Search and Response Drones with Young Entrepreneur Mihir Garimella | Global Summit

– Welcome back everyone.
I’m Alison Berman. Right now I’m sitting
here with Mihir Garimella. He is the winner of Google
Science Fair, which has some 10 to 15,000 entries. He’s
creating a search and response drone for emergency situations and it is absolutely incredible. Welcome! – Thank-you. Glad to be here. – So, tell me about the
drone that you’re creating. – Sure, yeah so, this
is my latest prototype. And basically the idea is
we can send these drones into really dangerous environments. So, you can think of like a
burning building, during a fire or a building that’s
damaged in an earthquake. Or even like, a large factory
with a gas leak and the robot would use some sensors
onboard to, like, find things. So it could find a person
who’s trapped or it could find the source of the fire
or where the gas leak is. – And so this is really
important for 3 basic reasons. Right, so the first is that
it prevents first responders from having to go into
this environment and risk their own lives. – Definitely. – And so like, in a burning
building, right, obviously if we can prevent firefighters
from having to go in and spend a lot of time
there, that’s really useful. And the 2nd thing is that,
by adding some intelligence sensors and some cool
algorithms to these, we actually make them faster than
first responders. Right? – [Alison] Wow. – And so, you know, also
based on the fact they’re really tiny, they can fly fast. And so we can carry out a
search and rescue mission in potentially 1 to 2 minutes.
Rather than half an hour. – Faster, keeping humans out
of dangerous situations– – Yeah, and then the third
thing is that we’re trying to make them really cheap. You know, existing flying
robots especially ones that have any degree of autonomy cost
tens of thousands of dollars. – Right. – And so, trying to get this
into the 250 to 500 dollar price range, so that’s
really useful for developing countries where they could
benefit from these a lot. And then also, as we try to
use these robots in swarms, so the cheaper they are,
the more we can have. – And when you say swarms,
would that be like, with this smaller one, and a fleet of
them going out searching? – For sure, yeah. Yeah. So,
swarms are really useful because rather than putting all
of the sensors on one robot, you can sort of distribute them around. So that we could definitely
shrink the robots. And then, they can also
cover a lot more area. Right? And so if you have
like a huge building, rather than one robot having
to fly, we could put like 10 or 15 robots and they
could carry that out. – [Alison] It’s amazing. And you were saying that this
one was built a bit to be like how a fly would get away
from someone swatting it, can you talk to me a bit
about why that was the design? – Yeah, so, this was my
Google Science Fair project a few years ago and this was
like one of the first steps towards this. – And so how old were
you for the first one? – 14. – 14 and now he’s 17, that’s amazing. – Thanks. So, basically
the idea here is that like, if we’re trying to do one
of those search and rescue missions that I talked
about, we’re in these really dangerous environments, right? – Right. – So, a ceiling could
collapse, an object could fall and we really have to be
able to deal with that within about half a second to
continue what we’re doing. But drones suck at basically
responding to different things like that. So they’re
really bad at flying… – [Alison] So, immediate shifts
in an environment you mean? – Yeah. Yeah. So, drones
really work in very predictable environments, but they’re
not really good at responding to like really dangerous,
unpredictable environments. So what I wanted to do, was
basically draw inspiration from the fruit fly’s escape. Because fruit flies, as
we all know, are brilliant at escaping. And so draw
inspiration from that escape to basically build some instincts
into these flying robots to make them better at working
in these really dangerous, real world environments. So that was my Google Science Fair project and since then, I’ve been looking
at creating some different core capabilities. That’s
one core capability, right? Being able to escape from stuff. I’ve been also working on
creating algorithms again based on biology, that can help a
drone avoid static obstacles. So, things like tables
and walls and chairs. And then also, find things. So that’s one of the most
important things that this drone has to do, is find people
or find different fires or find fire, stuff like that. So using some sensors and
some intelligent algorithms to be able to actually do that. And then one of the most recent
things I did, over the past summer at MIT, was basically,
creating an algorithm that could teach a flying robot
how to fly just like people learn how to walk. – [Alison] Amazing. – [Mihir] And so the idea is,
like, if we’re trying to fly in an environment with a lot
of smoke or with a lot of wind, stuff that could knock the
robot off course, it’s really hard to control the robot. And
so, if the robot can actually learn how to control itself,
that makes it a lot easier. – And so, you view this
self-learning as maybe the next generation of this? – For sure, yes. So, I think
autonomy and machine learning and self-learning robots,
drones, are the future. There’s a lot of drones on the
market today, are obviously remote-controlled drones, right? – Right. – So you really need a
skilled pilot who’s constantly telling them where to go. And, at least from personal
experience, I think piloting these things is really hard. – Okay. – So, you know, trying to
build some autonomy into them and also trying to teach them
how to control themselves, so trying to build that learning in, is really helpful as well. – And so you talked
about building algorithms that mimic biology, so it sounds
like biomimicry is playing a role. Can you talk to me a
bit about biomimicry for people who might not know what
it is and the principles? – Sure. So, biomimicry is
basically looking at biology for inspiration to solve
some engineering challenges. The reason it’s really useful
here, is because over millions of years, nature has, through
evolution, basically optimized all these kinds of things. Right? So, the fruit fly’s escape, for example. Fruit flies actually use this
really cool escape behavior. So, there’s this lab at
Caltech that basically found that whenever they’re escaping,
they first jump to the side, using their legs and
then they use their wings to move vertically. So that
behavior has been tuned over millions of years of evolution to be, basically perfect. Right? And so by looking
to biology for inspiration when designing these
drones, we can cut out a lot of the R&D work. – [Alison] Right. – [Mihir] So, nature’s already
done a lot of that testing and development. – And already tested it and become better over time. Right. It’s amazing. – Yeah – So, talk to me a bit more
about the functionality of this, and the specifics of the
technology that you have embedded here. I think people are probably curious, like what exactly is
the technology in here? – Sure. So, first, basically
my research over the past few years, in science fairs
and university labs, focused on building these core
capabilities, I talked about. This one is like, avoiding moving threats. – Right. – Right? So, avoiding things
like collapsing ceilings, falling objects. I also worked, in a lab at
Carnegie Mellon, which is close to where I live in Pittsburgh,
on avoiding static obstacles. So that’s the second thing. The
third thing is finding stuff and the fourth thing is
teaching a robot how to fly. – Yeah. – So, those are some of the
capabilities and so what I’m trying to do now is combine
those puzzle pieces, right? So those different pieces of the puzzle. Trying to combine those
to create this robot. So, the idea is– – And, what am I looking at right here? – Yeah, so, I built this robot
from scratch and the idea is that it has some basic,
pretty low-cost base hardware. So that’s motors and different
components like that. And then, you could plug in
different sensor modules, based on whatever task
you’re trying to accomplish. – Okay. – So, first responders, if
they’re trying to find people, they plug in a thermal camera.
If they’re trying to find a gas leak, they plug
in a gas sensor. Right? And then the robot can
intelligently use some of those algorithms that I developed
to be able to carry out that mission autonomously. – And when you were creating
this, what was your inspiration behind this idea? – So, it’s actually a funny story. The way I started was through
that, looking at fruit flies. So, a few years ago my family
went on vacation to India and when we got back we realized
that we’d left some bananas on our kitchen counter and
so our house was filled with fruit flies when we got back. So, I kept trying to swat
them and I got really mad when they kept escaping but I
also started to become really curious about like, how can
fruit flies escape so well? Right? Like, they’re really
tiny creatures, they have really small brains, they have awful
vision, so what could they possibly be doing to escape? So, I did some research. I
Googled a lot, found some papers online and basically, the
key to the fruit fly’s escape is the fact that they
have really bad eyesight. So, their eyes, they can’t see
in very much detail at all. But, that means they can
process whatever they can see and react to that really quickly. So, they can actually see
10 times faster than we can. – [Alison] Wow. – And that’s what enables
them to escape so effectively. Around the same time I had
started reading about drones, and so I realized, drones
have tremendous potential to help out in those kind of
applications I talked about. So, search and rescue
and things like that. But, they were really terrible
at dealing with unpredictable environments and so I wanted
to see whether we could simulate what I learned
about the fruit fly’s escape to make drones better. And then, after that initial
project, seeing it work really well, I got really
curious how else we could draw from biology to makes drones better. Because as I said, biomimicry,
I think, is a really cool approach to engineering that
can help us cut out a lot of the design time when sort
of, building these products. And so now, basically trying
to combine all that research to create a product that can
actually help save lives. – It’s amazing. – Thanks. Yeah. – And yeah, so you started
tinkering with robotics at the age of 10. – Yeah. – What sparked that? – Yeah. So, when I was really
little, so probably like 2 or 3, I had a robotic dog
and so, that was my pet because I didn’t have a real
dog, so that was my pet. – A robotic dog that you built
or that someone gave you? – No, that someone gave
me. I was 2. I was 2, so. – Oh, you were 2. (laughs) – Yeah, so. So, I would play with that
all the time and I thought it was really fascinating how
electronics and you know, some hardware and software
could literally bring something to life. So, you know, I loved to
play with that and that sort of sparked my interest in
robotics and computer science. So, I remember one of my first
projects, you know, stuff that I had built when I was 10. So, I was in my middle school
orchestra and I played violin, I would always play out of
tune because I didn’t know how to tune my violin and so my
teacher would always yell at me for like, playing badly. And
so what I did, I basically created a robot that could
tune my violin for me. And so, you’d play it and
it would analyze you know, the sound, determine how
out of tune the violin was and then it would use some
motors to like, turn the fine tuners to put it in tune. And so basically, working
on a lot of cool projects like that, and realizing that
robotics is a really powerful way, combining hardware and
software at that intersection, is sort of a really powerful
way to help solve a lot of problems. – And it seems like you’re
just very naturally curious in this area. Have you
just always felt that way? – Yeah. So, I think part
of it is my parents, right? They always inspired me
to wonder how things work. So you know, seeing the fruit
flies in my kitchen, seeing them escape, sort of
sparked that question, how do they do it? – Right, because a lot of
people would see fruit flies and just think, how do
I kill these things? – Yeah! Yeah. – But for you, you’ve got,
“Oh, fruit flies, what makes “these unique? Why can’t
I catch them?” Yeah. – Yeah, so, definitely really
curious and I think curiosity is a great skill for inventors
because it lets you find inspiration in a lot of places
that you might not look. So, you know, if I had
started the other way, trying to build an escape algorithm
for these drones, I would have no way to, no way to like, I
wouldn’t know where to start. Right? But looking at fruit
flies and getting sort of inspired by them, gave me,
helped me, gave me a really good place to look for inspiration. – Yeah. And so, you’re 17.
Technology is rapidly maturing and accelerating and there’s
new emerging technologies quite often. How do you think
current technology is going to impact the generation even below you? – Sure. So, I think it’s
becoming increasingly important for people in all kinds of
fields to know how to use and build technology. So for
example, things like law, medicine, finance, they’re
all being transformed by computer science and robotics. And so, I think the really
cool thing about growing up with technology is that
first, it makes us really good at working with it. And I
think, personally, growing up with technology has sort of
helped me learn how to build it. – [Alison] Yeah. – And so it’s– – You don’t have to backtrack at all. – Exactly! Yeah. So, like
growing up with that robotic dog toy, sort of showed
me what’s possible. So it gave me that initial
inspiration and so I think that, you know, going forward
it’s really important to be able to, for people in all
kinds of fields, to have those technology skills. To not only be able to work
with technology but actually to build it. And so I
think, for kids growing up with technology, the cool
thing is that they’re already there, because it’s all around them. – Yeah. And are there any
other projects that you’re also working on right now? – Yeah. So, this is my
main project right now. So, mainly working on trying
to see whether I can actually make this into a product that
can be used in the field. I’m also working at a
start-up this summer. So, back where I live in
Pittsburgh, they basically make autonomous drones for construction sites. So, the idea is, analyzing
piles of dirt and trying to find volume and stuff like that. – We covered that technology recently. – Yeah! Yeah. So, that is, I
think construction is one field that’s gonna be transformed
by robotics and specifically by drones. That’s part of a
bigger team there, obviously. But, I think it’s a really
cool experience to work in a small company, its around
like 30 people, to be able to work in that
environment is really cool. And, I’ve also worked on a
bunch of projects in the past that I really love. So, recently, for example, I
built something called HeadsUp. And so, the idea is like,
in youth sports for example, there’s this concussion problem. – [Alison] Definitely. – Where, a lot of players are
getting concussions and either coaches don’t know how to
diagnose those, so they don’t know that the players have concussions
or they try to push them back in the game because
they think it’s no big deal. – [Alison] Right. – Right. To actually diagnose
the concussions, the reason that they don’t get diagnosed
is because it’s a really long process. So a player has to
be taken out of play, sent to a hospital and a doctor
actually has to do some tests and diagnose them. With some friends at a hackathon,
I developed this device, it’s around $50 and basically,
uses some eye tracking technology to determine
whether people have concussions using this 30 second test. – Wow. – The idea there, is like,
whenever doctors are testing for concussions, they
move their finger in front of your eyes. – Right and they look
for the pupil dilation and all of that. – Yeah. So they actually, the
main thing they’re looking for is whether your eyes are
pointing in the same place at the same time. – [Alison] Okay. – Right. And so, if they’re
not, if one eye is slower than the other, that’s a sign of a concussion. So, using some eye tracking
technology, so basically, we built this headset and it
would flash a pattern on some LEDs, on some lights, around the device. And then it would see
how your eyes track that to determine whether or
not you have a concussion. So, it’s this really cheap,
easy way to test for that. So, that’s one project that I worked on. Another project that I really,
that I worked on in the past, that I really like is called TMAScan. It’s basically a set of image
processing algorithms that can help doctors diagnose brain tumors. So, the idea is like,
whenever doctors are, the way doctors currently
diagnose them is by looking at these huge images and so
you’ve got to pour through them over and over again of images
of tissue samples to determine whether there’s a tumor
and how bad that tumor is. And so, trying to automate
that process to make it a lot more subjective and also a
lot faster so they don’t have to do that manually because we
can write programs to do that for them. And so, actually I
had an opportunity to coauthor a paper with some doctors who
used my work to sort of get experimental results for
pediatric brain tumors. – [Alison] Incredible. – Thanks. – And so, you’ve already done
so much at your young age, you’re starting at Stanford
in a month. Congratulations. – Thanks. Yeah, I’m super excited. – You should be. And do you
have a dream of where you see yourself in 10 years? – Sure. So, I definitely
wanna start a company in the future. And… I think, specifically companies
that are focusing on solving really hard problems. So taking,
commercializing research. So taking some new science
is actually pushing the boundaries and trying to
build that into a product, is really exciting. So hopefully in, sooner than
10 years, maybe five years, working on some projects to
actually start companies around some really hard technical problems. And so I think that’s really
cool. I think that can be really impactful. – Amazing. Well, thank-you
so much for being here today. – Thank-you for having me, yeah. – And I am just absolutely
excited for your future.