1
00:00:00,080 --> 00:00:01,660
The following
content is provided

2
00:00:01,660 --> 00:00:03,800
under a Creative
Commons license.

3
00:00:03,800 --> 00:00:06,540
Your support will help MIT
OpenCourseWare continue

4
00:00:06,540 --> 00:00:10,150
to offer high quality
educational resources for free.

5
00:00:10,150 --> 00:00:12,690
To make a donation or to
view additional materials

6
00:00:12,690 --> 00:00:16,600
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:16,600 --> 00:00:17,260
at ocw.mit.edu.

8
00:00:27,340 --> 00:00:30,220
PROFESSOR: So I'll be sitting
down here and showing you

9
00:00:30,220 --> 00:00:33,970
a whole lot of really
cool videos on YouTube,

10
00:00:33,970 --> 00:00:38,380
and battlecode players that
I have programmed especially

11
00:00:38,380 --> 00:00:39,440
for tonight.

12
00:00:39,440 --> 00:00:41,730
So here we go.

13
00:00:41,730 --> 00:00:44,220
What's the point here?

14
00:00:44,220 --> 00:00:47,950
The point here is to talk
about the things that

15
00:00:47,950 --> 00:00:52,350
have been done before in swarm,
and thinking about swarm.

16
00:00:52,350 --> 00:00:55,660
So let's go straight
to a YouTube video

17
00:00:55,660 --> 00:00:59,600
and look at this animation.

18
00:00:59,600 --> 00:01:04,379
So you can see here, there
will be these birds that

19
00:01:04,379 --> 00:01:10,510
are following the green blob,
and they avoid the red blob.

20
00:01:10,510 --> 00:01:15,460
And what they're trying to do is
match one another's direction,

21
00:01:15,460 --> 00:01:18,920
try to get a certain distance
from the other birds,

22
00:01:18,920 --> 00:01:20,647
and try to avoid obstacles.

23
00:01:20,647 --> 00:01:22,230
And so you can see
how the combination

24
00:01:22,230 --> 00:01:26,640
of a few simple rules can
create a behavior that's

25
00:01:26,640 --> 00:01:29,590
sort of complicated looking,
or a behavior that's

26
00:01:29,590 --> 00:01:30,450
somewhat fluid.

27
00:01:30,450 --> 00:01:32,110
Whereas if you
had manually tried

28
00:01:32,110 --> 00:01:35,060
to get them all to go
in the same direction,

29
00:01:35,060 --> 00:01:36,870
it might have looked
somewhat more rigid.

30
00:01:36,870 --> 00:01:38,870
And it might not have
been able to handle things

31
00:01:38,870 --> 00:01:40,010
like obstacles.

32
00:01:40,010 --> 00:01:44,230
So let's look at another video
of how this is done in nature.

33
00:01:44,230 --> 00:01:46,880
Now, this is a massive
number of birds.

34
00:01:46,880 --> 00:01:51,110
I think in the video description
it says this is 300,000 birds.

35
00:01:51,110 --> 00:01:51,895
Wow.

36
00:01:51,895 --> 00:01:58,450
And it's so pretty how it
undulates and smushes around.

37
00:01:58,450 --> 00:02:00,340
Here's another view
of the same thing.

38
00:02:00,340 --> 00:02:02,900
I think they'll all take
off in the distance.

39
00:02:02,900 --> 00:02:04,530
Look at how pretty that is.

40
00:02:04,530 --> 00:02:06,890
And the all sort of do it
once, but there's certainly

41
00:02:06,890 --> 00:02:09,240
nobody in control.

42
00:02:09,240 --> 00:02:13,340
What a pretty-- ah, you've got
to love pictures of nature,

43
00:02:13,340 --> 00:02:15,980
in the evening,
in a lecture hall.

44
00:02:15,980 --> 00:02:18,120
We're about as far from
nature as people can get.

45
00:02:18,120 --> 00:02:19,786
My goodness.

46
00:02:19,786 --> 00:02:21,050
Ah, that's so pleasing.

47
00:02:21,050 --> 00:02:23,010
When are they going
to show the birds?

48
00:02:23,010 --> 00:02:25,240
I'm referring to the birds
that aren't in the water.

49
00:02:25,240 --> 00:02:27,110
The birds in the air, yeah.

50
00:02:27,110 --> 00:02:30,470
So what we want to
do today is just

51
00:02:30,470 --> 00:02:33,940
talk about the things
that are recognized

52
00:02:33,940 --> 00:02:35,720
as elements of swarm behavior.

53
00:02:35,720 --> 00:02:38,190
And then we'll
show how our robots

54
00:02:38,190 --> 00:02:41,670
can do much better than the
enemy robots if we do it right.

55
00:02:41,670 --> 00:02:42,320
Look at this.

56
00:02:42,320 --> 00:02:45,480
It's almost like two separate
swarms in certain areas.

57
00:02:45,480 --> 00:02:48,980
And it's an interesting
problem to say, at what point

58
00:02:48,980 --> 00:02:51,542
is it possible to
split into two swarms?

59
00:02:51,542 --> 00:02:53,750
Because maybe one of them
got too far from the other,

60
00:02:53,750 --> 00:02:56,350
or maybe it's just
too gosh darn big.

61
00:02:56,350 --> 00:02:59,280
Or maybe one half of them
are Democrats, the other half

62
00:02:59,280 --> 00:03:00,650
Republicans.

63
00:03:00,650 --> 00:03:02,550
There are a lot of
reasons that they

64
00:03:02,550 --> 00:03:05,140
might want to split into parts.

65
00:03:05,140 --> 00:03:08,410
So we'll talk about that too.

66
00:03:08,410 --> 00:03:10,250
Next, we'll talk about crowds.

67
00:03:10,250 --> 00:03:12,550
I mean, this is a thing
that you deal with every day

68
00:03:12,550 --> 00:03:15,680
when you go to eat
battlecode lecture dinner.

69
00:03:15,680 --> 00:03:19,990
You come up here, and there's
people trying to get to food.

70
00:03:19,990 --> 00:03:21,240
And I guess the food is there.

71
00:03:21,240 --> 00:03:23,060
No, this is a bad example,
because these guys are just

72
00:03:23,060 --> 00:03:24,900
going to walk
through one another.

73
00:03:24,900 --> 00:03:29,370
And the goal is trying
to get them to anticipate

74
00:03:29,370 --> 00:03:31,730
who isn't going to be in
the way, and to form lanes.

75
00:03:31,730 --> 00:03:35,520
That was the problem that these
guys were trying to address.

76
00:03:35,520 --> 00:03:37,640
And I guess nowadays
it's becoming

77
00:03:37,640 --> 00:03:40,600
popular to post your
thesis material online

78
00:03:40,600 --> 00:03:42,630
and to animate it in 3D.

79
00:03:42,630 --> 00:03:44,380
I don't see why this
is animated in 3D.

80
00:03:44,380 --> 00:03:45,796
You could just
show the same thing

81
00:03:45,796 --> 00:03:47,470
with a two dimensional map.

82
00:03:47,470 --> 00:03:50,570
Now here you're going
to see ripples form.

83
00:03:50,570 --> 00:03:53,590
And that's the expectation path.

84
00:03:53,590 --> 00:03:57,110
And so a small perturbation
in the position of someone

85
00:03:57,110 --> 00:04:00,780
is amplified into
like avoid this lane.

86
00:04:00,780 --> 00:04:04,050
And here he shows it
again with some lines.

87
00:04:04,050 --> 00:04:06,840
He does it with more people,
because every computer

88
00:04:06,840 --> 00:04:09,030
scientist knows that more
is better, especially

89
00:04:09,030 --> 00:04:10,710
when it comes to simulations.

90
00:04:10,710 --> 00:04:14,595
And so he shows a bunch of
them, and how these guys

91
00:04:14,595 --> 00:04:16,519
are going up, those guys
are going down there.

92
00:04:16,519 --> 00:04:19,050
And they sort of
make this vortex.

93
00:04:19,050 --> 00:04:20,230
And I think-- oh, yeah.

94
00:04:20,230 --> 00:04:20,730
I remember.

95
00:04:20,730 --> 00:04:22,079
This is really silly.

96
00:04:22,079 --> 00:04:24,010
The end of this
video, you're not

97
00:04:24,010 --> 00:04:25,450
going to see it coming at all.

98
00:04:25,450 --> 00:04:28,170
You're thinking, all right,
where can he go from here?

99
00:04:28,170 --> 00:04:30,220
He can show-- so
there's their model,

100
00:04:30,220 --> 00:04:32,030
and there's some
alternative models

101
00:04:32,030 --> 00:04:34,619
showing that their model is
a lot better because people

102
00:04:34,619 --> 00:04:35,410
aren't bunching up.

103
00:04:35,410 --> 00:04:38,900
So there's these high ES low
and high distance models.

104
00:04:38,900 --> 00:04:41,420
Which basically, the guys are
just avoiding one another.

105
00:04:41,420 --> 00:04:44,070
But in their model, they're
sort of creating a density

106
00:04:44,070 --> 00:04:47,770
field for an expected position.

107
00:04:47,770 --> 00:04:50,110
I think this is supposed
to be simulating armies.

108
00:04:50,110 --> 00:04:52,760
I guess they want a job
at Pixar or whatever.

109
00:04:52,760 --> 00:04:54,812
And so they just sort
of put this one in here,

110
00:04:54,812 --> 00:04:56,770
like oh, look at how
realistic your armies will

111
00:04:56,770 --> 00:04:59,200
look if people at the back
move 10 times as fast.

112
00:04:59,200 --> 00:05:01,660
[LAUGHTER]

113
00:05:01,660 --> 00:05:03,980
PROFESSOR: And then
another army is over here.

114
00:05:03,980 --> 00:05:06,960
Because just like I said,
these computer science people

115
00:05:06,960 --> 00:05:09,560
like nothing more than
to have a lot of things.

116
00:05:09,560 --> 00:05:12,190
I once saw a simulation.

117
00:05:12,190 --> 00:05:13,250
Oh, this was crazy.

118
00:05:13,250 --> 00:05:16,130
It was a simulation
of neutronics.

119
00:05:16,130 --> 00:05:18,202
So this is like
a college campus.

120
00:05:18,202 --> 00:05:20,660
I mean, you could really tell
where this guy's coming from.

121
00:05:20,660 --> 00:05:21,507
Movies, campus.

122
00:05:21,507 --> 00:05:23,590
And it's like people are
coming from certain doors

123
00:05:23,590 --> 00:05:24,310
and going to others.

124
00:05:24,310 --> 00:05:26,934
And I think the next shot he'll
show the paths of these people,

125
00:05:26,934 --> 00:05:29,610
showing that they're not bumping
into each other that much,

126
00:05:29,610 --> 00:05:31,560
and that they're kind
of straightish paths.

127
00:05:31,560 --> 00:05:33,290
Which is supposed to
be very impressive.

128
00:05:33,290 --> 00:05:34,665
But back to my
other story, I was

129
00:05:34,665 --> 00:05:36,350
watching a neutronics
simulation.

130
00:05:36,350 --> 00:05:38,840
And it had like 2
billion particles in it

131
00:05:38,840 --> 00:05:39,966
that were all in a lattice.

132
00:05:39,966 --> 00:05:42,381
OK, I'm going to pause this
here when it gets to the city.

133
00:05:42,381 --> 00:05:43,270
I mean, my goodness.

134
00:05:43,270 --> 00:05:44,720
How more complete do you need?

135
00:05:44,720 --> 00:05:45,850
Once you've seen it,
you've kind of seen it.

136
00:05:45,850 --> 00:05:48,230
But anyway, this 2 billion
particle lattice, which took

137
00:05:48,230 --> 00:05:50,690
like 10 hours on a
giant supercomputer

138
00:05:50,690 --> 00:05:54,390
to do this simulation of
damage from radiation,

139
00:05:54,390 --> 00:05:57,010
from like high energy
particle radiation.

140
00:05:57,010 --> 00:06:03,090
And the final result of this
massive simulation that I saw,

141
00:06:03,090 --> 00:06:06,810
it was to show that this
one incoming particle had

142
00:06:06,810 --> 00:06:09,540
ricocheted in a couple of ways.

143
00:06:09,540 --> 00:06:11,794
And the diagram of
this final result

144
00:06:11,794 --> 00:06:13,210
from this giant
computation, which

145
00:06:13,210 --> 00:06:14,990
was like the very end
of his presentation,

146
00:06:14,990 --> 00:06:15,950
because for some
reason people don't

147
00:06:15,950 --> 00:06:17,610
like to show the graphics first.

148
00:06:17,610 --> 00:06:19,399
They like to tell you
all the words first.

149
00:06:19,399 --> 00:06:20,940
In any case, he
showed me this thing.

150
00:06:20,940 --> 00:06:23,990
And it's like a little
hook or something.

151
00:06:23,990 --> 00:06:25,460
It was really stupid looking.

152
00:06:25,460 --> 00:06:26,740
Oh, my goodness.

153
00:06:26,740 --> 00:06:27,740
Can't believe it.

154
00:06:27,740 --> 00:06:29,180
But no, this guy's
final result is

155
00:06:29,180 --> 00:06:30,650
going to be much more exciting.

156
00:06:30,650 --> 00:06:31,750
So he's got these
people, and he's

157
00:06:31,750 --> 00:06:33,240
like, well, this
isn't realistic.

158
00:06:33,240 --> 00:06:35,630
In a real city, there are cars.

159
00:06:35,630 --> 00:06:38,470
So you've got to only
walk on the crosswalks.

160
00:06:38,470 --> 00:06:41,840
And you've got to
avoid certain things.

161
00:06:41,840 --> 00:06:44,550
Maybe you form lanes of traffic
on these crosswalks kind

162
00:06:44,550 --> 00:06:45,400
of automatically.

163
00:06:45,400 --> 00:06:46,530
Oh, now he's getting more.

164
00:06:46,530 --> 00:06:49,220
He's like, OK, I've got
too much time on my hands.

165
00:06:49,220 --> 00:06:52,034
I better implement stop lights.

166
00:06:52,034 --> 00:06:53,450
But it's still not
that realistic,

167
00:06:53,450 --> 00:06:55,720
because what if people
get hit by cars?

168
00:06:55,720 --> 00:06:57,910
There are no cars to
hit people, or to almost

169
00:06:57,910 --> 00:06:59,430
hit them, or to
scare them and get

170
00:06:59,430 --> 00:07:00,790
them to jump out of the way.

171
00:07:00,790 --> 00:07:05,730
So now we've got a 3D flyover,
because the only thing

172
00:07:05,730 --> 00:07:07,940
cooler than having
a 3D environment

173
00:07:07,940 --> 00:07:11,080
is a slow pan and zoom.

174
00:07:11,080 --> 00:07:12,980
So that's what he's
sort of doing here.

175
00:07:12,980 --> 00:07:15,310
And he puts the cars in.

176
00:07:15,310 --> 00:07:16,830
Obviously, he has
excellent taste.

177
00:07:16,830 --> 00:07:19,870
All the cars are from 1970.

178
00:07:19,870 --> 00:07:21,660
And I guess there's jaywalking.

179
00:07:21,660 --> 00:07:22,160
Oh, right.

180
00:07:22,160 --> 00:07:26,460
This is like a chemical
evacuation or something,

181
00:07:26,460 --> 00:07:29,210
so all these people are
covered in yellow stuff.

182
00:07:29,210 --> 00:07:30,600
And they're running away.

183
00:07:30,600 --> 00:07:34,900
And that's supposed to be really
useful for planning evacuations

184
00:07:34,900 --> 00:07:36,580
and such.

185
00:07:36,580 --> 00:07:37,080
Yeah.

186
00:07:37,080 --> 00:07:39,290
And everybody else
is just like, OK,

187
00:07:39,290 --> 00:07:43,320
I'm going to worry about
that three, two, one, now.

188
00:07:43,320 --> 00:07:47,810
And then the other people
start running away, too.

189
00:07:47,810 --> 00:07:49,650
They have really
bad reaction time.

190
00:07:49,650 --> 00:07:52,840
And then this is like, people
are walking in the street,

191
00:07:52,840 --> 00:07:54,950
and how it's messing up traffic.

192
00:07:54,950 --> 00:07:56,840
And it's like, wow,
this traffic is

193
00:07:56,840 --> 00:07:59,640
almost as bad as New
York City traffic.

194
00:07:59,640 --> 00:08:01,010
Almost as bad.

195
00:08:01,010 --> 00:08:02,900
And then, I mean,
what else can this guy

196
00:08:02,900 --> 00:08:04,950
do after he's done
this many things?

197
00:08:04,950 --> 00:08:07,055
Well, the last thing
that he has is aliens.

198
00:08:07,055 --> 00:08:08,910
[LAUGHTER]

199
00:08:08,910 --> 00:08:10,920
PROFESSOR: So there you go.

200
00:08:10,920 --> 00:08:13,300
Just to show that
everybody likes

201
00:08:13,300 --> 00:08:17,040
to have a little bit of fun
with pathing and navigation.

202
00:08:17,040 --> 00:08:19,820
So let's talk a little
bit about the things

203
00:08:19,820 --> 00:08:22,430
that we saw in those videos.

204
00:08:22,430 --> 00:08:24,720
and how they apply
to battlecode.

205
00:08:24,720 --> 00:08:27,160
So there's a whole
question of centralized

206
00:08:27,160 --> 00:08:28,800
versus decentralized control.

207
00:08:28,800 --> 00:08:31,086
And in the ones that we saw,
they weren't centralized.

208
00:08:31,086 --> 00:08:32,460
So I'm going to
show you examples

209
00:08:32,460 --> 00:08:34,980
in battlecode of both of those.

210
00:08:34,980 --> 00:08:38,299
Now, those can have
impacts on whether there's

211
00:08:38,299 --> 00:08:41,620
a radio attack possible,
whether your strategy's going

212
00:08:41,620 --> 00:08:43,990
to be vulnerable to that.

213
00:08:43,990 --> 00:08:46,880
Whether it's like a
map dependent strategy,

214
00:08:46,880 --> 00:08:50,030
because maybe on a
centralized controller,

215
00:08:50,030 --> 00:08:51,940
it works really well
if it's a tiny map.

216
00:08:51,940 --> 00:08:53,440
But maybe if it's
a huge map, you've

217
00:08:53,440 --> 00:08:56,060
got to start having multiple
swarms that are independent.

218
00:08:56,060 --> 00:09:00,060
And talking about
distributing information

219
00:09:00,060 --> 00:09:03,380
is another thing
to worry about it.

220
00:09:03,380 --> 00:09:07,110
So let's consider that
in the boyds model, which

221
00:09:07,110 --> 00:09:10,520
is supposed to sound like birds
if you have a Brooklyn accent,

222
00:09:10,520 --> 00:09:15,460
they considered several aspects
for computing the next position

223
00:09:15,460 --> 00:09:18,550
for each actor in the swarm.

224
00:09:18,550 --> 00:09:22,110
The separation of each
one to its neighbors.

225
00:09:22,110 --> 00:09:25,750
The pushing together of each
one, which they call cohesion,

226
00:09:25,750 --> 00:09:27,720
because the smarter
you sound, the more

227
00:09:27,720 --> 00:09:31,430
Ph.D. Money and grant
money you can make.

228
00:09:31,430 --> 00:09:34,367
They talk about alignment
so that everybody's

229
00:09:34,367 --> 00:09:35,450
facing the same direction.

230
00:09:35,450 --> 00:09:37,180
Now, that's somewhat
different this year

231
00:09:37,180 --> 00:09:39,780
than it has been in
past battlecode years.

232
00:09:39,780 --> 00:09:43,220
And then I guess they don't
talk too much about it,

233
00:09:43,220 --> 00:09:49,090
but they considered the later
effects of group splitting

234
00:09:49,090 --> 00:09:53,100
and lane creation,
which we saw today.

235
00:09:53,100 --> 00:09:54,930
So let's see here.

236
00:09:54,930 --> 00:09:57,420
I showed some examples
of these things.

237
00:09:57,420 --> 00:09:59,660
Alignment this year
is worth noting

238
00:09:59,660 --> 00:10:05,450
that there used to be a method
where you could take the robot

239
00:10:05,450 --> 00:10:07,210
info and do dot direction.

240
00:10:07,210 --> 00:10:10,650
You could get the direction
of any robot that you can see.

241
00:10:10,650 --> 00:10:13,140
Now, in this year's battlecode,
there's no direction.

242
00:10:13,140 --> 00:10:14,640
So what are you going to do?

243
00:10:14,640 --> 00:10:16,350
So let's see here.

244
00:10:16,350 --> 00:10:18,760
You can get the distances
for separation and cohesion.

245
00:10:18,760 --> 00:10:20,520
I mean, these are
just a fancy way

246
00:10:20,520 --> 00:10:23,200
of just saying there's
like a spring between you

247
00:10:23,200 --> 00:10:23,870
and the others.

248
00:10:23,870 --> 00:10:28,130
And this one is just saying that
you're headed the right way.

249
00:10:28,130 --> 00:10:30,446
There's really almost
nothing complicated to it.

250
00:10:30,446 --> 00:10:32,820
And we're going to end up
seeing some pretty neat looking

251
00:10:32,820 --> 00:10:35,780
behavior from very simple rules.

252
00:10:35,780 --> 00:10:37,900
So yeah, let's talk
about how you're

253
00:10:37,900 --> 00:10:40,570
going to know that you're
facing the right direction.

254
00:10:40,570 --> 00:10:42,690
I mean, you could check
all the neighbors.

255
00:10:42,690 --> 00:10:46,210
Let's see here, face
the right direction.

256
00:10:46,210 --> 00:10:48,690
You could check
all the neighbors

257
00:10:48,690 --> 00:10:50,700
against all the other ones.

258
00:10:50,700 --> 00:10:54,120
So you'd be checking
to see if robot

259
00:10:54,120 --> 00:10:57,150
ID number 97 was from this
position, and then that one.

260
00:10:57,150 --> 00:10:59,290
And then you would try
to find where he was,

261
00:10:59,290 --> 00:11:01,210
and correlate that,
and get the direction.

262
00:11:01,210 --> 00:11:03,910
And then take an average for
all the robots that you can see.

263
00:11:03,910 --> 00:11:06,140
That would be really
computationally expensive.

264
00:11:06,140 --> 00:11:08,410
Another option is
to say, all right,

265
00:11:08,410 --> 00:11:12,950
take average of
all allied robots.

266
00:11:12,950 --> 00:11:16,500
Or all within a range.

267
00:11:16,500 --> 00:11:20,700
And then see how
that average moves.

268
00:11:20,700 --> 00:11:22,280
So that's like an
implicit direction

269
00:11:22,280 --> 00:11:25,520
for the group of allied robots.

270
00:11:25,520 --> 00:11:28,500
And then an individual
robot, by doing that,

271
00:11:28,500 --> 00:11:30,030
could see where all of them are.

272
00:11:30,030 --> 00:11:32,430
Another option is just
to radio broadcast

273
00:11:32,430 --> 00:11:35,720
so that somebody somewhere is
deciding what direction to go.

274
00:11:35,720 --> 00:11:38,650
Or maybe all of them vote on the
direction that they want to go,

275
00:11:38,650 --> 00:11:40,608
because maybe some of
them are close to enemies

276
00:11:40,608 --> 00:11:42,260
and other ones are
standing on mines.

277
00:11:42,260 --> 00:11:44,600
And maybe they want
to put it up to a vote

278
00:11:44,600 --> 00:11:46,180
so that they can
wait for 50 turns

279
00:11:46,180 --> 00:11:48,750
before finding out
which direction they're

280
00:11:48,750 --> 00:11:51,639
going to go, and end up going
in the wrong direction anyway.

281
00:11:51,639 --> 00:11:53,430
Or they could have a
recount, and anything.

282
00:11:53,430 --> 00:11:54,800
It could be really fun.

283
00:11:54,800 --> 00:11:58,710
You could do like a whole
political simulation.

284
00:11:58,710 --> 00:12:01,970
So here are some
considerations when

285
00:12:01,970 --> 00:12:05,140
you're talking about central
control versus the group mind.

286
00:12:05,140 --> 00:12:09,070
Where the central controller,
you'd need a leader.

287
00:12:09,070 --> 00:12:12,510
And so leaders have
to have elections,

288
00:12:12,510 --> 00:12:17,640
and they also can die, and
sometimes they get lost.

289
00:12:17,640 --> 00:12:20,400
So those are things to worry
about when you have a leader.

290
00:12:20,400 --> 00:12:23,670
That means that the
headquarters is a good choice.

291
00:12:26,640 --> 00:12:29,700
And so if you can do all of
your computations in one place,

292
00:12:29,700 --> 00:12:31,380
that's much more efficient.

293
00:12:35,020 --> 00:12:37,710
Imagine if you had
to-- every robot

294
00:12:37,710 --> 00:12:39,170
has all the same information.

295
00:12:39,170 --> 00:12:41,519
But if all of them
make the decisions,

296
00:12:41,519 --> 00:12:43,310
then they all have to
do a lot of thinking.

297
00:12:43,310 --> 00:12:46,900
And in this game, if you
use up 10,000 byte codes,

298
00:12:46,900 --> 00:12:49,500
each of those robots is going
to have twice the power upkeep.

299
00:12:49,500 --> 00:12:51,380
It's going to be
pretty significant.

300
00:12:51,380 --> 00:12:56,610
So you could have the leader
broadcast all kinds of things.

301
00:12:56,610 --> 00:13:00,371
And later on we're going to talk
about some more little details

302
00:13:00,371 --> 00:13:02,870
of what the leader can broadcast
besides just what direction

303
00:13:02,870 --> 00:13:06,490
you're going in that can really
save you a lot of trouble.

304
00:13:06,490 --> 00:13:09,430
If you have a group mind,
it's a little different.

305
00:13:09,430 --> 00:13:12,510
You can have a
better site range,

306
00:13:12,510 --> 00:13:14,330
which is to say that
in previous years

307
00:13:14,330 --> 00:13:16,465
you were limited to just
one robot site range.

308
00:13:16,465 --> 00:13:18,090
So having a group
mind sort of gave you

309
00:13:18,090 --> 00:13:21,109
more access to the
edges of your cloud.

310
00:13:21,109 --> 00:13:23,150
And you could also make
it so that the front guys

311
00:13:23,150 --> 00:13:23,990
are leading.

312
00:13:23,990 --> 00:13:26,600
Which would be really
useful so that you

313
00:13:26,600 --> 00:13:28,360
could have them be the
first ones to note.

314
00:13:28,360 --> 00:13:30,760
This year it's not so important.

315
00:13:30,760 --> 00:13:32,821
But in like a real flock
of birds or something

316
00:13:32,821 --> 00:13:34,570
you might develop after
this, it would be.

317
00:13:34,570 --> 00:13:37,801
You can also broadcast of the
enemy locations in a group

318
00:13:37,801 --> 00:13:38,300
mind.

319
00:13:38,300 --> 00:13:40,596
There's no reason why a group
mind can't use broadcast.

320
00:13:40,596 --> 00:13:43,220
It's just a question of whether
you expect to receive something

321
00:13:43,220 --> 00:13:44,860
from a single entity,
or whether you

322
00:13:44,860 --> 00:13:47,020
expect to integrate
a bunch of broadcasts

323
00:13:47,020 --> 00:13:49,210
from a lot of entities.

324
00:13:49,210 --> 00:13:51,540
Broadcasting is becoming
cheaper and cheaper

325
00:13:51,540 --> 00:13:55,734
now that the read cost has
been decreased from the 0.01

326
00:13:55,734 --> 00:13:58,559
to 0.003.

327
00:13:58,559 --> 00:14:00,350
We've also increased
the number of channels

328
00:14:00,350 --> 00:14:02,940
in the latest release so
that your broadcasts are

329
00:14:02,940 --> 00:14:05,170
more secure against
radio attacks.

330
00:14:05,170 --> 00:14:07,240
At least those spanning attacks.

331
00:14:07,240 --> 00:14:09,800
It is worth mentioning
that you should

332
00:14:09,800 --> 00:14:12,567
be moving your channel somehow.

333
00:14:12,567 --> 00:14:14,900
Come up with your own algorithm
for shifting the channel

334
00:14:14,900 --> 00:14:15,880
that you're using.

335
00:14:15,880 --> 00:14:18,850
Something, for example, that
uses the clock round number

336
00:14:18,850 --> 00:14:21,422
so that you could always be
looking in a different place,

337
00:14:21,422 --> 00:14:23,130
and the enemy won't
locate which channels

338
00:14:23,130 --> 00:14:25,150
you're using on the radio.

339
00:14:25,150 --> 00:14:27,360
Which is to say,
which frequencies.

340
00:14:27,360 --> 00:14:32,510
So you might also want
some other features.

341
00:14:32,510 --> 00:14:36,280
So these boyds things with
their separation and such.

342
00:14:36,280 --> 00:14:40,260
What other features do you
want that aren't in boyds?

343
00:14:40,260 --> 00:14:42,630
So you might want
some in and out motion

344
00:14:42,630 --> 00:14:46,240
so that everybody takes a
uniform amount of damage.

345
00:14:46,240 --> 00:14:48,220
You might want some
cycling behavior.

346
00:14:48,220 --> 00:14:52,600
You might want to heal the weak
by going back to the med bays.

347
00:14:52,600 --> 00:14:57,310
You might want to apply
shields at shield generators.

348
00:14:57,310 --> 00:15:00,520
Those are encampments
that you could build.

349
00:15:00,520 --> 00:15:03,675
And you might want to make
sure that each robot knows

350
00:15:03,675 --> 00:15:04,550
the size of the pack.

351
00:15:04,550 --> 00:15:08,350
And if each one only has so
much information, especially

352
00:15:08,350 --> 00:15:11,180
about the enemy, then they
need to be able to get around.

353
00:15:11,180 --> 00:15:15,180
I mean, this year it's not
so much of a problem again.

354
00:15:15,180 --> 00:15:16,240
So let's see here.

355
00:15:16,240 --> 00:15:17,980
There's sort of an
interesting trade

356
00:15:17,980 --> 00:15:21,106
off here when you think
about the density of a group.

357
00:15:21,106 --> 00:15:23,480
But we're doing too much
talking and not enough watching.

358
00:15:23,480 --> 00:15:25,700
Let's watch an example.

359
00:15:25,700 --> 00:15:27,600
So I have some players here.

360
00:15:27,600 --> 00:15:32,200
And swarm one, I'm just going
to pick any old map here.

361
00:15:32,200 --> 00:15:33,970
Let's go to chunky.

362
00:15:33,970 --> 00:15:38,090
Chunky is a map I made which has
just these big chunks of mines.

363
00:15:38,090 --> 00:15:40,930
So here you can see, I've
got my swarm one player.

364
00:15:40,930 --> 00:15:42,360
It's very simple.

365
00:15:42,360 --> 00:15:45,306
It's quite close to the other
players that I showed you.

366
00:15:45,306 --> 00:15:47,430
It's not that exciting,
because the robots are just

367
00:15:47,430 --> 00:15:49,930
sort of-- they rush
at one another.

368
00:15:49,930 --> 00:15:51,980
They come together
in the middle.

369
00:15:51,980 --> 00:15:53,660
They go toward one another.

370
00:15:53,660 --> 00:15:55,800
There's not that much
swarm-like behavior.

371
00:15:55,800 --> 00:15:57,300
And when they come
together, they're

372
00:15:57,300 --> 00:15:59,490
not really doing anything
that intelligent.

373
00:15:59,490 --> 00:16:01,730
Like at this
moment, neither side

374
00:16:01,730 --> 00:16:05,500
is really being that smart
about what to do with the other.

375
00:16:05,500 --> 00:16:07,440
Here I'm going to do it
slightly differently.

376
00:16:07,440 --> 00:16:10,790
I'm going to have the robots,
instead of just choosing

377
00:16:10,790 --> 00:16:15,030
a rally point at random at the
very beginning of the game,

378
00:16:15,030 --> 00:16:18,220
and having each robot
compute it individually, here

379
00:16:18,220 --> 00:16:20,080
in swarm one, I'm
going to show how

380
00:16:20,080 --> 00:16:22,960
in swarm two,
which is this code,

381
00:16:22,960 --> 00:16:25,680
we've included the ability
for the headquarters

382
00:16:25,680 --> 00:16:28,450
to tell the robots where to go.

383
00:16:28,450 --> 00:16:30,820
So you could see--
let's see here,

384
00:16:30,820 --> 00:16:32,600
let's change the
font size so that you

385
00:16:32,600 --> 00:16:36,510
can see what I'm saying and
follow along in the code.

386
00:16:36,510 --> 00:16:39,550
Font 16, OK.

387
00:16:43,000 --> 00:16:46,480
The headquarter code here is
what the headquarters is doing.

388
00:16:46,480 --> 00:16:48,440
So it does some
spawning soldiers.

389
00:16:48,440 --> 00:16:51,350
And I've got a nice
code documentation here.

390
00:16:51,350 --> 00:16:53,090
And it moves the
rally point from time

391
00:16:53,090 --> 00:16:55,965
to time, which is to say that
at round greater than 500,

392
00:16:55,965 --> 00:17:00,180
it just sets the rally point to
the enemy headquarter location.

393
00:17:00,180 --> 00:17:02,410
And it messages allies
about where to go.

394
00:17:02,410 --> 00:17:04,609
So first it gets the
channel from my function

395
00:17:04,609 --> 00:17:05,859
called getChannel.

396
00:17:05,859 --> 00:17:11,050
And when I do that, I
say that if I'm team b,

397
00:17:11,050 --> 00:17:12,569
I'm going to have
this multiplier.

398
00:17:12,569 --> 00:17:14,910
If I'm team a, I'm going
to have this multiplier.

399
00:17:14,910 --> 00:17:17,280
So that when I
calculate the channel,

400
00:17:17,280 --> 00:17:21,630
I do the clock round number
times the multiplier,

401
00:17:21,630 --> 00:17:24,589
modulo the game constants,
broadcast max channels.

402
00:17:24,589 --> 00:17:27,200
So that means that the channel
that I'm broadcasting on

403
00:17:27,200 --> 00:17:29,650
is continually rotating,
but because it's just

404
00:17:29,650 --> 00:17:33,339
based on my team and the
clock round number, that

405
00:17:33,339 --> 00:17:35,630
means that every robot can
still read off that channel.

406
00:17:35,630 --> 00:17:37,650
Still knows where to
look for that channel.

407
00:17:37,650 --> 00:17:40,350
And rotating frequencies is
actually a common practice

408
00:17:40,350 --> 00:17:42,590
in modern military
communications.

409
00:17:42,590 --> 00:17:45,980
After all, it might not
have occurred to you

410
00:17:45,980 --> 00:17:50,440
that in the modern military,
they don't just flip open

411
00:17:50,440 --> 00:17:54,400
their cell phones and then
say, hey General Saxony,

412
00:17:54,400 --> 00:17:57,350
do you want to sort
of attack tonight?

413
00:17:57,350 --> 00:18:00,560
And the general
goes, yeah, maybe.

414
00:18:00,560 --> 00:18:01,590
I might.

415
00:18:01,590 --> 00:18:02,260
I'm not sure.

416
00:18:02,260 --> 00:18:04,374
Maybe I'll send you an email.

417
00:18:04,374 --> 00:18:05,790
They often use
radio, because it's

418
00:18:05,790 --> 00:18:08,090
there's not internet
connection out there

419
00:18:08,090 --> 00:18:10,110
in the middle of nowhere.

420
00:18:10,110 --> 00:18:12,094
And they have these
rotating radio frequencies.

421
00:18:12,094 --> 00:18:14,010
So yeah, that's a good
thing that you'll have.

422
00:18:14,010 --> 00:18:16,468
So once you've got the channel,
you need to have a message.

423
00:18:16,468 --> 00:18:19,220
So here I've written
the message builder

424
00:18:19,220 --> 00:18:21,880
that converts a map
location to an integer.

425
00:18:21,880 --> 00:18:23,850
So you can see
that function here.

426
00:18:23,850 --> 00:18:26,650
All I do is I multiply
the x value by 1,000

427
00:18:26,650 --> 00:18:28,120
and I add the y value.

428
00:18:28,120 --> 00:18:31,740
So then both x and y
values are included.

429
00:18:31,740 --> 00:18:33,550
Finally, on that
unpacking side, I

430
00:18:33,550 --> 00:18:36,190
use this function to
turn the integer back

431
00:18:36,190 --> 00:18:37,560
into a map location.

432
00:18:37,560 --> 00:18:39,850
It makes life easier.

433
00:18:39,850 --> 00:18:42,930
So then at the robot side,
you can see the result.

434
00:18:42,930 --> 00:18:47,270
So we're going to show
swarm one versus swarm two.

435
00:18:47,270 --> 00:18:48,990
Swarm two has two advantages.

436
00:18:48,990 --> 00:18:50,860
So here swarm two is in blue.

437
00:18:50,860 --> 00:18:53,590
And swarm two has two
advantages, the first of which

438
00:18:53,590 --> 00:18:58,550
is that it receives the
messages from the headquarters.

439
00:18:58,550 --> 00:19:01,490
So it'll go attack at
round 500, rather than

440
00:19:01,490 --> 00:19:04,430
the round that these guys are
attacking, which is round 200.

441
00:19:04,430 --> 00:19:06,304
So I guess it has sort
of an advantage there.

442
00:19:06,304 --> 00:19:09,850
And second advantage is, it
doesn't defuse mines in combat.

443
00:19:09,850 --> 00:19:11,940
Some of you may be
familiar with World

444
00:19:11,940 --> 00:19:13,370
of Warcraft or those type games.

445
00:19:13,370 --> 00:19:16,660
Defusing mines in combat
is like looting in combat.

446
00:19:16,660 --> 00:19:17,840
It's definitely a no-no.

447
00:19:17,840 --> 00:19:20,970
You don't want to-- that's
not how to be a team player.

448
00:19:20,970 --> 00:19:23,440
You could see the result is
that just because you aren't

449
00:19:23,440 --> 00:19:26,320
defusing mines in combat
where the other team was,

450
00:19:26,320 --> 00:19:28,990
you end up with almost
your whole army remaining.

451
00:19:28,990 --> 00:19:30,190
What a huge difference.

452
00:19:30,190 --> 00:19:31,820
And then quickly
at round 500, you

453
00:19:31,820 --> 00:19:35,360
can see that these guys are
successfully receiving a goal.

454
00:19:35,360 --> 00:19:39,450
And here in size
0.001 font, you can

455
00:19:39,450 --> 00:19:41,810
see that it knows
its current location.

456
00:19:41,810 --> 00:19:43,620
And it has a goal
location which has

457
00:19:43,620 --> 00:19:46,110
been broadcast from this guy.

458
00:19:46,110 --> 00:19:47,880
So there you can see
a little example,

459
00:19:47,880 --> 00:19:51,310
but that's not as
exciting as it might be.

460
00:19:51,310 --> 00:19:54,320
Let's consider adding
some distance tolerance

461
00:19:54,320 --> 00:19:56,210
in go to location.

462
00:19:56,210 --> 00:20:00,900
So let's do that so that
when they go to a location,

463
00:20:00,900 --> 00:20:03,400
they don't just keep
bobbling around in there.

464
00:20:03,400 --> 00:20:05,540
So that they can actually
start to get work done.

465
00:20:05,540 --> 00:20:07,110
So here, they're
going to location.

466
00:20:07,110 --> 00:20:08,750
And then they just calm down.

467
00:20:08,750 --> 00:20:11,810
And they take their
spot, and they sit there.

468
00:20:11,810 --> 00:20:12,800
Really nice.

469
00:20:12,800 --> 00:20:14,264
And if they're on
a certain tile,

470
00:20:14,264 --> 00:20:15,430
I even have they lay a mine.

471
00:20:15,430 --> 00:20:17,360
So this is a neato
thing that I thought

472
00:20:17,360 --> 00:20:22,000
of is if you lay mines not
in a fully dense pattern,

473
00:20:22,000 --> 00:20:26,270
then when the enemy attacks,
you have the opportunity

474
00:20:26,270 --> 00:20:28,620
to sort of whittle them down.

475
00:20:28,620 --> 00:20:30,430
Because you might
have this situation.

476
00:20:30,430 --> 00:20:32,380
It's kind of hard
to see on this map.

477
00:20:32,380 --> 00:20:34,680
I might take a screenshot
and then work on it.

478
00:20:34,680 --> 00:20:36,970
So let's say that we've
got these mines here,

479
00:20:36,970 --> 00:20:41,470
and that we intend to use
them in a really clever way.

480
00:20:41,470 --> 00:20:42,090
Yeah.

481
00:20:42,090 --> 00:20:43,410
So let's go here.

482
00:20:47,120 --> 00:20:49,310
Super leet.

483
00:20:49,310 --> 00:20:49,930
Super leet.

484
00:20:54,790 --> 00:20:57,585
So maybe the enemy
is coming this way.

485
00:20:57,585 --> 00:20:59,210
Now, if they come
here, they can either

486
00:20:59,210 --> 00:21:01,130
start to defuse the
mines, and they'll

487
00:21:01,130 --> 00:21:05,050
be just waiting on that corner,
or they can infiltrate inside.

488
00:21:05,050 --> 00:21:07,140
Because it doesn't really
present a giant obstacle

489
00:21:07,140 --> 00:21:07,940
to them.

490
00:21:07,940 --> 00:21:09,810
And once they're
in here, I'm going

491
00:21:09,810 --> 00:21:13,050
to have a big advantage
because I can confront them

492
00:21:13,050 --> 00:21:16,000
with a lot more area.

493
00:21:16,000 --> 00:21:18,730
Because they're all
so sparsely laid out.

494
00:21:18,730 --> 00:21:20,880
So I think this will
work really well.

495
00:21:20,880 --> 00:21:22,380
And if they want
to retreat, there's

496
00:21:22,380 --> 00:21:24,380
definitely no way that's
going to work for them.

497
00:21:24,380 --> 00:21:26,929
Because they can't do any real
cycling in this configuration.

498
00:21:26,929 --> 00:21:28,720
I think it's going to
totally mess them up.

499
00:21:31,260 --> 00:21:34,310
So here in this one, I've got
these guys who are cleverly

500
00:21:34,310 --> 00:21:40,810
not-- they're not going
to defuse mines in combat.

501
00:21:40,810 --> 00:21:44,050
And they'll even sort of back
up and fight over the mines.

502
00:21:44,050 --> 00:21:47,570
So here they're backing
up to the mines.

503
00:21:47,570 --> 00:21:52,680
And they're trying to bring
the fight back to home.

504
00:21:52,680 --> 00:21:57,100
Let's show it with swarm
one versus swarm three,

505
00:21:57,100 --> 00:22:00,610
because then they won't
both have that part.

506
00:22:00,610 --> 00:22:01,180
Oh, wait.

507
00:22:01,180 --> 00:22:02,560
That was pretty one-sided.

508
00:22:02,560 --> 00:22:03,060
Right.

509
00:22:03,060 --> 00:22:04,480
So this is something
that happens.

510
00:22:04,480 --> 00:22:07,690
So here the enemy
is going to show up

511
00:22:07,690 --> 00:22:11,170
and I'm going to retreat
off the minefield.

512
00:22:11,170 --> 00:22:13,115
So then they're sort of
busy, and I come back.

513
00:22:13,115 --> 00:22:14,910
And then I retreat
a little bit more.

514
00:22:14,910 --> 00:22:17,100
And they sort of start
defusing and they

515
00:22:17,100 --> 00:22:18,250
think that's a good idea.

516
00:22:18,250 --> 00:22:21,580
And then I sort of go this
back and forth type motion.

517
00:22:21,580 --> 00:22:24,290
And that kind of thing
can help to string out

518
00:22:24,290 --> 00:22:25,266
the enemies in a line.

519
00:22:25,266 --> 00:22:26,390
Give me a little advantage.

520
00:22:26,390 --> 00:22:32,030
But even still, it's not
looking that much like swarming.

521
00:22:32,030 --> 00:22:34,520
Now, I want to make the point
that a lot of these maps

522
00:22:34,520 --> 00:22:39,420
are absolutely filled
with these neutral mines,

523
00:22:39,420 --> 00:22:40,710
especially to begin with.

524
00:22:40,710 --> 00:22:42,016
In this map.

525
00:22:42,016 --> 00:22:43,890
there's so many neutral
mines to get through.

526
00:22:43,890 --> 00:22:46,265
If you were really intelligent,
you could take this path.

527
00:22:46,265 --> 00:22:48,540
And I'll show an
example later of how

528
00:22:48,540 --> 00:22:51,560
you could write a player
that really efficiently paths

529
00:22:51,560 --> 00:22:54,570
through this space for
a whole mess of troops.

530
00:22:54,570 --> 00:22:57,320
So here I'm retreating
past my little minefield.

531
00:22:57,320 --> 00:22:59,830
And then I'm sort of
bopping on forward again.

532
00:22:59,830 --> 00:23:01,900
And he's sort of
poking the dragon here.

533
00:23:01,900 --> 00:23:03,470
He's making me irritated.

534
00:23:03,470 --> 00:23:07,081
And so then-- oh, I'm
just going to kill him.

535
00:23:07,081 --> 00:23:07,580
Hmm.

536
00:23:07,580 --> 00:23:09,954
But the killing him was less
effective than it might have

537
00:23:09,954 --> 00:23:12,670
been because a lot of my
guys got stuck back there.

538
00:23:12,670 --> 00:23:14,940
And they didn't know what
to do because I told them,

539
00:23:14,940 --> 00:23:17,062
please don't defuse
mines in combat.

540
00:23:17,062 --> 00:23:18,520
And it was like
combat because they

541
00:23:18,520 --> 00:23:20,760
were going to attack
the enemy headquarters.

542
00:23:20,760 --> 00:23:22,540
But nevertheless,
these guys got stuck.

543
00:23:22,540 --> 00:23:23,885
So what am I going to do?

544
00:23:23,885 --> 00:23:25,260
Well, here's an
alternative where

545
00:23:25,260 --> 00:23:29,290
you could say each guy has a
certain amount of patience.

546
00:23:29,290 --> 00:23:31,280
And if he's stuck in
one spot, then he's

547
00:23:31,280 --> 00:23:35,390
going to go ahead and defuse
those mines in combat.

548
00:23:35,390 --> 00:23:37,870
So here, once again, this
pulling back and going forward

549
00:23:37,870 --> 00:23:41,230
strategy is very effective
against an equally sized enemy.

550
00:23:41,230 --> 00:23:43,590
These minefields are a
wonderful defensive advantage

551
00:23:43,590 --> 00:23:45,930
that you should definitely use.

552
00:23:45,930 --> 00:23:49,310
So then here, when
these guys get annoyed,

553
00:23:49,310 --> 00:23:50,874
they can start digging through.

554
00:23:50,874 --> 00:23:53,290
And eventually, they'll make
it through to the other side.

555
00:23:53,290 --> 00:23:56,840
Which will be pretty
useful in that case.

556
00:23:56,840 --> 00:24:00,730
Let's do another example
versus basic player.

557
00:24:00,730 --> 00:24:04,590
I think this example
demonstrates that tunneling

558
00:24:04,590 --> 00:24:06,225
effect a little bit better.

559
00:24:06,225 --> 00:24:08,600
Now, basic player is building
all kinds of useful things,

560
00:24:08,600 --> 00:24:10,480
like suppliers and generators.

561
00:24:10,480 --> 00:24:12,712
And that's going to make
him especially powerful.

562
00:24:12,712 --> 00:24:14,670
So it's going to be--
no, actually, I beat him.

563
00:24:14,670 --> 00:24:17,490
[LAUGHTER]

564
00:24:17,490 --> 00:24:19,290
PROFESSOR: So that's
a really good example.

565
00:24:19,290 --> 00:24:23,370
Let's move on, and I'm going
to try to beat swarm four.

566
00:24:23,370 --> 00:24:25,550
So swarm four was
like-- well, maybe

567
00:24:25,550 --> 00:24:27,345
I'll continue beating swarm one.

568
00:24:27,345 --> 00:24:28,720
What I'm going to
do is I'm going

569
00:24:28,720 --> 00:24:32,120
to use guided swarm, which
is a new player that I made.

570
00:24:32,120 --> 00:24:35,340
So this guided
swarm's really neat.

571
00:24:35,340 --> 00:24:37,250
This is a lot more like birds.

572
00:24:37,250 --> 00:24:43,930
And the guided swarm's goal, his
goal is to be a little bit more

573
00:24:43,930 --> 00:24:45,810
puffy.

574
00:24:45,810 --> 00:24:47,020
Yeah, that's him up there.

575
00:24:47,020 --> 00:24:47,519
Look at him.

576
00:24:47,519 --> 00:24:48,850
He's like all cloud like.

577
00:24:48,850 --> 00:24:51,380
And he's keeping it real.

578
00:24:51,380 --> 00:24:52,960
What a guy.

579
00:24:52,960 --> 00:24:55,980
So you can see here that the
guided swarm's kind of OK,

580
00:24:55,980 --> 00:24:59,590
but he's so spread out that
he doesn't do well in combat.

581
00:24:59,590 --> 00:25:02,240
What can we do to make
this guided swarm better?

582
00:25:02,240 --> 00:25:05,950
What can we do-- I think one
thing that works really well

583
00:25:05,950 --> 00:25:09,660
is to put him on a map
that's better for him.

584
00:25:09,660 --> 00:25:11,160
So I'm going to put
him on this one.

585
00:25:11,160 --> 00:25:13,200
Now he's a bit farther
away from the enemy.

586
00:25:13,200 --> 00:25:15,200
And my hope is that
he'll have time

587
00:25:15,200 --> 00:25:16,850
to produce enough
units that he'll

588
00:25:16,850 --> 00:25:18,610
start laying a big minefield.

589
00:25:18,610 --> 00:25:21,441
Yes, I think that will
help him out significantly.

590
00:25:21,441 --> 00:25:21,940
Oh, come on.

591
00:25:21,940 --> 00:25:23,710
Please lay that minefield.

592
00:25:23,710 --> 00:25:25,620
Please lay that minefield.

593
00:25:25,620 --> 00:25:28,544
Oh, he didn't have time
to lay the minefield.

594
00:25:28,544 --> 00:25:30,460
I think what I'll do is
I'll tell him to build

595
00:25:30,460 --> 00:25:32,824
the minefield just a
little bit earlier.

596
00:25:32,824 --> 00:25:34,240
I'm going to make
this-- right now

597
00:25:34,240 --> 00:25:36,960
I've got it-- ooh, don't let
me change the wrong code.

598
00:25:36,960 --> 00:25:38,147
It's so easy to do that.

599
00:25:38,147 --> 00:25:40,730
You think robot player got Java
is the code you're working on,

600
00:25:40,730 --> 00:25:42,194
and it is, but you're wrong.

601
00:25:42,194 --> 00:25:45,040
[LAUGHTER]

602
00:25:45,040 --> 00:25:47,580
So I was working on
guided swarm one,

603
00:25:47,580 --> 00:25:54,010
and I want the headquarters code
to start building these robots.

604
00:25:54,010 --> 00:25:54,880
Oh, yeah.

605
00:25:54,880 --> 00:25:56,910
I want him to research pickaxe.

606
00:25:56,910 --> 00:25:58,560
And once he's
researched pickaxe,

607
00:25:58,560 --> 00:26:01,850
I told him to tell the other
guys on channel plus 1.

608
00:26:01,850 --> 00:26:04,390
So here you can use your
get channel function

609
00:26:04,390 --> 00:26:05,890
and then add one,
and now you've got

610
00:26:05,890 --> 00:26:08,710
sort of a set of
channels you can use.

611
00:26:08,710 --> 00:26:11,490
So I want him to tell the others
to please start laying mines

612
00:26:11,490 --> 00:26:14,190
as soon as I have
the pickaxe upgrade.

613
00:26:14,190 --> 00:26:15,180
All right.

614
00:26:15,180 --> 00:26:17,260
So that's going to happen.

615
00:26:17,260 --> 00:26:21,080
I'll just make it at this
point, and hopefully that

616
00:26:21,080 --> 00:26:22,630
will give him enough
time to do it.

617
00:26:27,210 --> 00:26:27,710
Yeah.

618
00:26:27,710 --> 00:26:30,020
So now he's going to
build fewer allied units

619
00:26:30,020 --> 00:26:31,790
before he starts researching.

620
00:26:31,790 --> 00:26:34,510
Gosh, this is a tiny map.

621
00:26:34,510 --> 00:26:38,460
If you can see this accurately,
you don't need glasses.

622
00:26:38,460 --> 00:26:39,941
Yes, that's probably true.

623
00:26:39,941 --> 00:26:40,440
Ah, right.

624
00:26:40,440 --> 00:26:41,648
So they started laying mines.

625
00:26:41,648 --> 00:26:43,030
And look at that pattern.

626
00:26:43,030 --> 00:26:44,580
Look at the pattern they used.

627
00:26:44,580 --> 00:26:45,640
Oh, my goodness.

628
00:26:45,640 --> 00:26:47,520
They must be somewhat skilled.

629
00:26:47,520 --> 00:26:48,630
Can't believe that.

630
00:26:48,630 --> 00:26:49,440
Ah, ooh.

631
00:26:49,440 --> 00:26:50,746
[LAUGHTER]

632
00:26:50,746 --> 00:26:52,120
PROFESSOR: But
you see the point.

633
00:26:52,120 --> 00:26:55,140
The point that I made was
that I found a really neat way

634
00:26:55,140 --> 00:26:57,330
to make them lay mines
in the right pattern.

635
00:26:57,330 --> 00:26:59,120
Now, maybe you can't
really see that well.

636
00:26:59,120 --> 00:27:01,920
So I'm going to open up
good fashioned Mathematica.

637
00:27:01,920 --> 00:27:03,820
Not that.

638
00:27:03,820 --> 00:27:05,705
Good old fashioned-- not that.

639
00:27:05,705 --> 00:27:07,080
Good old fashioned--
there we go.

640
00:27:07,080 --> 00:27:07,780
Mathematica.

641
00:27:07,780 --> 00:27:11,230
So you can see that this
arrangement, this disposition

642
00:27:11,230 --> 00:27:14,680
of mine positions, which is
like over one, up two-- I've

643
00:27:14,680 --> 00:27:17,430
got to zoom in a little bit.

644
00:27:17,430 --> 00:27:19,630
Yeah, if you go over one,
up two, over one, up two,

645
00:27:19,630 --> 00:27:22,330
this arrangement is going
to totally fill space.

646
00:27:22,330 --> 00:27:24,922
So this is where I've colored
all the adjacent ones in.

647
00:27:24,922 --> 00:27:26,880
And this is going to mean
that if you lay mines

648
00:27:26,880 --> 00:27:29,590
at these centers, and you
have the pickaxe upgrade,

649
00:27:29,590 --> 00:27:31,100
then you will be
fully tiling space.

650
00:27:31,100 --> 00:27:32,300
Really useful.

651
00:27:32,300 --> 00:27:35,430
And here I'm going to
give away the trick.

652
00:27:35,430 --> 00:27:40,150
You'll be on the right spot
if mod of 2 plus your x val,

653
00:27:40,150 --> 00:27:44,230
plus 1 times your
y val equals zero.

654
00:27:44,230 --> 00:27:48,480
If mod, this number 5-- see
if I change this to like a 7,

655
00:27:48,480 --> 00:27:49,780
then the whole pattern changes.

656
00:27:49,780 --> 00:27:52,400
And you can investigate
different patterns

657
00:27:52,400 --> 00:27:55,314
by playing around with it.

658
00:27:55,314 --> 00:27:57,230
And these patterns may
have different effects,

659
00:27:57,230 --> 00:27:59,810
depending on your upgrades and
the positioning of your mines.

660
00:27:59,810 --> 00:28:02,110
And it made me get
kind of a neat idea.

661
00:28:02,110 --> 00:28:04,340
So this is fully dense.

662
00:28:04,340 --> 00:28:06,670
And you could do partly
dense, which is really simple.

663
00:28:06,670 --> 00:28:09,870
You could just do mod i plus j,
where i and j are your x and y

664
00:28:09,870 --> 00:28:10,370
values.

665
00:28:10,370 --> 00:28:11,750
You just do mod 2.

666
00:28:11,750 --> 00:28:15,140
And if that mod 2 is
equal to 1, then that'll

667
00:28:15,140 --> 00:28:16,182
get you every other tile.

668
00:28:16,182 --> 00:28:17,764
So no matter where
you are on the map,

669
00:28:17,764 --> 00:28:19,980
you don't have to start
comparing to your allied tile

670
00:28:19,980 --> 00:28:20,830
locations.

671
00:28:20,830 --> 00:28:22,390
You don't have to start
doing a whole bunch of really

672
00:28:22,390 --> 00:28:23,348
complicated operations.

673
00:28:23,348 --> 00:28:26,980
Just say, if my robot
is on this nice tile,

674
00:28:26,980 --> 00:28:29,370
then go ahead and lay a mine.

675
00:28:29,370 --> 00:28:30,330
Really nice.

676
00:28:30,330 --> 00:28:32,680
And here's an
interesting alternative.

677
00:28:32,680 --> 00:28:36,011
So this one I slightly farther
spaced apart than the one

678
00:28:36,011 --> 00:28:36,510
above.

679
00:28:36,510 --> 00:28:39,220
This is sort of like knights in
chess, where they can move up

680
00:28:39,220 --> 00:28:43,240
and up and up, or they can
move over and over and over.

681
00:28:43,240 --> 00:28:46,660
Well here, I've gone and done
it a little bit farther apart.

682
00:28:46,660 --> 00:28:49,650
And the result is this
neato looking structure.

683
00:28:49,650 --> 00:28:52,590
So it's like the
one we did before,

684
00:28:52,590 --> 00:28:55,480
where we had only done
a checkerboard pattern.

685
00:28:55,480 --> 00:28:58,430
Only now, we have a
lot fewer mines to lay,

686
00:28:58,430 --> 00:29:01,200
and we still get sort of
this interpenetrated mix.

687
00:29:01,200 --> 00:29:04,040
This can be really useful
for fooling the enemy team.

688
00:29:04,040 --> 00:29:09,080
The formula for this
system is 3x plus 1y.

689
00:29:09,080 --> 00:29:11,020
And as long as you
mod that with 8,

690
00:29:11,020 --> 00:29:13,350
you'll get the answer
you're looking for.

691
00:29:13,350 --> 00:29:15,310
Now, this is sort of
an interesting system.

692
00:29:15,310 --> 00:29:18,810
If I change this to 1,
and I change this to 5,

693
00:29:18,810 --> 00:29:23,040
I get the same result
but sort of mirror image.

694
00:29:23,040 --> 00:29:25,640
Just playing around with
this kind of problem

695
00:29:25,640 --> 00:29:29,200
can be quite enlightening
if you're not

696
00:29:29,200 --> 00:29:33,630
the kind of person who instantly
sees how these systems work.

697
00:29:33,630 --> 00:29:37,577
So let's do some
summaries, and then

698
00:29:37,577 --> 00:29:39,160
at the end of the
summaries, I'm going

699
00:29:39,160 --> 00:29:42,410
to show you a really
awesome player.

700
00:29:42,410 --> 00:29:45,120
A really amazing version
of this thing that

701
00:29:45,120 --> 00:29:47,067
has the headquarters
telling it where to go.

702
00:29:47,067 --> 00:29:49,150
So this is going to be at
the very end of lecture,

703
00:29:49,150 --> 00:29:50,066
I'm going to show you.

704
00:29:50,066 --> 00:29:51,450
It's the same as
this player that

705
00:29:51,450 --> 00:29:54,130
just lost this match,
this one up here,

706
00:29:54,130 --> 00:29:56,740
only instead of being dumb
when the enemy arrives,

707
00:29:56,740 --> 00:29:58,970
it's going to be smart
about where it puts itself

708
00:29:58,970 --> 00:30:00,110
with respect to the enemy.

709
00:30:00,110 --> 00:30:02,200
Which is to say that
it's going to avoid

710
00:30:02,200 --> 00:30:04,620
the enemy unless the
number of nearby allies

711
00:30:04,620 --> 00:30:06,360
is a really high number.

712
00:30:06,360 --> 00:30:08,890
And at the same time,
the central headquarters

713
00:30:08,890 --> 00:30:10,930
is going to tell it
to capture encampments

714
00:30:10,930 --> 00:30:13,630
and to build suppliers
and generators on them.

715
00:30:13,630 --> 00:30:15,610
So you're going to start
to have a player that,

716
00:30:15,610 --> 00:30:18,670
just from few little pieces,
is starting to do everything

717
00:30:18,670 --> 00:30:21,630
it needs to do to
start winning matches.

718
00:30:21,630 --> 00:30:23,360
So I'll show that
at the very end

719
00:30:23,360 --> 00:30:25,605
after I summarize some
of things that I've done,

720
00:30:25,605 --> 00:30:27,980
and some of the things that
I think you ought to try out.

721
00:30:27,980 --> 00:30:29,780
Because I can put
this code together

722
00:30:29,780 --> 00:30:31,370
in a certain amount
of time, but I

723
00:30:31,370 --> 00:30:33,480
can't try every
possible combination.

724
00:30:33,480 --> 00:30:36,460
So let's go straight
on over to here,

725
00:30:36,460 --> 00:30:39,130
and I want to just talk
about two categories.

726
00:30:39,130 --> 00:30:42,300
Let's talk about the shape
of your system, which

727
00:30:42,300 --> 00:30:44,210
is to say the cloud.

728
00:30:44,210 --> 00:30:47,970
Let's talk about how
you achieve the shape.

729
00:30:47,970 --> 00:30:52,960
And you can achieve it
either using a headquarters,

730
00:30:52,960 --> 00:30:54,710
and I'll talk about
all the amazing things

731
00:30:54,710 --> 00:30:56,293
that you can have
the headquarter tell

732
00:30:56,293 --> 00:30:58,180
the group that you might
not have thought of.

733
00:30:58,180 --> 00:31:03,880
And you can achieve it
using this group mind.

734
00:31:03,880 --> 00:31:07,330
So let's do the
group mind first,

735
00:31:07,330 --> 00:31:09,920
where you could
check the neighbors

736
00:31:09,920 --> 00:31:13,080
and go to the area
with the fewest.

737
00:31:13,080 --> 00:31:14,490
I sent around some code.

738
00:31:14,490 --> 00:31:16,600
I don't know how
many of you used it.

739
00:31:16,600 --> 00:31:17,860
And it looked like this.

740
00:31:17,860 --> 00:31:22,520
Let me show you my lecture
three robot player release.

741
00:31:22,520 --> 00:31:27,410
It's not incredibly good code,
but you can see the point.

742
00:31:27,410 --> 00:31:32,640
So when these guys get
close to one another,

743
00:31:32,640 --> 00:31:34,730
each one has this list.

744
00:31:34,730 --> 00:31:36,950
And this list shows the
number of adjacent enemy

745
00:31:36,950 --> 00:31:37,780
and allied units.

746
00:31:37,780 --> 00:31:41,620
So that list is
absolutely gosh darn tiny.

747
00:31:41,620 --> 00:31:44,770
So I'm going to go ahead and
do a fancy way of zooming in.

748
00:31:44,770 --> 00:31:45,510
Oh, my goodness.

749
00:31:45,510 --> 00:31:46,970
That's even less legible.

750
00:31:46,970 --> 00:31:52,670
That says 22, 21, 2,
0, 0, 1, 12, 23, me 12.

751
00:31:52,670 --> 00:31:56,490
So me 12 means that I
have two allies adjacent

752
00:31:56,490 --> 00:31:58,150
and one enemy adjacent.

753
00:31:58,150 --> 00:32:01,700
And adjacent this number, these
are eight different numbers

754
00:32:01,700 --> 00:32:07,160
indicating north,
northeast, east, southeast,

755
00:32:07,160 --> 00:32:10,880
and so on, around
the clock directions.

756
00:32:10,880 --> 00:32:13,420
Although not with
12, but with eight.

757
00:32:13,420 --> 00:32:14,620
Yeah.

758
00:32:14,620 --> 00:32:17,040
And it has this many enemies
and this many allies.

759
00:32:17,040 --> 00:32:19,600
And so I've written the code
for you that will do this.

760
00:32:19,600 --> 00:32:21,370
And so the robot
can sort of start

761
00:32:21,370 --> 00:32:23,820
to make decisions about
who is adjacent to it.

762
00:32:23,820 --> 00:32:26,200
All you've got to
do is start sorting

763
00:32:26,200 --> 00:32:28,350
how good it is to be next
to allies or enemies.

764
00:32:28,350 --> 00:32:29,920
Like maybe it's
really good to be

765
00:32:29,920 --> 00:32:32,980
next to a lot of allies
and very few enemies.

766
00:32:32,980 --> 00:32:34,880
Otherwise you want to
be close to few allies

767
00:32:34,880 --> 00:32:35,838
so that you spread out.

768
00:32:35,838 --> 00:32:39,920
There are a lot of things
that you can build on that.

769
00:32:39,920 --> 00:32:42,564
So that's why I
included that code.

770
00:32:42,564 --> 00:32:45,230
So you could check neighbors and
go to the area with the fewest.

771
00:32:45,230 --> 00:32:47,800
You could compare to
neighbor repulsion.

772
00:32:47,800 --> 00:32:52,830
So the repulsion is accomplished
in the following way.

773
00:32:52,830 --> 00:32:57,320
So I've written this code
here in guided swarm.

774
00:32:57,320 --> 00:33:00,999
And here's where you are
repelling from the enemies.

775
00:33:00,999 --> 00:33:02,540
So it isn't just
one way of doing it.

776
00:33:02,540 --> 00:33:03,870
A lot of ways of doing it.

777
00:33:03,870 --> 00:33:06,090
But I've called this
function freeGo,

778
00:33:06,090 --> 00:33:09,180
and it lets you
sort of freely float

779
00:33:09,180 --> 00:33:13,047
among your various
alternatives of places to go.

780
00:33:13,047 --> 00:33:14,880
What it does is it says,
all right, here I'm

781
00:33:14,880 --> 00:33:16,890
starting at my location.

782
00:33:16,890 --> 00:33:19,520
I'll get a direction
to the target location,

783
00:33:19,520 --> 00:33:22,090
which is like the waypoint.

784
00:33:22,090 --> 00:33:24,370
And now I'm going to
apply a target weight,

785
00:33:24,370 --> 00:33:26,980
because maybe I really want
to get there if I'm far away.

786
00:33:26,980 --> 00:33:29,530
But if I'm kind of close,
I don't really care.

787
00:33:29,530 --> 00:33:31,835
So now I have this function
called target weight, where

788
00:33:31,835 --> 00:33:34,230
if my distance squared
is greater than 100,

789
00:33:34,230 --> 00:33:35,352
I'll call the weighting 5.

790
00:33:35,352 --> 00:33:37,060
If my distance squared
is greater than 9,

791
00:33:37,060 --> 00:33:38,250
I'll call the weighting 2.

792
00:33:38,250 --> 00:33:41,520
And if I'm really close, then
I'll call the weighting 1.

793
00:33:41,520 --> 00:33:44,580
So that's sort of giving me
this spring-like behavior

794
00:33:44,580 --> 00:33:46,240
to get to where I want to go.

795
00:33:46,240 --> 00:33:48,926
And it's going to give
me this gassy behavior.

796
00:33:48,926 --> 00:33:51,050
So what I'll do is I'll
make a goal location, which

797
00:33:51,050 --> 00:33:53,060
is not the same as the target.

798
00:33:53,060 --> 00:33:56,822
So that goal location, I'm
going to just add to my location

799
00:33:56,822 --> 00:33:58,030
this direction to the target.

800
00:33:58,030 --> 00:34:02,570
So let's do it on Paint
to give an example.

801
00:34:02,570 --> 00:34:06,600
Because saying words is not
always incredibly clear.

802
00:34:06,600 --> 00:34:08,739
So here's me.

803
00:34:08,739 --> 00:34:10,379
Here's my waypoint.

804
00:34:13,060 --> 00:34:16,290
And maybe here is
the closest ally.

805
00:34:16,290 --> 00:34:20,739
So what I'm going to do is I'm
going to repel from the ally,

806
00:34:20,739 --> 00:34:22,480
and I'm going to
attract to the waypoint.

807
00:34:22,480 --> 00:34:24,938
Which means I'm going to end
up having a bunch of guys that

808
00:34:24,938 --> 00:34:27,260
are having competing
requirements.

809
00:34:27,260 --> 00:34:30,210
Competing requirements
is like everything

810
00:34:30,210 --> 00:34:33,310
in these kinds of
dynamic systems.

811
00:34:33,310 --> 00:34:35,380
Because in the end,
you'll want to expect

812
00:34:35,380 --> 00:34:37,989
them to find a balance between
these competing things.

813
00:34:37,989 --> 00:34:40,962
Because frankly, they want to
do things that are inconsistent.

814
00:34:40,962 --> 00:34:42,670
They want to be at
the waypoint, but they

815
00:34:42,670 --> 00:34:44,295
don't want to be next
to their friends.

816
00:34:44,295 --> 00:34:47,594
So that automatically means
that they can't be satisfied,

817
00:34:47,594 --> 00:34:48,760
like a lot of people I know.

818
00:34:51,699 --> 00:34:55,380
So you could see here
the repulsion term.

819
00:34:55,380 --> 00:34:59,436
So if there are allies, then
it will locate the closest ally

820
00:34:59,436 --> 00:35:00,810
using this function
that I wrote.

821
00:35:00,810 --> 00:35:01,800
Very simple.

822
00:35:01,800 --> 00:35:05,110
And then it will add to the
goal location negative 3

823
00:35:05,110 --> 00:35:06,800
of the direction to
the closest ally.

824
00:35:06,800 --> 00:35:08,170
so what the heck does that mean?

825
00:35:08,170 --> 00:35:10,375
That's just saying repel.

826
00:35:10,375 --> 00:35:14,330
So my goal location
started here.

827
00:35:14,330 --> 00:35:18,240
And then I moved it toward
the waypoint some amount.

828
00:35:18,240 --> 00:35:20,420
And it would be more
or less, depending

829
00:35:20,420 --> 00:35:23,890
on where the waypoint
was with respect to me.

830
00:35:23,890 --> 00:35:26,310
And now I'm going to take
the ally direction, which

831
00:35:26,310 --> 00:35:29,910
is this direction, and I'm going
to subtract that from this.

832
00:35:29,910 --> 00:35:31,610
So now I'm going
away from the ally

833
00:35:31,610 --> 00:35:35,120
and toward the waypoint,
which gets me this x.

834
00:35:35,120 --> 00:35:39,510
So in the final analysis,
this is my new goal location,

835
00:35:39,510 --> 00:35:43,380
which is the sum of a bunch
of weighted added directions.

836
00:35:43,380 --> 00:35:45,040
And now I'm going to
go that way, which

837
00:35:45,040 --> 00:35:47,130
is going to get me both
toward the waypoint

838
00:35:47,130 --> 00:35:48,610
and away from my ally.

839
00:35:48,610 --> 00:35:53,040
I mean, on a grid,
this direction

840
00:35:53,040 --> 00:35:55,640
may have been quite
close to or the same as

841
00:35:55,640 --> 00:35:58,020
the direction directly
to the waypoint.

842
00:35:58,020 --> 00:36:00,970
But when you keep
doing it over and over,

843
00:36:00,970 --> 00:36:04,020
those little edge cases where
a small difference makes you

844
00:36:04,020 --> 00:36:08,530
go one way or the other
start to become significant.

845
00:36:08,530 --> 00:36:11,810
And so you can do it just
simply like this, where you only

846
00:36:11,810 --> 00:36:15,080
care about where the
ally that's closest is.

847
00:36:15,080 --> 00:36:17,710
And you only care
about the weighting

848
00:36:17,710 --> 00:36:20,970
of where you're going.

849
00:36:20,970 --> 00:36:22,549
But wait just a second.

850
00:36:22,549 --> 00:36:24,340
There are so many things
you can add to it,

851
00:36:24,340 --> 00:36:28,110
and that I did add when I
wrote my guided swarm two,

852
00:36:28,110 --> 00:36:29,540
which is going to
end the lecture.

853
00:36:29,540 --> 00:36:36,030
So let's do all the things
that you can repel or attract.

854
00:36:36,030 --> 00:36:36,940
There's enemies.

855
00:36:36,940 --> 00:36:39,630
And you repel or attract to them
based on how many there are.

856
00:36:39,630 --> 00:36:41,060
There's mines .

857
00:36:41,060 --> 00:36:42,649
And if they're
yours, do you want

858
00:36:42,649 --> 00:36:43,940
to be right next to your mines?

859
00:36:43,940 --> 00:36:45,210
Do you want to be behind them?

860
00:36:45,210 --> 00:36:47,202
Do you want to be sort
of in a clump of them?

861
00:36:47,202 --> 00:36:48,660
Do want to keep
laying mi-- I mean,

862
00:36:48,660 --> 00:36:50,580
if you're in a
laying mines mode,

863
00:36:50,580 --> 00:36:52,660
then you don't really
want to stand on a mine,

864
00:36:52,660 --> 00:36:54,870
because then you can't lay one.

865
00:36:54,870 --> 00:36:56,390
You're on one already.

866
00:36:56,390 --> 00:36:59,999
So [INAUDIBLE] on
your current goals.

867
00:36:59,999 --> 00:37:01,790
So you could imagine
making the following--

868
00:37:01,790 --> 00:37:02,990
this would be so cool.

869
00:37:02,990 --> 00:37:04,750
Somebody in this room
needs to do this.

870
00:37:04,750 --> 00:37:07,835
I was going to do it for
today but ran out of time.

871
00:37:07,835 --> 00:37:08,710
You do the following.

872
00:37:08,710 --> 00:37:11,290
So here's what always
happens when I write players,

873
00:37:11,290 --> 00:37:13,710
is I've got a center,
and people start

874
00:37:13,710 --> 00:37:15,720
laying mines around the center.

875
00:37:15,720 --> 00:37:17,120
I think you saw
this today, where

876
00:37:17,120 --> 00:37:18,860
they were laying these mines.

877
00:37:18,860 --> 00:37:21,750
And the mines reached
a certain amount,

878
00:37:21,750 --> 00:37:25,110
and the robots were
only out to here.

879
00:37:25,110 --> 00:37:26,940
And so that was it.

880
00:37:26,940 --> 00:37:28,380
I didn't get any more mines.

881
00:37:28,380 --> 00:37:29,910
But here's what you could do.

882
00:37:29,910 --> 00:37:32,500
You could say, have
the headquarters,

883
00:37:32,500 --> 00:37:35,150
which might be, say, down here.

884
00:37:35,150 --> 00:37:38,130
Have it say, if-- let's do this.

885
00:37:38,130 --> 00:37:38,880
Oh, my goodness.

886
00:37:38,880 --> 00:37:40,010
Watch this.

887
00:37:40,010 --> 00:37:40,879
Oh, uh, uh.

888
00:37:40,879 --> 00:37:41,670
It's going to work.

889
00:37:41,670 --> 00:37:43,290
It's going to work.

890
00:37:43,290 --> 00:37:45,770
Headquarters.

891
00:37:45,770 --> 00:37:52,000
If there are a lot of mines
near the rally point-- and it

892
00:37:52,000 --> 00:37:53,600
can do this very easily.

893
00:37:53,600 --> 00:37:55,000
You know how to do it?

894
00:37:55,000 --> 00:37:57,220
There's a really,
really handy function.

895
00:37:57,220 --> 00:37:58,730
Don't do the following.

896
00:37:58,730 --> 00:38:01,140
Here's years what you
were probably thinking.

897
00:38:01,140 --> 00:38:05,510
What you might have done is
sense all the allied mines,

898
00:38:05,510 --> 00:38:07,464
or whatever the function
name is, and then

899
00:38:07,464 --> 00:38:09,380
parse through that list
and find the ones that

900
00:38:09,380 --> 00:38:10,260
are close the waypoint.

901
00:38:10,260 --> 00:38:11,380
You don't have to do that.

902
00:38:11,380 --> 00:38:12,950
You don't have to
do that at all.

903
00:38:12,950 --> 00:38:16,930
You can see here in the
documentation under Robot

904
00:38:16,930 --> 00:38:19,830
Controller, there's a
method that really helps.

905
00:38:19,830 --> 00:38:23,410
There's
senseMineLocations and it

906
00:38:23,410 --> 00:38:28,050
will let you specify a
center, a radius, and a team.

907
00:38:28,050 --> 00:38:29,860
So let's put that on here.

908
00:38:29,860 --> 00:38:31,940
So you definitely
want to be using this

909
00:38:31,940 --> 00:38:35,050
so that all you've got to
do is have the headquarters.

910
00:38:35,050 --> 00:38:36,940
And why would you
use the headquarters?

911
00:38:36,940 --> 00:38:39,273
Just so that it doesn't have
to be done again and again.

912
00:38:39,273 --> 00:38:42,930
So the headquarters is going
say, how many mines are there

913
00:38:42,930 --> 00:38:49,170
when this x is the center, and
I'm looking within some radius?

914
00:38:49,170 --> 00:38:51,650
And you could just expand
the radius turn by turn

915
00:38:51,650 --> 00:38:55,210
to see how many mines there
are until it starts to drop Off

916
00:38:55,210 --> 00:38:57,760
or you could just like
set it at a given radius,

917
00:38:57,760 --> 00:38:59,640
or the previous radius.

918
00:38:59,640 --> 00:39:00,226
So yeah.

919
00:39:00,226 --> 00:39:01,100
Let's do it that way.

920
00:39:01,100 --> 00:39:04,040
And you would say, how many of
my mines are in this circle?

921
00:39:04,040 --> 00:39:08,649
And so you'd say, if there
are full mines in the circle--

922
00:39:08,649 --> 00:39:10,940
and you can do that pretty
easily because the circle is

923
00:39:10,940 --> 00:39:13,290
defined by a radius
squared, right?

924
00:39:13,290 --> 00:39:18,300
So expected mines is
like pi r squared, right?

925
00:39:18,300 --> 00:39:19,420
That's a number of tiles.

926
00:39:19,420 --> 00:39:22,770
But r squared we already
have, and pi as an integer

927
00:39:22,770 --> 00:39:24,580
is equal to 3.

928
00:39:24,580 --> 00:39:27,800
And so we can just say the
number of expected mines

929
00:39:27,800 --> 00:39:31,040
equals 3 times the
radius squared.

930
00:39:31,040 --> 00:39:33,990
And so that's also nice, because
it's a little bit less than pi.

931
00:39:33,990 --> 00:39:37,360
And so if you're like mostly
filled of mines in that circle,

932
00:39:37,360 --> 00:39:43,110
then-- so if the
detected mines is

933
00:39:43,110 --> 00:39:47,830
greater than or equal to
expected mines in a full area,

934
00:39:47,830 --> 00:39:50,665
then expand the mine radius.

935
00:39:50,665 --> 00:39:51,920
And you might say, well, wait.

936
00:39:51,920 --> 00:39:53,730
How am I expanding
the mine radius?

937
00:39:53,730 --> 00:39:54,480
Well, here you go.

938
00:39:54,480 --> 00:39:55,410
Here you go.

939
00:39:55,410 --> 00:39:58,770
Right now, the function
looks like this

940
00:39:58,770 --> 00:40:00,710
for the repulsion
and attraction.

941
00:40:00,710 --> 00:40:03,330
It looks like the following.

942
00:40:03,330 --> 00:40:04,660
Oh, I'm going to use this.

943
00:40:04,660 --> 00:40:05,390
Yeah.

944
00:40:05,390 --> 00:40:06,220
Like that.

945
00:40:06,220 --> 00:40:09,340
So if this is the
distance x from center,

946
00:40:09,340 --> 00:40:11,600
and this is the
repulsive force--

947
00:40:11,600 --> 00:40:12,890
or I think we were attracting.

948
00:40:12,890 --> 00:40:14,620
Yeah, we were attracting
to the center.

949
00:40:14,620 --> 00:40:17,560
The attractive force
was greater the farther

950
00:40:17,560 --> 00:40:21,860
you got from the center.

951
00:40:21,860 --> 00:40:23,720
But you could imagine--
So that's going

952
00:40:23,720 --> 00:40:25,500
to push everybody
toward the middle.

953
00:40:25,500 --> 00:40:27,160
But you don't have to do that.

954
00:40:27,160 --> 00:40:30,210
You could just as easily
have the headquarters--

955
00:40:30,210 --> 00:40:33,450
the headquarters could
be telling everybody

956
00:40:33,450 --> 00:40:36,140
by message systems
what this looks like.

957
00:40:36,140 --> 00:40:38,390
And they could
very well tell them

958
00:40:38,390 --> 00:40:40,240
to make it a little
bit more like this.

959
00:40:40,240 --> 00:40:44,670
So let's say that you already
put mines out to this range.

960
00:40:44,670 --> 00:40:47,510
So you've already put
mines here and you've

961
00:40:47,510 --> 00:40:49,060
detected them using
the headquarters.

962
00:40:49,060 --> 00:40:51,420
So now you want everybody
to show up here.

963
00:40:51,420 --> 00:40:53,020
Well, you could
very simply do that.

964
00:40:53,020 --> 00:40:56,770
All you've got to do is make
the forces look like this.

965
00:40:56,770 --> 00:40:57,709
Oh, is that right?

966
00:40:57,709 --> 00:40:59,000
No, we would want-- Yeah, yeah.

967
00:40:59,000 --> 00:41:00,040
Well, that's the energy.

968
00:41:00,040 --> 00:41:01,520
And the force is this.

969
00:41:01,520 --> 00:41:03,110
The attractive force is this.

970
00:41:03,110 --> 00:41:05,130
So this is exactly a spring.

971
00:41:05,130 --> 00:41:07,070
That's a spring from physics.

972
00:41:07,070 --> 00:41:10,000
And that will make them go
toward it if they are inside,

973
00:41:10,000 --> 00:41:12,480
because they will go opposite
the direction of this vector

974
00:41:12,480 --> 00:41:14,460
from the origin.

975
00:41:14,460 --> 00:41:16,940
And they'll make
it go away if they

976
00:41:16,940 --> 00:41:20,082
are in the other direction,
and you get the idea.

977
00:41:20,082 --> 00:41:21,790
So that's a pretty
neat idea, because you

978
00:41:21,790 --> 00:41:24,000
could have them
spread out in a circle

979
00:41:24,000 --> 00:41:27,450
and maintain that sort
of positional authority,

980
00:41:27,450 --> 00:41:28,890
as it were.

981
00:41:28,890 --> 00:41:30,650
Very, very cool.

982
00:41:30,650 --> 00:41:34,220
We are also going to talk
about some other awesome things

983
00:41:34,220 --> 00:41:37,070
the headquarter
can tell them, even

984
00:41:37,070 --> 00:41:39,120
though our mouths are watering.

985
00:41:39,120 --> 00:41:41,100
Check the neighbors, OK.

986
00:41:41,100 --> 00:41:42,220
We can avoid this.

987
00:41:42,220 --> 00:41:44,180
We can avoid mines.

988
00:41:44,180 --> 00:41:45,400
We can avoid the enemy.

989
00:41:45,400 --> 00:41:48,200
We can do those things.

990
00:41:48,200 --> 00:41:49,350
There's no more direction.

991
00:41:49,350 --> 00:41:50,590
We've talked about that.

992
00:41:50,590 --> 00:41:52,740
So now let's go
back a little bit

993
00:41:52,740 --> 00:41:54,370
and talk about how
the headquarters can

994
00:41:54,370 --> 00:41:57,670
help you achieve a given shape.

995
00:41:57,670 --> 00:42:03,660
The headquarters can signal
when the enemy has artillery.

996
00:42:03,660 --> 00:42:06,040
Now, how's the
headquarters going to know?

997
00:42:06,040 --> 00:42:08,730
Well, it could
receive the message

998
00:42:08,730 --> 00:42:10,540
from somebody else
that notices his hit

999
00:42:10,540 --> 00:42:12,550
points have dropped by 20 or 40.

1000
00:42:12,550 --> 00:42:15,240
Or it could just
locate an artillery.

1001
00:42:15,240 --> 00:42:17,472
Like if you actually
physically walk past one,

1002
00:42:17,472 --> 00:42:19,430
then you could say there
is an artillery there.

1003
00:42:19,430 --> 00:42:21,737
I should start repelling
my other units.

1004
00:42:21,737 --> 00:42:24,070
They should start spreading
out, because if they are all

1005
00:42:24,070 --> 00:42:26,600
clumped together, they're going
to take a lot more damage.

1006
00:42:26,600 --> 00:42:29,220
They're going to take, instead
of 40 damage for just one hit,

1007
00:42:29,220 --> 00:42:31,532
they're going to take eight
splash damage as well.

1008
00:42:31,532 --> 00:42:32,990
And so that's going
to end up being

1009
00:42:32,990 --> 00:42:35,270
an additional five times worse.

1010
00:42:35,270 --> 00:42:36,420
Yeah.

1011
00:42:36,420 --> 00:42:38,950
Well, it's going to
be four times worse.

1012
00:42:38,950 --> 00:42:42,897
Five times as bad, as it were.

1013
00:42:42,897 --> 00:42:43,980
That's definitely no good.

1014
00:42:43,980 --> 00:42:46,330
So yeah, I mean,
that's how you're

1015
00:42:46,330 --> 00:42:51,384
changing the shape of your
cloud using the communication

1016
00:42:51,384 --> 00:42:53,050
that you're giving
from the headquarters

1017
00:42:53,050 --> 00:42:55,100
or from somewhere else.

1018
00:42:55,100 --> 00:42:56,340
So you could do that.

1019
00:42:56,340 --> 00:43:00,370
You can say, when
the group-- and you

1020
00:43:00,370 --> 00:43:04,070
could say like average
position-- has arrived

1021
00:43:04,070 --> 00:43:07,540
at a waypoint, like at a
waypoint that the headquarters

1022
00:43:07,540 --> 00:43:11,580
has specified, then
go to the next one.

1023
00:43:11,580 --> 00:43:15,010
And right there you can start
to fill the whole map that way.

1024
00:43:15,010 --> 00:43:16,690
Now, this is a
really crazy trick

1025
00:43:16,690 --> 00:43:18,730
that I think is the bee's knees.

1026
00:43:18,730 --> 00:43:21,750
And it's just going to make the
difference between amazing code

1027
00:43:21,750 --> 00:43:22,740
and non-amazing code.

1028
00:43:22,740 --> 00:43:25,760
So everybody is going to
have had this situation.

1029
00:43:25,760 --> 00:43:27,930
And I've had this
situation for a long time,

1030
00:43:27,930 --> 00:43:29,910
but I very rarely--
in fact, I've

1031
00:43:29,910 --> 00:43:32,369
never come up with a
solution until just recently.

1032
00:43:32,369 --> 00:43:33,410
And it was the following.

1033
00:43:33,410 --> 00:43:34,910
OK, so here's me.

1034
00:43:34,910 --> 00:43:38,350
Here's me and here's the goal.

1035
00:43:38,350 --> 00:43:40,200
OK I, want to go from
one to the other.

1036
00:43:40,200 --> 00:43:41,720
And a lot of times
what happens is

1037
00:43:41,720 --> 00:43:43,450
I've got a large
group of people,

1038
00:43:43,450 --> 00:43:44,810
and they end up splatting.

1039
00:43:44,810 --> 00:43:46,880
They just splat, and
then they trickle.

1040
00:43:46,880 --> 00:43:50,380
Like this is a little thin
line of trickled units.

1041
00:43:50,380 --> 00:43:52,400
And everybody else is
just splatting there,

1042
00:43:52,400 --> 00:43:53,525
and they're really unhappy.

1043
00:43:56,320 --> 00:43:58,870
Yeah, it's no good.

1044
00:43:58,870 --> 00:44:01,325
I think this guy here
has gone to sleep.

1045
00:44:04,200 --> 00:44:05,940
I mean, this guy
started to tunnel.

1046
00:44:05,940 --> 00:44:07,800
And so he's like
mining in there.

1047
00:44:07,800 --> 00:44:11,480
And then when these guys
arrived, he's like, I'm tired,

1048
00:44:11,480 --> 00:44:13,340
I'm going on a workers strike.

1049
00:44:13,340 --> 00:44:14,710
So I mean, it's just terrible.

1050
00:44:14,710 --> 00:44:17,410
So here's an
alternative to that that

1051
00:44:17,410 --> 00:44:19,869
may make the difference
between winning and losing.

1052
00:44:19,869 --> 00:44:21,660
This may just become
the dominant strategy.

1053
00:44:21,660 --> 00:44:24,200
I couldn't be more
infatuated with my idea.

1054
00:44:24,200 --> 00:44:25,886
And it works like this.

1055
00:44:25,886 --> 00:44:28,660
It works like this,
where I'd say, OK.

1056
00:44:28,660 --> 00:44:29,930
I'm headquarters.

1057
00:44:29,930 --> 00:44:31,720
I'm headquarters back here.

1058
00:44:31,720 --> 00:44:33,780
And I'm telling them what to do.

1059
00:44:33,780 --> 00:44:36,694
All I've got it do
is look here, here,

1060
00:44:36,694 --> 00:44:38,610
and here, and I can use
the senseNeutralMines.

1061
00:44:41,204 --> 00:44:42,870
I'll use the
senseNeutralMines, and I'll

1062
00:44:42,870 --> 00:44:44,430
put these as the centers.

1063
00:44:44,430 --> 00:44:47,710
One, two, three, four.

1064
00:44:47,710 --> 00:44:48,950
OK, five.

1065
00:44:48,950 --> 00:44:51,590
And these are the directions
that the whole swarm

1066
00:44:51,590 --> 00:44:52,250
could move.

1067
00:44:52,250 --> 00:44:54,920
And I'll say, OK, this has
seven mines and this has two.

1068
00:44:54,920 --> 00:44:55,980
No, this isn't realistic.

1069
00:44:55,980 --> 00:44:56,870
They all have zero.

1070
00:44:56,870 --> 00:44:58,140
OK, they all have zero.

1071
00:44:58,140 --> 00:44:58,640
Right?

1072
00:44:58,640 --> 00:45:02,857
But then once I move down a
little bit-- come on Paint,

1073
00:45:02,857 --> 00:45:03,440
come on Paint.

1074
00:45:03,440 --> 00:45:04,240
Help me out here.

1075
00:45:04,240 --> 00:45:04,420
OK.

1076
00:45:04,420 --> 00:45:06,210
Now I moved down a
little bit, and now

1077
00:45:06,210 --> 00:45:08,020
some directions really look bad.

1078
00:45:08,020 --> 00:45:10,020
So what I can do is
tell the whole swarm

1079
00:45:10,020 --> 00:45:13,500
to follow a waypoint set that
looks like a giant robot.

1080
00:45:13,500 --> 00:45:15,670
Not like a gundam, no.

1081
00:45:15,670 --> 00:45:17,600
Like one unit of robot.

1082
00:45:17,600 --> 00:45:20,040
So the whole swarm
becomes one big thing

1083
00:45:20,040 --> 00:45:22,040
that can all path
around the object,

1084
00:45:22,040 --> 00:45:24,390
because I've down
sampled the terrain.

1085
00:45:24,390 --> 00:45:26,610
This down sampling
idea can extend

1086
00:45:26,610 --> 00:45:30,330
to full search and
pathing algorithms.

1087
00:45:30,330 --> 00:45:32,920
So you'd split it up like this
or so, something like this.

1088
00:45:32,920 --> 00:45:34,670
They can be overlapping
or non-overlapping

1089
00:45:34,670 --> 00:45:36,000
down sample boxes.

1090
00:45:36,000 --> 00:45:38,170
You can use blur
or other features

1091
00:45:38,170 --> 00:45:41,180
as standard concepts
in computer science.

1092
00:45:41,180 --> 00:45:42,780
So that you can
start to do pathing

1093
00:45:42,780 --> 00:45:44,440
on a much simpler matrix.

1094
00:45:44,440 --> 00:45:47,211
And this can get you to
your enemy like leagues

1095
00:45:47,211 --> 00:45:48,710
faster than you
would otherwise get.

1096
00:45:48,710 --> 00:45:51,544
And if you use a heuristic,
then it doesn't even matter.

1097
00:45:51,544 --> 00:45:53,210
Let's say there's
like a few mines here.

1098
00:45:53,210 --> 00:45:55,770
Well, maybe you'll just
barrel right on through.

1099
00:45:55,770 --> 00:45:58,630
If you can just compare
to the amount of mines

1100
00:45:58,630 --> 00:46:01,350
that you're going to
find here versus here,

1101
00:46:01,350 --> 00:46:03,330
what better way to
make that comparison

1102
00:46:03,330 --> 00:46:06,130
than to count mines in
a giant area like this?

1103
00:46:06,130 --> 00:46:09,090
And these giant areas are going
to be so computationally simple

1104
00:46:09,090 --> 00:46:11,700
for you to compute, because
it's just one function call.

1105
00:46:11,700 --> 00:46:14,610
And I believe that function
call only costs you like $100.

1106
00:46:14,610 --> 00:46:17,140
And by dollars, I
mean byte codes.

1107
00:46:17,140 --> 00:46:21,490
So this has been a summary
of a bunch of types of swarm,

1108
00:46:21,490 --> 00:46:23,630
but it hasn't been a
complete summary yet, has it?

1109
00:46:23,630 --> 00:46:25,970
Because I told you I was going
to show you that amazing code

1110
00:46:25,970 --> 00:46:26,580
that I wrote.

1111
00:46:26,580 --> 00:46:30,180
And this time I may or may
not be being facetious.

1112
00:46:30,180 --> 00:46:34,540
And that amazing code where
I've made this guided swarm two,

1113
00:46:34,540 --> 00:46:37,390
and I'm going to play it
against a basic player on maybe

1114
00:46:37,390 --> 00:46:38,720
a couple of maps.

1115
00:46:38,720 --> 00:46:42,790
Basic player is not
horrible, I guess.

1116
00:46:42,790 --> 00:46:44,490
I don't really know
how good or bad

1117
00:46:44,490 --> 00:46:46,510
we are expecting to
claim that it is.

1118
00:46:46,510 --> 00:46:48,890
But let's go ahead
and show this match.

1119
00:46:48,890 --> 00:46:52,860
So here I've got
guided swarm in red.

1120
00:46:52,860 --> 00:46:54,270
And he's starting
out by building

1121
00:46:54,270 --> 00:46:57,234
a checkerboard pattern of mines.

1122
00:46:57,234 --> 00:46:59,650
And what he's going to do is
when he encounters the enemy,

1123
00:46:59,650 --> 00:47:02,219
he's sometimes going to just
rush straight at the enemy.

1124
00:47:02,219 --> 00:47:04,510
Other times he's going to
sort of back up and then only

1125
00:47:04,510 --> 00:47:06,560
fight at the right time.

1126
00:47:06,560 --> 00:47:10,311
In this example, it looks
like he was destroy-- oh.

1127
00:47:10,311 --> 00:47:12,560
Well, the enemy built artillery
right next to my base.

1128
00:47:12,560 --> 00:47:13,268
It doesn't count.

1129
00:47:13,268 --> 00:47:15,750
[LAUGHTER]

1130
00:47:15,750 --> 00:47:16,500
PROFESSOR: OK, OK.

1131
00:47:16,500 --> 00:47:18,920
This one, here I am
down in the bottom left.

1132
00:47:18,920 --> 00:47:20,850
I'm going to get ready
to research pickaxe.

1133
00:47:20,850 --> 00:47:22,456
I'm going to do it.

1134
00:47:22,456 --> 00:47:23,830
And then when I
research pickaxe,

1135
00:47:23,830 --> 00:47:25,497
their mining pattern
is going to change.

1136
00:47:25,497 --> 00:47:27,038
They're going to
spread out, I think.

1137
00:47:27,038 --> 00:47:28,870
Or maybe-- I might not
have implemented it.

1138
00:47:28,870 --> 00:47:30,400
Well in any case, it's there.

1139
00:47:30,400 --> 00:47:34,150
It sort of exists in my mind
that I'll do that sometime.

1140
00:47:34,150 --> 00:47:37,190
So what I'll do is
when they arrive here,

1141
00:47:37,190 --> 00:47:38,560
I'll totally mess them up.

1142
00:47:38,560 --> 00:47:39,620
They're going to show up.

1143
00:47:39,620 --> 00:47:41,300
I'll back them up
just a little bit.

1144
00:47:41,300 --> 00:47:43,040
And then they're going
to be on my mines

1145
00:47:43,040 --> 00:47:45,123
and they're going to have
nothing to say about it.

1146
00:47:45,123 --> 00:47:48,960
They're going to just-- they're
going to-- oh, so there I go.

1147
00:47:48,960 --> 00:47:49,460
There I go.

1148
00:47:49,460 --> 00:47:51,080
I'm starting up to move out.

1149
00:47:51,080 --> 00:47:53,210
I'm starting to
continue to build mines.

1150
00:47:53,210 --> 00:47:55,872
And I'm going forward.

1151
00:47:55,872 --> 00:47:58,455
I don't know that-- I think he
doesn't have really any robots.

1152
00:48:01,510 --> 00:48:02,200
Yeah.

1153
00:48:02,200 --> 00:48:03,690
Yeah, super effective.

1154
00:48:03,690 --> 00:48:04,850
Oh my goodness.

1155
00:48:04,850 --> 00:48:06,120
That was fantastic.

1156
00:48:06,120 --> 00:48:08,260
I'm actually going
to do-- I wonder

1157
00:48:08,260 --> 00:48:10,460
if it is working
as I was expecting.

1158
00:48:10,460 --> 00:48:12,115
So in this one,
I'm going across.

1159
00:48:12,115 --> 00:48:15,960
And when the enemy shows,
what I'm hoping is-- oop.

1160
00:48:15,960 --> 00:48:16,830
It already ended.

1161
00:48:16,830 --> 00:48:17,360
Yes.

1162
00:48:17,360 --> 00:48:19,840
I'm hoping to sort of
back up just a little bit,

1163
00:48:19,840 --> 00:48:22,420
and then take them on.

1164
00:48:22,420 --> 00:48:24,282
No, it looks like I
just run straight in.

1165
00:48:24,282 --> 00:48:25,637
[LAUGHTER]

1166
00:48:25,637 --> 00:48:27,720
PROFESSOR: Well in any
case, I sort of spread out.

1167
00:48:27,720 --> 00:48:29,240
I made this neat little pattern.

1168
00:48:29,240 --> 00:48:31,210
I'm starting to take
these encampments.

1169
00:48:31,210 --> 00:48:32,250
It's getting done.

1170
00:48:32,250 --> 00:48:33,100
I'm doing it up.

1171
00:48:33,100 --> 00:48:34,520
Making it happen.

1172
00:48:34,520 --> 00:48:35,620
I take the other side.

1173
00:48:35,620 --> 00:48:37,289
I take more encampments.

1174
00:48:37,289 --> 00:48:39,330
And then I just sort of--
because these are like,

1175
00:48:39,330 --> 00:48:42,250
now I've got better
production than he has.

1176
00:48:42,250 --> 00:48:44,511
I'm going to cast
these games this well

1177
00:48:44,511 --> 00:48:45,510
when we have the finals.

1178
00:48:45,510 --> 00:48:46,562
It's going to be great.

1179
00:48:49,264 --> 00:48:50,180
I'm making more units.

1180
00:48:50,180 --> 00:48:51,030
Oh, look at this.

1181
00:48:51,030 --> 00:48:52,490
Do you see this like micro?

1182
00:48:52,490 --> 00:48:55,802
I'm backing up a little
bit, letting them do it.

1183
00:48:55,802 --> 00:48:57,260
I'm do some micro
here, making sure

1184
00:48:57,260 --> 00:48:59,599
that I outnumber the
enemy before I go in.

1185
00:48:59,599 --> 00:49:01,640
I'm not keeping it so
spread out as I was before.

1186
00:49:01,640 --> 00:49:03,556
Some guys are doing some
important philosophy.

1187
00:49:03,556 --> 00:49:05,180
[LAUGHTER]

1188
00:49:05,180 --> 00:49:07,100
PROFESSOR: And those
philosophical results

1189
00:49:07,100 --> 00:49:08,930
just very well may
make the difference

1190
00:49:08,930 --> 00:49:12,350
between our salvation
as a species and not.

1191
00:49:12,350 --> 00:49:15,540
Thank you very much for
attending this swarm lecture.

1192
00:49:15,540 --> 00:49:17,040
Next lecture will
be about strategy.

1193
00:49:17,040 --> 00:49:19,890
And I hope you enjoy
the Indian food.