1
00:00:00,210 --> 00:00:02,550
The following content is
provided under a Creative

2
00:00:02,550 --> 00:00:03,970
Commons license.

3
00:00:03,970 --> 00:00:06,180
Your support will help
MIT OpenCourseWare

4
00:00:06,180 --> 00:00:10,270
continue to offer high quality
educational resources for free.

5
00:00:10,270 --> 00:00:12,810
To make a donation or to
view additional materials

6
00:00:12,810 --> 00:00:16,770
from hundreds of MIT courses,
visit MIT OpenCourseWare

7
00:00:16,770 --> 00:00:17,970
at ocw.mit.edu.

8
00:00:25,700 --> 00:00:28,670
PROFESSOR: Let's talk
about the V-Model, here.

9
00:00:28,670 --> 00:00:32,450
We are-- interesting, we have
a little bit of a lag here.

10
00:00:32,450 --> 00:00:38,240
So the lifecycle management
is our last topic for today

11
00:00:38,240 --> 00:00:41,210
and that sort of
completes the V. And then

12
00:00:41,210 --> 00:00:45,840
next [AUDIO OUT] manufacturing.

13
00:00:45,840 --> 00:00:48,800
So what I'd like to cover
today is essentially

14
00:00:48,800 --> 00:00:53,900
a definition of
[INAUDIBLE] mean by that,

15
00:00:53,900 --> 00:00:56,945
and then focus on the lifecycle
[INAUDIBLE] also known

16
00:00:56,945 --> 00:01:00,020
as the [? ilities. ?]
Discuss a little bit

17
00:01:00,020 --> 00:01:03,050
where they come from, how
much we know about them.

18
00:01:03,050 --> 00:01:04,580
The bulk of the
discussion is going

19
00:01:04,580 --> 00:01:06,980
to be a case study
called Reconfiguration

20
00:01:06,980 --> 00:01:09,890
of Communication
Satellite Constellations.

21
00:01:09,890 --> 00:01:13,460
And this sort of explains
the idea that system,

22
00:01:13,460 --> 00:01:16,440
once you brought a system
into operation, that's

23
00:01:16,440 --> 00:01:18,010
not the end of the story.

24
00:01:18,010 --> 00:01:21,530
The system is going to continue
to live and change and be

25
00:01:21,530 --> 00:01:23,754
modified over its life.

26
00:01:23,754 --> 00:01:25,670
I'll try to summarize
some of the key concepts

27
00:01:25,670 --> 00:01:28,140
over the whole semester.

28
00:01:28,140 --> 00:01:31,230
And then I have one
slide on career and study

29
00:01:31,230 --> 00:01:33,350
recommendations, if
you want to go further

30
00:01:33,350 --> 00:01:37,280
into systems engineering.

31
00:01:37,280 --> 00:01:40,720
So lifecycle management.

32
00:01:40,720 --> 00:01:44,260
So this is my definition of
what lifecycle management is.

33
00:01:44,260 --> 00:01:48,790
It's essentially the active
engagement of-- it's the active

34
00:01:48,790 --> 00:01:52,000
engagement of all the
stakeholders with the system,

35
00:01:52,000 --> 00:01:54,490
between the time when
it starts to operate--

36
00:01:54,490 --> 00:01:56,020
so you've done
the commissioning,

37
00:01:56,020 --> 00:01:57,910
you've done the operations--

38
00:01:57,910 --> 00:02:01,040
until you decide to decommission
and retire the system.

39
00:02:01,040 --> 00:02:03,700
So it's during the
whole operation life.

40
00:02:03,700 --> 00:02:05,200
And what you want
to do, is you want

41
00:02:05,200 --> 00:02:07,320
to maximize the
value that you gain

42
00:02:07,320 --> 00:02:09,139
from the system's existence.

43
00:02:09,139 --> 00:02:11,740
And so life cycle
management starts

44
00:02:11,740 --> 00:02:14,170
from the very, very, very start.

45
00:02:14,170 --> 00:02:16,140
And so I listed
here some activities

46
00:02:16,140 --> 00:02:18,980
that are included in
lifecycle management.

47
00:02:18,980 --> 00:02:21,770
So obviously daily operations,
you know monitoring,

48
00:02:21,770 --> 00:02:25,510
is the system
working as expected?

49
00:02:25,510 --> 00:02:28,750
Training and certification
of the operators.

50
00:02:28,750 --> 00:02:30,860
Sometimes, you know,
in long live systems,

51
00:02:30,860 --> 00:02:34,960
the people who were the early
operators, generation one,

52
00:02:34,960 --> 00:02:37,300
are not necessarily the
people who are going

53
00:02:37,300 --> 00:02:38,770
to operate for the whole life.

54
00:02:38,770 --> 00:02:42,130
So you have to have
new people operate.

55
00:02:42,130 --> 00:02:44,260
Maybe one of the most
extreme examples of that

56
00:02:44,260 --> 00:02:46,532
is an airplane
known as the b-52.

57
00:02:46,532 --> 00:02:47,490
Have you heard of this?

58
00:02:47,490 --> 00:02:50,530
This airplane might be the
first plane that actually has

59
00:02:50,530 --> 00:02:52,990
a 100 year operational life.

60
00:02:52,990 --> 00:02:55,240
And there are apparently Navy--

61
00:02:55,240 --> 00:02:58,330
Air Force guys at MIT,
you can confirm this,

62
00:02:58,330 --> 00:03:00,040
that there's actually like--

63
00:03:00,040 --> 00:03:03,580
the grandfather was
the first generation

64
00:03:03,580 --> 00:03:06,190
and there's like third
generation pilots now flying

65
00:03:06,190 --> 00:03:07,690
on the b-52.

66
00:03:07,690 --> 00:03:09,270
That's what I've been told.

67
00:03:09,270 --> 00:03:12,554
Did you hear this as well?

68
00:03:12,554 --> 00:03:15,670
AUDIENCE: Yeah, That's correct.

69
00:03:15,670 --> 00:03:18,150
PROFESSOR: Do some
of these people?

70
00:03:18,150 --> 00:03:20,080
AUDIENCE: No, I've
never actually worked

71
00:03:20,080 --> 00:03:23,440
on that airframe.

72
00:03:23,440 --> 00:03:24,250
PROFESSOR: OK.

73
00:03:24,250 --> 00:03:25,960
But anyway, so
that's what I mean

74
00:03:25,960 --> 00:03:29,590
by [INAUDIBLE] not just
the initial operators,

75
00:03:29,590 --> 00:03:32,200
but over multiple generations.

76
00:03:32,200 --> 00:03:36,040
Then the third thing
here is servicing.

77
00:03:36,040 --> 00:03:38,500
And servicing comes
in two flavors,

78
00:03:38,500 --> 00:03:40,210
of preventive
maintenance, you know

79
00:03:40,210 --> 00:03:43,390
regular maintenance, and
then corrective maintenance,

80
00:03:43,390 --> 00:03:45,820
which we also call repair.

81
00:03:45,820 --> 00:03:49,920
Dealing with small and large
failures, recalls, anomalies.

82
00:03:49,920 --> 00:03:52,030
This is a big deal in
the automotive industry,

83
00:03:52,030 --> 00:03:53,440
as we know.

84
00:03:53,440 --> 00:03:55,540
Increasingly
protecting the system

85
00:03:55,540 --> 00:03:58,480
from either random
or targeted attacks.

86
00:03:58,480 --> 00:04:03,380
Cyber, physical, right,
physical security but also cyber

87
00:04:03,380 --> 00:04:06,430
security is becoming
a huge topic.

88
00:04:06,430 --> 00:04:08,470
Sharing and archiving
the data that's

89
00:04:08,470 --> 00:04:09,820
being produced by the system.

90
00:04:09,820 --> 00:04:12,762
More and more systems produce
larger amounts of data.

91
00:04:12,762 --> 00:04:13,970
Well, what do you do with it?

92
00:04:13,970 --> 00:04:15,850
How do you store that data?

93
00:04:15,850 --> 00:04:18,670
Upgrading and retrofitting
the system as needed.

94
00:04:18,670 --> 00:04:21,670
So retrofitting means
we're physically

95
00:04:21,670 --> 00:04:23,710
changing the configuration
of the system--

96
00:04:23,710 --> 00:04:27,370
right?-- over its life
to add new capabilities.

97
00:04:27,370 --> 00:04:31,480
And upgrading; upgrading
can be done sometimes

98
00:04:31,480 --> 00:04:33,640
without having to
change the hardware.

99
00:04:33,640 --> 00:04:36,910
So you're doing a software
upgrade, or software update,

100
00:04:36,910 --> 00:04:38,380
software patching.

101
00:04:38,380 --> 00:04:40,255
All of this would
go under upgrades.

102
00:04:42,850 --> 00:04:45,770
Cross strapping the
system with other system

103
00:04:45,770 --> 00:04:47,680
in a federation of systems.

104
00:04:47,680 --> 00:04:49,960
So what I mean by
this is a system

105
00:04:49,960 --> 00:04:53,410
may have been designed
just to operate on its own,

106
00:04:53,410 --> 00:04:55,540
but now you're connecting
it with another system.

107
00:04:55,540 --> 00:04:58,351
This happened, for example,
the electrical grid.

108
00:04:58,351 --> 00:04:58,850
Right?

109
00:04:58,850 --> 00:05:03,160
The early electrical grids were
local, regional power grids.

110
00:05:03,160 --> 00:05:05,610
And now we've connected
them, in the US

111
00:05:05,610 --> 00:05:09,340
we have three major grids, the
eastern, the western, and then

112
00:05:09,340 --> 00:05:10,900
Texas has its own grid.

113
00:05:10,900 --> 00:05:13,030
Texas does a lot
of its own stuff.

114
00:05:13,030 --> 00:05:14,220
It's called ERCCOT.

115
00:05:14,220 --> 00:05:16,090
And then here in
Europe, you know

116
00:05:16,090 --> 00:05:19,960
the European electrical grid
is now continental-wide.

117
00:05:19,960 --> 00:05:21,010
It didn't used to be.

118
00:05:21,010 --> 00:05:25,720
And of course, that has big
implications for operations.

119
00:05:25,720 --> 00:05:27,355
Reducing the
resource consumption

120
00:05:27,355 --> 00:05:31,060
and environmental burden
of the system over time,

121
00:05:31,060 --> 00:05:32,800
and then finally,
decommissioning

122
00:05:32,800 --> 00:05:35,420
the system when
it's time to do so.

123
00:05:35,420 --> 00:05:38,670
And in many systems,
this is a big challenge

124
00:05:38,670 --> 00:05:41,020
as to finding when
is the right time

125
00:05:41,020 --> 00:05:43,060
to decommission the system.

126
00:05:43,060 --> 00:05:45,100
And that's often the
case when, you know,

127
00:05:45,100 --> 00:05:48,910
the operating costs exceed the
value that the system delivers.

128
00:05:48,910 --> 00:05:50,920
So this is a pretty
long list here.

129
00:05:50,920 --> 00:05:57,410
And a lot of tasks, a lot
of decisions to be made.

130
00:05:57,410 --> 00:06:00,010
Any questions about this list?

131
00:06:00,010 --> 00:06:04,180
I don't think it's
complete, but this

132
00:06:04,180 --> 00:06:07,510
is a lot of things you need to
take care of during operations.

133
00:06:07,510 --> 00:06:10,200
Here's a more
graphical view of this.

134
00:06:10,200 --> 00:06:13,930
So this is my sketch of the
systems engineering lifecycle.

135
00:06:13,930 --> 00:06:16,150
So this is part one, right?

136
00:06:16,150 --> 00:06:20,290
Conceptual design-- conception,
design, and implementation.

137
00:06:20,290 --> 00:06:23,440
You start the lifecycle, you
do something like a system

138
00:06:23,440 --> 00:06:26,740
requirements review, understand
the mission, the requirements,

139
00:06:26,740 --> 00:06:28,090
the constraints.

140
00:06:28,090 --> 00:06:31,510
You do conceptual design,
that's where you do creativity,

141
00:06:31,510 --> 00:06:33,430
architecting, trade studies.

142
00:06:33,430 --> 00:06:35,127
As a PDR, you choose--

143
00:06:35,127 --> 00:06:36,710
so in this case, you
know, we're going

144
00:06:36,710 --> 00:06:38,760
to go through this
triangle concept

145
00:06:38,760 --> 00:06:41,970
and then you design all
the details within it.

146
00:06:41,970 --> 00:06:48,570
That's where we do modeling,
simulation, experiments, MDO.

147
00:06:48,570 --> 00:06:51,000
And we iterate--
sometimes you have

148
00:06:51,000 --> 00:06:54,000
to iterate this
[? dashed ?] duration means

149
00:06:54,000 --> 00:06:57,210
abandoning the concept you chose
and go for a different concept.

150
00:06:57,210 --> 00:06:59,940
That's usually undesirable.

151
00:06:59,940 --> 00:07:02,820
After the CDR, we
implement the system,

152
00:07:02,820 --> 00:07:05,670
we turn information to
matter, and this all

153
00:07:05,670 --> 00:07:09,310
happens in a technological,
economic, social context.

154
00:07:09,310 --> 00:07:14,220
So then, the second part of
it, which in many systems--

155
00:07:14,220 --> 00:07:18,560
if this is 10 years, this
could be 30 years or more--

156
00:07:18,560 --> 00:07:20,940
is the actual operational space.

157
00:07:20,940 --> 00:07:24,300
So the system has now been
tested, validated, verified,

158
00:07:24,300 --> 00:07:25,530
and deployed.

159
00:07:25,530 --> 00:07:28,830
And now, you know, we
operate the system.

160
00:07:28,830 --> 00:07:33,340
Things break, we need to service
it, and then I have this--

161
00:07:33,340 --> 00:07:35,230
you see, there's two
versions of the system.

162
00:07:35,230 --> 00:07:38,220
One is solid, and the
other one is kind of faint.

163
00:07:38,220 --> 00:07:40,270
That's the virtual
version of the system.

164
00:07:40,270 --> 00:07:43,740
So one of the big concepts
that's being pushed now,

165
00:07:43,740 --> 00:07:47,700
certainly in the US by the
DOD, but also other places.

166
00:07:47,700 --> 00:07:50,280
Is the idea of a digital twin.

167
00:07:50,280 --> 00:07:52,500
The idea of a digital
twin is that there's

168
00:07:52,500 --> 00:07:53,850
a physical system--

169
00:07:53,850 --> 00:07:55,650
right?-- that exists,
and then there's

170
00:07:55,650 --> 00:07:58,530
a digital version, a
digital twin of that system,

171
00:07:58,530 --> 00:08:01,890
somewhere on a computer
that mirrors exactly

172
00:08:01,890 --> 00:08:03,820
what the real system is doing.

173
00:08:03,820 --> 00:08:06,300
So if something breaks
on the physical system,

174
00:08:06,300 --> 00:08:09,810
well that same component
fails in the digital twin.

175
00:08:09,810 --> 00:08:13,230
And then, before you actually
do a repair or any actions,

176
00:08:13,230 --> 00:08:15,970
you execute those actions
on the digital twin

177
00:08:15,970 --> 00:08:17,970
to see whether the system
will operate properly.

178
00:08:20,686 --> 00:08:24,840
And so this is kind of
the latest thinking,

179
00:08:24,840 --> 00:08:29,130
is that for any system we
should have a digital twin doing

180
00:08:29,130 --> 00:08:30,390
operations.

181
00:08:30,390 --> 00:08:31,630
Upgrading the system.

182
00:08:31,630 --> 00:08:34,440
So in this case, we're
adding things to it,

183
00:08:34,440 --> 00:08:35,820
we're connecting it.

184
00:08:35,820 --> 00:08:38,940
You know, and then at some
point, the system does degrade.

185
00:08:38,940 --> 00:08:41,520
Because, for example,
you know, materials, age,

186
00:08:41,520 --> 00:08:43,289
they get brittle.

187
00:08:43,289 --> 00:08:46,050
Technology becomes obsolete.

188
00:08:46,050 --> 00:08:49,710
You know, it gets harder and
harder to maintain spare parts

189
00:08:49,710 --> 00:08:52,740
or the suppliers of the
original spare parts

190
00:08:52,740 --> 00:08:54,060
went out of business.

191
00:08:54,060 --> 00:08:56,220
So these really
old legacy systems

192
00:08:56,220 --> 00:08:58,350
can be very
expensive to operate.

193
00:08:58,350 --> 00:09:00,490
And then at some
point you liquidate

194
00:09:00,490 --> 00:09:03,810
and it's the end
of the life cycle.

195
00:09:03,810 --> 00:09:06,790
So I've already told you
about one example of this,

196
00:09:06,790 --> 00:09:08,910
which is the space shuttle.

197
00:09:08,910 --> 00:09:11,920
And so I'm not going
to belabor this again,

198
00:09:11,920 --> 00:09:14,530
but I think it's a great
opportunity to learn.

199
00:09:14,530 --> 00:09:19,650
The space shuttle had a 10 year
design life, roughly from 1971

200
00:09:19,650 --> 00:09:21,750
until '81, first flight.

201
00:09:21,750 --> 00:09:25,020
And then we operated
the shuttle for 30 years

202
00:09:25,020 --> 00:09:28,710
and we had a total of 135
launches during that time.

203
00:09:28,710 --> 00:09:33,390
We spent $192 billion dollars
on the shuttle program.

204
00:09:33,390 --> 00:09:36,600
And if you average
that, it's $1.5 billion

205
00:09:36,600 --> 00:09:40,890
per flight, which is a
lot more than was planned.

206
00:09:40,890 --> 00:09:43,530
But just yesterday
here at the EPFL,

207
00:09:43,530 --> 00:09:47,910
we had the annual meeting
of the Swiss Space Center.

208
00:09:47,910 --> 00:09:52,380
And one of the astronauts
here, Claude Niccolier

209
00:09:52,380 --> 00:09:55,290
and his colleague,
Professor [? Mehron ?]

210
00:09:55,290 --> 00:09:57,630
gave a great talk
about the Hubble

211
00:09:57,630 --> 00:10:00,900
and the servicing of the Hubble
and the amazing things we've

212
00:10:00,900 --> 00:10:02,160
learned through it.

213
00:10:02,160 --> 00:10:05,490
And you guys at
MIT, Jeff Hoffman,

214
00:10:05,490 --> 00:10:09,040
was part of at least the first
servicing mission as well.

215
00:10:09,040 --> 00:10:14,290
So there's incredible things
that were done by the shuttle.

216
00:10:14,290 --> 00:10:16,860
And so I think we need to
acknowledge that, not just

217
00:10:16,860 --> 00:10:21,330
the fact that it
was very expensive.

218
00:10:21,330 --> 00:10:24,300
It accomplished great
things during its life.

219
00:10:24,300 --> 00:10:27,300
But you know, if we had a chance
to do it again, would we--

220
00:10:27,300 --> 00:10:29,490
would we come up with
the exact same system,

221
00:10:29,490 --> 00:10:31,890
would we make all the
same design decisions?

222
00:10:31,890 --> 00:10:33,240
You know, we lost two of them.

223
00:10:33,240 --> 00:10:35,820
Probably not, we could probably
do something-- something

224
00:10:35,820 --> 00:10:36,810
different.

225
00:10:36,810 --> 00:10:38,940
So here's what we wanted,
and then finally, we

226
00:10:38,940 --> 00:10:41,770
got a much more complex
system in the end.

227
00:10:41,770 --> 00:10:45,720
And particularly my sense is
that a lot of the things that

228
00:10:45,720 --> 00:10:49,410
made the shuttle expensive
to operate during its life

229
00:10:49,410 --> 00:10:51,731
were things related
to maintainability.

230
00:10:51,731 --> 00:10:52,230
Right?

231
00:10:52,230 --> 00:10:54,220
Reliability and so forth.

232
00:10:54,220 --> 00:10:56,160
So the lifecycle
properties, which I

233
00:10:56,160 --> 00:10:57,310
want to talk about next.

234
00:10:57,310 --> 00:10:59,470
So I'm just going to
repeat the questions,

235
00:10:59,470 --> 00:11:01,240
so you guys can hear as well.

236
00:11:01,240 --> 00:11:03,870
So the question was
about reusability.

237
00:11:03,870 --> 00:11:05,580
So the shuttle was--

238
00:11:05,580 --> 00:11:08,460
the orbiter was fully reusable.

239
00:11:08,460 --> 00:11:11,020
The external fuel
tank was not reusable.

240
00:11:11,020 --> 00:11:11,520
Right?

241
00:11:11,520 --> 00:11:13,200
And then the solid
rocket boosters

242
00:11:13,200 --> 00:11:15,180
were partially
reusable because they

243
00:11:15,180 --> 00:11:17,000
had to be stripped
down, you know,

244
00:11:17,000 --> 00:11:20,520
and rebuilt, essentially,
for every launch.

245
00:11:20,520 --> 00:11:24,750
The idea of Blue Origin
and then Space-X Stage One,

246
00:11:24,750 --> 00:11:27,390
flying it back to
the launch site,

247
00:11:27,390 --> 00:11:31,320
is absolutely the fact that
if you can make it reusable,

248
00:11:31,320 --> 00:11:35,490
then you can amortize
the capex in that element

249
00:11:35,490 --> 00:11:36,820
over multiple flights.

250
00:11:36,820 --> 00:11:38,500
And it should drop,
you know, the cost

251
00:11:38,500 --> 00:11:41,780
by a factor of
five or 10 or more.

252
00:11:41,780 --> 00:11:44,380
Now, the devil's in
the details, right?

253
00:11:44,380 --> 00:11:48,130
So if you start and restart
an engine multiple times,

254
00:11:48,130 --> 00:11:50,960
you can have a big
transient every time.

255
00:11:50,960 --> 00:11:53,350
So was this system,
in fact, designed

256
00:11:53,350 --> 00:11:55,360
to withstand these transients?

257
00:11:55,360 --> 00:11:58,420
Can it handle multiple starts?

258
00:11:58,420 --> 00:12:00,040
What does it do
to the materials?

259
00:12:00,040 --> 00:12:01,530
And so forth.

260
00:12:01,530 --> 00:12:05,200
And so to really model
this ahead of time,

261
00:12:05,200 --> 00:12:08,290
and then test it over
the cycles is the key.

262
00:12:08,290 --> 00:12:11,800
So in the shuttle, the
big problems-- the two

263
00:12:11,800 --> 00:12:14,260
major subsystems in the
shuttle, in the orbiter,

264
00:12:14,260 --> 00:12:18,760
in particular that
made it expensive

265
00:12:18,760 --> 00:12:20,740
and the difference
between the top picture

266
00:12:20,740 --> 00:12:24,130
and the bottom picture was a,
the TPS, the thermal protection

267
00:12:24,130 --> 00:12:25,120
system.

268
00:12:25,120 --> 00:12:29,290
You know, every tile has
hit a different curvature

269
00:12:29,290 --> 00:12:33,310
in geometry, so really
inspecting every tile,

270
00:12:33,310 --> 00:12:36,460
replacing tiles, making
sure that the TPS--

271
00:12:36,460 --> 00:12:39,100
because just one
weakness in the TPS

272
00:12:39,100 --> 00:12:41,710
could basically be
fatal on re-entry.

273
00:12:41,710 --> 00:12:45,422
And so, TPS was
very difficult and--

274
00:12:45,422 --> 00:12:47,380
you know, as opposed to
an ablative heat shield

275
00:12:47,380 --> 00:12:50,334
that you just sacrifice
the heat shield completely.

276
00:12:50,334 --> 00:12:52,000
And then the other
was the main engine--

277
00:12:52,000 --> 00:12:53,560
the shuttle main engine.

278
00:12:53,560 --> 00:12:56,890
Originally, the idea was to
only do very deep inspections

279
00:12:56,890 --> 00:12:59,440
and disassemble-- the
disassembly of the shuttle

280
00:12:59,440 --> 00:13:02,170
main engine after the
first few test flights,

281
00:13:02,170 --> 00:13:05,040
and then fly multiple times
without having to really

282
00:13:05,040 --> 00:13:07,990
re-inspect or
disassemble the engine.

283
00:13:07,990 --> 00:13:10,422
So they did the
disassembly of the engines

284
00:13:10,422 --> 00:13:12,130
after the test flights
and then they just

285
00:13:12,130 --> 00:13:16,210
kept doing it for every flight,
where the original intent was

286
00:13:16,210 --> 00:13:18,190
to only do it for
the test flights.

287
00:13:18,190 --> 00:13:22,210
And so the shuttle main engines
were the second big cost driver

288
00:13:22,210 --> 00:13:23,570
in operations.

289
00:13:23,570 --> 00:13:28,180
So re-usability is a great
idea, but to the degree

290
00:13:28,180 --> 00:13:30,220
to which re-usability
actually happens

291
00:13:30,220 --> 00:13:34,660
in reality, that's really
about detailed design decisions

292
00:13:34,660 --> 00:13:36,490
you make.

293
00:13:36,490 --> 00:13:37,430
OK.

294
00:13:37,430 --> 00:13:39,860
So let me move on then.

295
00:13:43,440 --> 00:13:48,470
When it comes to really
understanding lifecycle,

296
00:13:48,470 --> 00:13:54,900
I do want to point you to
the 15288 standard, which

297
00:13:54,900 --> 00:14:01,380
is the ISO 15288 standard,
which has a fairly complete list

298
00:14:01,380 --> 00:14:03,520
of system lifecycle processes.

299
00:14:03,520 --> 00:14:06,660
So this is, essentially
the system and software

300
00:14:06,660 --> 00:14:09,630
engineering, system
lifecycle process standard,

301
00:14:09,630 --> 00:14:13,770
that has a whole range of
processes that are described

302
00:14:13,770 --> 00:14:14,970
in quite some detail.

303
00:14:14,970 --> 00:14:17,670
And they're not just the
technical processes shown here

304
00:14:17,670 --> 00:14:21,270
on the right side, we focused
a lot on those technical

305
00:14:21,270 --> 00:14:24,717
processes in this class, the
stakeholders, the requirements,

306
00:14:24,717 --> 00:14:27,300
the architectural design-- which
is essentially the conceptual

307
00:14:27,300 --> 00:14:27,960
design--

308
00:14:27,960 --> 00:14:30,420
but there's also the
project processes, right?

309
00:14:30,420 --> 00:14:33,540
Executing the project
agreement processes.

310
00:14:33,540 --> 00:14:36,490
So this would be
negotiating contracts,

311
00:14:36,490 --> 00:14:39,780
you know supply contracts,
acquisition contracts, and then

312
00:14:39,780 --> 00:14:41,340
all the organizational
things you

313
00:14:41,340 --> 00:14:44,340
need to do to create the
right organization to execute

314
00:14:44,340 --> 00:14:45,730
these projects.

315
00:14:45,730 --> 00:14:49,350
So what's nice about
the 15288 standard

316
00:14:49,350 --> 00:14:52,691
is that it's a pretty
broad standard.

317
00:14:52,691 --> 00:14:53,190
OK.

318
00:14:53,190 --> 00:14:56,520
So let me talk about the ilities
of the lifecycle properties

319
00:14:56,520 --> 00:14:57,970
in particular.

320
00:14:57,970 --> 00:15:02,160
What they are and how we
can describe them and how,

321
00:15:02,160 --> 00:15:04,960
especially how they
relate to each other.

322
00:15:04,960 --> 00:15:08,250
And so the paper
that underlie this

323
00:15:08,250 --> 00:15:11,310
lecture that you need
to get some more detail.

324
00:15:11,310 --> 00:15:12,390
And this is one of them.

325
00:15:12,390 --> 00:15:15,600
So this paper is called
Investigating the Relationships

326
00:15:15,600 --> 00:15:18,840
and Semantic Sets Among
Lifecycle Properties.

327
00:15:18,840 --> 00:15:22,590
And we published this about
three years ago in Delft

328
00:15:22,590 --> 00:15:25,180
at the CESUN Conference.

329
00:15:25,180 --> 00:15:29,410
So the background on
here is the following,

330
00:15:29,410 --> 00:15:33,600
is that as you've seen,
complex engineering systems

331
00:15:33,600 --> 00:15:34,900
live for decades--

332
00:15:34,900 --> 00:15:36,660
some of them even
for centuries--

333
00:15:36,660 --> 00:15:39,330
and the ilities--
by ilities, we mean

334
00:15:39,330 --> 00:15:43,380
properties of systems that
are not the primary functional

335
00:15:43,380 --> 00:15:44,209
properties.

336
00:15:44,209 --> 00:15:45,750
In software engineering
they're often

337
00:15:45,750 --> 00:15:47,880
called nonfunctional properties.

338
00:15:47,880 --> 00:15:50,700
And the thing that's
tricky about the ilities

339
00:15:50,700 --> 00:15:54,660
is that you often only can
observe them during operations.

340
00:15:54,660 --> 00:15:57,360
So you know, you can test
systems in the short term,

341
00:15:57,360 --> 00:15:59,560
you can see does it work?

342
00:15:59,560 --> 00:16:00,930
Does it fulfill its function?

343
00:16:00,930 --> 00:16:03,660
But whether certain
ilities are present,

344
00:16:03,660 --> 00:16:07,800
it often only shows
itself over time.

345
00:16:07,800 --> 00:16:12,450
And so most of the
research that's been done

346
00:16:12,450 --> 00:16:14,850
and sort of quantifying
the ilities,

347
00:16:14,850 --> 00:16:17,610
has been looking at these
properties one at a time.

348
00:16:17,610 --> 00:16:20,340
So the questions we
wanted to go here,

349
00:16:20,340 --> 00:16:25,260
wanted to answer is, which
of these lifecycle properties

350
00:16:25,260 --> 00:16:27,210
are more prevalent than others?

351
00:16:27,210 --> 00:16:28,680
You know, the top 20.

352
00:16:28,680 --> 00:16:30,630
And then especially,
what's the relationship

353
00:16:30,630 --> 00:16:32,010
among lifecycle properties?

354
00:16:32,010 --> 00:16:34,650
Do they form what we
call semantic sets?

355
00:16:34,650 --> 00:16:37,680
And then how could you
use this information?

356
00:16:37,680 --> 00:16:39,840
And so here, we're
going to do this

357
00:16:39,840 --> 00:16:41,650
using two different methods.

358
00:16:41,650 --> 00:16:45,060
The first method is what I
call prevalence analysis.

359
00:16:45,060 --> 00:16:47,910
We're going to look in the
literature and on the internet,

360
00:16:47,910 --> 00:16:51,390
how frequently these
lifecycle properties show up,

361
00:16:51,390 --> 00:16:53,890
are mentioned, how much
we know about them.

362
00:16:53,890 --> 00:16:56,610
And then the second method
is a cognitive method.

363
00:16:56,610 --> 00:17:00,450
Where we ask people to
give their understanding

364
00:17:00,450 --> 00:17:04,589
of the lifecycle properties
and put them into a hierarchy,

365
00:17:04,589 --> 00:17:07,240
and then we'll
compare the results.

366
00:17:07,240 --> 00:17:10,500
So here are
essentially the results

367
00:17:10,500 --> 00:17:12,329
from the prevalence analysis.

368
00:17:12,329 --> 00:17:17,490
And this is ranked according to
the number of journal articles.

369
00:17:17,490 --> 00:17:20,520
This is scientific
papers written,

370
00:17:20,520 --> 00:17:24,480
where this particular key word,
by quality or reliability,

371
00:17:24,480 --> 00:17:28,470
shows up in the title or in
the abstract of the paper.

372
00:17:28,470 --> 00:17:30,800
And you can see this is
in units of thousands.

373
00:17:30,800 --> 00:17:31,530
OK?

374
00:17:31,530 --> 00:17:34,590
So quality is number one.

375
00:17:34,590 --> 00:17:35,310
Right?

376
00:17:35,310 --> 00:17:36,720
Quality is number one.

377
00:17:36,720 --> 00:17:40,110
The most scientific papers
are written about quality.

378
00:17:40,110 --> 00:17:43,380
And you can see, it's almost
a million journal articles.

379
00:17:43,380 --> 00:17:45,480
This is-- I'll show
this to you over time,

380
00:17:45,480 --> 00:17:49,170
but since 1884 is
the first year.

381
00:17:49,170 --> 00:17:52,710
And the databases that we
use for this are called it's

382
00:17:52,710 --> 00:17:54,961
Compendex and Inspect.

383
00:17:54,961 --> 00:17:56,460
These are actually
combined-- if you

384
00:17:56,460 --> 00:17:59,550
go to a website called
engineeringvillage.com,

385
00:17:59,550 --> 00:18:04,680
this is sort of a master
database for scientific papers

386
00:18:04,680 --> 00:18:06,320
and engineering.

387
00:18:06,320 --> 00:18:08,460
That was the basis for this.

388
00:18:08,460 --> 00:18:10,650
Number two, reliability.

389
00:18:10,650 --> 00:18:12,270
Number three, safety.

390
00:18:12,270 --> 00:18:15,730
Then flexibility,
robustness, and so forth.

391
00:18:15,730 --> 00:18:21,120
And the other bar, the gray bar,
is if you Google essentially

392
00:18:21,120 --> 00:18:25,530
for this particular keyword
and-- you know, keep in mind,

393
00:18:25,530 --> 00:18:27,750
this was done about
five years ago--

394
00:18:27,750 --> 00:18:30,150
this is the number--
the millions of hits

395
00:18:30,150 --> 00:18:31,380
that you will get.

396
00:18:31,380 --> 00:18:34,830
And, so you know,
there's a factor

397
00:18:34,830 --> 00:18:37,890
of 1,000 difference here
between journal articles

398
00:18:37,890 --> 00:18:39,420
and hits on internet.

399
00:18:39,420 --> 00:18:42,650
But still, you can sort
of compare the two.

400
00:18:42,650 --> 00:18:46,710
And so, in some
sense, the black bar

401
00:18:46,710 --> 00:18:49,890
is the totality of
scientific knowledge

402
00:18:49,890 --> 00:18:51,900
about this lifecycle property.

403
00:18:51,900 --> 00:18:56,390
And then the gray bar would
be the amount of information

404
00:18:56,390 --> 00:18:57,090
or uses--

405
00:18:57,090 --> 00:18:59,130
usage of that knowledge,
at least as far

406
00:18:59,130 --> 00:19:01,410
as it's represented
on the internet.

407
00:19:01,410 --> 00:19:04,030
And what's interesting
is, in some cases

408
00:19:04,030 --> 00:19:10,260
the black bar is smaller
than the gray bar, which

409
00:19:10,260 --> 00:19:12,600
means that common
usage is [? leading. ?]

410
00:19:12,600 --> 00:19:15,210
So sustainability
would be an example.

411
00:19:15,210 --> 00:19:17,340
Everybody talks
about sustainability,

412
00:19:17,340 --> 00:19:18,900
companies have--

413
00:19:18,900 --> 00:19:21,660
they say, our system is
sustainable because, you know,

414
00:19:21,660 --> 00:19:25,200
it uses less resources, it
produces less emissions.

415
00:19:25,200 --> 00:19:28,800
But the actual amount
of scientific knowledge,

416
00:19:28,800 --> 00:19:31,530
the actual amount of
research as to what

417
00:19:31,530 --> 00:19:34,650
really is sustainability and
how do you design for it,

418
00:19:34,650 --> 00:19:37,642
is actually smaller
than the usage.

419
00:19:37,642 --> 00:19:39,600
And then there's areas
where it's the opposite.

420
00:19:39,600 --> 00:19:41,260
Like for example, modularity.

421
00:19:41,260 --> 00:19:41,760
Right?

422
00:19:41,760 --> 00:19:44,040
You see that in modularity.

423
00:19:44,040 --> 00:19:46,650
So in the relative sense,
academic interest is leading.

424
00:19:46,650 --> 00:19:50,280
We see there's quite a lot
of literature on modularity.

425
00:19:50,280 --> 00:19:53,040
In mathematics, you know,
modularity and software,

426
00:19:53,040 --> 00:19:55,770
modularity and design,
but most people--

427
00:19:55,770 --> 00:19:59,010
you know, modularity per
se is not so interesting

428
00:19:59,010 --> 00:20:01,990
and so exciting to
the general public.

429
00:20:01,990 --> 00:20:04,890
So keep in mind that
for some ilities,

430
00:20:04,890 --> 00:20:06,510
there's an imbalance
essentially,

431
00:20:06,510 --> 00:20:09,189
between our understanding of--

432
00:20:09,189 --> 00:20:10,980
the scientific knowledge,
our understanding

433
00:20:10,980 --> 00:20:14,040
to be design, how to design
for it, and you know,

434
00:20:14,040 --> 00:20:16,440
how frequently at least
a keyword is used.

435
00:20:19,290 --> 00:20:21,720
Now here, this is
a little bit more--

436
00:20:21,720 --> 00:20:23,700
this is a little
bit harder to see.

437
00:20:23,700 --> 00:20:27,850
This is essentially looking
at these lifecycle properties

438
00:20:27,850 --> 00:20:29,980
over time.

439
00:20:29,980 --> 00:20:32,220
So what this graph
those you know

440
00:20:32,220 --> 00:20:37,980
is starting here in 1884,
cumulatively the number

441
00:20:37,980 --> 00:20:42,690
of journal articles published
about each lifecycle property.

442
00:20:42,690 --> 00:20:45,780
And roughly what way
you can think about this

443
00:20:45,780 --> 00:20:49,710
is that there's some lifecycle
property, the top four--

444
00:20:49,710 --> 00:20:53,490
five that we've been actually
working on for a long time.

445
00:20:53,490 --> 00:20:56,940
Even safety, you know, there's
some interesting articles

446
00:20:56,940 --> 00:21:01,680
in the 1890s about safety
for example, in mines.

447
00:21:01,680 --> 00:21:03,870
You know, coal mines.

448
00:21:03,870 --> 00:21:05,790
There's an article
about how does

449
00:21:05,790 --> 00:21:09,360
the impact of lighting--
better lighting-- on safety

450
00:21:09,360 --> 00:21:12,660
and productivity in 1890.

451
00:21:12,660 --> 00:21:15,420
And so they show that just
providing better lighting,

452
00:21:15,420 --> 00:21:18,280
actually has a dual
benefit; fewer accidents,

453
00:21:18,280 --> 00:21:21,340
fewer fatalities, and
better production output.

454
00:21:21,340 --> 00:21:25,170
And so that's-- and this was a
scientific study that was done

455
00:21:25,170 --> 00:21:29,040
by comparing data in different
mines and actually performing

456
00:21:29,040 --> 00:21:31,890
some experiments.

457
00:21:31,890 --> 00:21:35,070
Then you have a group
of life cycle properties

458
00:21:35,070 --> 00:21:38,310
that we only started really
thinking about and publishing

459
00:21:38,310 --> 00:21:41,010
about I would say,
around World War Two.

460
00:21:41,010 --> 00:21:44,400
Such as usability,
maintainability,

461
00:21:44,400 --> 00:21:48,330
and my interpretation of
this is that especially

462
00:21:48,330 --> 00:21:53,180
during World War Two, a big
difference made logistics.

463
00:21:53,180 --> 00:21:57,360
How easy was it was it
to use different weapons?

464
00:21:57,360 --> 00:22:00,180
How easy was it to
maintain equipment?

465
00:22:00,180 --> 00:22:02,730
It became a huge
determining factor

466
00:22:02,730 --> 00:22:04,050
in the outcome of the war.

467
00:22:04,050 --> 00:22:06,780
You know, for example, in North
Africa in the North African

468
00:22:06,780 --> 00:22:13,560
theater, you know tanks, trucks
being exposed to the sand

469
00:22:13,560 --> 00:22:14,320
and so forth.

470
00:22:14,320 --> 00:22:16,290
So people really
started thinking

471
00:22:16,290 --> 00:22:18,330
about the military,
started thinking heavily

472
00:22:18,330 --> 00:22:21,300
about the importance
of maintainability

473
00:22:21,300 --> 00:22:23,310
in the design of these systems.

474
00:22:23,310 --> 00:22:28,020
And after the war then, a lot
of these concepts like usability

475
00:22:28,020 --> 00:22:30,600
and maintainability
started spreading

476
00:22:30,600 --> 00:22:34,410
into general civilian life
and products and so forth.

477
00:22:37,980 --> 00:22:41,310
The third type of ilities
are the newer ones

478
00:22:41,310 --> 00:22:44,580
that we've done research
on just since the 70s,

479
00:22:44,580 --> 00:22:46,240
in the last 30 years.

480
00:22:46,240 --> 00:22:48,210
So here in this category
I would put things

481
00:22:48,210 --> 00:22:52,920
like sustainability,
recyclability, evolvability,

482
00:22:52,920 --> 00:22:56,010
even interoperability-- which
means the ability of systems

483
00:22:56,010 --> 00:22:59,790
to connect together and work
across system boundaries.

484
00:22:59,790 --> 00:23:02,820
Those are pretty recent
lifecycle properties

485
00:23:02,820 --> 00:23:06,350
and we're still actively
researching them.

486
00:23:06,350 --> 00:23:11,610
So another way to show this is
by essentially making a network

487
00:23:11,610 --> 00:23:13,800
of these lifecycle properties.

488
00:23:13,800 --> 00:23:17,690
And so the way-- this is still
all using this prevalence data.

489
00:23:17,690 --> 00:23:21,390
Essentially what you see
here is a network diagram

490
00:23:21,390 --> 00:23:24,030
that has these lifecycle
properties in a network

491
00:23:24,030 --> 00:23:25,360
relationship.

492
00:23:25,360 --> 00:23:29,370
Let me just explain, so
the size of the nodes

493
00:23:29,370 --> 00:23:32,580
relates to how much
knowledge we have.

494
00:23:32,580 --> 00:23:35,730
Essentially the height of
the bars that we saw earlier.

495
00:23:35,730 --> 00:23:38,700
And then the strength
the thickness of the line

496
00:23:38,700 --> 00:23:40,890
relates to the
strength of-- really

497
00:23:40,890 --> 00:23:43,890
how closely are these
two ilities related.

498
00:23:43,890 --> 00:23:45,570
And the way this
is calculated it

499
00:23:45,570 --> 00:23:48,750
is using the so-called
2-tuple correlation.

500
00:23:48,750 --> 00:23:51,900
So you take, essentially,
articles that have--

501
00:23:51,900 --> 00:23:55,050
you look at articles that
have, for example reliability

502
00:23:55,050 --> 00:23:58,440
and maintainability
in the same article.

503
00:23:58,440 --> 00:23:59,070
Right?

504
00:23:59,070 --> 00:24:00,960
And then you divide
that by the total number

505
00:24:00,960 --> 00:24:05,250
of articles on reliability
and or maintainability.

506
00:24:05,250 --> 00:24:07,410
And that's a ratio
between 0 and 1.

507
00:24:07,410 --> 00:24:10,815
And then this graph was produced
with a cutoff strength of so.

508
00:24:10,815 --> 00:24:12,030
1, right?

509
00:24:12,030 --> 00:24:15,600
So if there's more
than 10% of articles

510
00:24:15,600 --> 00:24:18,670
list these two
properties together,

511
00:24:18,670 --> 00:24:20,760
then there will be a line here.

512
00:24:20,760 --> 00:24:23,160
And the stronger the line,
the thicker the line is,

513
00:24:23,160 --> 00:24:26,280
the more closely, the
more often these concepts

514
00:24:26,280 --> 00:24:28,890
are mentioned in
the same article

515
00:24:28,890 --> 00:24:30,100
or the same piece of work.

516
00:24:30,100 --> 00:24:33,780
So that reflects the
strength of relationship.

517
00:24:33,780 --> 00:24:38,160
Now, what's interesting is
when you first look at this,

518
00:24:38,160 --> 00:24:39,566
you don't see much.

519
00:24:39,566 --> 00:24:41,190
But after you look
at this for a while,

520
00:24:41,190 --> 00:24:43,330
you start to realize
a few things.

521
00:24:43,330 --> 00:24:45,960
First of all, in the
center of the graph

522
00:24:45,960 --> 00:24:48,690
we have the classic
ilities of engineering.

523
00:24:48,690 --> 00:24:53,250
Quality, safety, reliability,
and I would argue flexibility

524
00:24:53,250 --> 00:24:53,920
as well.

525
00:24:53,920 --> 00:24:56,410
Those are the top four
that we saw before.

526
00:24:56,410 --> 00:24:58,560
And then around the
periphery of the graph,

527
00:24:58,560 --> 00:25:01,980
we have lifecycle properties
that are more recent.

528
00:25:01,980 --> 00:25:05,550
We haven't really thought
about them too much,

529
00:25:05,550 --> 00:25:09,120
we don't fully know how
to design for them yet.

530
00:25:09,120 --> 00:25:10,930
And if you look
at them in groups,

531
00:25:10,930 --> 00:25:13,630
I will argue that there are
three major groups here.

532
00:25:13,630 --> 00:25:15,990
So the first group
in the upper left

533
00:25:15,990 --> 00:25:21,280
is things like maintainability,
durability, reliability,

534
00:25:21,280 --> 00:25:24,230
quality, and so forth.

535
00:25:24,230 --> 00:25:26,870
So this is all about,
is the system made well,

536
00:25:26,870 --> 00:25:29,780
with high quality,
particularly early in its life.

537
00:25:29,780 --> 00:25:31,420
Right?

538
00:25:31,420 --> 00:25:36,670
Then we have a group here on
the right, and this is about,

539
00:25:36,670 --> 00:25:39,000
is the system easy to change.

540
00:25:39,000 --> 00:25:42,850
Is it easy to change the system
configuration-- flexibility,

541
00:25:42,850 --> 00:25:49,930
extensibility, modularity and
scalability, interoperability.

542
00:25:49,930 --> 00:25:53,410
Those are all different
sub-flavors, if you want,

543
00:25:53,410 --> 00:25:57,490
of being able to modify
the system over its life.

544
00:25:57,490 --> 00:26:01,180
And so that's a group of
ilities that that goes together.

545
00:26:01,180 --> 00:26:05,260
And then the third one is
resilience, robustness,

546
00:26:05,260 --> 00:26:08,920
quality, safety.

547
00:26:08,920 --> 00:26:11,950
And so this is really
related to performance

548
00:26:11,950 --> 00:26:14,800
of the system under different
types of uncertainty.

549
00:26:14,800 --> 00:26:17,410
You know, either
environmental variability

550
00:26:17,410 --> 00:26:21,640
or failures in the system,
and the ability of the system

551
00:26:21,640 --> 00:26:24,330
to withstand, or
at least perform--

552
00:26:24,330 --> 00:26:26,260
you know, have good
residual performance

553
00:26:26,260 --> 00:26:28,600
even in the face of failures.

554
00:26:28,600 --> 00:26:32,740
So that's, I think, an
important way to think about it.

555
00:26:32,740 --> 00:26:35,080
And when you're writing
requirements for systems--

556
00:26:35,080 --> 00:26:37,180
like the system
should be resilient.

557
00:26:37,180 --> 00:26:39,010
Then this helps
understand well, what

558
00:26:39,010 --> 00:26:41,560
are the other properties
that are linked to it

559
00:26:41,560 --> 00:26:43,780
and that maybe support that?

560
00:26:43,780 --> 00:26:47,860
OK, any questions
here or at MIT about--

561
00:26:47,860 --> 00:26:50,500
this is sort of method
one, is get data

562
00:26:50,500 --> 00:26:53,470
about the lifecycle property
and put them in relation

563
00:26:53,470 --> 00:26:54,670
to each other.

564
00:26:54,670 --> 00:26:56,180
Any questions?

565
00:27:02,050 --> 00:27:04,824
Johanna, any question there?

566
00:27:04,824 --> 00:27:05,990
AUDIENCE: I have a question.

567
00:27:05,990 --> 00:27:08,620
So is this chart that you
have now actually used

568
00:27:08,620 --> 00:27:12,130
in the initial design and
like conception [? con-ops ?]

569
00:27:12,130 --> 00:27:16,160
to like try to tease out
the interdependencies,

570
00:27:16,160 --> 00:27:19,760
or is it too abstract and just
more of an educational tool.

571
00:27:22,057 --> 00:27:23,640
PROFESSOR: It's
really at this point--

572
00:27:23,640 --> 00:27:25,150
you know this is fairly recent.

573
00:27:25,150 --> 00:27:27,740
And this was just done
in the last few years.

574
00:27:27,740 --> 00:27:31,110
So I don't think this has fully
penetrated system engineering

575
00:27:31,110 --> 00:27:32,110
practice yet.

576
00:27:32,110 --> 00:27:35,970
But the point-- the point that
I want to make here is that

577
00:27:35,970 --> 00:27:36,840
there's a huge--

578
00:27:36,840 --> 00:27:40,230
there's a huge gap right
now between when people

579
00:27:40,230 --> 00:27:44,310
in briefings you know,
either at the Pentagon

580
00:27:44,310 --> 00:27:47,190
or at a corporate
headquarters say,

581
00:27:47,190 --> 00:27:49,620
you know we want a
sustainable product

582
00:27:49,620 --> 00:27:52,590
or we want-- we're
a software company,

583
00:27:52,590 --> 00:27:55,680
we do optical
networking and we want

584
00:27:55,680 --> 00:27:57,810
we want to have the
most resilient design

585
00:27:57,810 --> 00:27:59,070
of the industry.

586
00:27:59,070 --> 00:28:01,890
You know, they'll put that
as a goal for the project.

587
00:28:01,890 --> 00:28:04,140
The question then is,
well OK, that's fine.

588
00:28:04,140 --> 00:28:05,760
But what does that really mean?

589
00:28:05,760 --> 00:28:07,500
What does it really
mean, resilient?

590
00:28:07,500 --> 00:28:10,320
You have to operationalize
that definition such

591
00:28:10,320 --> 00:28:13,316
that you can derive from
it lower level requirements

592
00:28:13,316 --> 00:28:14,940
that you can actually
go and design to,

593
00:28:14,940 --> 00:28:16,590
that you can test for.

594
00:28:16,590 --> 00:28:20,820
So what this helps you
to do is understand

595
00:28:20,820 --> 00:28:24,240
what may be supporting--

596
00:28:24,240 --> 00:28:27,960
what are supporting
elements that

597
00:28:27,960 --> 00:28:30,120
will be related to resilience.

598
00:28:30,120 --> 00:28:34,260
You know, what are supporting
concepts, supporting

599
00:28:34,260 --> 00:28:37,800
lifecycle properties that
this property that you're

600
00:28:37,800 --> 00:28:39,970
looking for is linked to?

601
00:28:39,970 --> 00:28:40,470
Right?

602
00:28:40,470 --> 00:28:44,400
So it's an evolution-- you
know, we know how to design now

603
00:28:44,400 --> 00:28:48,780
for speed and energy
efficiency, and you know,

604
00:28:48,780 --> 00:28:51,070
the sort of things that
were really hard to do 20,

605
00:28:51,070 --> 00:28:52,630
30, 50 years ago.

606
00:28:52,630 --> 00:28:54,600
It's pretty standard
practice now.

607
00:28:54,600 --> 00:28:57,480
You know, how do you design
for optimal interoperability?

608
00:28:57,480 --> 00:29:00,630
We don't quite know yet,
but we're finding our way.

609
00:29:00,630 --> 00:29:05,760
And so what I believe
is that especially

610
00:29:05,760 --> 00:29:09,840
the lifecycle properties on the
outer periphery of this chart,

611
00:29:09,840 --> 00:29:12,660
those are the ones that
need more work and those are

612
00:29:12,660 --> 00:29:16,122
the ones that we're really
learning how to operationalize.

613
00:29:18,900 --> 00:29:21,952
Does that make sense?

614
00:29:21,952 --> 00:29:22,910
AUDIENCE: Yeah, thanks.

615
00:29:22,910 --> 00:29:25,630
That makes sense.

616
00:29:25,630 --> 00:29:28,630
PROFESSOR: OK, good.

617
00:29:28,630 --> 00:29:30,760
Let me talk about the
[? message ?] too.

618
00:29:30,760 --> 00:29:35,590
So this was basically trying
to get to-- how do people,

619
00:29:35,590 --> 00:29:36,790
how do humans--

620
00:29:36,790 --> 00:29:39,471
there's semantics, semantics
means the meaning of words,

621
00:29:39,471 --> 00:29:39,970
right?

622
00:29:39,970 --> 00:29:44,000
Semantics is the science
of the meaning of words.

623
00:29:44,000 --> 00:29:48,060
How do they interpret
these lifecycle properties?

624
00:29:48,060 --> 00:29:51,580
And so humans have a
deep and possibly varied

625
00:29:51,580 --> 00:29:55,180
understanding of the semantics
of these lifecycle properties.

626
00:29:55,180 --> 00:29:59,050
So what was done here was
that a list of 15 life cycle

627
00:29:59,050 --> 00:30:02,230
properties, many of them are
overlapping with the ones you

628
00:30:02,230 --> 00:30:08,750
saw earlier, were presented
and then the challenge was to--

629
00:30:08,750 --> 00:30:12,650
the question that
these four groups--

630
00:30:12,650 --> 00:30:15,050
12 participants, four groups--

631
00:30:15,050 --> 00:30:17,810
had to say, is there
a hierarchy here?

632
00:30:17,810 --> 00:30:21,590
Are some higher level lifecycle
properties are some of them

633
00:30:21,590 --> 00:30:23,690
lower level properties
that support these.

634
00:30:23,690 --> 00:30:27,410
So there was a round one, find
the parent-child relationships,

635
00:30:27,410 --> 00:30:28,670
describe these.

636
00:30:28,670 --> 00:30:31,260
Interviews, and then
the second round.

637
00:30:31,260 --> 00:30:38,390
So here's essentially what was
the results of the first round.

638
00:30:38,390 --> 00:30:41,120
So four different
groups, so each of them

639
00:30:41,120 --> 00:30:45,530
basically came up with its
own version of a hierarchy.

640
00:30:45,530 --> 00:30:47,030
They didn't talk
to each other, they

641
00:30:47,030 --> 00:30:48,980
were firewalled from each other.

642
00:30:48,980 --> 00:30:52,700
And in all cases this
notion of value robustness

643
00:30:52,700 --> 00:30:53,540
came out on top.

644
00:30:53,540 --> 00:30:57,470
So value robustness means the
system should deliver value,

645
00:30:57,470 --> 00:30:59,300
right, to the stakeholders.

646
00:30:59,300 --> 00:31:03,490
Despite failures you know,
environmental changes,

647
00:31:03,490 --> 00:31:07,190
so value delivery of the system
is the top, the most important

648
00:31:07,190 --> 00:31:07,860
thing.

649
00:31:07,860 --> 00:31:12,240
Then what's different here
is how do you achieve that?

650
00:31:12,240 --> 00:31:14,750
So for example, you
see group one had

651
00:31:14,750 --> 00:31:17,870
robustness and changeability.

652
00:31:17,870 --> 00:31:20,060
So robustness
typically means even

653
00:31:20,060 --> 00:31:23,230
if you don't make any changes to
the system, it should continue,

654
00:31:23,230 --> 00:31:24,900
you know, it should
be survivable.

655
00:31:24,900 --> 00:31:26,960
It should be versatile.

656
00:31:26,960 --> 00:31:29,000
Changeable means
you actually modify

657
00:31:29,000 --> 00:31:30,860
the system over its life.

658
00:31:30,860 --> 00:31:33,530
And then we have lower level
properties like modularity

659
00:31:33,530 --> 00:31:34,880
or reconfigurability.

660
00:31:34,880 --> 00:31:37,760
And you know, there were
differences between the groups

661
00:31:37,760 --> 00:31:42,270
but not as big as
you might think.

662
00:31:42,270 --> 00:31:46,070
So in the second round, as a
result of the second round,

663
00:31:46,070 --> 00:31:51,020
then this so-called means to
end hierarchy was constructed.

664
00:31:51,020 --> 00:31:54,440
So means means, you know, these
are the enabling lifecycle

665
00:31:54,440 --> 00:31:57,470
properties and ends means this
is the final, sort of this

666
00:31:57,470 --> 00:31:59,730
is really what you
want to achieve.

667
00:31:59,730 --> 00:32:02,010
And so this was
the result of that.

668
00:32:02,010 --> 00:32:04,190
So again, at the top
we have this notion

669
00:32:04,190 --> 00:32:06,020
of value robustness.

670
00:32:06,020 --> 00:32:08,030
And this is achieved
by a combination

671
00:32:08,030 --> 00:32:11,900
of survivability robustness
and changeability.

672
00:32:11,900 --> 00:32:15,020
And then these, in turn,
are achieved by lower level

673
00:32:15,020 --> 00:32:16,610
lifecycle properties.

674
00:32:16,610 --> 00:32:18,830
And then at the
lowest level we have

675
00:32:18,830 --> 00:32:22,340
things like
interoperability, modularity,

676
00:32:22,340 --> 00:32:25,410
reconfigurability, and so forth.

677
00:32:25,410 --> 00:32:28,070
And the difference
between the solid lines

678
00:32:28,070 --> 00:32:32,420
and the dashed lines here
is that if it's a solid line

679
00:32:32,420 --> 00:32:35,490
and these are
directed arrows, that

680
00:32:35,490 --> 00:32:40,930
means that three or four
out of four groups, right?

681
00:32:40,930 --> 00:32:43,230
So the majority
of groups had this

682
00:32:43,230 --> 00:32:46,110
is a particular
parent-child relationship.

683
00:32:46,110 --> 00:32:49,120
And if it's a dashed line it
means only two out of four.

684
00:32:49,120 --> 00:32:51,850
And if it's only one out
of 4, it's not shown here.

685
00:32:51,850 --> 00:32:52,350
OK?

686
00:32:52,350 --> 00:32:57,660
So this is kind of a combined
result across the four groups.

687
00:32:57,660 --> 00:33:01,290
So you know, what can
we take from this?

688
00:33:01,290 --> 00:33:03,420
So first of all,
lifecycle properties

689
00:33:03,420 --> 00:33:05,460
are absolutely critical.

690
00:33:05,460 --> 00:33:08,760
You'll find us in
mission statements,

691
00:33:08,760 --> 00:33:11,050
you know, you'll find it
in even in advertising.

692
00:33:11,050 --> 00:33:11,550
Right?

693
00:33:11,550 --> 00:33:14,640
A lot of companies say we have
we have a robust solution,

694
00:33:14,640 --> 00:33:18,420
we have a sustainable system,
we have resilient networks.

695
00:33:18,420 --> 00:33:21,870
So it really is a huge,
huge selling point.

696
00:33:21,870 --> 00:33:23,580
It's very critical.

697
00:33:23,580 --> 00:33:25,740
What I encourage you to
do as system engineers is

698
00:33:25,740 --> 00:33:29,520
take a critical look at this
and say, well how resilient

699
00:33:29,520 --> 00:33:30,240
is resilient?

700
00:33:30,240 --> 00:33:34,440
You know, how many subsequent
failures of the system

701
00:33:34,440 --> 00:33:36,000
can you tolerate?

702
00:33:36,000 --> 00:33:37,590
Sustainability;
what does that mean

703
00:33:37,590 --> 00:33:42,130
in terms of kilowatt
hours per hour of usage?

704
00:33:42,130 --> 00:33:43,860
You have to get
down to the details

705
00:33:43,860 --> 00:33:46,710
to start quantifying what
these lifecycle properties mean

706
00:33:46,710 --> 00:33:50,410
and compare them among systems.

707
00:33:50,410 --> 00:33:54,120
Despite differences, the two
methods that we just looked

708
00:33:54,120 --> 00:33:57,210
at-- so one is the prevalence
analysis and the other one is

709
00:33:57,210 --> 00:34:01,410
the human semantic exercise--

710
00:34:01,410 --> 00:34:05,170
despite differences, the high
level conclusions were similar.

711
00:34:05,170 --> 00:34:08,489
So some ilities are closely
related to each other and form

712
00:34:08,489 --> 00:34:11,989
semantic sets, meaning
they're tied together by both

713
00:34:11,989 --> 00:34:15,050
[? synonymie ?] or
polysemi relationships.

714
00:34:15,050 --> 00:34:17,736
[? Synonymie ?] means they
essentially mean the same--

715
00:34:17,736 --> 00:34:19,860
they're synonymous-- they
essentially mean the same

716
00:34:19,860 --> 00:34:20,820
thing.

717
00:34:20,820 --> 00:34:23,969
polysemi means-- it
is one word, but it

718
00:34:23,969 --> 00:34:26,790
means you can have
possible sub meanings.

719
00:34:26,790 --> 00:34:27,480
Right?

720
00:34:27,480 --> 00:34:31,320
So the idea of groups
of semantic sets.

721
00:34:31,320 --> 00:34:37,800
And those groups
essentially are robustness--

722
00:34:37,800 --> 00:34:39,989
so this is the
ability of the system

723
00:34:39,989 --> 00:34:44,670
to perform its job despite
either internal or exogenous

724
00:34:44,670 --> 00:34:45,429
disturbances.

725
00:34:45,429 --> 00:34:45,929
Right?

726
00:34:45,929 --> 00:34:47,489
That's robustness.

727
00:34:47,489 --> 00:34:52,110
Flexible or changeable,
which means that you

728
00:34:52,110 --> 00:34:54,330
can modify the system easily.

729
00:34:54,330 --> 00:34:56,370
You know, if you operate
the system for a while

730
00:34:56,370 --> 00:34:59,070
and then you realize,
Ah, I need the system

731
00:34:59,070 --> 00:35:00,110
to do something else.

732
00:35:00,110 --> 00:35:03,270
Or I need to adapt it, or make
it bigger, or make it smaller,

733
00:35:03,270 --> 00:35:04,890
or add some function.

734
00:35:04,890 --> 00:35:06,930
So you would modify
the system and that's

735
00:35:06,930 --> 00:35:09,420
flexibility or changeability.

736
00:35:09,420 --> 00:35:12,750
And then resilient
and survivable

737
00:35:12,750 --> 00:35:16,080
is very specifically the
ability of this system

738
00:35:16,080 --> 00:35:20,400
to continue performing
despite failures or attacks.

739
00:35:20,400 --> 00:35:20,910
Right?

740
00:35:20,910 --> 00:35:23,220
That's what resilience
really means.

741
00:35:23,220 --> 00:35:27,420
And so those seem to be the big
three semantic clusters that we

742
00:35:27,420 --> 00:35:29,620
see in life cycle properties.

743
00:35:29,620 --> 00:35:31,530
And then the third
point here is there

744
00:35:31,530 --> 00:35:35,160
appears to be a hierarchy of
life cycle properties with two

745
00:35:35,160 --> 00:35:36,510
or three levels.

746
00:35:36,510 --> 00:35:39,150
Where we have the lower
level properties of systems

747
00:35:39,150 --> 00:35:41,290
like modularity, for example.

748
00:35:41,290 --> 00:35:43,350
Really your customer
probably does not

749
00:35:43,350 --> 00:35:45,480
care about modularity, right?

750
00:35:45,480 --> 00:35:49,200
If you advertise the
modularity of the product, some

751
00:35:49,200 --> 00:35:52,620
of the more educated, some
of the more technically

752
00:35:52,620 --> 00:35:55,830
savvy customers, they may
understand what that means.

753
00:35:55,830 --> 00:35:57,210
But most of your
customers really

754
00:35:57,210 --> 00:35:59,430
won't appreciate
modularity because it's

755
00:35:59,430 --> 00:36:02,400
kind of a low level
form-oriented technical

756
00:36:02,400 --> 00:36:04,300
property of the system.

757
00:36:04,300 --> 00:36:07,110
But what they will appreciate
is the ability of the system

758
00:36:07,110 --> 00:36:11,670
to be reconfigured or
adaptive for different pieces,

759
00:36:11,670 --> 00:36:14,460
so interoperability,
modularity, et cetera

760
00:36:14,460 --> 00:36:17,910
are low level
lifecycle properties

761
00:36:17,910 --> 00:36:22,500
that act as enablers of a higher
level lifecycle properties.

762
00:36:22,500 --> 00:36:24,480
And so future work--

763
00:36:24,480 --> 00:36:27,970
future work in this
area is to both just

764
00:36:27,970 --> 00:36:30,570
to apply these methods
to a broader set

765
00:36:30,570 --> 00:36:34,410
abilities, larger group of
test subjects and more data,

766
00:36:34,410 --> 00:36:36,900
but also from a
practical standpoint

767
00:36:36,900 --> 00:36:40,060
to operationalize better.

768
00:36:40,060 --> 00:36:41,600
How do we write requirements?

769
00:36:41,600 --> 00:36:44,460
How do we actually design
for the lifecycle properties

770
00:36:44,460 --> 00:36:47,910
like resilience,
flexibility, changeability.

771
00:36:47,910 --> 00:36:49,080
How do you really design it?

772
00:36:49,080 --> 00:36:53,540
So go from the keyword
to real engineering

773
00:36:53,540 --> 00:36:57,870
by operationalizing the
sub-attributes or factors

774
00:36:57,870 --> 00:36:59,360
in the system.

775
00:36:59,360 --> 00:37:00,130
OK?

776
00:37:00,130 --> 00:37:02,430
So that's a quick
summary of life--

777
00:37:02,430 --> 00:37:05,910
you know, I could go on
for hours about this.

778
00:37:05,910 --> 00:37:08,130
I'm very passionate
about this topic

779
00:37:08,130 --> 00:37:12,450
and I will say that
people who have

780
00:37:12,450 --> 00:37:15,090
dealt with large complex
systems you know, whether you're

781
00:37:15,090 --> 00:37:18,600
operating the transportation
system of a city

782
00:37:18,600 --> 00:37:22,110
or in airline
operations, or you're

783
00:37:22,110 --> 00:37:26,010
running the IP infrastructure
of a major corporation.

784
00:37:26,010 --> 00:37:29,440
These words, these
words are real.

785
00:37:29,440 --> 00:37:33,510
These words are you know,
real dollars, real challenges.

786
00:37:33,510 --> 00:37:35,560
This is really
where the action is

787
00:37:35,560 --> 00:37:37,515
in a lot of these
large complex systems.

788
00:37:40,720 --> 00:37:41,860
OK.

789
00:37:41,860 --> 00:37:45,340
Any comments or questions
about lifecycle properties?

790
00:37:45,340 --> 00:37:47,770
What they are, how they
relate to each other,

791
00:37:47,770 --> 00:37:48,710
why they're important.

792
00:37:48,710 --> 00:37:51,820
Voelker, did you want
to maybe say something?

793
00:37:51,820 --> 00:37:52,450
OK.

794
00:37:52,450 --> 00:37:53,570
All right.

795
00:37:53,570 --> 00:37:55,480
So let me talk about--

796
00:37:55,480 --> 00:37:57,970
try to make this a
little more real.

797
00:37:57,970 --> 00:38:00,730
Let me talk about a case
study, a very specific case

798
00:38:00,730 --> 00:38:05,230
study about communications
satellite constellations.

799
00:38:05,230 --> 00:38:09,910
And this is the second paper
that underlies today's lecture.

800
00:38:09,910 --> 00:38:14,110
And so let me give you a
little context for this first.

801
00:38:14,110 --> 00:38:18,760
So this work here was originally
published about a decade ago,

802
00:38:18,760 --> 00:38:20,650
in March 2004.

803
00:38:20,650 --> 00:38:25,500
And that was a few years after
the Iridium and Globalstar--

804
00:38:25,500 --> 00:38:28,660
these the two well known
communication satellite

805
00:38:28,660 --> 00:38:29,890
constellations--

806
00:38:29,890 --> 00:38:30,820
had been launched.

807
00:38:30,820 --> 00:38:33,340
And I have to give you a
little context, when I first

808
00:38:33,340 --> 00:38:36,630
came to MIT in the
mid 1990s, there

809
00:38:36,630 --> 00:38:40,660
was a huge interest in this area
of satellite constellations.

810
00:38:40,660 --> 00:38:45,790
You know commercially,
scientifically, in fact,

811
00:38:45,790 --> 00:38:49,360
the impression was there were so
many applications for satellite

812
00:38:49,360 --> 00:38:51,927
compilations filed,
that we're going

813
00:38:51,927 --> 00:38:54,010
to have so many satellites
up there you won't even

814
00:38:54,010 --> 00:38:55,370
see the sun anymore.

815
00:38:55,370 --> 00:38:59,050
It was just like thousands
and thousands of satellites.

816
00:38:59,050 --> 00:39:01,750
And the Iridium and
Globalstar were really

817
00:39:01,750 --> 00:39:04,330
the first two constellations
that were fully

818
00:39:04,330 --> 00:39:07,660
developed, launched,
and both of them

819
00:39:07,660 --> 00:39:10,810
failed commercially
within a very short time,

820
00:39:10,810 --> 00:39:12,190
a couple of years.

821
00:39:12,190 --> 00:39:15,450
And so after this happened
the whole market and interest

822
00:39:15,450 --> 00:39:20,530
in satellite constellations
collapsed for a long time,

823
00:39:20,530 --> 00:39:24,000
until about two years ago,
two or three years ago.

824
00:39:24,000 --> 00:39:27,250
Now people talk about
these constellations again.

825
00:39:27,250 --> 00:39:32,890
You know, constellations of
[INAUDIBLE],, Iridium Next.

826
00:39:32,890 --> 00:39:35,530
You know, there's sort
of new enthusiasm ,

827
00:39:35,530 --> 00:39:39,730
new wave of enthusiasm
for constellations.

828
00:39:39,730 --> 00:39:43,640
So this paper, and this case
that I want to tell you about

829
00:39:43,640 --> 00:39:46,540
is really about the first wave.

830
00:39:46,540 --> 00:39:49,570
And in terms of the
ilities, the one

831
00:39:49,570 --> 00:39:53,590
that that I'd like to
explain to you about here

832
00:39:53,590 --> 00:39:55,630
is flexibility and scalability.

833
00:39:55,630 --> 00:39:58,300
Rather than thinking about
a system as something

834
00:39:58,300 --> 00:40:01,230
that you design and
build all at once,

835
00:40:01,230 --> 00:40:04,300
how do you design a system
such that you can gradually

836
00:40:04,300 --> 00:40:05,200
deploy it?

837
00:40:05,200 --> 00:40:07,510
We call that staged
deployment approach.

838
00:40:07,510 --> 00:40:09,890
And what's the benefit of that?

839
00:40:09,890 --> 00:40:14,360
So here's some pictures
you see on the right side,

840
00:40:14,360 --> 00:40:17,680
this is what the original
iridium satellites look like.

841
00:40:17,680 --> 00:40:21,202
They actually use
phased array antennas--

842
00:40:21,202 --> 00:40:23,035
and actually both use
phased array antennas.

843
00:40:23,035 --> 00:40:25,510
So you have individual
elements here.

844
00:40:25,510 --> 00:40:29,590
And by differentially
phasing the signal,

845
00:40:29,590 --> 00:40:31,610
you can actually steer the beam.

846
00:40:31,610 --> 00:40:34,480
This was a very new
technology at time.

847
00:40:34,480 --> 00:40:36,970
And then the
Globalstar satellite

848
00:40:36,970 --> 00:40:40,690
shown at the lower picture.

849
00:40:40,690 --> 00:40:43,050
Here is a little
bit of that data.

850
00:40:43,050 --> 00:40:46,490
So both of these were
launched in the late 1990s.

851
00:40:46,490 --> 00:40:50,920
Iridium has 66
satellites, Globalstar 48.

852
00:40:50,920 --> 00:40:52,930
Iridium is a polar
constellation,

853
00:40:52,930 --> 00:40:57,230
so the satellites go almost
directly over the poles.

854
00:40:57,230 --> 00:41:00,520
Globalstar is a walker
constellation, so inclined.

855
00:41:00,520 --> 00:41:02,590
Doesn't quite give you
full global coverage,

856
00:41:02,590 --> 00:41:05,170
it's about plus or
minus [? 78 ?] degrees,

857
00:41:05,170 --> 00:41:08,470
so you can't use
Globalstar at the poles.

858
00:41:08,470 --> 00:41:10,960
The altitudes are a
bit different too.

859
00:41:10,960 --> 00:41:16,240
Iridium is at 780 kilometers,
while Globalstar is at 1,400.

860
00:41:16,240 --> 00:41:20,650
Which-- I know at least one
or two of you are working

861
00:41:20,650 --> 00:41:22,060
on the Van Allen--

862
00:41:22,060 --> 00:41:24,070
the Van Allen Belt Commission.

863
00:41:24,070 --> 00:41:27,490
Right? [? The cube stats ?] to
measure the Van Allen Belts.

864
00:41:27,490 --> 00:41:30,110
Somebody mentioned that
today, who was that?

865
00:41:30,110 --> 00:41:30,610
No?

866
00:41:30,610 --> 00:41:32,870
Did I hear that wrong?

867
00:41:32,870 --> 00:41:34,870
It was mentioned, right?

868
00:41:34,870 --> 00:41:35,800
So who was that?

869
00:41:38,620 --> 00:41:41,460
Arnold, and he's not here.

870
00:41:41,460 --> 00:41:42,190
Ah, see.

871
00:41:42,190 --> 00:41:46,950
So 1,400 kilometers, are
you actually pushing,

872
00:41:46,950 --> 00:41:50,140
you're starting to push the
lower edges of the Van Allen

873
00:41:50,140 --> 00:41:53,360
Belt. So one of the big,
you know, in some sense,

874
00:41:53,360 --> 00:41:56,440
it's easier to be higher because
you need fewer satellites

875
00:41:56,440 --> 00:41:57,840
to cover the whole earth.

876
00:41:57,840 --> 00:41:59,740
But the higher you
go, the more exposed

877
00:41:59,740 --> 00:42:01,780
you are to the
radiation environment.

878
00:42:01,780 --> 00:42:03,460
You get closer to
the Van Allen Belt,

879
00:42:03,460 --> 00:42:05,890
so that's the big
trade-off there.

880
00:42:05,890 --> 00:42:07,590
You know, the mass
of the satellites,

881
00:42:07,590 --> 00:42:12,820
they're between 450 and 700
kilograms, transmitter power,

882
00:42:12,820 --> 00:42:14,417
around 400 watts.

883
00:42:14,417 --> 00:42:16,750
You know, which is not that
much, if you think about it,

884
00:42:16,750 --> 00:42:17,250
right?

885
00:42:17,250 --> 00:42:21,120
That's four very strong
light bulbs, the old style

886
00:42:21,120 --> 00:42:21,695
light bulbs.

887
00:42:24,610 --> 00:42:26,710
And then what's very
different, again,

888
00:42:26,710 --> 00:42:28,860
is the multi-access schemes.

889
00:42:28,860 --> 00:42:34,780
So Iridium used time division
multiplexing and Globalstar,

890
00:42:34,780 --> 00:42:38,620
which was supported by
Qualcomm, used essentially CDMA.

891
00:42:38,620 --> 00:42:41,530
So you don't chop your frequency
band into separate channels,

892
00:42:41,530 --> 00:42:44,780
you use the whole
frequency band and then

893
00:42:44,780 --> 00:42:47,660
you use a pseudo
random access code

894
00:42:47,660 --> 00:42:54,280
that essentially de-convolves
the signal for each channel.

895
00:42:54,280 --> 00:42:57,040
The number of
channels, about 72,000,

896
00:42:57,040 --> 00:43:01,330
about 120,000 duplex channels so
that you can actually-- duplex

897
00:43:01,330 --> 00:43:04,360
mean you can carry on a two
way conversation as opposed

898
00:43:04,360 --> 00:43:07,100
to just be asynchronous.

899
00:43:07,100 --> 00:43:09,180
And then you can see the
data rates, quite low,

900
00:43:09,180 --> 00:43:10,090
like per channel.

901
00:43:10,090 --> 00:43:16,510
4.8 or kilobits per second,
2.4, 4.8, 9.6 for Globalstar

902
00:43:16,510 --> 00:43:19,870
it's enough for
having a conversation.

903
00:43:19,870 --> 00:43:25,330
And total system cost, Iridium
was about $5.7 billion dollars

904
00:43:25,330 --> 00:43:27,910
and Globalstar
about $3.3 million.

905
00:43:27,910 --> 00:43:30,910
Not including the cost
of the ground stations.

906
00:43:30,910 --> 00:43:35,560
Both went bankrupt relatively
quickly after they launched.

907
00:43:35,560 --> 00:43:38,350
However, they've been operating
really since then, right?

908
00:43:38,350 --> 00:43:39,460
Since this time.

909
00:43:39,460 --> 00:43:41,890
Iridium Next is currently
under development

910
00:43:41,890 --> 00:43:45,550
and is scheduled
to launch in 2017.

911
00:43:45,550 --> 00:43:49,030
Globalstar is publicly
traded and actually valued

912
00:43:49,030 --> 00:43:52,210
as a company at $1.9
billion dollars.

913
00:43:52,210 --> 00:43:54,280
So they're actually both--

914
00:43:54,280 --> 00:43:57,400
you could almost
argue that they were

915
00:43:57,400 --> 00:44:00,590
a decade ahead of their time.

916
00:44:00,590 --> 00:44:01,090
And

917
00:44:01,090 --> 00:44:03,900
I want to tell you a little bit
about the story of, especially

918
00:44:03,900 --> 00:44:04,870
Iridium.

919
00:44:04,870 --> 00:44:09,280
So here's a couple of press
releases, so things that

920
00:44:09,280 --> 00:44:11,170
have been written in the press.

921
00:44:11,170 --> 00:44:15,170
So look at this one,
26th of June, 1990.

922
00:44:15,170 --> 00:44:17,350
Motorola unveiled
a new concept for

923
00:44:17,350 --> 00:44:19,480
global personal
communication. [? Bases ?]

924
00:44:19,480 --> 00:44:24,550
a constellation of low earth
orbit cellular satellites.

925
00:44:24,550 --> 00:44:30,460
August 18th 1999, nine years
later, last week Iridium LLC

926
00:44:30,460 --> 00:44:33,790
filed bankruptcy court
protection last investments

927
00:44:33,790 --> 00:44:36,610
are estimated at $5 billion.

928
00:44:36,610 --> 00:44:39,800
So the question is
why did it happen?

929
00:44:39,800 --> 00:44:42,250
The technology actually
worked quite well, it

930
00:44:42,250 --> 00:44:46,360
was not a technological failure,
it was a business failure--

931
00:44:46,360 --> 00:44:48,700
but I would argue
a system's failure.

932
00:44:48,700 --> 00:44:51,000
To think about the
problem differently,

933
00:44:51,000 --> 00:44:55,810
so the fundamental challenge
is to properly size

934
00:44:55,810 --> 00:44:57,625
the capacity of a large system.

935
00:44:57,625 --> 00:44:59,000
So if you're
designing, you know,

936
00:44:59,000 --> 00:45:03,325
a car factory, a
new power system

937
00:45:03,325 --> 00:45:06,310
and for a future
uncertain demand,

938
00:45:06,310 --> 00:45:09,610
it's very difficult to do this
because demand is uncertain.

939
00:45:09,610 --> 00:45:13,330
So is it better to oversized the
system or are you conservative

940
00:45:13,330 --> 00:45:15,100
and you make it smaller?

941
00:45:15,100 --> 00:45:17,560
Market assumptions
can change, right?

942
00:45:17,560 --> 00:45:21,430
In a seven to eight years.

943
00:45:21,430 --> 00:45:23,080
So essentially the v, right?

944
00:45:23,080 --> 00:45:25,720
Like getting back to the v.
Each of those two systems,

945
00:45:25,720 --> 00:45:30,070
Iridium and Globalstar, it
took them essentially a decade,

946
00:45:30,070 --> 00:45:32,860
almost 10 years-- right?--
to go across the whole v.

947
00:45:32,860 --> 00:45:34,970
And when you make
your requirements,

948
00:45:34,970 --> 00:45:37,540
your stakeholders, all the
stuff in the upper left portion

949
00:45:37,540 --> 00:45:42,430
of the v, there's so many years
that elapse between when you

950
00:45:42,430 --> 00:45:44,770
make those assumptions, you
write those requirements,

951
00:45:44,770 --> 00:45:47,350
and when you actually go to
market a lot of things can

952
00:45:47,350 --> 00:45:48,500
change.

953
00:45:48,500 --> 00:45:52,870
And that's fundamentally
the challenge here.

954
00:45:52,870 --> 00:45:57,670
Just to illustrate this
for you, showing some data,

955
00:45:57,670 --> 00:46:00,790
this is cellular this
is not space space,

956
00:46:00,790 --> 00:46:02,170
this is on the ground.

957
00:46:02,170 --> 00:46:06,070
Cellular subscribers
for mobile phones.

958
00:46:06,070 --> 00:46:08,170
I know this is hard
for your generation

959
00:46:08,170 --> 00:46:10,330
to sort of understand
this, but we actually

960
00:46:10,330 --> 00:46:13,450
didn't have mobile phones or
there were these clunky bricks

961
00:46:13,450 --> 00:46:14,740
in your car.

962
00:46:14,740 --> 00:46:18,460
You know, it's really
remarkable what happened.

963
00:46:18,460 --> 00:46:21,880
So look at this
data, in 1991 which

964
00:46:21,880 --> 00:46:24,810
is when the system was
just being developed,

965
00:46:24,810 --> 00:46:30,310
there were less than 10 million
mobile phone users in the US.

966
00:46:30,310 --> 00:46:32,200
It just wasn't-- it
was very expensive,

967
00:46:32,200 --> 00:46:34,110
it was just not widely--

968
00:46:34,110 --> 00:46:37,090
that technology hadn't been,
the networks weren't there.

969
00:46:37,090 --> 00:46:43,000
You know, so the green bars
where the forecasts that was

970
00:46:43,000 --> 00:46:48,430
done in 1991, as to how
quickly the mobile user

971
00:46:48,430 --> 00:46:50,530
market would evolve in the US.

972
00:46:50,530 --> 00:46:54,340
So their prediction was that
by 2000, a decade later,

973
00:46:54,340 --> 00:46:57,530
there'd be just shy
of 40 million users.

974
00:46:57,530 --> 00:46:58,390
OK?

975
00:46:58,390 --> 00:47:03,790
Now the dark blue bar
is the actual evolution.

976
00:47:03,790 --> 00:47:10,240
So actually by 2000 you know
the US now has 310 or so million

977
00:47:10,240 --> 00:47:12,580
inhabitants.

978
00:47:12,580 --> 00:47:17,020
So by 2000 there were
120 million users.

979
00:47:17,020 --> 00:47:19,450
So the forecast that was
done 10 years earlier

980
00:47:19,450 --> 00:47:21,660
was off by a factor of three.

981
00:47:21,660 --> 00:47:26,710
That grounds, the
terrestrial mobile networks,

982
00:47:26,710 --> 00:47:29,920
developed three times faster
than had been predicted.

983
00:47:29,920 --> 00:47:33,700
Now that's great for the
terrestrial people, right?

984
00:47:33,700 --> 00:47:38,240
AT&T and Comcast-- not
Comcast, Qualcomm, et cetera.

985
00:47:38,240 --> 00:47:42,310
But the problem is, of course,
that because you know ground

986
00:47:42,310 --> 00:47:44,870
based communications
was so much easier,

987
00:47:44,870 --> 00:47:47,050
a lot of the market that
had been anticipated

988
00:47:47,050 --> 00:47:50,110
for the satellite
based communications

989
00:47:50,110 --> 00:47:56,140
was essentially eaten
away by this competitor,

990
00:47:56,140 --> 00:47:57,440
by the ground based competitor.

991
00:47:57,440 --> 00:48:01,070
VOELKER: [INAUDIBLE]

992
00:48:01,070 --> 00:48:03,160
PROFESSOR: Yeah, OK.

993
00:48:03,160 --> 00:48:06,480
Can you still hear it,
MIT, are you still with us?

994
00:48:06,480 --> 00:48:07,211
AUDIENCE: Yes.

995
00:48:07,211 --> 00:48:08,210
PROFESSOR: Yes, you are.

996
00:48:08,210 --> 00:48:08,709
OK.

997
00:48:08,709 --> 00:48:12,710
Voelker wants to say something.

998
00:48:12,710 --> 00:48:17,150
VOELKER: Actually, the figures
that you showed [INAUDIBLE]

999
00:48:17,150 --> 00:48:21,290
and for once, Americans
did not look over the pond.

1000
00:48:21,290 --> 00:48:24,200
Because at that time in
Europe, the GSM systems

1001
00:48:24,200 --> 00:48:26,570
came really strong
and very quickly.

1002
00:48:26,570 --> 00:48:29,420
In the early 90s, you could
have already had your Sim card

1003
00:48:29,420 --> 00:48:32,430
in your cell phone [INAUDIBLE]
and in fact in America

1004
00:48:32,430 --> 00:48:35,390
then, you could only buy the
telephone with all the Sim card

1005
00:48:35,390 --> 00:48:37,532
in-- you had to buy
the whole system.

1006
00:48:37,532 --> 00:48:38,990
And so as the system
was controlled

1007
00:48:38,990 --> 00:48:41,400
by the large companies
and not by the users,

1008
00:48:41,400 --> 00:48:43,310
the large companies,
Motorola, thought

1009
00:48:43,310 --> 00:48:46,010
that they could
impose their satellite

1010
00:48:46,010 --> 00:48:47,510
system on the markets.

1011
00:48:47,510 --> 00:48:51,650
And in this case,
the country to the--

1012
00:48:51,650 --> 00:48:56,180
video recorders, hi fi
systems earlier, Europe

1013
00:48:56,180 --> 00:48:57,770
went much faster than America.

1014
00:48:57,770 --> 00:48:59,960
And actually
European GSM market--

1015
00:48:59,960 --> 00:49:01,730
cell phones, as you know them--

1016
00:49:01,730 --> 00:49:02,840
went much quicker.

1017
00:49:02,840 --> 00:49:05,080
And then Nokia, which
is actually European,

1018
00:49:05,080 --> 00:49:08,540
overtook and it took probably
the next decade for Americans

1019
00:49:08,540 --> 00:49:09,800
to catch up.

1020
00:49:09,800 --> 00:49:12,490
So finally here this was,
a combination of people

1021
00:49:12,490 --> 00:49:15,320
had the product and they
just couldn't get in the US.

1022
00:49:15,320 --> 00:49:17,540
So they were limited
to their own markets,

1023
00:49:17,540 --> 00:49:20,210
and not to sell their products
to the rest of the world

1024
00:49:20,210 --> 00:49:22,420
as they had done in the past.

1025
00:49:22,420 --> 00:49:23,790
PROFESSOR: Yeah, good point.

1026
00:49:23,790 --> 00:49:28,700
And in fact, you know
one of the things that--

1027
00:49:28,700 --> 00:49:31,820
I think both Globalstar
and Iridium had to do--

1028
00:49:31,820 --> 00:49:34,310
and this was a very
late decision, just

1029
00:49:34,310 --> 00:49:36,390
in the last couple of
years here before launch,

1030
00:49:36,390 --> 00:49:41,740
97, 98, is to make
their handsets dual use.

1031
00:49:41,740 --> 00:49:45,080
Like that if you
were in an area where

1032
00:49:45,080 --> 00:49:47,540
there was a cellular
network, the phone

1033
00:49:47,540 --> 00:49:49,780
would switch to that because
that would be cheaper,

1034
00:49:49,780 --> 00:49:51,560
and if you didn't
have cellular network

1035
00:49:51,560 --> 00:49:53,720
access it would automatically
try to communicate

1036
00:49:53,720 --> 00:49:54,740
to the satellite.

1037
00:49:54,740 --> 00:49:56,420
So there was a lot of--

1038
00:49:56,420 --> 00:49:58,580
there were a lot of
issues that came up

1039
00:49:58,580 --> 00:50:03,520
by essentially the markets, I
think in Europe, your point,

1040
00:50:03,520 --> 00:50:06,770
well for-- about the
GSM, and in the US just

1041
00:50:06,770 --> 00:50:10,040
developing quite differently
than had been anticipated.

1042
00:50:10,040 --> 00:50:13,310
OK so I want to give you
a little bit about sort

1043
00:50:13,310 --> 00:50:14,900
of economics.

1044
00:50:14,900 --> 00:50:19,700
In the end a lot of this is
driven by money, by economics.

1045
00:50:19,700 --> 00:50:24,260
And so this is
satellite economics 101.

1046
00:50:24,260 --> 00:50:29,030
The key question here
is how expensive is it--

1047
00:50:29,030 --> 00:50:30,710
what's the cost,
and then what is

1048
00:50:30,710 --> 00:50:33,560
the price that you can charge
for one minute of service,

1049
00:50:33,560 --> 00:50:34,060
right?

1050
00:50:34,060 --> 00:50:37,030
One minute, one unit of service.

1051
00:50:37,030 --> 00:50:39,780
And so look at
this equation here.

1052
00:50:39,780 --> 00:50:43,800
This is CPF, stands
for cost per function.

1053
00:50:43,800 --> 00:50:47,690
And so in the numerator we
have the lifecycle cost, which

1054
00:50:47,690 --> 00:50:50,490
is your initial investment.

1055
00:50:50,490 --> 00:50:52,820
And then we will
essentially capitalize

1056
00:50:52,820 --> 00:50:55,160
that with some interest rate k.

1057
00:50:55,160 --> 00:50:57,410
Plus then, for each
year of operation,

1058
00:50:57,410 --> 00:51:00,590
you have the operational costs
for that year you have to add.

1059
00:51:00,590 --> 00:51:01,700
So that's your initial--

1060
00:51:01,700 --> 00:51:04,790
that's your development costs,
your manufacturing costs

1061
00:51:04,790 --> 00:51:08,540
for the satellite,
including the launch costs.

1062
00:51:08,540 --> 00:51:11,750
And then the ops costs would be
operating your ground stations,

1063
00:51:11,750 --> 00:51:14,380
your networks, any
replenishment costs,

1064
00:51:14,380 --> 00:51:16,550
they would be in your ops costs.

1065
00:51:16,550 --> 00:51:18,380
And then you divide
this by what's

1066
00:51:18,380 --> 00:51:21,219
called here the number
of billable minutes.

1067
00:51:21,219 --> 00:51:23,510
That doesn't mean you're
actually going to bill people,

1068
00:51:23,510 --> 00:51:27,150
it means just they are
potentially billable minutes.

1069
00:51:27,150 --> 00:51:28,910
So that's the capacity
of the system,

1070
00:51:28,910 --> 00:51:32,930
[? see ?] the best times
you know, 365 days,

1071
00:51:32,930 --> 00:51:35,330
times 24 hours,
times 60 minutes,

1072
00:51:35,330 --> 00:51:37,780
times what's the load factor.

1073
00:51:37,780 --> 00:51:39,320
So the load factor
is essentially

1074
00:51:39,320 --> 00:51:44,090
the capacity utilization of
the system that you anticipate.

1075
00:51:44,090 --> 00:51:44,930
OK.

1076
00:51:44,930 --> 00:51:48,560
So that's the basic
equation for calculating

1077
00:51:48,560 --> 00:51:50,210
CPF, cost per function.

1078
00:51:50,210 --> 00:51:52,500
Plug-in some numbers here.

1079
00:51:52,500 --> 00:51:54,590
These are the numbers
that have been assumed,

1080
00:51:54,590 --> 00:51:57,410
you know a $3 billion
investment, 5%

1081
00:51:57,410 --> 00:52:02,350
interest rate, $300 million per
year of ops cost, 15 year life

1082
00:52:02,350 --> 00:52:03,120
cycle--

1083
00:52:03,120 --> 00:52:05,840
this is the capital
T, over 15 years.

1084
00:52:05,840 --> 00:52:09,860
100,000 channels--
that's your capacity.

1085
00:52:09,860 --> 00:52:14,210
Number of users in this
case, so the load factor

1086
00:52:14,210 --> 00:52:18,740
is simply the number of users
times the average activity

1087
00:52:18,740 --> 00:52:19,520
per user.

1088
00:52:19,520 --> 00:52:24,230
In this case, it seems to
be 1,200 minutes per year,

1089
00:52:24,230 --> 00:52:25,760
about 100 minutes per month.

1090
00:52:25,760 --> 00:52:27,140
OK?

1091
00:52:27,140 --> 00:52:28,850
So that gets you--

1092
00:52:28,850 --> 00:52:32,300
that gets you a CPF
of $0.20 per minute.

1093
00:52:32,300 --> 00:52:35,780
And based on that you
can-- the business case

1094
00:52:35,780 --> 00:52:39,220
was made based on
these kind of numbers.

1095
00:52:39,220 --> 00:52:44,690
And so 3 million users, just
3 million subscribers, right?

1096
00:52:44,690 --> 00:52:48,790
Not 3 million users
at the same time,

1097
00:52:48,790 --> 00:52:50,330
that number would
be much smaller.

1098
00:52:50,330 --> 00:52:52,460
That number can't be
bigger than 100,000

1099
00:52:52,460 --> 00:52:54,200
because that's your capacity.

1100
00:52:54,200 --> 00:52:56,390
So for example, if you run--

1101
00:52:56,390 --> 00:52:59,450
if you run a
fitness club, right?

1102
00:52:59,450 --> 00:53:01,490
At any given time
in your fitness club

1103
00:53:01,490 --> 00:53:05,625
you can have 50 or 100
people actually working out,

1104
00:53:05,625 --> 00:53:09,520
but your number of
customers, your subscribers,

1105
00:53:09,520 --> 00:53:12,474
should be 1,000 or 2,000.

1106
00:53:12,474 --> 00:53:14,140
And if they all show
up at the same time

1107
00:53:14,140 --> 00:53:15,610
you're in big trouble, right?

1108
00:53:15,610 --> 00:53:19,240
So that's the difference
between number of users

1109
00:53:19,240 --> 00:53:21,280
or the number of
subscribers and number

1110
00:53:21,280 --> 00:53:24,280
of active users
at any given time.

1111
00:53:24,280 --> 00:53:28,730
That's a big part of managing
these kind of systems.

1112
00:53:28,730 --> 00:53:32,520
However, what actually happened
is the number of actual users

1113
00:53:32,520 --> 00:53:34,250
grew much slower.

1114
00:53:34,250 --> 00:53:36,700
So if you plug-in some
different numbers,

1115
00:53:36,700 --> 00:53:38,890
let's keep all the
numbers the same,

1116
00:53:38,890 --> 00:53:43,000
except for the subscriber
base-- the number of users.

1117
00:53:43,000 --> 00:53:47,030
In this case we're going to
assume 50,000 users instead of

1118
00:53:47,030 --> 00:53:48,010
the three million.

1119
00:53:48,010 --> 00:53:51,010
This is closer to what
they had after about

1120
00:53:51,010 --> 00:53:52,240
a year of operation.

1121
00:53:52,240 --> 00:53:55,400
Now your CPF goes
to $12 per minute,

1122
00:53:55,400 --> 00:53:57,490
which is noncompetitive.

1123
00:53:57,490 --> 00:53:58,030
Right?

1124
00:53:58,030 --> 00:54:04,030
Except for some extremely, like,
military applications or making

1125
00:54:04,030 --> 00:54:07,420
some emergency phone
calls on an oil rig,

1126
00:54:07,420 --> 00:54:08,800
you know, on the ocean.

1127
00:54:08,800 --> 00:54:11,320
Most people at the time--

1128
00:54:11,320 --> 00:54:12,700
now and at that time--

1129
00:54:12,700 --> 00:54:17,920
would never pay that for
one minute of service.

1130
00:54:17,920 --> 00:54:20,980
And so that was the
fundamental problem,

1131
00:54:20,980 --> 00:54:25,240
is that the user base did not
materialize as fast as planned.

1132
00:54:25,240 --> 00:54:28,210
Therefore this cost per
function was way higher

1133
00:54:28,210 --> 00:54:31,510
and they did charge,
you know, $3 to $5

1134
00:54:31,510 --> 00:54:34,930
per minute of usage
of the system, which

1135
00:54:34,930 --> 00:54:38,920
as you try to squeeze
your existing users more,

1136
00:54:38,920 --> 00:54:42,190
you're not going to get the ramp
up and scale up in the system

1137
00:54:42,190 --> 00:54:42,970
that you need.

1138
00:54:42,970 --> 00:54:46,450
That was the fundamental
problem with the economics

1139
00:54:46,450 --> 00:54:49,490
of the system.

1140
00:54:49,490 --> 00:54:52,790
So let me talk a little bit
about the design decisions

1141
00:54:52,790 --> 00:54:56,740
the conceptual design of what
this design space looks like.

1142
00:54:56,740 --> 00:55:00,160
So fundamental-- oh, the other
thing I should mention to you,

1143
00:55:00,160 --> 00:55:04,570
what was interesting is after
the bankruptcy, both of Iridium

1144
00:55:04,570 --> 00:55:08,020
and Globalstar, both of the
chief engineers for both

1145
00:55:08,020 --> 00:55:12,460
of these systems
took refuge at MIT.

1146
00:55:12,460 --> 00:55:14,050
Essentially, they came to like--

1147
00:55:14,050 --> 00:55:17,400
Joel Schindall, who was the
chief engineer for Globalstar,

1148
00:55:17,400 --> 00:55:21,250
great guy, both very competent
people, came to MIT--

1149
00:55:21,250 --> 00:55:23,560
still there, still
a professor there.

1150
00:55:23,560 --> 00:55:26,490
And then Ray Leopold was
one of the three architects

1151
00:55:26,490 --> 00:55:29,980
for Iridium, also came
to MIT during the time.

1152
00:55:29,980 --> 00:55:33,070
So I had extensive
discussions with them,

1153
00:55:33,070 --> 00:55:35,890
and what you see
here on slide 25

1154
00:55:35,890 --> 00:55:38,170
is one result of
those discussions.

1155
00:55:38,170 --> 00:55:40,840
Which is, the key
design decisions

1156
00:55:40,840 --> 00:55:44,800
that they have to make,
fundamentally, there is--

1157
00:55:44,800 --> 00:55:47,050
you won't be
surprised to see this,

1158
00:55:47,050 --> 00:55:49,930
this is the magic
number seven, right?

1159
00:55:49,930 --> 00:55:52,030
Seven key design decisions.

1160
00:55:52,030 --> 00:55:55,360
When you design a satellite
consultation for communication

1161
00:55:55,360 --> 00:55:57,970
purposes, constellation type--

1162
00:55:57,970 --> 00:56:01,330
polar or walker,
orbital altitude,

1163
00:56:01,330 --> 00:56:03,710
minimum elevation angle
above the horizon--

1164
00:56:03,710 --> 00:56:06,580
that you can communicate--

1165
00:56:06,580 --> 00:56:09,820
satellite transmitting
power, the size

1166
00:56:09,820 --> 00:56:13,100
of your primary antenna,
your multi-access scheme.

1167
00:56:13,100 --> 00:56:17,080
So this is time division or
code division multiplexing.

1168
00:56:17,080 --> 00:56:19,840
And then the last one is about
the network architecture--

1169
00:56:19,840 --> 00:56:22,600
do you have inter-satellite
links, yes or no?

1170
00:56:22,600 --> 00:56:24,640
Inter-satellite links
means satellites

1171
00:56:24,640 --> 00:56:27,250
can talk to each other.

1172
00:56:27,250 --> 00:56:31,210
And Iridium chose that,
Globalstar did not.

1173
00:56:31,210 --> 00:56:34,870
So in Globalstar, the Globalstar
satellites cannot talk to each

1174
00:56:34,870 --> 00:56:36,370
other directly in space.

1175
00:56:36,370 --> 00:56:37,450
They can only talk--

1176
00:56:37,450 --> 00:56:41,560
it's a bent pipe system
to a ground station.

1177
00:56:41,560 --> 00:56:42,060
OK?

1178
00:56:42,060 --> 00:56:45,360
So if you this is like
the morphological matrix

1179
00:56:45,360 --> 00:56:49,350
that you learned about, you
just pick design decisions

1180
00:56:49,350 --> 00:56:52,015
in this morphological
matrix and you come up

1181
00:56:52,015 --> 00:56:52,890
with an architecture.

1182
00:56:52,890 --> 00:56:56,550
In fact, a full factorial
search of the space

1183
00:56:56,550 --> 00:57:02,370
would reveal 1,440 different
satellite architectures.

1184
00:57:02,370 --> 00:57:05,960
So that's on the input side,
what's the output vector?

1185
00:57:05,960 --> 00:57:09,720
What do you care about in
terms of output of the system?

1186
00:57:09,720 --> 00:57:12,600
Well first of all, performance.

1187
00:57:12,600 --> 00:57:16,500
So the performance-- this
is for voice communications.

1188
00:57:16,500 --> 00:57:20,310
In a sense it applies to
data communications as well.

1189
00:57:20,310 --> 00:57:23,070
It is your data rate
per channel, right?

1190
00:57:23,070 --> 00:57:27,750
4.8 Kbps, your bit
error rate, like

1191
00:57:27,750 --> 00:57:30,990
what is the average number
of bits that are wrong?

1192
00:57:30,990 --> 00:57:32,400
10 to the minus 3.

1193
00:57:32,400 --> 00:57:35,360
That's actually
pretty-- not a very--

1194
00:57:35,360 --> 00:57:37,350
that's not a very
stringent requirement.

1195
00:57:37,350 --> 00:57:40,560
And the reason is this is
for voice communication,

1196
00:57:40,560 --> 00:57:44,280
this is not for sending, you
know, commands to a spacecraft

1197
00:57:44,280 --> 00:57:45,200
going to Mars.

1198
00:57:45,200 --> 00:57:49,170
If you send commands, it would
have to be 10 to the minus 10,

1199
00:57:49,170 --> 00:57:51,060
or 10 to the minus 9.

1200
00:57:51,060 --> 00:57:53,050
Much, much better
bit error rate.

1201
00:57:53,050 --> 00:57:55,650
But this is OK for voice.

1202
00:57:55,650 --> 00:57:59,220
And then the link
fading margin, 16 DB.

1203
00:57:59,220 --> 00:58:01,570
This is the strength
of the signal, which

1204
00:58:01,570 --> 00:58:03,420
which will dictate
whether or not

1205
00:58:03,420 --> 00:58:07,300
you can use the phone under
trees or in buildings.

1206
00:58:07,300 --> 00:58:09,900
So you want this number to be
higher, but the higher it is,

1207
00:58:09,900 --> 00:58:11,490
you know given
the power you have

1208
00:58:11,490 --> 00:58:13,860
on the satellite, the
fewer channels you have.

1209
00:58:13,860 --> 00:58:15,270
So there's trade-offs.

1210
00:58:15,270 --> 00:58:19,520
So what was done here is to
keep the performance foxed,

1211
00:58:19,520 --> 00:58:22,200
and so that you can
compare architectures,

1212
00:58:22,200 --> 00:58:24,900
compare apples to apples.

1213
00:58:24,900 --> 00:58:26,790
Then we have capacity,
which is the number

1214
00:58:26,790 --> 00:58:29,490
of simultaneous duplex channels.

1215
00:58:29,490 --> 00:58:31,290
And then finally,
the lifecycle cost

1216
00:58:31,290 --> 00:58:34,560
of the system, which includes
research development, test,

1217
00:58:34,560 --> 00:58:38,760
and evaluation, satellite
construction and test, launch

1218
00:58:38,760 --> 00:58:41,540
and orbital insertion,
and then the operations

1219
00:58:41,540 --> 00:58:43,060
and replenishment.

1220
00:58:43,060 --> 00:58:45,000
So if the satellite
fails in orbit,

1221
00:58:45,000 --> 00:58:47,820
either you have to already
have prepositioned to spare,

1222
00:58:47,820 --> 00:58:50,310
or you're going to
launch a replacement.

1223
00:58:50,310 --> 00:58:53,220
And this actually
happened in both cases,

1224
00:58:53,220 --> 00:58:56,290
both constellations.

1225
00:58:56,290 --> 00:59:00,270
So in order to then, connect
the input to the output,

1226
00:59:00,270 --> 00:59:04,620
you need to build a simulator
or a model of this system.

1227
00:59:04,620 --> 00:59:07,920
And this is a high-level view
of what that model looks like.

1228
00:59:07,920 --> 00:59:11,010
So we take our input,
our design decisions,

1229
00:59:11,010 --> 00:59:14,850
and certain constants that we
assume-- a constant vector.

1230
00:59:14,850 --> 00:59:19,470
And cascade this information
from the constellation module,

1231
00:59:19,470 --> 00:59:22,560
takes the altitude and the
minimum elevation angle,

1232
00:59:22,560 --> 00:59:23,940
and produces--

1233
00:59:23,940 --> 00:59:26,250
t is the number of
satellites and p

1234
00:59:26,250 --> 00:59:28,710
is the number of
orbital planes, so this

1235
00:59:28,710 --> 00:59:30,510
is sort of orbital dynamics.

1236
00:59:30,510 --> 00:59:35,580
The spacecraft module
calculates the satellite mass,

1237
00:59:35,580 --> 00:59:40,500
the satellite network builds
essentially the communication

1238
00:59:40,500 --> 00:59:41,730
pathways.

1239
00:59:41,730 --> 00:59:46,200
The link budget will calculate
the capacity of the system.

1240
00:59:46,200 --> 00:59:48,930
You know, given those
performance constraints.

1241
00:59:48,930 --> 00:59:52,530
The launch module will
determine how many launches

1242
00:59:52,530 --> 00:59:54,330
do you need, from where.

1243
00:59:54,330 --> 00:59:56,790
And then the cost module
basically calculates

1244
00:59:56,790 --> 00:59:59,680
the total cost of the
system, the lifecycle cost,

1245
00:59:59,680 --> 01:00:03,390
and finally you get essentially
a trade-off of lifecycle cost

1246
01:00:03,390 --> 01:00:05,265
versus capacity of the system.

1247
01:00:07,996 --> 01:00:09,870
Let me just show you
some, you know, you say,

1248
01:00:09,870 --> 01:00:12,940
well that's tying together
a lot of information

1249
01:00:12,940 --> 01:00:14,520
into a multi-disciplinary model.

1250
01:00:14,520 --> 01:00:17,380
So what kind of
information do you have?

1251
01:00:17,380 --> 01:00:19,540
You have a mix, really
of physics based models--

1252
01:00:19,540 --> 01:00:22,800
so this is a very well known
equation, the eB/N0, that this

1253
01:00:22,800 --> 01:00:25,380
is the energy bit
over noise ratio.

1254
01:00:25,380 --> 01:00:28,890
This is a closed form
physics-based equation that

1255
01:00:28,890 --> 01:00:31,600
tells you how much
energy is there per bit--

1256
01:00:31,600 --> 01:00:35,010
what's the signal to noise
ratio on a per bit basis.

1257
01:00:35,010 --> 01:00:38,580
And this is a function of
transmitter power, receiver

1258
01:00:38,580 --> 01:00:42,000
and transmission gains,
various losses in the system.

1259
01:00:42,000 --> 01:00:43,740
And then some of the equations--

1260
01:00:43,740 --> 01:00:45,390
some of the information
is empirical.

1261
01:00:45,390 --> 01:00:48,300
For example, the relationship
between spacecraft

1262
01:00:48,300 --> 01:00:50,910
[? wet ?] mass
and payload power.

1263
01:00:50,910 --> 01:00:54,300
You know, if wanted to have
a closed form or you wanted

1264
01:00:54,300 --> 01:00:56,700
to have more detail, you
would have to almost build

1265
01:00:56,700 --> 01:00:58,380
a CAD model, right?

1266
01:00:58,380 --> 01:00:59,700
A separate model.

1267
01:00:59,700 --> 01:01:03,090
You'd have to-- for each
of the 1,440 architectures,

1268
01:01:03,090 --> 01:01:07,740
you'd have to manually construct
an individual detailed design.

1269
01:01:07,740 --> 01:01:10,440
And that's not feasible,
so what you do instead

1270
01:01:10,440 --> 01:01:14,280
is you use some prior
data-- and you can see here

1271
01:01:14,280 --> 01:01:16,650
a scaling relationship.

1272
01:01:16,650 --> 01:01:21,180
It's not perfect, but
we have error bars,

1273
01:01:21,180 --> 01:01:22,660
we know how good it is.

1274
01:01:22,660 --> 01:01:25,080
Satellite mass,
satellite wet mass

1275
01:01:25,080 --> 01:01:28,770
is a function of transmitter
power and, in this case,

1276
01:01:28,770 --> 01:01:31,080
propellant mass.

1277
01:01:31,080 --> 01:01:35,890
So two kinds of equations.

1278
01:01:35,890 --> 01:01:38,110
Benchmarking.

1279
01:01:38,110 --> 01:01:41,020
So once you have this model,
you need to ask the question,

1280
01:01:41,020 --> 01:01:42,950
how can I trust this model?

1281
01:01:42,950 --> 01:01:45,850
Does it give me a
reasonable answers?

1282
01:01:45,850 --> 01:01:48,340
And in this case,
benchmarking the process

1283
01:01:48,340 --> 01:01:50,620
of validating
simulation by comparing

1284
01:01:50,620 --> 01:01:53,230
the predictive response
against reality.

1285
01:01:53,230 --> 01:01:57,080
So just quickly here, showing
you four kinds of data.

1286
01:01:57,080 --> 01:01:59,860
So one is the simultaneous
channels of the constellation,

1287
01:01:59,860 --> 01:02:03,340
this is the prediction
of capacity.

1288
01:02:03,340 --> 01:02:05,140
You can see it's pretty good.

1289
01:02:05,140 --> 01:02:08,080
In this case, the model is
actually a little conservative.

1290
01:02:08,080 --> 01:02:11,530
So the blue bars are the
actual planned capacities.

1291
01:02:11,530 --> 01:02:14,560
The red ones or
magenta is simulated,

1292
01:02:14,560 --> 01:02:16,990
and you can see that the
simulation under-predicts,

1293
01:02:16,990 --> 01:02:20,450
slightly, the true
capacity of the system.

1294
01:02:20,450 --> 01:02:24,010
Life cycle costs, you saw that
Iridium was just a bit more

1295
01:02:24,010 --> 01:02:25,570
than $5 billion.

1296
01:02:25,570 --> 01:02:28,870
Globalstar was between $3
million and $4 million.

1297
01:02:28,870 --> 01:02:32,080
And so here we're in
the right ballpark.

1298
01:02:32,080 --> 01:02:34,630
And then in terms
of satellite mass

1299
01:02:34,630 --> 01:02:38,350
and the number of satellites
in the constellation required,

1300
01:02:38,350 --> 01:02:40,310
we matched that very closely.

1301
01:02:40,310 --> 01:02:42,340
And the reason for
this is, fundamentally

1302
01:02:42,340 --> 01:02:43,960
this is just geometry.

1303
01:02:43,960 --> 01:02:44,860
Right?

1304
01:02:44,860 --> 01:02:48,040
If I can tell you the altitude
and the minimum elevation

1305
01:02:48,040 --> 01:02:51,280
angle, and I'm telling you I
need I need global coverage

1306
01:02:51,280 --> 01:02:53,500
and I want dual redundancy,
so you can always

1307
01:02:53,500 --> 01:02:56,410
see at least two
satellites, it's

1308
01:02:56,410 --> 01:02:59,290
just geometry to be able to
figure out how many satellites

1309
01:02:59,290 --> 01:03:01,240
and how many orbital
planes you need.

1310
01:03:01,240 --> 01:03:03,670
And that's why the
model, or the simulation,

1311
01:03:03,670 --> 01:03:06,280
and reality match
very, very closely.

1312
01:03:06,280 --> 01:03:09,730
So this gives you
some confidence

1313
01:03:09,730 --> 01:03:11,960
that this model is reasonable.

1314
01:03:11,960 --> 01:03:15,070
So what you can
do with it is now

1315
01:03:15,070 --> 01:03:17,020
what we call Trade
Space exploration.

1316
01:03:17,020 --> 01:03:20,470
Which is, in a sense, what
you did for the [INAUDIBLE]

1317
01:03:20,470 --> 01:03:21,740
competition.

1318
01:03:21,740 --> 01:03:25,960
So this picture here, this
graph shows you the life cycle

1319
01:03:25,960 --> 01:03:28,990
costs over those
15 years, versus

1320
01:03:28,990 --> 01:03:31,060
global capacity of a system.

1321
01:03:31,060 --> 01:03:33,880
Each of these blue
dots represents one

1322
01:03:33,880 --> 01:03:36,880
of those 1,440 architectures.

1323
01:03:36,880 --> 01:03:41,080
And you can even see on
this, where Iridium actual

1324
01:03:41,080 --> 01:03:44,920
versus simulated falls and
where Globalstar actual

1325
01:03:44,920 --> 01:03:46,020
versus simulated falls.

1326
01:03:46,020 --> 01:03:49,210
So both of them were actually
off the Pareto frontier,

1327
01:03:49,210 --> 01:03:52,540
which was interesting and
led to quite some discussion.

1328
01:03:52,540 --> 01:03:55,790
And you can-- the
Pareto frontier itself

1329
01:03:55,790 --> 01:03:58,240
was, of course, very useful.

1330
01:03:58,240 --> 01:04:02,470
One of the reasons that Iridium
is not on the Pareto frontier

1331
01:04:02,470 --> 01:04:07,126
is that fundamentally, polar
constellations are inefficient.

1332
01:04:07,126 --> 01:04:09,250
And, if you think about
it, when the satellites are

1333
01:04:09,250 --> 01:04:13,190
crossing over the poles, they're
very close to each other.

1334
01:04:13,190 --> 01:04:14,670
And they're actually
not crossing

1335
01:04:14,670 --> 01:04:17,660
exactly over the poles, or you
could actually collide, right?

1336
01:04:17,660 --> 01:04:21,350
So you offset them slightly
to avoid collisions.

1337
01:04:21,350 --> 01:04:25,320
But some of the
satellites actually

1338
01:04:25,320 --> 01:04:28,170
get turned off when
they cross the poles.

1339
01:04:28,170 --> 01:04:31,740
So you're not utilizing your
assets super efficiently

1340
01:04:31,740 --> 01:04:33,490
in the polar constellation.

1341
01:04:33,490 --> 01:04:36,270
Which is one of the
reasons why they are not

1342
01:04:36,270 --> 01:04:38,130
on Pareto frontier.

1343
01:04:38,130 --> 01:04:41,670
Now, the way you would use
this in a traditional system

1344
01:04:41,670 --> 01:04:44,700
engineering approach, which is
traditional system engineering

1345
01:04:44,700 --> 01:04:47,180
approach means give
me the requirement,

1346
01:04:47,180 --> 01:04:50,595
write down the requirements,
and then find the minimum cost

1347
01:04:50,595 --> 01:04:52,110
design for that requirement.

1348
01:04:52,110 --> 01:04:57,750
So a requirement could be we
need a capacity of 50,000,

1349
01:04:57,750 --> 01:05:00,780
for example, and then you
find this intersection

1350
01:05:00,780 --> 01:05:03,690
with the Pareto
front, and that's

1351
01:05:03,690 --> 01:05:07,440
the minimum lifecycle
cost design that

1352
01:05:07,440 --> 01:05:08,580
gives you that capacity.

1353
01:05:08,580 --> 01:05:09,120
Right?

1354
01:05:09,120 --> 01:05:11,400
And that's the system
you pick, and that's

1355
01:05:11,400 --> 01:05:12,630
what you go and build.

1356
01:05:12,630 --> 01:05:14,430
And that's essentially
what they did.

1357
01:05:14,430 --> 01:05:17,460
The problem with this if
there is high uncertainty

1358
01:05:17,460 --> 01:05:21,090
is, the true requirement
is kind of unknown.

1359
01:05:21,090 --> 01:05:24,150
The market will tell
us in the future what

1360
01:05:24,150 --> 01:05:26,130
the true capacity should be.

1361
01:05:26,130 --> 01:05:28,270
So how do you deal with this?

1362
01:05:28,270 --> 01:05:30,180
Well, you could be--

1363
01:05:30,180 --> 01:05:32,180
we're going to be on
the safe side, right?

1364
01:05:35,730 --> 01:05:38,760
We're going to oversize--
so if the true demand,

1365
01:05:38,760 --> 01:05:40,610
the true capacity
that you would need,

1366
01:05:40,610 --> 01:05:42,840
is higher than
what you just had,

1367
01:05:42,840 --> 01:05:45,810
your system is going to
be under capacity, right?

1368
01:05:45,810 --> 01:05:47,430
Your system is going
to be undersized

1369
01:05:47,430 --> 01:05:48,700
for what it should be.

1370
01:05:48,700 --> 01:05:52,140
And so what happens in practice?

1371
01:05:52,140 --> 01:05:55,493
What do you think happens in
an under capacity situation?

1372
01:05:59,357 --> 01:06:00,323
What would you say?

1373
01:06:05,160 --> 01:06:09,821
The system is too small relative
to the market that you're in.

1374
01:06:13,190 --> 01:06:15,090
You're not capturing
as much as you can,

1375
01:06:15,090 --> 01:06:17,210
meaning a competitor
will probably capture.

1376
01:06:17,210 --> 01:06:19,970
I mean, I guess you can
change your price policy,

1377
01:06:19,970 --> 01:06:22,370
but fundamentally,
you're missing out

1378
01:06:22,370 --> 01:06:24,540
on market opportunity.

1379
01:06:24,540 --> 01:06:27,230
The other situation, which
is actually what happened,

1380
01:06:27,230 --> 01:06:31,220
is that the demand is much
less than we anticipated.

1381
01:06:31,220 --> 01:06:33,830
Your system is
oversized and remember,

1382
01:06:33,830 --> 01:06:36,440
once you launch a
satellite constellation,

1383
01:06:36,440 --> 01:06:38,420
it's not like a fleet
of taxis, right?

1384
01:06:38,420 --> 01:06:43,250
Where you just park
them in a parking lot.

1385
01:06:43,250 --> 01:06:47,210
You've already-- the
fixed cost is very high.

1386
01:06:47,210 --> 01:06:52,980
So if demand is below capacity,
you have all this investment

1387
01:06:52,980 --> 01:06:55,530
here, lifecycle cost has
been wasted because you

1388
01:06:55,530 --> 01:06:56,990
oversized the system.

1389
01:06:56,990 --> 01:06:59,630
And the challenge is
that the true requirement

1390
01:06:59,630 --> 01:07:02,810
is kind of like this-- there's
this probability, density

1391
01:07:02,810 --> 01:07:03,950
function.

1392
01:07:03,950 --> 01:07:04,720
Right?

1393
01:07:04,720 --> 01:07:07,600
It's a probabilistic
requirement, essentially.

1394
01:07:07,600 --> 01:07:11,120
And there's no set of
experiments because of all

1395
01:07:11,120 --> 01:07:13,100
the [INAUDIBLE] uncertainty.

1396
01:07:13,100 --> 01:07:15,170
It's not a systemic uncertainty.

1397
01:07:15,170 --> 01:07:17,090
There's no set of
experiments you

1398
01:07:17,090 --> 01:07:21,480
could do today that will help
you refine the requirements.

1399
01:07:21,480 --> 01:07:25,670
Now they did do market studies,
but again those market studies

1400
01:07:25,670 --> 01:07:28,990
were years ahead of when the
actual system was launched.

1401
01:07:28,990 --> 01:07:31,310
So they were unreliable,
fundamentally.

1402
01:07:31,310 --> 01:07:35,900
So what I'm arguing here
is in this kind of system

1403
01:07:35,900 --> 01:07:37,316
where you have
large uncertainties

1404
01:07:37,316 --> 01:07:40,520
as to the true requirements,
you shouldn't just

1405
01:07:40,520 --> 01:07:43,670
guess and then put billions
of dollars on a guess.

1406
01:07:43,670 --> 01:07:46,010
This is like playing
in the casino, right?

1407
01:07:46,010 --> 01:07:46,970
Essentially.

1408
01:07:46,970 --> 01:07:50,400
What you should do is think
about the problem differently.

1409
01:07:50,400 --> 01:07:53,540
And in this case, the
answer is staged deployment,

1410
01:07:53,540 --> 01:07:56,050
or one of the answers
is staged deployments.

1411
01:07:56,050 --> 01:07:59,690
So you build a system such
that it can adapt itself

1412
01:07:59,690 --> 01:08:01,130
to the uncertain demand.

1413
01:08:01,130 --> 01:08:04,280
So you build, initially,
a smaller, more affordable

1414
01:08:04,280 --> 01:08:04,990
system.

1415
01:08:04,990 --> 01:08:06,890
But-- and then the
system has already

1416
01:08:06,890 --> 01:08:10,160
built into it the
flexibility or scalability

1417
01:08:10,160 --> 01:08:11,510
to expand if needed.

1418
01:08:11,510 --> 01:08:13,730
But only if needed.

1419
01:08:13,730 --> 01:08:16,979
And there are two major
economic advantages to that.

1420
01:08:16,979 --> 01:08:20,060
One is that you don't need to
spend all the capital up front.

1421
01:08:20,060 --> 01:08:22,729
As so you're deferring
capital investment.

1422
01:08:22,729 --> 01:08:27,050
And you retain the decision
freedom to reconfigure or not,

1423
01:08:27,050 --> 01:08:30,170
and that's typically what
we call the real option.

1424
01:08:30,170 --> 01:08:34,490
You've created a real
option for the future.

1425
01:08:34,490 --> 01:08:38,460
So the question, then, is
well how valuable is that?

1426
01:08:38,460 --> 01:08:42,830
It is worthwhile doing this
staged deployment approach?

1427
01:08:42,830 --> 01:08:46,130
So there's probably
different ways of doing this,

1428
01:08:46,130 --> 01:08:47,660
and I just want to
share with you--

1429
01:08:47,660 --> 01:08:50,010
and this is from the
paper-- how this was done.

1430
01:08:50,010 --> 01:08:53,450
So step one is you partition
your design vector.

1431
01:08:53,450 --> 01:08:56,600
So you basically decide
which part of the system

1432
01:08:56,600 --> 01:08:58,790
is flexible-- because
you can't make everything

1433
01:08:58,790 --> 01:09:01,310
flexible typically--
what part is flexible

1434
01:09:01,310 --> 01:09:05,390
and what part is fixed, is
the base, you can't change it?

1435
01:09:05,390 --> 01:09:08,330
So in this case
usually the idea is

1436
01:09:08,330 --> 01:09:11,210
that the satellites
themselves, the design

1437
01:09:11,210 --> 01:09:13,700
of the satellite, the
transmitter power,

1438
01:09:13,700 --> 01:09:18,470
that their protocol, and
the network architecture

1439
01:09:18,470 --> 01:09:20,330
should be fixed, right?

1440
01:09:20,330 --> 01:09:23,370
Those things are
difficult to change.

1441
01:09:23,370 --> 01:09:26,810
We're going to allow the
astrodynamics, the actual shape

1442
01:09:26,810 --> 01:09:29,090
of the constellation,
to be flexible.

1443
01:09:29,090 --> 01:09:30,920
So keep the satellites
the same, only

1444
01:09:30,920 --> 01:09:33,020
change the arrangement in space.

1445
01:09:33,020 --> 01:09:35,450
So what you do is you
partition the design vector

1446
01:09:35,450 --> 01:09:40,180
into the flexible
and inflexible parts.

1447
01:09:40,180 --> 01:09:43,779
And when you do this,
in step two, what's

1448
01:09:43,779 --> 01:09:47,790
nice about this is you can
then actually find families

1449
01:09:47,790 --> 01:09:49,569
of designs, right?

1450
01:09:49,569 --> 01:09:53,569
You can find families
of designs where the--

1451
01:09:53,569 --> 01:09:55,240
we have a little lag here--

1452
01:09:55,240 --> 01:09:57,130
that you can find
families of design

1453
01:09:57,130 --> 01:09:58,750
that share the constant parts.

1454
01:09:58,750 --> 01:10:00,590
That have the common--

1455
01:10:00,590 --> 01:10:03,550
So what's shown in this
graph here-- this is again,

1456
01:10:03,550 --> 01:10:06,940
our design space, kind of
zoomed in a little bit more--

1457
01:10:06,940 --> 01:10:10,600
every one of these points that
are connected with these lines

1458
01:10:10,600 --> 01:10:14,580
uses the same types of
satellite [INAUDIBLE]..

1459
01:10:14,580 --> 01:10:17,890
Same transmitter power,
same antenna diameter,

1460
01:10:17,890 --> 01:10:20,350
and all of them have
inter-satellite links.

1461
01:10:20,350 --> 01:10:23,780
So effectively,
you're partitioning

1462
01:10:23,780 --> 01:10:27,580
the design say into subsets
that share common features,

1463
01:10:27,580 --> 01:10:31,180
and then we call this
a family of designs.

1464
01:10:31,180 --> 01:10:35,710
And the idea is to start
small, so on the left side,

1465
01:10:35,710 --> 01:10:38,410
and then you grow the
system over time right?

1466
01:10:38,410 --> 01:10:40,400
But only if needed.

1467
01:10:40,400 --> 01:10:48,820
So when you do this, you 1440-
strong design space turns out

1468
01:10:48,820 --> 01:10:52,330
to be decomposable into
40 different paths.

1469
01:10:52,330 --> 01:10:54,940
Now, which path would
you choose here?

1470
01:10:54,940 --> 01:10:58,630
Well you want to be as close
to the Pareto front possible.

1471
01:10:58,630 --> 01:11:02,670
So this is an example
of all of a path.

1472
01:11:02,670 --> 01:11:05,710
In this path, if we
started on the lower

1473
01:11:05,710 --> 01:11:08,530
we would start
with 24 satellites

1474
01:11:08,530 --> 01:11:13,090
and then we would actually
move them to a lower altitude.

1475
01:11:13,090 --> 01:11:17,320
And add 30 more in
step two, and then you

1476
01:11:17,320 --> 01:11:20,180
would gradually grow the
constellation over time.

1477
01:11:20,180 --> 01:11:20,850
You see that?

1478
01:11:20,850 --> 01:11:23,190
This is actually not
too different from how

1479
01:11:23,190 --> 01:11:24,460
GPS was done, right?

1480
01:11:24,460 --> 01:11:28,580
The GPS constellation
was deployed in phases,

1481
01:11:28,580 --> 01:11:30,610
but it was not market
driven, it was just

1482
01:11:30,610 --> 01:11:34,360
sort of phasing the
development for risk reduction,

1483
01:11:34,360 --> 01:11:37,084
and spreading off the capital.

1484
01:11:37,084 --> 01:11:38,500
The other thing
that's interesting

1485
01:11:38,500 --> 01:11:41,650
here is that as you
grow the constellation,

1486
01:11:41,650 --> 01:11:43,752
it moves further away.

1487
01:11:43,752 --> 01:11:46,650
It moves further
away from the front.

1488
01:11:46,650 --> 01:11:49,360
So when it's small, this
particular [INAUDIBLE]

1489
01:11:49,360 --> 01:11:52,060
is close to being optimal,
and as you scale it

1490
01:11:52,060 --> 01:11:54,230
up it becomes more sub-optimal.

1491
01:11:54,230 --> 01:11:57,550
And there's other paths that
have the opposite behavior;

1492
01:11:57,550 --> 01:12:00,520
where when small it's
kind of suboptimal,

1493
01:12:00,520 --> 01:12:03,880
and as you make it bigger, it
gets closer to the optimality.

1494
01:12:03,880 --> 01:12:05,150
Pretty interesting.

1495
01:12:05,150 --> 01:12:09,130
So that's step two,
find the paths.

1496
01:12:09,130 --> 01:12:13,210
Step three, you now have to
model the actual uncertainty.

1497
01:12:13,210 --> 01:12:15,400
And in this case there
are different ways

1498
01:12:15,400 --> 01:12:17,020
to model uncertainty.

1499
01:12:17,020 --> 01:12:19,850
In this case, what we
used was a GBM model--

1500
01:12:19,850 --> 01:12:21,900
geometric Brownian motion.

1501
01:12:21,900 --> 01:12:24,670
This is well-known in
physics and applied

1502
01:12:24,670 --> 01:12:27,530
to statistics, physics,
this has been applied

1503
01:12:27,530 --> 01:12:29,690
to the stock market, right?

1504
01:12:29,690 --> 01:12:32,890
And the idea here is
that in this case,

1505
01:12:32,890 --> 01:12:35,810
the man behaves like
a particle in a fluid.

1506
01:12:35,810 --> 01:12:36,310
Right?

1507
01:12:36,310 --> 01:12:40,020
That's kind of moving in
an unpredictable fashion.

1508
01:12:40,020 --> 01:12:43,210
So this is the basic
equation for-- this

1509
01:12:43,210 --> 01:12:47,200
is a discrete version of
geometric Brownian motion,

1510
01:12:47,200 --> 01:12:52,040
you have some trends
nu, nu times delta-t,

1511
01:12:52,040 --> 01:12:56,350
so this is the delta and
demand, or the change in demand,

1512
01:12:56,350 --> 01:13:00,850
divided by demand, so this is
normalized change in demand,

1513
01:13:00,850 --> 01:13:04,450
is nu times delta-t,
so your trend times

1514
01:13:04,450 --> 01:13:09,940
delta-t plus sigma times epsilon
times square root of delta-t.

1515
01:13:09,940 --> 01:13:11,980
Epsilon is a standard,
normally distributed

1516
01:13:11,980 --> 01:13:17,830
random variable between 0
and 1, and epsilon and sigma

1517
01:13:17,830 --> 01:13:20,800
is your volatility, essentially.

1518
01:13:20,800 --> 01:13:22,900
So here's some
examples, if you start

1519
01:13:22,900 --> 01:13:25,660
with an initial
demand of 50,000,

1520
01:13:25,660 --> 01:13:30,280
which is what they actually
saw as an initial demand early

1521
01:13:30,280 --> 01:13:34,430
on, you have a growth
of about 8% per year.

1522
01:13:34,430 --> 01:13:36,310
This is your trend.

1523
01:13:36,310 --> 01:13:40,030
Plus in this case,
a volatility of 40%

1524
01:13:40,030 --> 01:13:43,390
per year, which seems high
but in this kind of very

1525
01:13:43,390 --> 01:13:46,930
new technology, new system, is
actually not that far fetched.

1526
01:13:46,930 --> 01:13:50,440
You get three different-- these
are just three examples of how

1527
01:13:50,440 --> 01:13:51,730
demand might evolve.

1528
01:13:51,730 --> 01:13:53,350
These are just three scenarios.

1529
01:13:53,350 --> 01:13:57,190
So demand can go up,
it can go down, and--

1530
01:13:57,190 --> 01:14:00,610
GBM is very nice, but one
of the downsides of GBM

1531
01:14:00,610 --> 01:14:04,660
is there's infinitely many
scenarios that you can generate

1532
01:14:04,660 --> 01:14:07,090
because it's fundamentally--
even though it's

1533
01:14:07,090 --> 01:14:11,290
discretized in time, it's not
discretized and in demand.

1534
01:14:11,290 --> 01:14:15,520
So a simpler version of this,
a more discrete version,

1535
01:14:15,520 --> 01:14:18,860
is a so-called
binomial lattice model.

1536
01:14:18,860 --> 01:14:20,980
So the way this
works is again, you

1537
01:14:20,980 --> 01:14:23,660
start with some initial
demand here on the left

1538
01:14:23,660 --> 01:14:28,970
and then you discretize this
such that moving through time,

1539
01:14:28,970 --> 01:14:31,720
you can look at
different scenarios.

1540
01:14:31,720 --> 01:14:32,220
Right?

1541
01:14:32,220 --> 01:14:36,180
These scenarios could
be-- the best scenario

1542
01:14:36,180 --> 01:14:38,571
is things just keep growing.

1543
01:14:38,571 --> 01:14:39,070
Right?

1544
01:14:39,070 --> 01:14:40,900
Grow, grow, grow.

1545
01:14:40,900 --> 01:14:43,870
Here, time, the 15 years,
have been sub-divided

1546
01:14:43,870 --> 01:14:47,390
into five three-year periods.

1547
01:14:47,390 --> 01:14:50,800
So here's a sample scenario,
demand goes up, goes up,

1548
01:14:50,800 --> 01:14:52,480
and then it goes
down twice, and then

1549
01:14:52,480 --> 01:14:54,530
it goes up again in the period.

1550
01:14:54,530 --> 01:14:57,490
And so this is a
discretized random walk.

1551
01:14:57,490 --> 01:15:01,960
And you can choose the numbers,
so the sigma, the volatility,

1552
01:15:01,960 --> 01:15:05,560
the amount of the up and down
movement, and the probability,

1553
01:15:05,560 --> 01:15:08,230
p, of going up, and then
the probability of 1

1554
01:15:08,230 --> 01:15:10,960
minus p of going
down, are chosen such

1555
01:15:10,960 --> 01:15:14,870
that they're consistent
with the GBM model.

1556
01:15:14,870 --> 01:15:18,520
So this model is the
equivalent of the GBM model,

1557
01:15:18,520 --> 01:15:19,960
it's just discretized.

1558
01:15:19,960 --> 01:15:22,390
So the beauty of this
is that you can now

1559
01:15:22,390 --> 01:15:25,270
simplify this to 32
different scenarios

1560
01:15:25,270 --> 01:15:27,280
instead of infinitely many.

1561
01:15:27,280 --> 01:15:32,390
And each of those scenarios
is not equally probable.

1562
01:15:32,390 --> 01:15:35,410
There is actually probability
weighted scenarios,

1563
01:15:35,410 --> 01:15:39,700
depending on what nu and
sigma are that are underlined.

1564
01:15:39,700 --> 01:15:42,430
So now we have 40 paths, right?

1565
01:15:42,430 --> 01:15:44,800
We have 40 evolution
paths of the system

1566
01:15:44,800 --> 01:15:47,570
and we have 32 different
future scenarios.

1567
01:15:47,570 --> 01:15:50,290
So what we do now in
the next step, four,

1568
01:15:50,290 --> 01:15:53,470
is to put the two
together and calculate

1569
01:15:53,470 --> 01:15:55,780
the cost of each path.

1570
01:15:55,780 --> 01:15:58,060
Calculate the cost of
each path with respect

1571
01:15:58,060 --> 01:16:00,160
to each of the demand
scenarios, look

1572
01:16:00,160 --> 01:16:04,840
at the weighted average of all
the possible paths, and the one

1573
01:16:04,840 --> 01:16:05,530
thing--

1574
01:16:05,530 --> 01:16:07,180
the one tricky
thing about this is

1575
01:16:07,180 --> 01:16:09,510
you need to build
in a decision rule.

1576
01:16:09,510 --> 01:16:11,320
Such that, if the system--

1577
01:16:11,320 --> 01:16:14,330
if demand exceeds
capacity the system,

1578
01:16:14,330 --> 01:16:16,660
you're going to expand and
move to the next stage,

1579
01:16:16,660 --> 01:16:19,810
you're going to stage
deploy the next phase.

1580
01:16:19,810 --> 01:16:22,780
And there's many different ways
of doing this decision rule.

1581
01:16:22,780 --> 01:16:25,270
So the simplest one
was chosen here,

1582
01:16:25,270 --> 01:16:27,590
and of course costs
are discounted.

1583
01:16:27,590 --> 01:16:30,790
So let me explain to
you how this works.

1584
01:16:30,790 --> 01:16:37,070
So we start the simulation, so
the first initial deployment.

1585
01:16:37,070 --> 01:16:39,400
This is our initial stage one--

1586
01:16:39,400 --> 01:16:41,200
right-- for the constellation.

1587
01:16:41,200 --> 01:16:45,700
And then we start
operating for two years,

1588
01:16:45,700 --> 01:16:47,380
and demand in this
period goes up.

1589
01:16:47,380 --> 01:16:50,230
You see the ops cost in this
case goes slightly down,

1590
01:16:50,230 --> 01:16:51,790
this is due to the discounting.

1591
01:16:51,790 --> 01:16:53,290
This is a discounting effect.

1592
01:16:53,290 --> 01:16:56,200
We arrive at the end of
our first three year period

1593
01:16:56,200 --> 01:16:59,890
and then the question is here
is our capacity of the system.

1594
01:16:59,890 --> 01:17:01,692
What would you do
in this situation?

1595
01:17:01,692 --> 01:17:03,400
Let me ask somebody
at MIT, make sure you

1596
01:17:03,400 --> 01:17:05,920
guys are still with me.

1597
01:17:05,920 --> 01:17:10,580
So what would you do
in this situation?

1598
01:17:10,580 --> 01:17:12,796
You're now at the
end of year three.

1599
01:17:16,732 --> 01:17:18,450
Anybody?

1600
01:17:18,450 --> 01:17:20,200
AUDIENCE: We lost you
for a second halfway

1601
01:17:20,200 --> 01:17:23,680
through your statement,
could you repeat?

1602
01:17:23,680 --> 01:17:27,330
PROFESSOR: Yes, so you've
done your initial three years,

1603
01:17:27,330 --> 01:17:29,940
you've deployed your
initial constellation.

1604
01:17:29,940 --> 01:17:32,640
And you now have the
situation shown on this chart.

1605
01:17:32,640 --> 01:17:35,970
What's the right decision,
according to the decision rule?

1606
01:17:45,280 --> 01:17:46,750
What's the right decision?

1607
01:17:52,140 --> 01:17:54,900
AUDIENCE: OK, I'd
keep it the same.

1608
01:17:54,900 --> 01:17:55,400
Because--

1609
01:17:55,400 --> 01:17:56,400
PROFESSOR: Yes, exactly.

1610
01:17:56,400 --> 01:17:58,750
AUDIENCE: --it has not
exceeded demand yet.

1611
01:17:58,750 --> 01:18:00,090
But prepare to change--

1612
01:18:00,090 --> 01:18:01,590
PROFESSOR: Exactly,
you wait, right?

1613
01:18:01,590 --> 01:18:04,609
AUDIENCE: Because you have
to lower your capacity.

1614
01:18:04,609 --> 01:18:06,900
PROFESSOR: You keep it the
same, you don't do anything.

1615
01:18:06,900 --> 01:18:08,370
You just keep operating.

1616
01:18:08,370 --> 01:18:09,520
Exactly, right.

1617
01:18:09,520 --> 01:18:14,550
So you have another
three years, right?

1618
01:18:14,550 --> 01:18:17,550
During these three years,
demand keeps growing.

1619
01:18:17,550 --> 01:18:21,190
And you now, at year six,
arrive at this point here.

1620
01:18:21,190 --> 01:18:23,040
So what's the
right decision now?

1621
01:18:23,040 --> 01:18:25,248
AUDIENCE: Can I ask a
question about the decision

1622
01:18:25,248 --> 01:18:28,920
to stay the same?

1623
01:18:28,920 --> 01:18:31,920
PROFESSOR: Would
you say the same?

1624
01:18:31,920 --> 01:18:35,684
What's happening
to the demand line?

1625
01:18:35,684 --> 01:18:37,600
Veronica, what's happening
to the demand line?

1626
01:18:37,600 --> 01:18:39,391
AUDIENCE: Can I ask a
quick question first,

1627
01:18:39,391 --> 01:18:42,226
about the decision
to stay the same?

1628
01:18:42,226 --> 01:18:42,850
PROFESSOR: Yes.

1629
01:18:42,850 --> 01:18:46,060
AUDIENCE: I feel like by waiting
until we've exceeded demand

1630
01:18:46,060 --> 01:18:47,890
to choose to expand
the constellation,

1631
01:18:47,890 --> 01:18:51,310
we're introducing a lag
between demand and capability

1632
01:18:51,310 --> 01:18:53,890
that creates an inefficiency
that may actually drive users

1633
01:18:53,890 --> 01:18:54,890
away from the system.

1634
01:18:54,890 --> 01:18:56,890
And I see this kind of
oscillatory effect, where

1635
01:18:56,890 --> 01:18:59,530
you would expand, and
then people would meet,

1636
01:18:59,530 --> 01:19:01,750
and then the system
would be insufficient

1637
01:19:01,750 --> 01:19:03,820
and then people would move
to a different system.

1638
01:19:03,820 --> 01:19:06,940
And you kind of have a yo-yo
around the maximum carrying

1639
01:19:06,940 --> 01:19:07,787
capacity.

1640
01:19:07,787 --> 01:19:09,370
And that doesn't
seem efficient to me,

1641
01:19:09,370 --> 01:19:12,654
so I'm wondering if you
could speak to that.

1642
01:19:12,654 --> 01:19:13,320
PROFESSOR: Yeah.

1643
01:19:13,320 --> 01:19:15,910
AUDIENCE: So what
you're essentially

1644
01:19:15,910 --> 01:19:20,170
looking for is what I would
[INAUDIBLE] decision rule.

1645
01:19:20,170 --> 01:19:21,570
So you would actually--

1646
01:19:21,570 --> 01:19:24,030
so the decision here
is to deploy, right?

1647
01:19:24,030 --> 01:19:26,760
You've got to deploy your
second stage because you've

1648
01:19:26,760 --> 01:19:28,980
saturated the system.

1649
01:19:28,980 --> 01:19:33,520
What you're arguing for
is before this occurs,

1650
01:19:33,520 --> 01:19:36,540
you know like at 80%
saturation, that's

1651
01:19:36,540 --> 01:19:38,790
when you trigger the
next stage, right?

1652
01:19:38,790 --> 01:19:42,810
In order to anticipate
the saturation.

1653
01:19:42,810 --> 01:19:45,120
And so, absolutely,
you could do this.

1654
01:19:45,120 --> 01:19:47,550
There are many
different decision rules

1655
01:19:47,550 --> 01:19:52,470
that you can try when you
design a flexible, deployable,

1656
01:19:52,470 --> 01:19:53,410
scalable system.

1657
01:19:53,410 --> 01:19:55,320
And that's part of
the decision space.

1658
01:19:55,320 --> 01:19:58,830
So in this particular
example, we just

1659
01:19:58,830 --> 01:20:01,460
implemented the simplest
possible decision rule,

1660
01:20:01,460 --> 01:20:05,460
which was when saturation
has occurred you do deploy

1661
01:20:05,460 --> 01:20:08,050
the next stage, but not before.

1662
01:20:08,050 --> 01:20:09,646
AUDIENCE: OK, thank you.

1663
01:20:12,880 --> 01:20:15,160
PROFESSOR: Is that clear?

1664
01:20:15,160 --> 01:20:17,370
AUDIENCE: Yes, thank you.

1665
01:20:17,370 --> 01:20:20,650
PROFESSOR: OK, so we deploy
the second stage now,

1666
01:20:20,650 --> 01:20:24,240
and you can see there's another
spike here-- right-- of capital

1667
01:20:24,240 --> 01:20:25,290
that's needed.

1668
01:20:25,290 --> 01:20:27,960
It's not as high as the
first initial stage,

1669
01:20:27,960 --> 01:20:30,210
but it is substantial.

1670
01:20:30,210 --> 01:20:35,400
So now as we deploy
the next stage,

1671
01:20:35,400 --> 01:20:38,190
our capacity went
to the higher limit.

1672
01:20:38,190 --> 01:20:40,920
So we're now at
capacity level two.

1673
01:20:40,920 --> 01:20:42,640
We're now operating the system.

1674
01:20:42,640 --> 01:20:45,510
Demand keeps
growing, that's good.

1675
01:20:45,510 --> 01:20:47,800
Now we have year
nine, you can see.

1676
01:20:47,800 --> 01:20:50,410
But we haven't yet
saturated our new capacity.

1677
01:20:50,410 --> 01:20:52,860
So again, the
optimal decision is--

1678
01:20:52,860 --> 01:20:54,570
or I shouldn't say
the optimal decision,

1679
01:20:54,570 --> 01:20:56,340
but the decision
according to the rule

1680
01:20:56,340 --> 01:20:57,900
is you just wait, right?

1681
01:20:57,900 --> 01:20:59,850
You keep operating.

1682
01:20:59,850 --> 01:21:05,020
Ah, now from year nine to year
12, demand starts to go down.

1683
01:21:05,020 --> 01:21:05,520
Right?

1684
01:21:05,520 --> 01:21:07,400
And this happens in
some systems, right?

1685
01:21:07,400 --> 01:21:11,260
They are growing and they
peak, and then things go down.

1686
01:21:11,260 --> 01:21:12,900
So in this case, we wait.

1687
01:21:12,900 --> 01:21:15,690
And then in this scenario
it goes down again.

1688
01:21:15,690 --> 01:21:17,200
You see how that works?

1689
01:21:17,200 --> 01:21:21,180
So what's nice about
this is you can now--

1690
01:21:21,180 --> 01:21:26,320
in this sense, all the 32
scenarios of possible futures,

1691
01:21:26,320 --> 01:21:29,010
you can run them against
the 40 different families

1692
01:21:29,010 --> 01:21:31,590
of designs or evolution paths.

1693
01:21:31,590 --> 01:21:34,650
And out of that you can
find the best path, right?

1694
01:21:34,650 --> 01:21:38,040
This is the path, the
path that on average--

1695
01:21:38,040 --> 01:21:41,580
I should point out on average--
will satisfy your demand

1696
01:21:41,580 --> 01:21:45,270
at minimum lifecycle costs,
given the uncertainty models

1697
01:21:45,270 --> 01:21:46,588
that you have composed.

1698
01:21:46,588 --> 01:21:47,254
Yes, [INAUDIBLE]

1699
01:21:47,254 --> 01:21:55,010
AUDIENCE: [INAUDIBLE]

1700
01:21:55,010 --> 01:21:56,423
PROFESSOR: Why did it go down?

1701
01:21:56,423 --> 01:21:59,345
AUDIENCE: Like, [INAUDIBLE]

1702
01:21:59,345 --> 01:22:00,220
PROFESSOR: This here?

1703
01:22:00,220 --> 01:22:03,720
AUDIENCE: [INAUDIBLE]

1704
01:22:03,720 --> 01:22:05,880
PROFESSOR: No, the
capacity stays the same.

1705
01:22:05,880 --> 01:22:07,830
You don't go down in capacity.

1706
01:22:07,830 --> 01:22:11,590
The reason is-- this
is a good question--

1707
01:22:11,590 --> 01:22:14,910
this particular system,
a satellite constellation

1708
01:22:14,910 --> 01:22:17,130
is dominated by
fixed costs, right?

1709
01:22:17,130 --> 01:22:20,686
So if you wouldn't retire
half your satellite,

1710
01:22:20,686 --> 01:22:23,580
because then you would lose
coverage also, potentially.

1711
01:22:23,580 --> 01:22:25,910
Because you've moved
them through orbits

1712
01:22:25,910 --> 01:22:27,870
that give you the
right coverage,

1713
01:22:27,870 --> 01:22:29,790
you just increased
your capacity.

1714
01:22:29,790 --> 01:22:32,010
This is different
from a system that

1715
01:22:32,010 --> 01:22:34,170
is dominated by variable costs.

1716
01:22:34,170 --> 01:22:37,820
Like for example, if you
operate a fleet of taxis

1717
01:22:37,820 --> 01:22:41,490
and and for whatever reason,
demand goes way down,

1718
01:22:41,490 --> 01:22:43,410
you can go and park
half your taxis

1719
01:22:43,410 --> 01:22:45,240
and just not operate them.

1720
01:22:45,240 --> 01:22:47,670
You may still have to
pay a lease on them,

1721
01:22:47,670 --> 01:22:51,070
but essentially you
can downward that

1722
01:22:51,070 --> 01:22:52,920
the capacity of the system.

1723
01:22:52,920 --> 01:22:56,750
You can't do that here,
it doesn't make sense.

1724
01:22:56,750 --> 01:22:57,480
Right?

1725
01:22:57,480 --> 01:22:59,700
So that's the big
difference between systems

1726
01:22:59,700 --> 01:23:02,580
that are fundamentally
fixed cost dominated

1727
01:23:02,580 --> 01:23:04,370
versus variable cost.

1728
01:23:04,370 --> 01:23:07,830
So in this case, we just
don't use it as efficiently

1729
01:23:07,830 --> 01:23:08,910
in the last six years.

1730
01:23:13,500 --> 01:23:16,410
The answer here is that there's
an optimal path, and that's

1731
01:23:16,410 --> 01:23:22,210
shown here, that will, for
a given targeted capacity,

1732
01:23:22,210 --> 01:23:25,230
you can compare
essentially this blue path

1733
01:23:25,230 --> 01:23:27,370
against the traditional
fixed design.

1734
01:23:27,370 --> 01:23:30,915
And in this case, the
traditional architecture--

1735
01:23:30,915 --> 01:23:35,100
fixed architecture-- would
cost about $2 billion dollars

1736
01:23:35,100 --> 01:23:39,510
to build, both and on
average the red point is

1737
01:23:39,510 --> 01:23:44,800
the average lifecycle cost of
the evolvable system, is 1.36.

1738
01:23:44,800 --> 01:23:45,300
Right?

1739
01:23:45,300 --> 01:23:49,260
Lifecycle cost of the rigid
design versus the expected

1740
01:23:49,260 --> 01:23:53,580
lifecycle cost of
the best deployment

1741
01:23:53,580 --> 01:23:57,900
strategy or staged deployment
strategy, which is--

1742
01:23:57,900 --> 01:23:59,780
essentially the way
that's calculated

1743
01:23:59,780 --> 01:24:02,430
is the probability
weighted lifecycle

1744
01:24:02,430 --> 01:24:04,890
cost of each of the
scenarios, right?

1745
01:24:04,890 --> 01:24:06,330
Against this path.

1746
01:24:06,330 --> 01:24:10,520
And the difference between those
two numbers is about a third,

1747
01:24:10,520 --> 01:24:12,870
$650 million.

1748
01:24:12,870 --> 01:24:15,450
And that is the value
of the real option.

1749
01:24:15,450 --> 01:24:20,070
That's the value of designing
the system with scalability,

1750
01:24:20,070 --> 01:24:22,980
with flexibility.

1751
01:24:22,980 --> 01:24:23,726
OK?

1752
01:24:23,726 --> 01:24:24,226
Yeah?

1753
01:24:24,226 --> 01:24:37,060
AUDIENCE: [INAUDIBLE]

1754
01:24:37,060 --> 01:24:38,080
PROFESSOR: Yes.

1755
01:24:38,080 --> 01:24:39,010
That's true.

1756
01:24:39,010 --> 01:24:40,540
So the lifetime--
the question was,

1757
01:24:40,540 --> 01:24:42,670
is the life expectancy similar?

1758
01:24:42,670 --> 01:24:46,390
Of course-- so the 15 year
life of the whole system

1759
01:24:46,390 --> 01:24:48,130
is the same for both.

1760
01:24:48,130 --> 01:24:51,770
Of course, in the staged
deployment strategy,

1761
01:24:51,770 --> 01:24:54,400
some of the satellites
are going to be younger

1762
01:24:54,400 --> 01:24:55,840
at the end of the 15 years.

1763
01:24:55,840 --> 01:24:58,580
And so they may have
longer residual life left,

1764
01:24:58,580 --> 01:25:00,770
which is actually not included.

1765
01:25:00,770 --> 01:25:04,609
That's not even included
in this real option value.

1766
01:25:04,609 --> 01:25:21,242
AUDIENCE: [INAUDIBLE]

1767
01:25:21,242 --> 01:25:21,950
PROFESSOR: Right.

1768
01:25:21,950 --> 01:25:28,189
AUDIENCE: [INAUDIBLE]

1769
01:25:28,189 --> 01:25:29,480
PROFESSOR: That-- that's right.

1770
01:25:29,480 --> 01:25:33,800
So you could sort of refine
this model to include a more--

1771
01:25:33,800 --> 01:25:36,260
a staged decommissions, right?

1772
01:25:36,260 --> 01:25:39,630
Or staged transition to
a next generation system.

1773
01:25:39,630 --> 01:25:43,040
So in this case, in this
model, after 15 years--

1774
01:25:43,040 --> 01:25:45,940
boom-- you know everything,
you just finish.

1775
01:25:45,940 --> 01:25:48,410
No more revenues, you
just decommission.

1776
01:25:48,410 --> 01:25:51,560
A hard end, a hard
stop after 15 years.

1777
01:25:51,560 --> 01:25:55,010
But you could actually build
transition and decommissioning

1778
01:25:55,010 --> 01:25:56,210
models as well.

1779
01:25:56,210 --> 01:25:58,070
Good point.

1780
01:25:58,070 --> 01:26:00,890
So that's essentially
the case study

1781
01:26:00,890 --> 01:26:02,810
that I wanted to show you.

1782
01:26:02,810 --> 01:26:08,000
And, you know, you say
that's kind of like in the US

1783
01:26:08,000 --> 01:26:11,414
we talk about Monday morning
quarterbacking, you know?

1784
01:26:11,414 --> 01:26:13,580
You come in and all the
football games that happened

1785
01:26:13,580 --> 01:26:15,830
and the mistakes that
the coaches made,

1786
01:26:15,830 --> 01:26:18,230
and I would have done this
or I would have done that.

1787
01:26:18,230 --> 01:26:20,810
So yes, you know, this
is sort of like looking

1788
01:26:20,810 --> 01:26:23,420
at this problem in
hindsight, but the reality

1789
01:26:23,420 --> 01:26:29,960
is that this really has had a
big impact and a big dampening

1790
01:26:29,960 --> 01:26:32,570
effect and that the new
generation of systems--

1791
01:26:32,570 --> 01:26:34,670
I think-- are built
much more intelligently

1792
01:26:34,670 --> 01:26:37,340
with this kind of
evolution in mind.

1793
01:26:37,340 --> 01:26:42,230
I will also mention to you that
Iridium is actually the system

1794
01:26:42,230 --> 01:26:44,210
that went bankrupt first.

1795
01:26:44,210 --> 01:26:46,700
And the reason for
the bankruptcy-- well,

1796
01:26:46,700 --> 01:26:49,130
one of the immediate
reasons for the bankruptcy--

1797
01:26:49,130 --> 01:26:52,340
is because of the way that
the project was financed.

1798
01:26:52,340 --> 01:26:55,820
So about one third of the
funds-- the $5 billion--

1799
01:26:55,820 --> 01:26:58,340
one third of the
funds for Iridium

1800
01:26:58,340 --> 01:27:02,990
came from equity from Motorola.

1801
01:27:02,990 --> 01:27:05,150
So that was their own
money that they lost.

1802
01:27:05,150 --> 01:27:09,640
About one third was an IPO,
initial public offering,

1803
01:27:09,640 --> 01:27:11,830
you know, shares
sold to the public.

1804
01:27:11,830 --> 01:27:15,170
And about one third
was bank loans.

1805
01:27:15,170 --> 01:27:18,050
And these bank loans
were relatively expensive

1806
01:27:18,050 --> 01:27:20,270
and the banks
expect that they get

1807
01:27:20,270 --> 01:27:24,230
paid back at a certain speed,
depending on the market

1808
01:27:24,230 --> 01:27:25,270
evolution.

1809
01:27:25,270 --> 01:27:27,800
And it's essentially
the inability

1810
01:27:27,800 --> 01:27:30,695
to pay for and service
their loans, that

1811
01:27:30,695 --> 01:27:32,810
caused the Iridium bankruptcy.

1812
01:27:32,810 --> 01:27:35,810
And it turns out that the
loans were about one third

1813
01:27:35,810 --> 01:27:38,640
of the total capitalization
of the project.

1814
01:27:38,640 --> 01:27:40,670
So what I would argue--

1815
01:27:40,670 --> 01:27:45,340
I would argue that if they
had done this more flexible,

1816
01:27:45,340 --> 01:27:48,110
staged approach, they
could have saved one

1817
01:27:48,110 --> 01:27:50,540
third, which is about
the value of this option

1818
01:27:50,540 --> 01:27:53,840
and capitalization, and
just build the system only

1819
01:27:53,840 --> 01:27:57,260
with equity and with
the money from the IPO.

1820
01:27:57,260 --> 01:27:59,465
And that would have given
them a lot more time,

1821
01:27:59,465 --> 01:28:01,740
you know they wouldn't have
had to service those loans

1822
01:28:01,740 --> 01:28:03,890
and it would have given
them a lot more time

1823
01:28:03,890 --> 01:28:05,670
to wait for the
market to develop,

1824
01:28:05,670 --> 01:28:07,340
which it eventually did.

1825
01:28:07,340 --> 01:28:10,070
So the whole
financial architecture

1826
01:28:10,070 --> 01:28:11,770
was poorly done as well.

1827
01:28:11,770 --> 01:28:13,570
It's a whole other--

1828
01:28:13,570 --> 01:28:15,430
that's a whole other question.

1829
01:28:15,430 --> 01:28:16,930
AUDIENCE: Professor,
quick question?

1830
01:28:16,930 --> 01:28:17,596
PROFESSOR: Yeah.

1831
01:28:17,596 --> 01:28:36,280
AUDIENCE: Are there
situations where [INAUDIBLE]

1832
01:28:36,280 --> 01:28:37,170
PROFESSOR: Yes.

1833
01:28:37,170 --> 01:28:39,003
So the question, I'm
just going to repeat it

1834
01:28:39,003 --> 01:28:40,720
for you guys at MIT,
the question was

1835
01:28:40,720 --> 01:28:43,540
are there cases where the
flexible approach is not

1836
01:28:43,540 --> 01:28:44,500
the best?

1837
01:28:44,500 --> 01:28:45,620
And the answer is yes.

1838
01:28:45,620 --> 01:28:47,260
So for example, if you're in--

1839
01:28:47,260 --> 01:28:50,950
I know that you're studying
the energy market--

1840
01:28:50,950 --> 01:28:54,250
if you're in a situation
where you have guarantees,

1841
01:28:54,250 --> 01:28:57,520
a guaranteed customer,
so the government

1842
01:28:57,520 --> 01:29:00,040
or one of the big
utilities is entering

1843
01:29:00,040 --> 01:29:02,290
into a long-term
agreement with you,

1844
01:29:02,290 --> 01:29:04,150
that they will buy
the power that you

1845
01:29:04,150 --> 01:29:09,430
produce at a certain price
for the next 20 years.

1846
01:29:09,430 --> 01:29:12,390
And so you're not subject
to that market volatility,

1847
01:29:12,390 --> 01:29:16,570
but you have a long term
agreement, including clauses

1848
01:29:16,570 --> 01:29:19,900
that say well if the
government doesn't do this then

1849
01:29:19,900 --> 01:29:22,630
they'll compensate
you for any losses,

1850
01:29:22,630 --> 01:29:24,560
this is a financial contraction.

1851
01:29:24,560 --> 01:29:28,720
If you basically have eliminated
demand uncertainty, because

1852
01:29:28,720 --> 01:29:31,042
of the particular
contractual arrangements,

1853
01:29:31,042 --> 01:29:33,250
then there's no reason you
shouldn't build the system

1854
01:29:33,250 --> 01:29:35,560
for a fixed capacity.

1855
01:29:35,560 --> 01:29:41,440
And then you can build something
right on the pareto front,

1856
01:29:41,440 --> 01:29:43,150
because that
particular uncertainty

1857
01:29:43,150 --> 01:29:47,500
has been eliminated for
policy or contractual reasons.

1858
01:29:47,500 --> 01:29:49,880
So, good question.

1859
01:29:49,880 --> 01:29:52,730
OK, so we're almost there.

1860
01:29:52,730 --> 01:29:54,770
Let me just try to
summarize some--

1861
01:29:54,770 --> 01:29:57,940
AUDIENCE: Can I ask you
question on that last slide?

1862
01:29:57,940 --> 01:29:59,950
PROFESSOR: Please, go ahead.

1863
01:29:59,950 --> 01:30:02,290
AUDIENCE: On the
slide before, 37.

1864
01:30:02,290 --> 01:30:07,120
It seemed like the
probability weighting

1865
01:30:07,120 --> 01:30:10,570
was for a general capacity.

1866
01:30:10,570 --> 01:30:14,380
Like something that could be
adapted over the life cycle

1867
01:30:14,380 --> 01:30:18,220
but here we're pinning
down a specific capacity.

1868
01:30:18,220 --> 01:30:20,170
So why is it the sum
of probabilities, then?

1869
01:30:20,170 --> 01:30:21,836
Haven't, kind of, the
probabilities been

1870
01:30:21,836 --> 01:30:25,500
realized, don't you know
exactly what your capacity is

1871
01:30:25,500 --> 01:30:28,170
that you're operating at?

1872
01:30:28,170 --> 01:30:30,910
PROFESSOR: Yes, so just
to be clear on this.

1873
01:30:30,910 --> 01:30:34,430
So the probability,
those p's of i's,

1874
01:30:34,430 --> 01:30:37,410
they come from
the lattice model.

1875
01:30:37,410 --> 01:30:39,650
They are determined
by the lattice model.

1876
01:30:39,650 --> 01:30:43,960
So what you do is for those
32 scenarios of the future,

1877
01:30:43,960 --> 01:30:45,400
this is the pi, right?

1878
01:30:45,400 --> 01:30:49,270
The i is the index, is the--

1879
01:30:49,270 --> 01:30:51,490
for this particular
future scenario,

1880
01:30:51,490 --> 01:30:55,570
for each future
scenario, you always

1881
01:30:55,570 --> 01:30:58,720
start with a1, which is
your initial configuration

1882
01:30:58,720 --> 01:31:00,160
of the constellation.

1883
01:31:00,160 --> 01:31:02,700
You always start with a1.

1884
01:31:02,700 --> 01:31:05,290
And for each of those
future scenarios,

1885
01:31:05,290 --> 01:31:11,770
you then simulate that
future and for those futures,

1886
01:31:11,770 --> 01:31:14,630
where the demand doesn't
ever really take off

1887
01:31:14,630 --> 01:31:16,300
and materializes,
you never actually

1888
01:31:16,300 --> 01:31:18,500
trigger that next
expansion stage.

1889
01:31:18,500 --> 01:31:21,640
And it's actually that asymmetry
that gives you the advantage.

1890
01:31:21,640 --> 01:31:25,180
And it's essentially
not deploying capacity

1891
01:31:25,180 --> 01:31:26,660
when you really don't need it.

1892
01:31:26,660 --> 01:31:30,100
And so you do this
for all of this n

1893
01:31:30,100 --> 01:31:34,630
scenarios for the future,
and the pi, the probability,

1894
01:31:34,630 --> 01:31:36,400
weighting of each
future scenario

1895
01:31:36,400 --> 01:31:38,830
comes directly out
of the lattice model.

1896
01:31:38,830 --> 01:31:41,755
That's not something that you
have to manually determine.

1897
01:31:41,755 --> 01:31:43,380
AUDIENCE: Right, no,
I understand that.

1898
01:31:43,380 --> 01:31:46,720
But here, since we've kind of
pinned down a capacity already,

1899
01:31:46,720 --> 01:31:48,670
we know what endpoint we're at.

1900
01:31:48,670 --> 01:31:51,490
I'm just wondering how the
probabilities come into that,

1901
01:31:51,490 --> 01:31:54,100
or is it kind of working
backwards now, of all the paths

1902
01:31:54,100 --> 01:31:56,920
that go there on
this specific plot?

1903
01:31:56,920 --> 01:31:58,420
PROFESSOR: I see
what you're saying.

1904
01:31:58,420 --> 01:31:59,020
OK.

1905
01:31:59,020 --> 01:32:03,490
So the last point, the
a1 point, so the a1 point

1906
01:32:03,490 --> 01:32:06,490
is driven by-- you need
to make some assumptions

1907
01:32:06,490 --> 01:32:12,130
about initial capacity and
then the endpoint, the a4,

1908
01:32:12,130 --> 01:32:17,010
is driven by how large
you made the trade space.

1909
01:32:17,010 --> 01:32:19,080
So you could make
the trade space,

1910
01:32:19,080 --> 01:32:21,540
you could make the design
space even larger, right?

1911
01:32:21,540 --> 01:32:26,130
This is a discretized full
factorial design space,

1912
01:32:26,130 --> 01:32:28,020
you could make it even larger.

1913
01:32:28,020 --> 01:32:30,060
And then a4 would
not be the end form.

1914
01:32:30,060 --> 01:32:32,330
This is just sort of
a finite size effect,

1915
01:32:32,330 --> 01:32:34,994
given to the size
of the trade space.

1916
01:32:34,994 --> 01:32:36,160
AUDIENCE: All right, thanks.

1917
01:32:40,790 --> 01:32:47,210
PROFESSOR: All right, so the way
I want to do the summary here

1918
01:32:47,210 --> 01:32:49,280
is just go back to the
learning objectives.

1919
01:32:49,280 --> 01:32:51,590
You know, I always do this
in every class, you know,

1920
01:32:51,590 --> 01:32:53,150
let's close the loop.

1921
01:32:53,150 --> 01:32:56,570
What did I promise
you in lecture one

1922
01:32:56,570 --> 01:32:58,250
that you would
learn in this class?

1923
01:32:58,250 --> 01:33:01,150
And, you know,
each of you, you're

1924
01:33:01,150 --> 01:33:03,370
going to have to decide
for yourself, you know,

1925
01:33:03,370 --> 01:33:05,510
did I actually learn this?

1926
01:33:05,510 --> 01:33:06,510
Did we actually do this?

1927
01:33:06,510 --> 01:33:10,270
So this is sort of due diligence
on the learning objectives.

1928
01:33:10,270 --> 01:33:11,940
And I have one slide
on each of them.

1929
01:33:11,940 --> 01:33:14,010
So the first objective was to--

1930
01:33:14,010 --> 01:33:15,860
for you to really
understand the system

1931
01:33:15,860 --> 01:33:18,210
engineering standards
and best practices,

1932
01:33:18,210 --> 01:33:20,870
as well as more approaches.

1933
01:33:20,870 --> 01:33:24,350
Number two is understand the key
steps in the system engineering

1934
01:33:24,350 --> 01:33:25,640
process.

1935
01:33:25,640 --> 01:33:28,460
Number three was
analyze and understand

1936
01:33:28,460 --> 01:33:32,060
the role of humans in the
system as beneficiaries,

1937
01:33:32,060 --> 01:33:34,790
designers, operators,
and so forth.

1938
01:33:34,790 --> 01:33:38,150
Number four was being
honest and characterizing

1939
01:33:38,150 --> 01:33:41,480
the limitations of system
engineering as we practice it

1940
01:33:41,480 --> 01:33:42,620
today.

1941
01:33:42,620 --> 01:33:46,310
And then objectified
learning and objectifiable

1942
01:33:46,310 --> 01:33:49,550
was applying these methods
and tools to a real--

1943
01:33:49,550 --> 01:33:54,830
even if not so complex-- cyber
electromechanical system.

1944
01:33:54,830 --> 01:33:59,720
So I see one,
essentially, hopefully--

1945
01:33:59,720 --> 01:34:02,000
no, we didn't check
the readings but I'm

1946
01:34:02,000 --> 01:34:04,310
hoping that you did your
readings, that you really

1947
01:34:04,310 --> 01:34:08,240
feel like you understand
the NASA System Engineering

1948
01:34:08,240 --> 01:34:11,930
Handbook, which is our
quasi-textbook for this class.

1949
01:34:11,930 --> 01:34:14,580
But there's other
standards as well,

1950
01:34:14,580 --> 01:34:17,270
such as the
[INAUDIBLE] handbook,

1951
01:34:17,270 --> 01:34:19,970
which is actually the
basis for certification.

1952
01:34:19,970 --> 01:34:21,210
So several of you--

1953
01:34:21,210 --> 01:34:23,000
your [? EPS ?] fellow
mentioned to me

1954
01:34:23,000 --> 01:34:25,160
that you're interested
in certification,

1955
01:34:25,160 --> 01:34:27,230
so I encourage that.

1956
01:34:27,230 --> 01:34:30,710
Most of that is based
on the number two.

1957
01:34:30,710 --> 01:34:33,980
I did mention the
ISO standard 15288.

1958
01:34:33,980 --> 01:34:37,680
That's probably the best known
standard; ISO is a very--

1959
01:34:37,680 --> 01:34:39,750
they're located
right here in Geneva.

1960
01:34:39,750 --> 01:34:42,980
ISO is a very important
organization, you know?

1961
01:34:42,980 --> 01:34:46,790
I know ISO standards are not the
most exciting things to read,

1962
01:34:46,790 --> 01:34:51,190
but they have huge
impact and so 15288

1963
01:34:51,190 --> 01:34:57,440
is something that
worldwide-known and across all

1964
01:34:57,440 --> 01:34:58,850
industries really.

1965
01:34:58,850 --> 01:35:02,340
And then I want to mention
also the European Systems

1966
01:35:02,340 --> 01:35:03,650
Engineering Standard.

1967
01:35:03,650 --> 01:35:05,420
And Foster Voelker
and I had a discussion

1968
01:35:05,420 --> 01:35:07,200
about this earlier today.

1969
01:35:07,200 --> 01:35:11,000
This is issued by the
European Space Agency.

1970
01:35:11,000 --> 01:35:14,390
And it is a little-- there
are some subtle differences

1971
01:35:14,390 --> 01:35:17,570
between ISO approach
and the NASA approach.

1972
01:35:17,570 --> 01:35:21,140
And so for those of you on
this side of the Atlantic,

1973
01:35:21,140 --> 01:35:23,640
I do encourage you to take
a look at that [? ESES ?]

1974
01:35:23,640 --> 01:35:25,410
standard as well.

1975
01:35:25,410 --> 01:35:29,420
And then we augmented this with
different papers, and readings,

1976
01:35:29,420 --> 01:35:30,900
and so forth.

1977
01:35:30,900 --> 01:35:33,950
So I want to do a very quick--

1978
01:35:33,950 --> 01:35:37,260
this is our last concert
question for the class.

1979
01:35:37,260 --> 01:35:47,530
So here's the link,
[INAUDIBLE] and it's

1980
01:35:47,530 --> 01:35:50,560
essentially I want to
get some feedback what

1981
01:35:50,560 --> 01:35:51,880
you think about the standard.

1982
01:35:51,880 --> 01:35:55,540
All outdated and surpassed
by the digital revolution,

1983
01:35:55,540 --> 01:35:58,790
one of you thinks that's true.

1984
01:35:58,790 --> 01:36:05,250
90% of you think they're useful
codification and best practice.

1985
01:36:05,250 --> 01:36:08,260
14% think they can
be dangerous if you

1986
01:36:08,260 --> 01:36:11,920
adhere to them too closely.

1987
01:36:11,920 --> 01:36:14,650
About 60% of you think
they're essential reading

1988
01:36:14,650 --> 01:36:16,700
for any serious system engineer.

1989
01:36:16,700 --> 01:36:20,710
And 10% said that you
would never use them

1990
01:36:20,710 --> 01:36:22,150
as a daily reference book.

1991
01:36:22,150 --> 01:36:24,010
VOELKER: OK, so that's good.

1992
01:36:24,010 --> 01:36:26,050
I think that's-- what
do you think, Voelker,

1993
01:36:26,050 --> 01:36:28,040
pretty reasonable outcomes?

1994
01:36:28,040 --> 01:36:29,500
OK.

1995
01:36:29,500 --> 01:36:32,290
But what is true is
that they're going

1996
01:36:32,290 --> 01:36:34,120
to be referenced, in contracts.

1997
01:36:34,120 --> 01:36:37,540
You know if-- if you're doing--

1998
01:36:37,540 --> 01:36:41,560
you know, it depends, again,
on the industry that you're in.

1999
01:36:41,560 --> 01:36:43,120
But you're going
to adhere to some

2000
01:36:43,120 --> 01:36:45,940
of these things in contracts,
and if things don't go well,

2001
01:36:45,940 --> 01:36:48,760
people will actually check
whether, did you do this step,

2002
01:36:48,760 --> 01:36:50,330
did you do that step.

2003
01:36:50,330 --> 01:36:52,626
So it does have
real consequences.

2004
01:36:52,626 --> 01:36:53,500
PROFESSOR: OK, great.

2005
01:36:53,500 --> 01:36:54,500
Thank you very much.

2006
01:36:54,500 --> 01:36:57,170
That was that was good.

2007
01:36:57,170 --> 01:37:00,790
Step two, you know the
key steps, v model,

2008
01:37:00,790 --> 01:37:03,280
I think I'm not going
to explain it again.

2009
01:37:03,280 --> 01:37:05,260
You know, hopefully
this is something

2010
01:37:05,260 --> 01:37:06,610
that you will not forget.

2011
01:37:06,610 --> 01:37:09,070
And just one point
about the v model

2012
01:37:09,070 --> 01:37:13,380
is it does not imply that
system engineering is always

2013
01:37:13,380 --> 01:37:14,260
sequential.

2014
01:37:14,260 --> 01:37:17,980
You know, it's possible
to iterate between steps

2015
01:37:17,980 --> 01:37:21,490
and across the whole v as well.

2016
01:37:21,490 --> 01:37:24,830
SE3, stakeholders
and value of network.

2017
01:37:24,830 --> 01:37:27,760
So this is the role of humans.

2018
01:37:27,760 --> 01:37:30,820
We talked about this very
early on, several of you

2019
01:37:30,820 --> 01:37:33,880
talked about it during
the oral exam here today.

2020
01:37:33,880 --> 01:37:37,270
You know, the hub and spoke
versus the stakeholder value

2021
01:37:37,270 --> 01:37:38,290
network.

2022
01:37:38,290 --> 01:37:41,650
The method underlying
this is critical, but I do

2023
01:37:41,650 --> 01:37:43,970
want to mention, very quickly--

2024
01:37:43,970 --> 01:37:46,570
and this is something we
didn't spend a lot of time on--

2025
01:37:46,570 --> 01:37:47,690
is human factors.

2026
01:37:47,690 --> 01:37:48,200
Right?

2027
01:37:48,200 --> 01:37:51,400
How do you design systems so
that humans can effectively

2028
01:37:51,400 --> 01:37:52,750
and safely use them?

2029
01:37:52,750 --> 01:37:56,890
You know, interfaces,
procedures, and so this

2030
01:37:56,890 --> 01:37:59,460
is a couple of slides from
one of my colleagues, Missy

2031
01:37:59,460 --> 01:38:00,430
Cummings.

2032
01:38:00,430 --> 01:38:01,900
This is actually some--

2033
01:38:01,900 --> 01:38:03,030
in a nuclear plant--

2034
01:38:03,030 --> 01:38:04,900
Katya, you're going
to like this--

2035
01:38:04,900 --> 01:38:07,450
lots of these dials,
and this is actually

2036
01:38:07,450 --> 01:38:08,720
a Russian nuclear plant.

2037
01:38:08,720 --> 01:38:12,280
So it has it has a
very specific layout.

2038
01:38:12,280 --> 01:38:14,710
And so the important
questions in human factors

2039
01:38:14,710 --> 01:38:17,290
are, how do you best
display status information?

2040
01:38:17,290 --> 01:38:19,880
What tasks do humans do?

2041
01:38:19,880 --> 01:38:21,280
What level of automation?

2042
01:38:21,280 --> 01:38:23,230
What are the training
requirements?

2043
01:38:23,230 --> 01:38:26,110
And these are also very
important questions.

2044
01:38:26,110 --> 01:38:29,650
And so you can apply the
human systems engineering

2045
01:38:29,650 --> 01:38:32,860
process, very much
analogous to what

2046
01:38:32,860 --> 01:38:34,630
we covered in class so far.

2047
01:38:34,630 --> 01:38:39,010
So you do mission and scenario
analysis, function allocation.

2048
01:38:39,010 --> 01:38:42,970
And then in human factors, you
talk more about task analysis.

2049
01:38:42,970 --> 01:38:46,720
But essentially, eventually,
leads to system design.

2050
01:38:46,720 --> 01:38:49,090
Like what are the
buttons that you

2051
01:38:49,090 --> 01:38:52,390
push, what does the layout
look like, or the user

2052
01:38:52,390 --> 01:38:53,780
interface, and so forth.

2053
01:38:53,780 --> 01:38:56,770
So, you know, I'm not I'm
not presenting human factors

2054
01:38:56,770 --> 01:38:59,980
as a separate topic, but the
human factors requirements

2055
01:38:59,980 --> 01:39:03,460
should be built in
right from the start.

2056
01:39:03,460 --> 01:39:06,730
And essentially the
functional allocation

2057
01:39:06,730 --> 01:39:08,770
is sort of a key question here.

2058
01:39:08,770 --> 01:39:11,950
How much is done by automation,
how much is done by humans,

2059
01:39:11,950 --> 01:39:13,910
and how do you split
between the two?

2060
01:39:13,910 --> 01:39:16,240
And one of the most important
things to consider here

2061
01:39:16,240 --> 01:39:18,600
is this human performance curve.

2062
01:39:18,600 --> 01:39:20,440
And we know this
pretty well now,

2063
01:39:20,440 --> 01:39:23,440
that when humans are
extremely highly loaded--

2064
01:39:23,440 --> 01:39:26,920
you know, when you
have a huge workload--

2065
01:39:26,920 --> 01:39:28,870
your performance goes down.

2066
01:39:28,870 --> 01:39:32,110
But interestingly, if
you're under challenge

2067
01:39:32,110 --> 01:39:34,240
your performance
goes down as well.

2068
01:39:34,240 --> 01:39:36,190
And you know, so
for example, people

2069
01:39:36,190 --> 01:39:40,180
who are in power stations
or mission control

2070
01:39:40,180 --> 01:39:44,110
centers where nothing is
happening for days and weeks,

2071
01:39:44,110 --> 01:39:46,240
you know their
attention goes down,

2072
01:39:46,240 --> 01:39:48,130
and they don't perform
at a very high level.

2073
01:39:48,130 --> 01:39:50,850
Humans perform best at
a moderate workload,

2074
01:39:50,850 --> 01:39:52,450
and this is well-known.

2075
01:39:52,450 --> 01:39:54,460
And it should influence
how you define

2076
01:39:54,460 --> 01:39:57,520
the human interface and
the split among automation

2077
01:39:57,520 --> 01:39:58,670
and humans.

2078
01:39:58,670 --> 01:40:01,660
SE4, system complexity.

2079
01:40:01,660 --> 01:40:03,730
Last time I will
mention the magic number

2080
01:40:03,730 --> 01:40:06,100
seven, plus or minus two.

2081
01:40:06,100 --> 01:40:09,940
The real limitations come in
and when we operate at levels,

2082
01:40:09,940 --> 01:40:12,340
you know, three,
four, five, and six.

2083
01:40:12,340 --> 01:40:15,520
And the main reason for this
is because now a single human

2084
01:40:15,520 --> 01:40:17,890
cannot understand,
cannot remember,

2085
01:40:17,890 --> 01:40:20,530
cannot deal with all
the detailed components.

2086
01:40:20,530 --> 01:40:22,840
And you need to split
the work among teams,

2087
01:40:22,840 --> 01:40:26,350
among organizations, and
creates extra interfaces

2088
01:40:26,350 --> 01:40:28,280
and complexities.

2089
01:40:28,280 --> 01:40:29,500
OK?

2090
01:40:29,500 --> 01:40:33,580
And just to show you, Iridium,
we talked about Iridium.

2091
01:40:33,580 --> 01:40:37,120
This is a very recent news
story this is from October 29th.

2092
01:40:37,120 --> 01:40:40,240
The new Iridium Next has
again been pushed back

2093
01:40:40,240 --> 01:40:41,710
by four months.

2094
01:40:41,710 --> 01:40:47,770
This gentleman here is the CEO
of Iridium, chief executive

2095
01:40:47,770 --> 01:40:50,950
for Iridium Next, Mr. Desh.

2096
01:40:50,950 --> 01:40:55,300
And he's talking here about
a particular [INAUDIBLE]

2097
01:40:55,300 --> 01:40:58,870
and then Viasat is one
of the contractors.

2098
01:40:58,870 --> 01:41:02,470
There's a particular component
that's been giving trouble,

2099
01:41:02,470 --> 01:41:05,860
and that's sort of in level
three or four in the hierarchy,

2100
01:41:05,860 --> 01:41:08,200
and that's essentially
delaying the whole system.

2101
01:41:08,200 --> 01:41:13,560
So there's an example of how the
large complexity of the system

2102
01:41:13,560 --> 01:41:16,522
is causing issues.

2103
01:41:16,522 --> 01:41:21,680
AUDIENCE: [INAUDIBLE]

2104
01:41:21,680 --> 01:41:23,990
PROFESSOR: But they're
actually a supplier

2105
01:41:23,990 --> 01:41:29,020
to [INAUDIBLE] through
the whole chain, right?

2106
01:41:29,020 --> 01:41:32,540
Oh you're saying there's more
going on than meets the eye.

2107
01:41:32,540 --> 01:41:34,580
Yeah, who knows?

2108
01:41:34,580 --> 01:41:36,010
Who knows?

2109
01:41:36,010 --> 01:41:39,920
OK, and then finally here I
want to mention application

2110
01:41:39,920 --> 01:41:41,150
to a case study.

2111
01:41:41,150 --> 01:41:44,390
You know, we used the
CanSat 2016 competition

2112
01:41:44,390 --> 01:41:47,780
as a quote, unquote
safe case study.

2113
01:41:47,780 --> 01:41:50,960
And I have to say I'm
really pleased with how

2114
01:41:50,960 --> 01:41:52,130
this worked out.

2115
01:41:52,130 --> 01:41:55,490
You had the 47
requirements as a starter,

2116
01:41:55,490 --> 01:41:58,130
you approved, them
you group them,

2117
01:41:58,130 --> 01:42:01,520
and my comments about the
PDR that we had a week ago

2118
01:42:01,520 --> 01:42:05,570
are all teams passed
successfully, the PDR.

2119
01:42:05,570 --> 01:42:07,310
You know, if this
was a real PDR,

2120
01:42:07,310 --> 01:42:09,630
we would have issued a couple
of [? ridds, ?] I think.

2121
01:42:09,630 --> 01:42:12,560
There was some-- a couple
of teams were over budget,

2122
01:42:12,560 --> 01:42:16,370
or needed to work out their
aerodynamics in more detail.

2123
01:42:16,370 --> 01:42:19,370
But by and large I thought this
was an excellent application

2124
01:42:19,370 --> 01:42:21,170
of system engineering concepts.

2125
01:42:21,170 --> 01:42:23,745
Went beyond my expectations.

2126
01:42:23,745 --> 01:42:25,370
Several of you
mentioned the importance

2127
01:42:25,370 --> 01:42:27,680
of concept generation,
you know hybrid use

2128
01:42:27,680 --> 01:42:31,370
of structured and unstructured
creativity techniques.

2129
01:42:31,370 --> 01:42:34,010
I know at least one team
applied to actually go

2130
01:42:34,010 --> 01:42:35,520
to the competition.

2131
01:42:35,520 --> 01:42:37,410
And this shows you
a couple examples.

2132
01:42:37,410 --> 01:42:39,290
So we have one team
here at the EPSL

2133
01:42:39,290 --> 01:42:41,420
using a bio-inspired design.

2134
01:42:41,420 --> 01:42:46,070
This is an actual seed airfoil
that actually occurs in nature.

2135
01:42:46,070 --> 01:42:51,230
And then here's an example from
MIT, the Rogallo Wing Team 7.

2136
01:42:51,230 --> 01:42:54,680
So, I'm sorry I didn't
mention all the teams here,

2137
01:42:54,680 --> 01:42:58,940
but it was really great to
see the application of this

2138
01:42:58,940 --> 01:43:02,600
in the CanSat competition case.

2139
01:43:02,600 --> 01:43:05,510
So the last thing I want to
do-- and I only have one minute

2140
01:43:05,510 --> 01:43:08,180
left, actually I'm already
slightly over time--

2141
01:43:08,180 --> 01:43:11,090
but I just want to give
you some career and study

2142
01:43:11,090 --> 01:43:12,020
recommendations.

2143
01:43:12,020 --> 01:43:15,760
So first of all, you know, I
want to make sure that you--

2144
01:43:15,760 --> 01:43:18,440
and I said this in
the first lecture--

2145
01:43:18,440 --> 01:43:22,100
there's a lot more to the
system engineering and systems

2146
01:43:22,100 --> 01:43:24,650
research than we were able
to cover in this class.

2147
01:43:24,650 --> 01:43:27,830
So this class is really what I
call a door opener to the world

2148
01:43:27,830 --> 01:43:28,850
of system engineering.

2149
01:43:28,850 --> 01:43:32,630
If you want to go deeper,
there are deeper subjects.

2150
01:43:32,630 --> 01:43:36,790
You know, model based system
engineering is the big trend.

2151
01:43:36,790 --> 01:43:38,132
System safety.

2152
01:43:38,132 --> 01:43:39,590
And then in the
spring I'm teaching

2153
01:43:39,590 --> 01:43:42,350
a class called
Multi-Disciplinary System

2154
01:43:42,350 --> 01:43:43,630
Optimization.

2155
01:43:43,630 --> 01:43:47,340
And there is actually a
WebEx access to it as well.

2156
01:43:47,340 --> 01:43:49,240
So it's not going to
be officially offered

2157
01:43:49,240 --> 01:43:52,070
as a [INAUDIBLE] class, but if
individual students here are

2158
01:43:52,070 --> 01:43:54,050
interested, please contact me.

2159
01:43:54,050 --> 01:43:56,445
Self-study, you know the
System Engineering Journal,

2160
01:43:56,445 --> 01:43:59,390
there's IEEE journals,
there's also--

2161
01:43:59,390 --> 01:44:02,300
you know, for some of you this
is maybe a little too soon--

2162
01:44:02,300 --> 01:44:06,530
but there's professional
degrees in system engineering.

2163
01:44:06,530 --> 01:44:09,680
For example at MIT we have
the SDN program, System Design

2164
01:44:09,680 --> 01:44:10,460
and Management.

2165
01:44:10,460 --> 01:44:15,530
The average age of the
students in that program is 33,

2166
01:44:15,530 --> 01:44:17,120
but it is--

2167
01:44:17,120 --> 01:44:20,660
they're sort of coming back
to get their masters in system

2168
01:44:20,660 --> 01:44:22,970
design and management.

2169
01:44:22,970 --> 01:44:25,470
And there's quite a bit
of SLOAN content as well.

2170
01:44:25,470 --> 01:44:27,860
So finance, you know,
understanding the financial--

2171
01:44:27,860 --> 01:44:30,800
how to build a business
case around systems as well.

2172
01:44:30,800 --> 01:44:32,030
Professional experience.

2173
01:44:32,030 --> 01:44:34,370
You know, getting experience
in actual projects.

2174
01:44:34,370 --> 01:44:36,740
Like at MIT, for example Rexus.

2175
01:44:36,740 --> 01:44:38,870
You know, you've got the
Clean Space One project,

2176
01:44:38,870 --> 01:44:41,810
we heard about [? Optaneous ?]
One, Solar Impulse,

2177
01:44:41,810 --> 01:44:43,430
there are a lot
of opportunities.

2178
01:44:43,430 --> 01:44:46,490
Or-- we also had
a couple of people

2179
01:44:46,490 --> 01:44:49,210
here mention that you're
starting your own company,

2180
01:44:49,210 --> 01:44:50,060
your own venture.

2181
01:44:50,060 --> 01:44:52,220
And you know, when
you're doing, that you

2182
01:44:52,220 --> 01:44:53,750
have to be the system engineer.

2183
01:44:53,750 --> 01:44:56,570
You have to understand all
of these things, interfaces,

2184
01:44:56,570 --> 01:44:59,510
suppliers, requirements,
markets, all

2185
01:44:59,510 --> 01:45:01,400
that has to be integrated.

2186
01:45:01,400 --> 01:45:03,940
Finally, you did
hear about INCOSE.

2187
01:45:03,940 --> 01:45:07,280
We had a quick dial-in
with at an earlier session.

2188
01:45:07,280 --> 01:45:09,440
So if you're interested,
you can join either

2189
01:45:09,440 --> 01:45:11,570
as a student or a
professional member.

2190
01:45:11,570 --> 01:45:13,460
And also certification.

2191
01:45:13,460 --> 01:45:22,460
This class was not
[INAUDIBLE] for that,

2192
01:45:22,460 --> 01:45:25,460
but if you're interested
in this site, [INAUDIBLE]

2193
01:45:25,460 --> 01:45:28,440
number of years of experience
at the [? CCF ?] level.

2194
01:45:28,440 --> 01:45:30,980
And then finally,
please keep in touch.

2195
01:45:30,980 --> 01:45:33,110
And last but not
least, I want to thank

2196
01:45:33,110 --> 01:45:40,330
all of you, the students at MIT,
at EPFL, our TAs, [INAUDIBLE]

2197
01:45:40,330 --> 01:45:44,570
here at EPFL, Johana at
MIT, the technical staff

2198
01:45:44,570 --> 01:45:50,780
has been helping run the
technology, Voelker, thank you.

2199
01:45:50,780 --> 01:45:52,860
And-- and that's it.

2200
01:45:52,860 --> 01:45:54,560
So I want to thank all of you.

2201
01:45:54,560 --> 01:45:56,840
Next week we do have
a voluntary seminar

2202
01:45:56,840 --> 01:45:59,750
on sort of future
trends in manufacturing,

2203
01:45:59,750 --> 01:46:01,890
but again, it's not mandatory.

2204
01:46:01,890 --> 01:46:04,670
It's going to be in the
same place, but like I said,

2205
01:46:04,670 --> 01:46:06,870
it's not part of
the official class.

2206
01:46:06,870 --> 01:46:11,330
So with that, sorry for running
a little bit over time but

2207
01:46:11,330 --> 01:46:14,150
it's been a joy to
teach this class

2208
01:46:14,150 --> 01:46:16,510
in this kind of new format.