Good morning.

My name is Bill worked for and

I am chair of the Woodruff
school mechanical engineering.

And it is my privilege and honor to
welcome you to the annual Harold W.

Guggenheim or lecture on the innovation.

The Guggenheim or

lecture has been conducted on an annual
basis since one nine hundred ninety five

the lecture series was established
soon and dominant from Mr Harold W.

Guggenheim or Woodruff school
graduate class of thirty three.

To support student programs that encourage
creativity innovation and design.

Through the lecture series in support
of Capstone design projects students

are exposed to processes that simulate
creativity and lead to innovations and

patents this year Speaker is

so with her about her.

Yes we do is demonstrate it Murphy's Law
for the second year in a row.

We are really fortunate
to have a good time or

a lecture who happens to be Woodruff
school alumnus Robin got her undergraduate

degree in mechanical engineering back
in one thousand nine hundred in maybe

it's because we were in the building and
almost nothing worked in that building but

she spent some time in the industry and
then received both a master's and a Ph D.

in the in computer science in in
what's now the College of computing.

She has published more than one
hundred papers in the areas

of artificial intelligence.

And human robot interactions.

She's published a textbook to textbooks
as a matter of fact introduction to

AI robotics AI in Mobile
were mobile robotic.

Yes.

In two thousand and
eight she was awarded the now.

Oby outstanding contributor
award by the side foundation for

founding the field of rescue robotics.

She is a Fellow of the I simply and

serves on numerous governmental boards
including the Defense Science Board.

It is my distinct pleasure to
introduce Professor Robin Murphy

who will be presenting a lecture
in titled robots to the rescue.

Thank you for having me here and
howdy which is the official greeting at

Texas A and M. It's an honor to be
back in Georgia tacking to show what

how I've taken what I've
learned here as a student and

applied it to field of searching around
here and one thing that I've learned about

rescue robots is that it's not about
mobility are just the networks or

artificial intelligence but rather
it's a complex socio technical system

where every piece must work
together are nothing works at all.

So let's start by looking at some of the
disasters and robots we participate in.

That's for a start but this

is your sound not a street

fighter coming over in recent years.

You might want to drop

right.

For such a shock

like that.

Now I do my work based out of Texas A and
M.

I've been there for two years now and
it's all a lot like Georgia Tech.

Although we have a medical college and

a larger level arts program but
it's very much an engineering school.

What makes an unusual is that
it's part has a state agency

called teams that trains over eighty
one thousand responders a year and

per session all emergency
response has a fifty two acre

city called Disaster city which is
why your favorite place on earth but

let's go to talk about innovation and
let's talk about rescue robotics so

if you're going to innovate and

rescue robotics which I hope
I can entice you to do.

You're going to need the answers to
the following questions why robots.

What's it like if you're a robot.

What's it like if you're a user.

What's it like for the survivors and

why aren't there more robots why why
does our center the center of a robot

as a search and rescue have than
those deploying that's in the world.

Well.

Let's look at the idea of Haiti Haiti.

This was a standing the number of
the lifesavers that was found by

New York Task Force One They found
six survivors in eight days with

eighty two person team
that's a great number

of your responder but clearly that's not
going to scale that's a huge footprint.

So we need robots to change the way
we're doing things are to amplify.

Just to see how badly
the traditional methods

don't scale let's look at some statistics.

So if you were injured in a disaster
your mortality curve is about forty

eight hours about that point after
about fifty two hours the odds of

you surviving even though we may be able
to get you out may die in the hospital.

So you've got that now eighty
percent of the survivors

are surface are lightly trout So
here's a case in India.

There's a person right there.

So they were just out there in
the open but they were broken leg or

something in their trap.

So they couldn't do it.

So just the first responder
that's the regular farm and

the regular policeman was
able to go over and get that.

So that's not even the theme that teams
that we're thinking about now eighty

percent of those are and they're found
by when they're found by Mozart.

There we go.

But the that's not necessarily eighty
percent of the victims as just eighty

percent of people we find the twenty
percent that we find with a professional

rescue teams like you saw with
with Haiti those take about

twelve to eighteen hours for
the teens to or rock and deploy.

And they're mostly able to look in
the twenty to forty foot range of rubble.

So if you look right there.

You start off looking with dogs
the knowledge you have you use a camera on

a stick.

There are several different brands of
these A can actually will shut everything

down in periodic Lee use these.

Who stick sensors that you manually
move the transducers around which

is a D S P problem looking for a solution.

And it takes ten people for two an hour
stoops track to once we found you.

And only on about.

The last twenty minutes can
we get medical attention and

we actually touch you because you
see we're having to dig and so

if it was easy to get to somebody
would have gotten to you a nice idea.

Now you can see where robots can stand
go beyond that twenty to forty feet.

So you'd like to get in there.

Particularly because think of this
building collapse if they could only get

twenty feet where the core of
us the majority are further

in from the outside of the walls right so
we would not have a good chance

of survival as it was a large void and
we were just able to wait it out.

So that's kind of scary but

let's look at what have we have terrorism
like we have the Oklahoma City bombing.

Are we have Khobar Towers of the Libya
Libyan embassy that sort of thing.

Everybody who was our traditional
survivors right the surface vehicle

their dad.

OK.

The people in the twenty to forty
feet range are badly burned.

If they don't get immediate attention they
will pass away and then there's a large

number of people in the forty to sixty
foot range beyond what we've traditionally

been able to get to that are survivors but
we have to get to them and we can't.

So we need robots and
I think it's important also.

It's not robots not just searching.

It's information if we know
if we can find you and

we're contrary arch and say this person
and there's ten people over here and

one of the we can reduce deaths but
also robots.

Looking to help us celebrate damage
assessment and minimize economic downtime.

Here's a picture from the Italy
recent earthly Italy earthquake where

all the tent cities where they
were expecting to have to see.

Day in those tents for eight months
because there's not enough civil engineers

there's not enough waste to shore
up the buildings to go in and

have the civil engineers check to
see if it was in fact still safe

to go back in it could be
reoccupied a big thing in Florida.

You know we have all
the hurricanes in Florida

is can we reopen the schools you're not
recovered and so you have the schools open

because people can't go back to
work because their kids are home.

So why robots you think about the
structural damage from building collapses

also robots of them propose for
if there's no structural

damage those would be the cases of
can buy and those are very very rare.

And then you have a strategic or

geographically distributed activities
the wilderness search and rescue.

You know when some are somebody wanders
off shore or wildland firefighting and

I think the important thing is this is
the most important thing robots don't do

things that people are dogs can do
they don't replace people or dogs.

What they do is extend those abilities and
do things that they can't do.

Now if you look at so they're just giving
cheap abilities that we never had before.

If you're talking about him
Vaio what we've discovered

Edgewood chemical biological centers as
cover that allow us the robots are better

than a person sitting up in one of those
from Brenda's and then sits level A suits.

It's not worth it.

They have to be faster and better.

And here they don't have to be better but

they just more economical than than
hiring a guy in a helicopter or

getting the civilian air patrol
in the timing something that

you can throw up in the air when you need
it may be more efficient more economical.

So I'm going to how many of you
know about artificial intelligence.

So let me give you some heuristics
some rules of thumb about AI and

this will maybe help as
we go through the top.

What's easy for person.

Is hard for a computer or a robot.

And what's hard core person is easy for
a computer robot right so

what's hard you know a lot of
the math a lot of optimization leads

into where optimization I'm thinking
a computer can do that easily.

And so it's easy for a person things
are like mobility and perception.

We just look at things.

There's the visual perception
there's the ball things pop

up that is that's extraordinarily hard for
computers and robots of keep that in mind.

Now let's take that further
let's look at a rescue robotics.

We're not building a task of all ages.

We're not building Lieutenant Commander
Data We're not trying to do it in

a human equivalent.

We're doing what's called or

in the present standing where we're
trying to extend ourselves and

at the very naturally and looking very
actually through these remote devices and

actually we don't even care what they were
you know if you put a camera on that pitch

and we'd be you know and
it got to what you need to be happy.

You don't care that it's a robot and that
changes you'll see the way we think about

AI because it's no longer fully in one
place it's now shared between a person and

the platform the computer
the AI on the platform itself.

Let's talk about if you're a robot.

What's a disaster like for you.

Well here's.

Just what Tower Two at the World Trade
Center looked like after the collapse.

There's no need to search
on top of the rubble

people like I said are you
do the surface search.

How do you get into that what you're
seeing is that there's no voice there

voice of the biggest forwards we were
finding were on a characteristic diameter

last than two feet and so
these were the robots that were all

brought up traditionally people bring bomb
squad robots and then they just sit on

the sides the very the smallest of
all of these robots were there but

it was actually to big for
that for most of the day.

Now let's look at what happens in
mining this gives you an idea of

how what turbot conditions.

And that's caused all the facilities and

they don't realize what's going
happen it's miners were trapped or

killed in the collapse of
the Crandall Canyon Mine in Utah.

After Three rescuers were killed and three
others injured trying to enter the mine

on August seventeenth the Mine Safety and
Health Administration are in shock

considered waiving standards for
the operation of potentially explosive

equipment in coal mines in August
twenty first begin the process

of allowing a robot in or through
small bore holes drilled from a pup.

In three days the Nekton
service is a candidate

constructed a variant of their tele
operated waterproof robots to be

connected to a high flying international
tether and controller system.

The seventy robot had a camera
in front for navigation and

was deployed in the Latin condition
it then exited the hole and

raised a ten inch silver
spectrum nine hundred and

tilt zoom camera with two hundred watts
of why the robot had to be deployed

from abroad good conditions on a mountain
top ten thousand feet above sea level.

It had to travel two thousand feet
down and eight and seventy singe and

case borehole it then had to pass
the steel mesh that holds up mine moves.

Once on the floor the robot needed to
crawl up to one thousand feet dragging

the Tatar the robot experience
numerous difficulties as it went down

to bore holes for a total of four
insertions the robot caught within

ten feet of the mine opening a hole
three that could go no further due to

patricians in the hole approximately
eight a half gallons a minute of water.

Dhubri were flowing over the robot fouling
the camera at each hole that hole for

the robot was finally able to enter the
mine with both cameras covered in water

to Perri in drilling foam
it slid into place and

was able to show that this part of
the mine had collapsed due to micro.

Seismic activity there was
no reason to go further and

after death after the robot
reentered the bore hole.

On ascent.

However the robot was lost about fifty two
feet from the surface as the bore hole

routed and share with boulders on
the robot on Friday August thirty first.

We told the families we
had done our best but

the robot monk shot did not pay off and
shot in close the response

this effort highlights the need for
additional research and

development in self cleaning sensors and
scissor placement mobility.

Power in communications and
human robot interaction.

So you're seeing a lot of water you're
seeing rocks are seeing things like that

that also caused a problem.

The New Zealand mine disaster
where yeah it's an outdoor robot

wasn't expecting a half gallons of water.

If you think about things
underground groundwater.

And also it's very dangerous
if you're a robot.

It's very dangerous here in the minus
Goldline Here's a robot that we had

lowered down and it was getting rocks
as it went down through this void.

Was sort of repelling down would
land on and in damage that.

Fortunately that was the Navy's favors
robot that I borrowed we also lost

two robots the New Zealanders lost two of
their bomb squad robots in there you know

it's a methane producing coal mines and
exploded and the robots were in there.

Another thing to keep in mind is
wireless and G.P.S. just doesn't work

in these environments we know that a lot
of it doesn't work you know G.P.S.

doesn't work indoors wireless sort of
works indoors wireless really doesn't or

in a collapse because you're mostly
looking at commercial buildings which have

high I'm mounts of metal metal stuff
that's all the rebar everything compresses

and on much higher density because we
wouldn't have this volume of space.

Hoping for any propagation.

Here's from the World Trade Center
this is going into Tower two on

the Thursday of the disaster.

We have now gone at this point
we're thirty feet into it.

Look how much communications drop out and

this was a military robot out what's
now called the Talon series and

they say line of sight military
three miles and we eventually

lost lost all communications were really
helpful is that they had never thought

about what happens if you flip upside down
and your antenna grounds to the ground.

You know these things
are very difficult and

they were very little clearance for this
robot and so it eventually shorted out.

It's it sent on a piece
of rebar hold down.

So again very demanding a lot
of things we know how to do.

It's pretty late on the usual channels to
get up aerial vehicles now how they have

to operate closed.

You know out over the motel room so

they can get the same guy
to hold their scientists

are using this robotic helicopter to
see things they've never seen before.

Like what happened in Biloxi when
the storm tossed a casino barge more than

half a mile before slamming it into
a motel more were able to get in then is

zoom in very high resolution photographs.

They can be gotten from
somebody on the ground and

can be done from the man's helicopter.

Professor Robin Murphy used robots
on the ground to search in New York

after the nine eleven attacks.

They may help engineers discover
why some buildings collapsed and

others hold up after any type of disaster.

So they'll be looking at two things
One is what did they learn in

terms of the structure how did it.

Hold up to fail in unpredictable
ways chiller Bilton pilot is

the aircraft of this Gold Coast mission.

We are carrying a still camera that
has a video downlink as well so

that we can actually see there aren't
pieces was the still camera saying so

they can frame the shot and then we have
no control to trigger the camera this

technology alone
Engineer Scott Nekton a former

firefighter to beautiful oxy images from
nine hundred miles away in terms of the.

These ability I don't see any reason
why this equipment couldn't be used for

initial life safety evaluations well
looking at locations where people may be

trapped the National Science Foundation is
funding the work of this flying camera and

there is hope that some may be in
the field by next hurricane season.

Daniel Sieberg C.N.N. Atlanta.

And we'll talk about why those aren't
in the field worn in the field for

the next hurricane season.

Now let's think about water vehicle
service Cerner water vehicles

checking a bridge This is after
Hurricane Ike manual divers could only

do this fifteen minutes twice a day at the
changing of the Tods because the tides.

After all they are so
intense the currents are so there and

you don't know what debris is under there.

It's a very target.

So the manual divers can only go down for

fifteen minutes at a time because they can
only see about four feet in front of them.

And when you talk about currents it's not
necessarily easy for the robots either for

a while we had ours on a leash.

So that we could make sure
it didn't get away from us.

So let's go to recap with
the technical barriers.

We're talking about it to mean
that if you're on the ground.

We're looking at theory very small
sizes if it's a pancake collapse.

You're talking less than two inches.

We're talking just any kind of collapse
if it's more than two feet it's going

to be less than two feet that and
it's very three D.

very twisty So the obvious thing to get
more a better view is to be taller.

Which doesn't.

How do you get taller.

You know you're bumping up or
snagging something.

The sensors and
sensing is incredibly important.

We need them smaller there's
a lot of wonderful sensors but

they're just too big and to have a cell
cleaning all the water all the dirt

all the dust that gets here
both are tied by G.P.S.

and wireless denied operations mobility
notice there was an awful lot of

burdock coal stuff going on right here
you're actually doing a lot of Helling.

And.
By the way to avoid the wireless

you're using a tad or you're using fire
optics and a lot of times fiber optic and

putting a rope our aircraft
carrier Keeble around it so

you can get your robot back so
at some point we need to look at hybrid

tether wireless systems where you use
the tether and they get off of that for

short periods of time and
reliability is just such a huge problem.

My.
Well my robots waterproof well meaning all

of the half gallons can ever
sort of water resistant maybe.

Now let's talk about if you're a user and
we're going to calm users the operators.

But I mean expand them
out because who really

cares about the operators it's really
the decision makers who are getting

the information are hopefully getting
the information from the robots.

Well first off if you're in there.

There's a lot of physical barriers
Here's minus gold mine disaster and

we had to stay on the whole time
which you know if you got up to do

something with down you know that you're
always on the lie so that's a pain.

You have a lot of physical
discomfort of being there.

So a little bit of this.

I want to just show you this is
a simulation of a World Trade Center site

mission with the actual robots would it
looks like from this just to illustrate

how hard it is and you're going to say
I'm going to it's just that robot and

you know if there are some fundamental
ways about how we deal with

brains deal with Tootie but
our brains are really wired for three D.

and you have the mistakes we made.

OK So there's all robot going down and

locked in series robot we're
going down one of those box.

Beams that you see what kind of
like a ping a soda straw get into

the middle of it.

We're there.

About sixty feet long and they're trying
to penetrate and what we're trying to use

the robot we know there's going to be no
survivors in a space that's about one and

a half by two feet but
can we find an access to the.

Basement.

Because your only other choice at this
point is to you know just put crane

somewhere and just.

Hope you find a short cut to the base of
that which is where the only place you're

going to find survivors and they.

Now notice.

Did you know it looks most people think
we're moving this way we're actually

going down about a forty five degree
angle or now coming up to some debris.

There's not a lot of power planing
as you know it's only that but

whenever you have a choice.

You don't worry about optimal
what you do is you always choose

to the right into a white
wall following algorithm

because that's what all responders all
over the world are trained for and so

you've got that repeatability
you don't want to change much.

What's going on there right.

So there's a guy there who's not finding
any other way to go this to go further.

OK so now we're look.

Trying to get there.

See a little vibration in the mobility
challenger's loops that looks

like they could be churned out
boot from a responder that's

put many more against a piece of metal and
now now we're going.

Maybe both that there could be
something over the lip that

doesn't listen to look cool that extent.

So we're always polymorphic my favorite
robot for these situations would be

a ferret like robot because they're
they've got legs but they can also see.

Like a snake kind of slink around.

I mean it's I think in

this video there are at
least three sets of

remains we were able to
pick up the first one.

As if all of us saw because of
the slight coloration even though

everything's great because of that
dost you get a little tinge of red it.

You know that's usually are are are some
indicator like that but

what we saw later on.

And even you know experts
we looked at these over and

over again because we
lose depth perception.

We see things at the wrong size and so.

You'll see right in the middle right now
is is a something you know size of a rock

about this big.

If you enhance the image
that's not a rock.

That's that's a rock with hair.

So it's very demanding very different.

I now think about doing this when you've
had less than three hours sleep over

a fifty hour period and the international
stress we found that two heads are nine

times better than one putting people in
teams to work together they can complement

each other they can work around some of
the cognitive barriers the keyhole of that

of the dimension of the loss of depth
perception and these sorts of things.

Except you have to give them a protocol
because otherwise they'll do lots of

backseat driving which was what our first
study found out when we took these results

of that we've seen in
the World Trade Center repeated them so

this led us to conclude that there's
one formula that says if you're

starting out with a robot
a human has one rule at a time.

Usually people in these
conditions can only do one thing.

Well at a time.

So the number of rolls or the number of
people is equal to the number of vehicles

plus the number of payloads on the vehicle
plus one to be your safety officer.

So essentially you're getting one
person to one run one run one robot

which isn't that bad when you look at
the twenty to one they use Global Hawk or

the predator numbers but that did not make
me very popular because what about we're

specific one person one hundred robots.

OK but let's start looking at some
of the problems associated with that

assumption right now when we need to do
that but not the way were set things up

right now let's look at the human
out of the loop control program.

Now this is an aerial vehicle that
Chandler went back based on his

experiences with back any
developed hurricane Katrina.

That engineered it it's a ton of us.

He's an oversight and takes off.

Been.

Once it gets up here he says only one
person eats around I mean he's hungry.

And then he's going to switch to G.P.S.

now he's going to go get everything on
waypoints and he's going to go ahead.

Stan and then he just has to
fly the camera because he

knows the robot won't do anything.

OK Except now it's why
simple interview that way.

Look now it's going that way you

notice that the manufacturer is looking
at kind of looking around back and forth.

OK you know and
I'm taking the film here and

you don't see the safety officer inside.

OK So what are we seeing here.

We saw the robot adding
Murphy's Law right at the moment

when it got to a certain place it was
going to go to waypoint navigation G.P.S.

was engaged G.P.S. was showing great
at that moment at the exact moment

you swapped it over G.P.S. not so
good because we're in our urban canyon.

And so it began getting the signals and

humans don't switch from one
line of thought to the other

computer scientists you know it's a paging
problem aging in Beijing out very quickly.

And so what we really find out is that
even though we're delegating more and

more to robot which is great.

The human is always responsible.

So even though it was an honest
vehicle if something goes wrong

the human was expected to take over.

Well that if there's a human
trying to take over as a pilot or

has to be the backup pilot and

you have the mission specialist
who's often just heads down looking.

It's a very hard to swap
those roles in real time.

And so we have to start
thinking about these not as

we're just going to make this a ton on
this and we don't have to worry about you

need to think of it as what David would
sports and all NATO cognitive engineers if

you're from industrial engineering
you should know these guys.

A joint cognitive system.

We're not talking about whose autonomy
is but so how does the system more.

It's a shared autonomy paradigm.

And as networks will get better
are getting lots better.

That means that people.

The users the mission specialists
the decision makers are no longer

code located with the person using it.

They're all over the place and then you
have latencies but you have even more

problems with situational awareness trying
to figure out what you're looking at

when you're three hundred eight
hundred three thousand miles away.

And if you're a user I would just like to
throw him that is not as fun as it sounds

or looks Here's Dr Epstein while trying
to retrieve you to successfully retrieve

one of our underwater vehicles that
were using after Hurricane Ike And

that water was really cold and there
was a hole in his whites it destroys it

so technical barriers which
are beginning to see is that here.

The problem is this idea of
human robot interaction and

interfaces that there are some
fundamental cognitive barriers that

a third person you know first person
shooter interface doesn't solve

the way we mentally process
information and also we're trying to

do this in real time we're not saying go
out be like a Mars probe and come back and

tell us something we're also beginning
to see the need for users to manipulate.

And being having to manipulate when you
can't you know your body you know you have

no sensors to feel purpose up
to see what your goals are palm.

Now if you're a victim.

What is this lie and I think that right
now the state has a mind you think about

the miners and shell a you know
whether they're you know is that but

actually a better metaphor to
think about is more realistic.

That happens is is something like
this where you're highly can find

your way up there people or your way back
there and you're stuck there for at least

ten hours so we never say it took twelve
to eighteen hours before we found it.

OK so the simulated injuries
are something people

name you mean wonderful to be rescued by
a robot right as a lot of Washington vets

the robot is on its way the first
challenge is to renew it

by remote control through to
rob the difficulty of this.

When the robot finds me.

Rescuers need a good look at a
conversation we want to know if you're OK

or where does it hurt my have I have pain
in my left shoulder that my heart and

Marie my a pain in my own left leg but
there must have been a Zen robot and

right away by you get a good look at scene
United injuries so that rescuers who trio.

Get help to the most seriously injured
victims first in this simulation

they determine that my condition is stable
several rescue robots can be working

the disaster scene at the same
time another simulated disaster

victim Jamie is now being contacted
by a new generation of rescue robot

did this simulation she doesn't respond.

Is she alive and unconscious or
a she passed away and if she's not alive.

Rescuers don't want to use
valuable time that could

be critical of saving other victims

this robot carries a kind of nose a sensor
samples the air around the victim's mouth.

Exhale carbon dioxide if she were to
move in the green here shows that

there's enough C O two to show that
she's really a major breakthrough.

They want robots a tuba robot can deliver
water or even medicine to the hours or

days it takes rescuers to dig down
the victim the robot could be the only

lifeline how to get them water how to
keep them psychologically confort it.

Talk to them keep them
motivated to survive.

So you know thinking out but but again
you know the robot was in the US there

was that robot would you imagine
suddenly waking up in the dark

you have this nose like things sticking at
you like an insect is that a good thing or

not say anything like this is
going to be like that to me.

Well and you was but
rich Bowles took the Germany are great.

Terminator which is actually a two inch
hole so they can go through the boring

hole and then go in and then they'll do
the brushstroke just like the Terminator

to get in these places that the responders
are bored a hole into so again you

could literally have something that looks
like the Terminator coming from your.

Well we do some work with
Clifford now at Stanford

fabulous book Byron Ritz clip
now the media Quezon media equal

real why they have conclusive studies that
even though we believe it's not true.

People respond.

Unconsciously to communication media

as if it were human if it
moves you just think it's in

what we see that with puppets right you
know even other persons right there.

You wind up talking you know the puppet
has it's the one that's moving.

So let's talk about whether
this is hearing and

how this can be annoyingly
inconsistent best theory.

Thank you

so we have watched our medical
doctors we let them drive them around

how they would direct us to write the
robot So let's look at the different areas

of the city of put a blanket
over the very you know.

At the same time with all
the educational her and as our robots

making of the study and
this is how we drive robots

see a problem with this if you're on
the other end and you don't notice it.

If you're in the operator.

Yeah maybe we're not helping.

After about fifteen minutes we can imagine
people get really annoyed with this robot

and if you're already like in shock and
stuff it this is not going to help.

So since you went and looked also about
this more motive using what we know from

the social sciences they are a Animals
that turn the lights down be submissive

have more lighting so you can see what's
coming at you without whining you know

one

person actually retched reached out and
head at the robots and

we had to through that data point you
know it was very frustrating and so

you would use dogs that
indeed we didn't just Jerry.

That they and they were definitely
more mommy is you know long well

known humans everything that
you want to do those so

even though it sounds kind of silly to us
particular was as mechanical engineers you

know you know well just getting there
is not really the problem is you know

you get there but you're the one
thing that's not going to help is it.

Well let's look at what
happens with you know so

let's make it more human right OK Well
there's there's a problem there.

But you know again if you
have the thirty rockets.

There's trees.

Greatest courage of Dr Walt Disney
Larry Flynt the Japanese but

they can't do it because
of the uncanny valley.

Let me show you some Check
out this chart you see

as artificial representations of
humans become more and more realistic.

They reach a point where they stop being
enduring and become creepy to me and

stuff was all right with like R two D
two and c three feel they're nice and

up here we were real person and so he acts
like he doesn't care what he does but

down here we have a C.G.I. stormtrooper or
Tom Hanks in The Polar Express.

I just can't get me.

That's the problem.

You're the value now and
it's impossible to get out.

That's what your goal of boards
are designed a video game with care

before golden points.

My genius will not be denied.

I like won't shock you like that guy that
was always jealous of Mozart Sally Erin.

No thank you already eight.

So making something more human isn't
necessarily the right way to go and

if you think about form over function
having you know it doesn't necessarily

buy you a whole lot for
a function getting there three years

of work that we've been doing with
what we call Survivor by the lines for

situations where human is highly dependent
on the morning example of a dependent.

It's a traffic to using
a multimedia interface a rescue.

Robot to communicate with the outside
world watch T.V. or search the web or

waiting to be extricated clearing
the robot is acting as a medium for

the computers as social actors or
counts of model human robot interaction

suggest that since the robot moves
the dependent will respond to the robot as

the social actor but then it may
become distrustful as well as kind

until very confused by a robot that
sometimes acts like an intelligent being.

And sometimes like a stationary T.V.
the social medium idea is that

the robot should combine attributes of
the social actor with that of the media.

It is supported by current work
showing how knock on effect

of motions of the robot stressed.

Simulated victims in order to
study does not deny the robot.

There is a social media survivor but

it was designed to be capable of
wide range of Affleck to expressions

consist of poor drawings and
neck rules a plus or minus fifty degrees

speeds of up to two hundred ninety
degrees perception or not close or

minus ninety degrees in up to four hundred
eighty degrees per second and told

of negative eighty to ninety degrees in
two hundred seventy degrees per second and

negative five to ninety degrees for the
tour so tilt at eighty degrees per second.

This combination of joints allows
survivor buddy to produce a wide range

non-verbal effect of expressions

the expressions so all the things
you start seeing is that I'm going

to run with productively for Pixar to
get the basic emotions you are then with

it's like hand sense don't work fine don't
write actual waiters in the first time we

try to do it without that we've got got
some very clunky notions there but you

don't want necessarily to be humanoid but
you want to have the range of motion so

that you could be rigid and can be
very subtle and some of your motions.

I mean skip through victims do very
strange things they break your robots they

grab it.

We've had anytime we act you know we would
never think of manhandling a robot because

we know how much it cost usually and
increments of fifty thousand.

You know like so

one hundred thousand Sujal is
starting a hundred fifty thousand.

So technical barriers if
they are the victim is that

the human interaction isn't there for your
They're not thinking about you at all.

And there's a lot of need for
APIC to computing so

well all together we need to
start doing a better job of

tailoring this since seen the mobility
manipulation to actual environments.

Even though there's no vocabulary
to describe this environment so

if you ask me Well what is the train
compliance of mixed rubble and

the answer is I don't know I don't even
know how to measure it is so bizarre.

When you have carpet and and
harness and steel rebar and things and

there's also lots of human robot
interaction and this is because we're

seeing the barriers are it's formative
we've never done search and rescue robots.

It's doing things people and dogs aren't
doing so there's no way to cheat.

And it's as socio technical system.

There's a people probably
involved in it so

we can't just automate it already about
the mechanism and it's a system that.

If the whole thing doesn't work.

It doesn't work so
why not more rescue robots and

you know some of them look like they
were getting pretty good right.

There's First off there's
no monetary incentive

if you look particularly
if it's profit margin.

If you're in dear Dean making super them
high profit margin high profit margin and

consumer public safety very low and
better yet low profit low volume.

Yeah eighteen acquisition the people who
want to are responders and policemen.

But their acquisition goes through city
government state government sometimes

federal sometimes Department of Justice
and there's a high training problem.

Not that the training is hot.

I think it need be they think getting
a three day class on a new technology

is a huge incomprehensible deal and

to be able to do three days
a year on a technology.

So imagine how rusty how
quickly you can get in so

current ways don't work very well for
them.

There's only one bar rescue team that owns
a rescue robot in the United States and

that's New Jersey Task Force one.

None of the twenty thousand machines
do because they can't get the money to

purchase them because
there are no standards but

once the standards are in place.

I think the end of next year.

There is no money for them to buy anyway
so you know the teams use whatever

Jersey got it on a very crazy set of money
set aside right after World Trade Center.

Why not more rescue or
government regulations.

So if you just look very quickly here
these are cases where robots were not

allowed to be used in the aerial
vehicles in this case because of F.A.A.

regulations.

The Haitians there was a lot
of possibility to use it.

They are but they just OK it's just easier
to say no than yes to think about so

very frustrating and
in these very chicken and

when you think about acquisition
people because it's informative

formative nobody knows really
what unique type and eleven does.

Ousters and I'm only beginning to see some
things that look similar every disaster is

different so
nobody can really say what they need.

So whatever they get is not quite
right which is what you'd expect at

the beginning of an innovation
adoption cycle but

none of the government people want to
sign off the acquisition people and

in fact they want to wait longer because
maybe if they just wait another six months

it'll get better but
you can't get the improvements and

so the actual users have them in hand and
start using them in modifying them.

So they'll be adopted.

So that school every.

Another reason is that we've been
thinking about the innovation chasm.

I'm sure familiar with that
where you can get it here and

then you know when you actually get
the mainstream people to do it.

But it's not just the innovation part
you know talked about an acquisition.

There's an educational component what I
continue to see is that the people who

know about the social cognitive side
versus the technical mechanical

side are like ships passing in the not I
think we've just hitting our students and

graduates monitor vision.

You know and I wouldn't be doing this
work if I hit and taking courses.

When I was in mechanical engineering but
I took the lectern over and

I asked why I eat and then as a graduate
student because of the sims program market

take classes outside of my major that
turned out to be a huge help rather

than just simply being it and
fit my schedule.

Let's also talk about robotics and

ethics because I think that you know
we've seen a lot of discussions about

that lately I think Bill Joy in his Wired
article The future doesn't need as was

really the first serious voice to complain
about robots and that they were good.

You know we were there with
genetic engineering and

nanotechnology is things that people
should have never come up with.

And lately we've seen a shift toward the
military in fully autonomous robots was

the great work by Dr Ron Arkin to
Dr Nyce are there but I think that the.

While these are all important discussions
to have I think we're forgetting

the morality of not using technology will
we could that has to the T is as bad

as deliberately saying we are using
to help ourselves our neighbors.

Because we certainly could have used
workable robots at Oklahoma City

at will trade center at Banda Haiti and
any number of hurricanes and

as engineers and scientists we know
that no barrier is insurmountable.

And that many of the barriers you saw you
already have innovative solutions to.

It's not cheap or convenient but
robots robot systems

are needed and are possible and
so we need places

like Georgia Tech to innovate and educate
and keep us thinking along this way.

So thank you very much for
having me here and

I'm very proud to be a rambling rock
from Georgia Tech and how the engineer.

Was

very much a problem and
we have time for soon.

Q. and A We have I think microphone the
microphone in there's a third microphone.

I know some of you have to go to
twelve noon we understand we get it

but if you have a question please
proceed to a microphone man

that is the robots approach people and

it was often creepy scary when you
are in one school and you discussed it.

You did the doctors or you talked about
how doctors approach people with the Duke.

But you didn't discuss that there would
be a potential employer when I'm in

a doctor's It says I'm going
to listen to your heart.

Now she says I'm going to touch
you to look at your knee review.

He tells you that if we reached him from
Surely with a humor puts a stethoscope on

your chest was there a voice coming
from the computer too I mean

from the robot telling these people.

Let's come to great point and

you are absolutely right in that study
no because we had all of that but

what we discovered in working with
the medical doctors and we all take

turns doing victims and doing all the the
Taliban is saying is that after a while.

It's annoying.

How many found doctors to be rude.

You know let's not make eye contact which
OK You're lucky to get one look at you and

the up not just like caps.

So what we're finding is that and
we see over and

over that the lack of
eye contact is that so

now you start giving Ok let the structural
engineers in the medical people and

they can talk into whatever point the
robot the base of the robot wherever they

want to but have something that
tries to keep you there and I know

some of y'all were laughing about watching
T.V. What do you plan to do for ten hours.

I mean that the Chilean
miners all wanted to shoot

the two miners were trapped in
Tasmania and after about an hour.

You know they were fine.

They they checked and

they were trapped was all they were now
find evidence that they said we're really

tired of talking to you if you just
send down in P three player with the flu

fighters which is why there's
the ballad the Beaconsfield miners.

So you get charted that.

So it's very interesting but if the key
mansion to now where is that having

something be rude to you and
not being able to mentally clip now since

we're going up a wrong that
little of going yes but that way.

Thank you.
The new questions have to be

here and there were a bashful audience
who we have another question it sir.

You know they should like to know what
something that you believe in but

that not many other people believe
in that certainly is open and

I think the biggest thing that I have is
that many people don't believe him or

use the human robot interaction either for
interacting with the operators

in the session acres nor the difficulty if
you're the victim of that interacting with

you know I think I continually
continually see that over estimating so

the Crandall Canyon you know the robot
that was lowered down the whole game and

then it has to flip up like this.

So it goes the bore hole lands and
then that camera that can work around.

They built a force in three days and
not in services can a company very

great team doubts and the engineer with
us again we put it together and it's a.

And we're starting it.

You know testing the parking lot before
we take it up the mountain which

is a two hour drive and

is away when you when you start doing
it up like this count to two no.

Well OK down to ten because if you go
more than ten it will jam in place and

you will be able to lower it and
we're like you know stop.

Maybe your heart stopped on it maybe
an encoder Why didn't you give us at all.

No we tried it in the lab and the office
was it's get used to one thousand one one

thousand to one to get away at
that robot in the think for

over twenty hours we're
trying to is it down.

Is it that you know and but then
remembered that I get this all the time.

Let's give you a display and
let's put start putting writing over it.

Let's put it on e a Palm Pilot.

Ari I thought I'm doing
a visually directed task and

you're giving me smaller and smaller
screens thing you know this is helping and

people look at them but
in a science book or so.

I think people double and that is
the thing that I don't believe that when I

say I think he threw another question is.

Thank you for this talk a bit yeah
the options rest are about the teacher.

I personally don't know very
much about Christie very much.

But when I see that title and
if you agree.

It seems like even a good headway search
through what's his or actually rescue or

is that the one in the body bought or
whatever it's called Risky body

is there anything else that robots
are looking to do is actually rest.

Well certainly there is extra care.

So there's the binding here and
then what do we do when we found it.

Well one thing is to keep you amused and

it leaves not actively creep you out
are no idea but could we help do that.

Well it turns out one of the things
that they may use the robots for

is structural engineering because it's
like pick up stats one of the reasons why

it takes ten ten hours is
they don't really on average.

They don't know when
you know if I do this.

What's that going to cause here.

So now you have something inside that
can help us and see what shaking or

the little telltale sounds.

Now when you start thinking about
things the size of Chibok cells.

You know now seriously how much
manipulation are we going to be able

to do so what we really think you're going
to able to do in chickenshit is more

like hope is that was that full of

good fun you know dragging people out.

Unlike like the love
down the back the bear.

You know the bear robot
that's that's designed for

military where you're still
in a human habitable space.

And you've been shot.

If you try people are legally allowed to

go screw you up here you're
out of a burning building.

They have certain procedures
that they have to use and

mostly because you might have
a spinal damage or things like that.

So they it's unlikely and those are really
small spaces were not going to go.

Just dragging you out of the building and
if we could have dragged out a building

a person would have done that a long
time ago before the robots are right on

the scene because even the professional
are apart apartment will have the regular

first responders they don't have
a special operations if their major city

Atlanta has a big one and then you'll
have a regional one in the all share.

It'll take three to four hours
to get physically over there.

So if there was any person that neither
to carry out in Wouldn't carried out

things with a little like people
that one person said to me.

Six of the question over there you know.

Thank you very much for the bird work and

you have very mixed as you
background counsel amazing work

span different feuds of your
champions your structured into

computer sign speak on this
with us this is your signs.

So if you know from your experience want
to fly a twice suggestion you have to keep

our young engineering students here either
graduate or undergraduate read a student.

Watch a book you know what to make
you have this kind of amazing work.

Well I'm going to be
good think a suggestion I

would have is to just be aware that you
probably aren't the first person to have

ever thought of this before writing and
get people who are sending like

look you know I'm going to do
something with computer vision and

it was like guess what that was
done back in the seventy's.

Go get the best algorithm.

It's on the social and

the contents of engineering student
amounts of work have been done there.

So they just the kind of people love it.

They just hysterically because that
human out of the loop control program

has problems found by and by Wiccans
they've they've found all this back

in the seventy's and they just allow them
as if saying you know you guys an unmanned

aerial vehicles are determined
to replicate every single error.

Man why.

And autopilot just and you're going to do.

Yep they're going to crash for
this reason you know up and

then they're going you know and it's like
white people thought of this people.

It's just not not thinking to look and we
live in such a rich environment where we

can look and see access to books access to
other experts and farm and such a short

attack that encourages multi-disciplinary
work and we just don't think to do it.

That's the thing I think
the thing outside the box but

the more questions are yes Hi Yeah I
talked to the idea for the video

link it's it's really interesting because
you can give you know people are trapped

an update on their status from a rescue
worker where you keep entertained.

I was just wondering if you could comment
on the how to how would this thing.

How would such a system be you know
practically implemented because earlier

you lecture in the like you'd said that
wireless communication at such depths.

You know is pretty much be
possible just to see to comment.

Well the robot's going to be on a tiger
you just run it through the robots

tether as well.

The cop becomes Lincoln power.

OK but I mean people now practically all

the bots ground robots or
tether clearly the aerial vehicles and

they would it be said that most people are
attracted between forty and sixty feet.

He is going through
a lot of rubble right so

we're going to be using a tether you're
not putting a robot into forty or

sixty feet of rubble without a rope
wanting to get it back because it's not

going that way it's going this way and
maybe this way and that way nest way.

So you always have a tether.

And by the way to make the robot
this big you pretty much have

to have a power off or right because
if you have power on board and

it's working on the really dense heavy and
they have at their motors or

I have been around I have
to have bigger batteries.

So the Pack Bot series is
about the best you can do and

they're matter what they
tell you they only last.

And maybe twenty minutes in going through
very intense rubble going lot across

you know doing I.D.'s on one
side where it's very different.

So if you assume you have
a tether that gives you power

perfect power perfect cops except when
that doesn't work who broke in tatters and

just has a thing but you can now do that.

So now you've got your two way video
your two way audio you're your video

conferencing and things should be
pretty reliable going to get it so

you keep it you have the last question.

They get their Did you create
no pressure no pressure.

You mention that.

You mentioned that each robot cost about
fifty thousand dollars is that only

because data each of the custom belts and
the parts are so expensive or just that

the manufacturing process has not reached
a high enough level yet to make it cheap.

I think it's a little of so

which do you think is harder now that
you've seen the talk in all of this.

Do you think which is are the Mars rover.

Are a general purpose
underground rescue robot

with a general purpose
Mars rover very expensive.

How much cheaper Do you think a Mars
rover would be if we need a hundred after

an instead of two of them still
prime probably pretty pricey.

I mean probably have to get to
the extent the automobile industry I

think to make them feasible.

Are you accept the fact that
something's just costs more.

How much you know is it OK to
have a robot that cost as much as

your tricked out S.U.V. The response.

Probably if the tools useful
enough money isn't the problem.

It's getting tool expensive
tools that don't work.

That's the problem.

So if you need something that will work.

I think that we can then
a system that works and

is proven to work then we can get.

Or work on Bill from their doctor
killer to get more money to buy.

So I get down to a little
bit more market value.

It'll be a small market that public
safety is a small you know so

I don't send him orders are spots in
this business that markets right or

that you think the let's give
Dr Murray the round of applause.

We have a token of appreciation
in you being or Guggenheim or

lecture you may open it.

It's not a robot.

Maybe we need to change in the future and
change the mood to a robot.

But this is what's fragile

and then we also have a model of
school medallion for you guys.

So literally get it
super tall very good and

we will have a reception
in the galleries outside in

the lobby after we finish up here so
I invite all of you to stay.

I know there are a few of you who are a
little bashful and had great questions but

didn't want to get in front of
the microphone and I know Dr Murphy

would be delighted to take your
questions in the Richards gallery.

So thank you all again one
more great round of applause.

This is my favorite topic so I am happy
to talk about I really think you Richard.