1
00:00:00,871 --> 00:00:02,173
I was only nine

2
00:00:02,173 --> 00:00:04,384
when my grandfather first described to me

3
00:00:04,384 --> 00:00:07,120
the horrors he witnessed six years earlier

4
00:00:07,120 --> 00:00:09,989
when human stampedes killed 39 people

5
00:00:09,989 --> 00:00:12,736
in our hometown of Nashik, India.

6
00:00:12,736 --> 00:00:15,954
It was during the 2003 Nashik Kumbh Mela,

7
00:00:15,954 --> 00:00:19,450
one of the world's largest
religious gatherings.

8
00:00:19,450 --> 00:00:22,783
Every 12 years, over 30 million
Hindu worshippers

9
00:00:22,783 --> 00:00:24,674
descend upon our city,

10
00:00:24,674 --> 00:00:27,559
which is built only
for 1.5 million people,

11
00:00:27,559 --> 00:00:29,657
and stay for 45 days.

12
00:00:29,657 --> 00:00:32,711
The main purpose is
to wash away all their sins

13
00:00:32,711 --> 00:00:35,888
to bathing in the river Godavari,

14
00:00:35,888 --> 00:00:38,249
and stampedes can easily happen

15
00:00:38,249 --> 00:00:42,230
because a high-density crowd
moves at a slow pace.

16
00:00:42,230 --> 00:00:46,469
Apart from Nashik, this event happens
in three other places in India

17
00:00:46,469 --> 00:00:48,258
with varying frequency,

18
00:00:48,258 --> 00:00:50,738
and between 2001 and 2014,

19
00:00:50,738 --> 00:00:54,867
over 2,400 lives have been lost

20
00:00:54,867 --> 00:00:57,582
in stampedes at these events.

21
00:00:57,582 --> 00:00:59,515
What saddened me the most

22
00:00:59,515 --> 00:01:03,671
is seeing people around me
resigning to the city's fate

23
00:01:03,671 --> 00:01:07,413
in witnessing the seemingly
inevitable deaths of dozens

24
00:01:07,413 --> 00:01:09,523
at every Kumbh Mela.

25
00:01:09,523 --> 00:01:11,168
I sought to change this,

26
00:01:11,168 --> 00:01:14,449
and I thought, why can't I try
to find a solution to this?

27
00:01:14,449 --> 00:01:15,961
Because I knew it is wrong.

28
00:01:15,961 --> 00:01:17,637
Having learned coding at an early age

29
00:01:17,637 --> 00:01:19,122
and being a maker,

30
00:01:19,122 --> 00:01:21,806
I considered the wild idea.

31
00:01:21,806 --> 00:01:23,742
(Laughter)

32
00:01:23,742 --> 00:01:27,079
I considered the wild idea
of building a system

33
00:01:27,079 --> 00:01:29,353
that would help regulate
the flow of people

34
00:01:29,353 --> 00:01:32,258
and use it in the next Kumbh Mela in 2015,

35
00:01:32,258 --> 00:01:34,086
to have fewer stampedes

36
00:01:34,086 --> 00:01:36,461
and hopefully fewer deaths.

37
00:01:36,461 --> 00:01:38,877
It seemed like a mission impossible,

38
00:01:38,877 --> 00:01:40,754
a dream too big,

39
00:01:40,754 --> 00:01:42,764
especially for a 15-year old,

40
00:01:42,764 --> 00:01:44,924
yet that dream came through in 2015,

41
00:01:44,924 --> 00:01:49,758
when not only did we succeed

42
00:01:49,758 --> 00:01:52,527
in reducing the stampedes
and their intensity,

43
00:01:52,527 --> 00:01:55,306
but we marked 2015

44
00:01:55,306 --> 00:01:57,528
as the first Nashik Kumbh Mela

45
00:01:57,528 --> 00:01:59,848
to have zero stampedes.

46
00:01:59,848 --> 00:02:03,397
(Applause)

47
00:02:05,170 --> 00:02:07,648
It was the first time in recorded history

48
00:02:07,648 --> 00:02:10,772
that this event passed
without any casualties.

49
00:02:10,772 --> 00:02:12,803
How did we do it?

50
00:02:12,803 --> 00:02:15,692
It all started when I joined
an innovation workshop

51
00:02:15,692 --> 00:02:18,696
by MIT Media Lab in 2014

52
00:02:18,696 --> 00:02:20,493
called the Kumbhathon

53
00:02:20,493 --> 00:02:25,845
that aimed at solving challenging
faced at the grand scale of Kumbh Mela.

54
00:02:25,845 --> 00:02:30,568
Now, we figured out to solve
the stampede problem,

55
00:02:30,568 --> 00:02:32,400
we wanted to know only three things:

56
00:02:32,400 --> 00:02:36,074
the number of people, the location,

57
00:02:36,074 --> 00:02:39,738
and the rate of people per minute.

58
00:02:39,738 --> 00:02:43,891
So we started to look for technologies
that would help us get these three things.

59
00:02:43,891 --> 00:02:46,603
Can we distribute radio frequency tokens

60
00:02:46,603 --> 00:02:48,605
to identify people?

61
00:02:48,605 --> 00:02:52,016
We figured out that it would
be too expensive and impractical

62
00:02:52,016 --> 00:02:54,486
to distribute 30 million tags.

63
00:02:54,486 --> 00:02:58,130
Can you use CCTV cameras
with image-processing techniques?

64
00:02:58,130 --> 00:03:00,173
Again, too expensive for that scale,

65
00:03:00,173 --> 00:03:03,103
along with the disadvantages
of being non-portable

66
00:03:03,103 --> 00:03:04,781
and being completely useless

67
00:03:04,781 --> 00:03:05,996
in the case of rain,

68
00:03:05,996 --> 00:03:09,300
which is a common thing
to happen in Kumbh Mela.

69
00:03:09,300 --> 00:03:12,073
Can we use cell phone tower data?

70
00:03:12,073 --> 00:03:14,533
It sounds like the perfect solution,

71
00:03:14,533 --> 00:03:16,059
but the funny part is,

72
00:03:16,059 --> 00:03:18,608
most of the people
do not carry cell phones

73
00:03:18,608 --> 00:03:21,024
in events like Kumbh Mela.

74
00:03:21,024 --> 00:03:24,640
Also, the data wouldn't have been
granular enough for us.

75
00:03:24,640 --> 00:03:27,125
So we wanted something that was real-time,

76
00:03:27,125 --> 00:03:30,441
low-cost, sturdy and waterproof,

77
00:03:30,441 --> 00:03:33,522
and it was easy to get
the data for processing.

78
00:03:33,522 --> 00:03:36,011
So we built Ashioto,

79
00:03:36,011 --> 00:03:38,395
meaning "footstep" in Japanese,

80
00:03:38,395 --> 00:03:40,293
as it consists of a portable mat

81
00:03:40,293 --> 00:03:42,550
which has pressure sensors

82
00:03:42,550 --> 00:03:44,959
which can count the number
of people walking on it

83
00:03:44,959 --> 00:03:46,846
and sends the data over the internet

84
00:03:46,846 --> 00:03:50,466
to the advance data analysis
software we created.

85
00:03:50,466 --> 00:03:53,923
The possible errors,
like overcounting or double-stepping,

86
00:03:53,923 --> 00:03:57,182
were overcome using design interventions.

87
00:03:57,182 --> 00:04:00,558
The optimum breadth of the mat
was determined to be 18 inches,

88
00:04:00,558 --> 00:04:02,607
after we tested many different versions

89
00:04:02,607 --> 00:04:06,166
and observed the average
stride length of a person.

90
00:04:06,166 --> 00:04:09,268
Otherwise, people might step
over the sensor.

91
00:04:09,268 --> 00:04:11,786
We started with a proof of concept
built in three days

92
00:04:11,786 --> 00:04:14,916
made out of cardboard and aluminum foil.

93
00:04:14,916 --> 00:04:17,959
It worked for real.

94
00:04:17,959 --> 00:04:20,828
We built another one
with aluminum composite panels

95
00:04:20,828 --> 00:04:22,513
and piezoelectric plates,

96
00:04:22,513 --> 00:04:26,676
which are plates that generate a small
pulse of electricity under pressure.

97
00:04:26,676 --> 00:04:30,382
We tested this at 30
different pilots in public,

98
00:04:30,382 --> 00:04:33,504
in crowded restaurants,
in malls, in temples, etc.,

99
00:04:33,504 --> 00:04:36,051
to see how people reacted.

100
00:04:36,051 --> 00:04:38,683
And people let us run these pilots

101
00:04:38,683 --> 00:04:45,893
because they were excited to see locals
work on problems for the city.

102
00:04:45,893 --> 00:04:49,425
I was 15 and my team members
were in their early 20s.

103
00:04:49,425 --> 00:04:52,247
When the sensors were colored,

104
00:04:52,247 --> 00:04:57,377
people would get scared
and would ask us questions like,

105
00:04:57,377 --> 00:05:00,295
"Will I get electrocuted
if I step on this?"

106
00:05:00,295 --> 00:05:04,947
Or, if it was very obvious that it was
an electronic sensor on the ground,

107
00:05:04,947 --> 00:05:07,384
they would just jump over it.

108
00:05:07,384 --> 00:05:10,592
So we decided to design
a cover for the sensor

109
00:05:10,592 --> 00:05:13,787
so that people don't have to worry
what it is on the ground.

110
00:05:13,787 --> 00:05:15,610
So after some experimentation,

111
00:05:15,610 --> 00:05:18,283
we decided to use an industrial sensor

112
00:05:18,283 --> 00:05:22,386
used a safety trigger
in hazardous areas as the sensor,

113
00:05:22,386 --> 00:05:24,841
and we black neoprene rubber sheet

114
00:05:24,841 --> 00:05:26,626
as the cover.

115
00:05:26,626 --> 00:05:29,466
Now, another added benefit
of using black rubber

116
00:05:29,466 --> 00:05:33,116
was that dust naturally
accumulates over the surface,

117
00:05:33,116 --> 00:05:35,768
eventually camouflaging it
with the ground.

118
00:05:35,768 --> 00:05:40,904
We also had to make sure that the sensor
is no higher than 12 millimeters.

119
00:05:40,904 --> 00:05:43,596
Otherwise, people might trip over it,

120
00:05:43,596 --> 00:05:46,195
which in itself would cause stampedes.

121
00:05:46,195 --> 00:05:48,676
(Laughter)

122
00:05:49,966 --> 00:05:52,693
We don't want that.

123
00:05:52,693 --> 00:05:57,705
So we were able to design a sensor
which was only 10 millimeters thick.

124
00:05:57,705 --> 00:05:59,958
Now the data is sent
to the server in real time,

125
00:05:59,958 --> 00:06:01,435
and a heat map is plotted,

126
00:06:01,435 --> 00:06:04,617
taking into account all
the active devices on the ground.

127
00:06:04,617 --> 00:06:09,032
The authorities could be alerted
if the crowd movement slowed down

128
00:06:09,032 --> 00:06:13,228
or if the crowd density moved
beyond a desired threshold.

129
00:06:13,228 --> 00:06:15,565
We installed five of these mats

130
00:06:15,565 --> 00:06:17,995
in the Nashik Kumbh Mela 2015,

131
00:06:17,995 --> 00:06:20,602
and counted over half a million people

132
00:06:20,602 --> 00:06:22,382
in 18 hours,

133
00:06:22,382 --> 00:06:26,782
ensuring that the data was available
in realtime at various checkpoints,

134
00:06:26,782 --> 00:06:28,681
ensuring a safe flow of people.

135
00:06:28,681 --> 00:06:32,012
Now, this system, eventually,
with other innovations,

136
00:06:32,012 --> 00:06:35,909
is what helped prevent stampedes
altogether at that festival.

137
00:06:35,909 --> 00:06:38,610
The code used by Ashioto during Kumbh Mela

138
00:06:38,610 --> 00:06:41,803
will soon be made publicly available
free to use for anyone.

139
00:06:41,803 --> 00:06:43,743
I would be glad if someone used this,

140
00:06:43,743 --> 00:06:46,976
the code, to make
many more gatherings safer.

141
00:06:46,976 --> 00:06:49,228
Having succeeded at Kumbh Mela

142
00:06:49,228 --> 00:06:53,028
has inspired me to help others
who may also suffer from stampedes.

143
00:06:53,028 --> 00:06:55,757
The design of the system
makes it adaptable

144
00:06:55,757 --> 00:06:57,795
to pretty much any event

145
00:06:57,795 --> 00:07:01,615
that involves an organized
gathering of people,

146
00:07:01,615 --> 00:07:06,014
and my new dream is to improve,
adapt, and deploy the system

147
00:07:06,014 --> 00:07:11,262
all over the world to prevent loss of life
and ensure a safe flow of people,

148
00:07:11,262 --> 00:07:14,345
because every human soul is precious,

149
00:07:14,345 --> 00:07:16,746
whether at concerts or sporting events,

150
00:07:16,746 --> 00:07:18,656
the Maha Kumbh Mela in Allahabad,

151
00:07:18,656 --> 00:07:20,041
the Hajj in Mecca,

152
00:07:20,041 --> 00:07:21,840
the Shia procession to Karbala,

153
00:07:21,840 --> 00:07:23,930
or at the Vatican City.

154
00:07:23,930 --> 00:07:26,285
So what do you all think, can we do it?

155
00:07:26,285 --> 00:07:29,412
(Cheers)

156
00:07:29,412 --> 00:07:32,650
(Applause)