ ืขืฆืžืื™ ืœืืŸ ืœื ื•ืข.
ื ื ื™ื— ืฉื ืชื‘ืงืฉืชื™ ืœื”ื™ื›ื ืก ืœื‘ื ื™ื™ืŸ
13:22
So suppose I were to send this into a building,
278
802125
2207
ื•ืื™ืŸ ืœื™ ืžื•ืฉื’ ื›ื™ืฆื“ ื”ื‘ื ื™ื™ืŸ ื ืจืื” ืžื‘ืคื ื™ื,
13:24
and I had no idea what this building looked like.
279
804356
2302
ืื– ืื ื™ ื™ื›ื•ืœ ืœื‘ืงืฉ ืžื”ืจื•ื‘ื•ื˜ ืœื”ื™ื›ื ืก,
13:26
I can ask this robot to go in,
280
806682
1444
ืœื™ืฆื•ืจ ืžืคื”
13:28
create a map,
281
808150
1904
ื•ืื– ืœื—ื–ื•ืจ ืืœื™ื™ ื•ืœืกืคืจ ืœื™ ื›ื™ืฆื“ ื”ื‘ื ื™ื™ืŸ ื ืจืื” ืžื‘ืคื ื™ื.
13:30
and then come back and tell me what the building looks like.
282
810078
2865
13:32
So here, the robot is not only solving the problem
283
812967
3168
ืื– ื”ื ื”, ืœื ืจืง ืฉื”ืจื•ื‘ื•ื˜ ืคื•ืชืจ ืืช ื”ื‘ืขื™ื”
ืฉืœ ื›ื™ืฆื“ ืœืขื‘ื•ืจ ืžื ืงื•ื“ื” A ืœื ืงื•ื“ื” B ื‘ืžืคื” ื–ื•,
13:36
of how to go from point A to point B in this map,
284
816159
2634
13:38
but it's figuring out what the best point B is at every time.
285
818817
4399
ืืœื ื”ื•ื ื’ื ืžื—ืฉื‘
ืžื”ื™ ื ืงื•ื“ืช ื”-B ื”ืžื™ื˜ื‘ื™ืช ื‘ื›ืœ ืจื’ืข.
ื‘ืขื™ืงืจื•ืŸ ื”ื•ื ื™ื•ื“ืข ืœืืŸ ืœื ื•ืข ื›ื“ื™ ืœื—ืคืฉ
13:43
So essentially it knows where to go
286
823240
2054
13:45
to look for places that have the least information,
287
825318
2785
ืžืงื•ืžื•ืช ืฉื™ืฉ ืขืœื™ื”ื ืžื™ื“ืข ืžื™ื ื™ืžืœื™.
ื•ื›ืš ื”ื•ื ืžืื›ืœืก ืืช ื”ืžืคื”.
13:48
and that's how it populates this map.
288
828127
1968
13:50
So I want to leave you with one last application.
289
830944
2998
ื‘ืจืฆื•ื ื™ ืœื”ืฉืื™ืจื›ื
ืขื ืฉื™ืžื•ืฉ ืื—ื“ ืื—ืจื•ืŸ.
13:54
And there are many applications of this technology.
290
834982
2524
ื™ืฉื ื ื”ืจื‘ื” ืฉื™ืžื•ืฉื™ื ืœื˜ื›ื ื•ืœื•ื’ื™ื” ื–ื•.
13:57
I'm a professor, and we're passionate about education.
291
837823
2548
ืื ื™ ืคืจื•ืค' ื•ื”ื—ื™ื ื•ืš ื—ืฉื•ื‘ ืžืื•ื“ ืขื‘ื•ืจื ื• ื”ืคืจื•ืคืกื•ืจื™ื.
ืจื•ื‘ื•ื˜ื™ื ื›ืืœื” ื™ื›ื•ืœื™ื ื‘ืืžืช ืœืฉื ื•ืช ืืช ื”ื“ืจืš
14:00
Robots like this can really change the way we do K-12 education.
292
840395
3936
ื‘ื” ืื ื• ืžืœืžื“ื™ื ืžื”ื’ืŸ ืขื“ ืœื‘ื’ืจื•ืช.
ืื‘ืœ ืื ื• ื ืžืฆืื™ื ื‘ื“ืจื•ื-ืงืœื™ืคื•ืจื ื™ื”,
14:04
But we're in Southern California,
293
844355
1825
ืงืจื•ื‘ ืœืœื•ืก-ืื ื’'ืœืก,
14:06
close to Los Angeles,
294
846204
1833
ืœื›ืŸ ืขืœื™ื™ ืœืกื™ื™ื
14:08
so I have to conclude with something focused on entertainment.
295
848061
3250
ืขื ืžืฉื”ื• ื”ืงืฉื•ืจ ืœื‘ื™ื“ื•ืจ.
ื‘ืจืฆื•ื ื™ ืœืกื™ื™ื ืขื ืกืจื˜ื•ืŸ ืžื•ื–ื™ืงื”.
14:12
I want to conclude with a music video.
296
852407
1834
ืื ื™ ืจื•ืฆื” ืœื”ืฆื™ื’ ืืช ื”ื™ื•ืฆืจื™ื, ืืœื›ืก ื•ื“ื ื™ืืœ,
14:14
I want to introduce the creators, Alex and Daniel, who created this video.
297
854265
4714
ืืฉืจ ื™ืฆืจื• ื•ื™ื“ืื• ื–ื”.
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
14:19
(Applause)
298
859003
6878
14:25
So before I play this video,
299
865905
1434
ืœืคื ื™ ืฉืืจื™ืฅ ืืช ื”ื•ื™ื“ืื•,
14:27
I want to tell you that they created it in the last three days,
300
867363
3197
ืืกืคืจ ืœื›ื ืฉื”ื ื™ืฆืจื• ืื•ืชื• ื‘ืฉืœื•ืฉืช ื”ื™ืžื™ื ื”ืื—ืจื•ื ื™ื
14:30
after getting a call from Chris.
301
870584
1652
ืœืื—ืจ ืฉืงื™ื‘ืœื• ื˜ืœืคื•ืŸ ืžื›ืจื™ืก.
14:32
And the robots that play in the video are completely autonomous.
302
872743
3817
ื•ื”ืจื•ื‘ื•ื˜ื™ื ืฉืžื ื’ื ื™ื ื‘ื•ื™ื“ืื•
ื”ื ืขืฆืžืื™ื™ื ืœื—ืœื•ื˜ื™ืŸ.
14:36
You will see nine robots play six different instruments.
303
876584
3502
ืืชื ืชืจืื• 9 ืจื•ื‘ื•ื˜ื™ื ื”ืžื ื’ื ื™ื 6 ื›ืœื™ื ืฉื•ื ื™ื.
ื•ื›ืžื•ื‘ืŸ, ื–ื” ื ืขืฉื” ื‘ืžื™ื•ื—ื“ ืœืจื’ืœ TED 2012.
14:40
And of course, it's made exclusively for TED 2012.
304
880697
2971
ื”ื‘ื” ื ืฆืคื”.
14:44
Let's watch.
305
884533
1152
14:46
(Sound of air escaping from valve)
306
886824
4404
14:53
(Music)
307
893547
3336
14:56
(Whirring sound)
308
896907
5520
15:19
(Music)
309
919442
6629
(ืžื•ื–ื™ืงื”)
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
16:24
(Applause) (Cheers)
310
984029
5984
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7