(Music)
- Hey folks. Thank you for joining us. Go ahead,
don't be shy, jump in the chat, send us your questions but before we do the whole rigmarole,
Terry is already an alum of the podcast. So, for
folks if you are not familiar with the podcast, we
do the... we recorded an episode last year for -
- A year ago, yeah.
- For CES but we did a Throwback Thursday, it was
what the crazy kids do nowadays online. But that
episode is out there where we talked to Terry in
detail about his background and stuff, but Diana
this is the first time that you have joined us. We
always start off the same way with, just tell us
about yourself. How did you join NASA? What brought you to Silicon Valley?
- Yeah, so I never thought I would work at NASA.
It was not a dream as a child. Actually, NASA showed up on my radar back in 1997. I was
a high schooler, a young high schooler at the time
and NASA just sent the Pathfinder mission to Mars.
And so we had the Sojourner rover driving around
Mars and I thought it was fascinating. Not really
because of the science, like Jim and Greg last
time, but more the problem that they had to solve
to get there. NASA hadn't been to Mars successfully in 20 years and since then, we
have been back there eight times but for me I thought,
wow this is a really hard challenge that I want to
know how to solve. I want to be a part of that
problem.
- Like in high school or college?
- That was in high school.
- Oh wow.
- That was early high school. And so I didn't really know quite how to get in that world.
It wasn't necessarily the rover problem I wanted
to solve but I wanted to do something really
difficult and overcome that. So, I found myself in
engineering in college and wound up going to
aeronautical and aerospace engineering in college.
And as always looking for part of the field that
was changing, that was evolving, that was growing
and some of the things were, well, we figured it
out it. And it was just a matter of optimising and getting it just a little bit better. But
I wanted something that was rapidly changing
and with computers and control systems, that seemed
like the place to be.
- This is just like an internship or something that you ended up jumping in or were you working
in the field?
- I found a professor that I really liked working
with and so was working that, and involved with an
internship with JPL, down at JPL.
- Our friends in the south.
- Yeah, exactly. And then afterwards, I was just
looking at places that did the kind of algorithms and math and engineering that I had learned
to do in grad school and found my way here.
- Nice.
- I met people at a conference, came and gave a
presentation and received an offer.
- Awesome. We'll go a little bit more into some
of the stuff you are working on and some stuff that Terry is working on. And, of course,
we'll get questions from the chat. Folks, if you
are joining us, you are watching the second ever
episode of the NASA…NASA in Silicon Valley Live.
This is a conversational show on Twitch TV. With
various researchers, scientists, engineers and all
round cool people at NASA. Specifically, here at
NASA's Ames Research Center in Silicon Valley. So,
as I mentioned at our premier episode, we are
trying something new here. We are basically taking
the audio podcast and doing it live on Twitch. And
so, last time we had a lot of fun talking about
the moon and today, we are talking about self-driving robot, planes and automobiles.
- First and foremost a shout out to the live audience on the chat. We are going to kick
things off by talking with our guests and we're going
to try to answer as many questions as possible
from the chat and based off of last time, we're
going to try some rapid fire questions at the very
end, so don't be shy, send in as many questions
as you can. Or just feel free to just send emotes
and spam that at us non-stop in the chat, because
we're going to be looking at it. So, I am your
host, Matthew Buffington and this time my host
Abby Tabor will be looking and taking the questions from the chat. So good luck with
that Abby.
- Oh yes, thank you very much. I look forward to
the challenge. And first of all, we already have
some action. So let me say hello to [CafeMedfica]. I'm going to butcher your handle, sorry about
that. Hello from Sweden and [Rigaydee], hello there
to you too. And now, let me introduce our guests.
Right here next to me I have Terry Fong, chief roboticist at NASA Ames and the lead of the
Intelligent Robotics Group here, right?
-Yep!
- Excellent, we gotta hear what all of that that
means in a minute and right down there, Diana Acosta, aerospace engineer with the Intelligent
Systems Division at Ames, right?
- Right.
- And you're also working on innovation I believe.
The NASA Innovation Collaborative Initiative. That sounds intriguing. So, you'll tell us
more about that later?
- Absolutely.
- Great. Welcome .
- Before we get into the good stuff and talking about robots and self-driving cars, I want
to sort through some housekeeping. We're going to
do this podcast live and on Twitch TV/NASA, for the
next couple of weeks. We'll be back next Friday
at 2 o'clock, 2 p.m. Pacific time and we have a
special treat for that episode, where we are doing
a little bit of a "Let's Play." So, I'm sure
that Twitch audience is a little bit familiar with
some, there is a lot of space video games out
there. So, get ready for that for next week. But
for this week, just to let you know, we really want your feedback. We are figuring this all
out, trying to - any feedback, advice, stuff you
want to tell us, just let us know. If you cannot
catch us live, that's no big deal. You can catch
us on YouTube.com/NASAAmes afterwards or on podcast
services throughout the solar system and beyond. But right now, the plan is to have the versions
up, I think, on Monday, is we are going to have
those on demand versions up.
- But, now that we have got to know Diana and
Terry a little bit, we can just jump right into
the conversation. So whenever we are talking about
self-driving cars and in NASA-speak, we keep referring to it as autonomy. I purposely did
not put autonomy in the title of the show because
that was like, I don't know what that means to
most people. So I replaced it with self-driving.
But is that really fair or accurate? Let's talk
a little bit about what is autonomy? So, Terry..
- Autonomy means that you do things by yourself. I
mean, it's a simple as that. I mean, my cat's autonomous. My kids are autonomous. Probably
more autonomous than I want oftentimes. Robots
can be autonomous and that really just means that
they can go off and do things and achieve goals
or objectives that they are carrying out by
themselves. And whether it is self-driving or
autonomous, frankly, I don't care. It just means
that hey, they are off independent, I don't have
to be hands on, I am not joy sticking them, I say
hey go do something and hopefully they will get
something done.
- Yeah, a lot of people make references to joy
sticking it. So, this is literally like a video
game? You're driving the rover or you're driving the machine, you are operating it. That's
the idea?
- Yeah, I mean. People look at NASA, they think,
Oh, my God you have the most advanced robots out
there, but sometimes we are joy sticking it and
that literally means we have hands on the joy
sticks or we use the technical terms: hand controllers. And we control how robots move,
whether their arms, robot arms or or they're free
flying systems or rovers. A lot of what we try to
do these days, at least in research and development, is to go beyond that. We want
the robots to be, you know, more independent.
I don't want to joy stick my kids to say, go left,
go right, stop, come back. I say hey, go mow
the lawn or go to the store and get something for me.
And I want robots to do the same thing.
- I have a question here that might be relevant. [Radiateurs], maybe? Are you going to talk
about deep learning or neural networks? Genetic
algorithms? What kinds of things are controlling these robots that connect independently.
- Well, I mean there are lots of different things
you can use to make these robots or systems in
general more independent. There is a tremendous amount of research going on today involving
machine learning, deep learning, AI. All these different words you hear out there, but at
the end of the day it is trying to make the system
function more intelligently or in a way that seems
more intelligent. That is, you want it to be more
capable, more competent. You want it to do something in a way that seems to make sense.
I look at like a, like a… Lots of robot vacuum
cleaner out there.
- Like the Roombas and stuff.
- Like a Roombas you see. Roombas can do a great
job, but you look at them and you have no idea
what they are doing, because they are kind of
wandering around and bumping into stuff. It is
not really clear, you know, how they are cleaning. And so I look at that, is that intelligent?
Well, if I watch it, it certainly doesn't look
intelligent but it can still do a good job.
- It gets the job done.
- It gets the job done. But there are other systems out there, you can see they do these
lawnmower patterns and it's very obvious what they
are doing and you think, that looks more intelligent because it is doing this in a,
you know, very careful way. And a way that's very
efficient. But I look at a Roomba it is doing this
kind of stuff, yeah, I'm not sure how smart is
that.
- How does that match into some of the work that
you are doing Diana? Because I know you and Terry
work quite a bit, but as a fancy aerospace engineer. How does one go from aerospace engineer
to working on autonomous systems? I'm guessing that aviation… How does that work?
- Actually, when I was studying aerospace engineering, I focused on machine learning
in terms of being able to not collect an abundance
of data but be able to take the data that you are
receiving at the time from all the sensors that
on the aircraft or spacecraft or airship and be
able to utilise that in a smart way. Learn from
it and then control that vehicle in a way that
will be successful to achieve the goals. It is
quite a bit different than the big data that is
going on within industry, around us, especially in Silicon Valley, where they have a lot of
data that they are collecting and they are using
that to be able to, in general, assist the humans
and provide information to the humans or make
sound business decisions. In our cases, whether
it's a robot or a spacecraft or an aircraft, we are
looking at taking that data and being able to let
the robot or the aircraft or the spacecraft make
the decision itself. And act on that decision. So
I think that's the differentiator between what we
see outside the gate and what we are doing on the
inside.
- OK.
- I'm not sure the chat is blowing up..
- Yeah, it is. I'm already way behind in questions. [Valask] is saying" I can get behind
self-driving cars. With self-driving planes is it
the same idea of autonomy? is it a lesser degree
of independence for a plane than a car?"
- You know, airplanes have been self-flying for
decades.
-Okay. Interesting.
- We have been flying with auto pilots, and flight
management systems, and the pilot can get in, make
sure the company programed the route correctly, and go. Now, they are fully responsible, as
a pilot, when you're going on a trip, they are
fully responsible for the safety of that aircraft
and monitoring all the various systems. It is
incredibly complicated. But aircraft, they have a
nice safety buffer between them and other aircraft. It's pretty predictable. We understand
the impacts of weather and other contingencies. We
know how to handle that. So, in a way I see aircraft as being on the forefront of that
self-driving or self-flying area, and it's a lot
further ahead of the game. The car is so much more
challenging with all the dynamics. You know, cars
and kids and pedestrians and weather and sun and
oh, it is very complicated. And…
- I have a question about weather. What sort of
considerations have weather conditions had on
designs of self-driving vehicles? Do you know how
that's considered?
- I will ask Terry to take that one.
- Well, so we do a lot of work these days and
trying to make robots function more robustly across all kinds of conditions. Whether it's
weather, or frankly, traffic patterns or congestion, all kinds of different things.
In terms of cars, we have been working with some
companies in Silicon Valley that work on self-driving cars. A lot of the challenges
have to do with the fact they all rely on sensor data.
You know, cameras, radars, laser systems, to
understand what's around them. And if you have
weather, which fortunately we don't have is a
whole lot here in Silicon Valley, then you...
- You have the best weather.
-[Crosstalk] We have the best weather.
- We don't get rain and fog and snow and sleet, those are the kind of things which make all
the sensors stop functioning. It's lile any of
us trying to drive on the roads by saving "Hey,
okay, where am I going to go next?" The answer is,
it's really hard if you can't see. The same is
true about self-driving cars and robots.
- Okay, right. Along the lines but specifically about GPS, [Cafe Medfica] is asking, is NASA
working with big car makers to make GPS better for
self-driving cars? Is that an important part?
- I don't know if anyone across NASA is doing that. I mean, GPS is used a lot, not just
in terms of like cars, but terrestrial robots. Agricultural
robots, drones. They all rely on GPS for understanding where they are in the world
but the reality is for things like automobiles, it
is not sufficient just to rely on GPS. You need to
worry, for example, am I exactly right next to the
car, you will not tell that from GPS.
-It's a finer scale?
- It's a finer scale. So, you're going to need
other sensors. We rely on all kinds of things to
give us very precise positioning information, especially how close to objects that we might
want to stay away from. And so that's far beyond
GPS.
- Yeah.
- I will jump on in, because if you are just joining us now you are watching NASA in Silicon
Valley Live. A new conversational show that we are
trying out on Twitch.tv/NASA and are chatting with
Terry and with Diana about self-driving robots, planes and automobiles. So, we're going to
keep taking as many questions as we possibly can.
But I gotta jump in. Because there was one thing
where, when we the podcast, Terry and I were talking.
We were talking about the early days of autonomy,
you were talking about 3D mapping and how even
some of your early work with VR, helped kind of play
into some of that. So, maybe we can talk about
some of the early days and I think not like self-driving
cars and VR, very much buzzwords, anybody who is
into gaming, is hearing all about that. So, talk
about some of the early days and I'm pretty sure
Bill over in the back has some cool images that
we're going to pop on up.
- So, I mean, a lot of people think that VR is
something that just happened a couple of years ago
but this is actually the third or fourth wave. Actually, the picture that's up here was from
about 1990 or so. There was a lot of research and
development here at NASA Ames and Silicon Valley,
looking at different VR headsets. You look at
this, and it looks kind of clunky, but at the same
time it has a lot of things you see today. It's
just that today, they are a hundred times cheaper,
they are higher resolution. But, some of the basic fundamentals of how to put people in
these virtual worlds was done here at Ames. And
we used that…
- That's awesome!
- …a long time ago because we were interested of
transporting scientists to other worlds. The idea
that you could remotely explore Mars by using a
head mounted display, maybe you have some data
glove on. Maybe you're trying to reach out and
manipulate a virtual rock. These were the kinds of
things that we were really interested in, you
know, even back 30 years ago at the previous wave.
Or maybe it was two or three waves ago of VR. But
for us, at NASA, it's all about how can you better touch the data and immersive 3D, rendering
3D user interfaces, head-mounted displays. That's
all part of that. This is actually, the screen shot here is from a robot control interface
that we developed here back around 1992, it was
called VEVI. It was the Virtual Environment Vehicle
Interface. It wasn't a particularly good acronym, but the idea here was we could use it to remotely
operate robots by way of a VR interface. So we
control the robot in VR. It sends commands to the
actual robot, which might have been thousands of
miles away or even on other planets. The idea here
is that we interact with the data and that allows
us to better understand what's going on with the
robot and then the robot can go off and do its own
work.
- Alright. Abby, how is the chat going?
- Oh, my gosh.
- It is blowing up.
- Our friend [CafeMedfica]: "Dope show, thank you
very much." And Diana. How will, from [Jay Stubbles], how will automated aircraft handle
perilous situations such as bird strike? Will it
look like a landing solution? Like Sully Sullenberger's Hudson River landing? Like
in the movie Sully.
- Absolutely. Actually, one of our colleagues, David Smith, he's actually retiring today,
so shout out to David. And he developed an emergency
landing planner with some of our colleagues and
what it does is, that flight management system where the company can put in the route and
the aircraft can follow from one point to another
to another. The emergency landing planner helps
to look at where the aircraft is, given the
circumstance, whether it's a bird strike, an
engine out, fire, whatever the emergency might be,
and take into account all of the airports within
the area and what emergency services they have,
and any weather that's between the aircraft and
those landing points. And then it develops a
route that can also take into consideration the
ability of the aircraft. if you lose an engine, you don't want to turn in certain ways. You
want to turn one way or the other. If you lose
part of your tail, you certainly don't want to be
doing certain maneuvers. So it can take into the
maneuverability… take into account that maneuverability of the aircraft and provide
routes and suggested landing points for the pilots.
And give even explanations as to why this airport,
why not that one. I can see that one, that one
is over there, and so it helps drill down. So David
Smith along with colleagues in the Intelligent Systems
Division and the Human Systems Interaction or
Integration Division have worked that out and
yeah, absolutely, we're making progress, here at
Ames.
- Awesome. Cool to hear about. A couple of more
comments. [Noxum96] likes I'm smiling at the laptop all the time because this chat is crazy.
[Laughter]
- Thanks for noticing! [LaraBug] says "Thanks so
much for doing this stream." And here's an interesting one. [YoungReefer] asks "If these
self-driving things are using AI, can they take on
solutions they have created themselves? "
- Yeah, that's a great question. And I hope the
answer is yes. Because I don't believe that anyone, certainly not me or people necessarily
here at NASA have all the answers. In fact, what
we really would like are systems that can be
adaptive to the world. It is really difficult to
sort of like program a system for every single thing that could possibly happen. In fact,
we don't train ourselves to do that. When you
teach someone to drive a car, it is not like these
are the 100 things you you're ever going to encounter
and only these hundred. Instead, you try to teach
them how to deal with different things. Oh hey! A
tree fell down on the road. Or maybe, for some
reason there is a cow in the middle of the street.
At least maybe not around Mountain View, but in
some places, that could happen. And so we want are
systems that can really learn or at least adapt to
changing circumstances. That it's not stuff we
programmed.
- Yeah, it's gotta be. Especially if that's space exploration, right? We don't know what
we we're going to find on Mars.
- That's true.
- So the challenge for us, as developers, as
humans, is to really blow out that space of what
could be the possibilities so that the computer can really fill in its knowledge. And then
act on that knowledge.
- Isn't that not too different from the humans? What you actually physically see, your brain
fills in a lot of gaps. Because I mean, it is how
optical illusions and stuff work. I would imagine
that robots, the software has to take short cuts
as well, to be more efficient, I'm guessing?
- I think that humans make a lot of guesses and
some of those guesses are hopefully educated. I
mean, in general, we don't let drive until they
have learned how to drive and pass the test and
did a road test and that kind of thing. And maybe
at some point in time we going to have something similar for say, self driving cars or robots
as well. Because one of the, I guess real problems
we have, is that when you guess, usually you
will guess right. At least humans, we like to believe
we can guess right, but you're right. There are
times we guess wrong or we have incomplete information and we make a mistake. Those are
things that I do worry about as well for any sort
of autonomous vehicle or robot tjat we create. Whether or not… are they going to make the
wrong assumption? Are they going to leap to the
wrong conclusion? But we'll see.
- Do you have more, Abby?
- Yeah. I hesitated because was a long handle that I had trouble pronouncing. So, [TigerionDono]
was asking… Sorry, wait. I mixed up a couple of
people. [Post VT] wanted to know what are your
thoughts on AI and self learning tech. Do you
think this could cause problems in future? If so
why. I think we touched on that.
- It's always the SkyNet reference.
- Yeah, sure. SkyNet and whatever else is going to
come take over the world. I tell you, one of the
things we worry a lot about at NASA, because we
tend to build systems that are very expensive, then they go to space or they're flight systems,
is how do you really sort of test and make sure
they will work as predicted? You know, the terms
that we typically use are things like verification, validation, certification, all
those "-ation" kind of things. But that really means,
can you make sure they going to work as planned? And as soon as you let something learn, and
adapt, the question is how do you test that? That's
a really difficult thing and I am not sure we
really have the answers. Maybe Diana has some great
thoughts about that.
- No, we certainly don't have the answers there.
When we are not aiming for the highest goal, we
can certainly let the systems learn, let them make
decisions but then give them that boundary. Like
you might a child, putting the child in the play
yard, and the play yard is fenced. Go off, have
fun, I know you're not getting past that fence. So
we can do that with our robots and with our systems that we equip with AI, but once we
recognise that we want the robots to achieve more,
we want them to accomplish more and serve us
better for the NASA missions, we have to take down
the fence and trust them. So how do you establish that trust, if you are putting millions of
dollars into that mission? Or you are the scientist
who waited your whole life to get the data from
that mission? Verification and validation, certainly
important.
- Now, I do want to get to [TigerionDono's] question. It's a great question. Lots of science
produces lots of data. More than can retained. For example, CERN's Large Hadron Collider,
the LHC. How can you teach then self-controlling
vehicles what to send back for review and what to
just ignore?
- That that is a really great question. I think
one thing that people may not be aware of is that,
every single mission that NASA sends into space is
capable of acquiring even more data than we have
ever had before. And at the same time, we still
have a same narrow communication pipe to send things back to Earth. Which means, you know,
that oftentimes we have this real problem of how
do you sort through all the data that's collected
and figure out what to prioritise, what to send
back. So, it is true that it is a really challenging
question. And I think one way to kind of address that is to build systems that can do more
processing on board and try to do some interpretation of that. That's a kind of
information we like to have used on Mars, for
example. We have a lot of interest of being able
to track these things called dust devils. The kind
of swirling little cyclones because they have a
practical use for us with our solar powered Rovers. Spirit and Opportunity, we like to
get them in a place where the dust can come by
and clean off the solar panels so we have better
energy.
- But not too much!
- Not too much. A gentle blow dry kind of thing,
you know? But in order to do that, we don't want
to try to loop all the data back to Earth and make
decisions because it takes too long and by that
time the dust devil would have passed and the idea
is to track it on board. So, that means you have
to do more processing of the data on board the
spacecraft or on board the robot. I think that's
where we are headed.
- These are one of the things we talked about in
the podcast before. That is not just like Mars.
There is the speed of light. Information can only
travel back and forth so much and even when you
are out by Saturn and further out, there is always
going to be that delay. So you need autonomous systems or else, the delay is going to be
so extreme.
- And then I was going to add. We were talking about one system trying to communicate with
Earth. Now, if you add multiple spacecraft and you
need those spacecraft to communicate with each
other to be able to acquire the data, in the right
sequence, at the right time and be able to fill in
for anything that goes wrong. Prioritising that
data exchange, that knowledge generation across multiple spacecraft is a problem we are working
on.
- Awesome. I had a question, we have talked a lot
about this self-driving cars. OK. [ReynarTheConquerer] wants to know, I would
really like to know what is NASA doing with autonomous
vehicles other than the fact it would be required in a rover for a pilot-based system. Basically
you guys are probably looking forward to sending
people to Mars, so how big of a part of your work
is that? Human-controlled Mars rover versus what
else you are working on?
- A lot of what we are doing right now is trying
to make planetary robots more capable because we
want to send them places that are more difficult to get to. If we want to send a robot into
a lava tube, for example, not so easy to have continuous
communication when you are down underneath the
surface.
- That's like a cave?
- Like a cave robot. If you want to send robots that are far away, as Diana was saying, if
you go out, even just Mars, for example is 20, 40
minutes delay, if you go beyond Mars, it is even worse.
And so, you can't just have humans and sort of
like real-time control kind of situations. And so
the key is for us to create systems that can function by themselves. You know, maybe they
are not going to make all decisions by themselves,
maybe they will make decisions in some way that's
achieving a particular purpose, like driving from
point A to point B is a good example of that. Rather than just trying to say, go discover
stuff, you know. We might say, hey, at least go from
this point to this point and then we will make
decisions about the science we are trying to
actually carry out. But all of that means that we
need the systems to be more reliable, more autonomous, more able to make their own decisions.
And that's a lot that we are trying to do here.
- And so if you think about human exploration, you
can also make a parallel back to your home when
you go to work every day or you go to school, when
you leave your home, you can turn the thermostat on and trust that you will come home to a
warm comfy place. Or if you want to have stew,
you might turn on the crockpot and trust that
all things will go well. And if you want clean
clothes, you can go to bed turning on the washer
and throw it in the dryer when you wake up. There
is other things like your oven, you might not want
to leave on while you run to the grocery store to
get that extra thing.
- [Laughter] Maybe not.
- And when we are going to be sending humans to
the moon or Mars, we will want to send their habitat first. And there's many different
mission operations, but one of the main concepts is
have a precursor mission where we are sending the
habitat ahead. Maybe assembling the habitat ahead
and getting it running and operating autonomously
before we send the humans there. Alleviate risk
and I know this pod, or this broadcast is geared
towards robots and airoplanes and automobiles but
think about your habitat. Your house. It is then
self, taking care of itself. You have to monitor all the systems, all the water, life support,
everything. Because those humans rely on everything that we send. And that requires
autonomy.
- Fascinating. That's a cool comparison to real
life.
- There's one thing you touched on and we briefly
started talking about before we even started, you
mentioned multiple systems and like even like swarms of satellites and that's another thing
that like not only just NASA but like the scientific
community and makers and like universities have
been putting small sats up. Talk a little bit how
you could use SmallSats as a swarm? How does autonomy play into that? And why would NASA
be involved in this?
- Right. So, some scientists are looking at studying the sun and I'm not an expert in
heliophysics but people are.
- But I have friends.
- We have friends.
- We know people.
- And they want to be able to study the magnetosphere or the stuff that comes out
of the sun and bombards the Earth all the time or
gets closer to Earth and then we are protected.
But they are look at sending multiple spacecraft.
You can't just send one and get the data from
one to get the big picture of the kind of the tide
that the sun is sending out. It is really a shifting
and dynamic tide. So, they want data points across
that whole tide
A mosaic…
- And so you think of a string or string of satellites orbiting in this tide, and can
we collect that data. And do that autonomously.
Now, the reason to push towards autonomy though,
is not that we have so many spacecraft, because we
know how to operate one. You might say, well sure,
let's operate 100 the same way. It does come down
to cost though. If we have dozens of people to
operate one spacecraft, what is it going to take
to operate 100? And if we can have the same sized
team for that swarm of a hundred spacecraft, then
we will be able to accomplish the mission. If we
have to multiply the team by 100 for every spacecraft, it is beyond the budget. We won't
be able to do it.
- Well, let me interrupt because [BrooklynKnightz] will like to be reminded who we all are.
- Well, it is about that time. So if you are just
joining us, you are watching NASA in Silicon Valley Live. This is a new conversational
show that we are trying out on Twitch dot TV slash
NASA and are chat being self-driving robots, planes
and automobiles with Terry Fong and Diana Acosta.
So, that's who we are.
- I have a couple of comments to share. [monkaS] says "Thank you for this information." People
are appreciating this and [PostVT} says "This
is actually the coolest Twitch stream, freaking
NASA is talking to people on the internet. Like,
what the hell." I love that.
- Nice.
- And then, here is a great question I want to ask
from [Sunny_Deity]. "How would something like a
self-driving car or rover deal in a catch 22
situation? Like someone in danger from walking across the street on a red light?"
- That's a great question also. And it is probably
the single most challenging thing about making cars or robots autonomous. How do you deal
with the unusual situations? Especially the ones
that are have lots of life and death consequences.
You pull up, there is a car, you know, blocking
everything that you can see, and all of a sudden
you see the ball rolling to the street. Is that
going to mean that there's a kid running after it
right away? Do you slam on the brakes or you just
keep going? Because that's a ball, run over it.
Those are things that really difficult to deal
with and I think that's one of the reasons why we
see self-driving car development still taking a
lot longer than some people had thought. People are yeah, next year we will self-driving cars
everywhere. Here in Silicon Valley you see lots of
self-driving cars but you still also see safety drivers because there are all these difficult
situations. Well, some of the situations are ones
that you have to react right away. The ball and
maybe a kid coming out on the street. Other situations are still things that are unplanned.
The fact you turn down a road and Oh, my God there
is a tree in the middle of the road, what do I do?
Do I drive on the wrong side of the road? Do I
drive on the curb? Do I back up and go around? For
those kinds of situations, some of the work we
have done here at NASA actually provides a good
solution. And that is the idea of having somebody that you can phone home to. At NASA. we call
it mission control. I know a number of self-driving
car companies with looking at the call centers, like support centers. Tech support, if you
want to think of it that way. So the car might
get in a situation, and it phones home and then some
human will just, sort of, pop into the car, you
know, via some 4G network data transfer, and say
"Hey, what's going on? So, I see there is a tree
here. Well we'll tell you that you should drive
on the shoulder." That's an acceptable thing here
and that's how you solve that kind of problem.
- I imagine over time, even the system would still
learn…
- Yes.
- …in the unique anomalies and these unique situations, over time it can probably continue
to learn what… like the correct way. - Sure.
Exactly. So the next time you see a tree in the road,
you think "Oh, last time it was OK to drive on
the shoulder, so maybe we should do that again."
Or at least, that will be the start of a possible
solution around that problem.
- Cool. So I think this point we can do a little
bit of what I have just made up in my head called
"Video Roulette" because I know both Diana and
Terry brought videos we were going to talk on
over. So Bill and David over in the back, they
going to do video. And then whoever's video it is,
they will talk about why they chose this one. This
is –
- Who brought us this one?
- All right, this is my video. This is a mission concept that we have had here at NASA Ames
and it's not just about the spacecraft here, it
is about the interesting thing that will blow
up, which is well… expand, I should say. This
is called super-ball. It's a robot that doesn't
look like any robot you have seen. It's a basically
a collection of rods and cables and here we
are showing how the system, which is, in technical
terms a dynamic tensegrity system, can land all
the way from orbit and roll around. Basically, by
controlling the length of the cables we can change
the overall size and shape of the robot. Here is a
table top model of super-ball. So you can see we
have the rods and the cables connect all the different end points here and by changing
the lengths, we can compress, and we can expand.
I thought you see these sold as baby or cat
toys. But what we are trying to do is take these
and make them from being the things that you just
handle yourself to be ones that are true robots. And maybe Bill… yeah, here's a picture of
one of our current prototypes. This is super-ball
number two here at NASA Ames and showing it can change
its shape. It can squat all the way down. It can
become very, very flat, which will allow it to
scoot under things. It can change its shape to
become larger. You can imagine putting an instrument or rocket motor in the center,
so it can do all kinds of things. Here we have the
3D printed little camera as an example of how
you would put an instrument inside of this thing.
And it is really cool. We have been doing all
kinds of fun - this passes for work sometimes. Drop
tests here. Trying to say hey can we throw a robot
off the top of building and see how well it survives?
This is the kind of thing you wouldn't do with a
more traditional robot but allows us to to really
explore the whole new space of what robots can do
for future missions.
- Awesome.
- Cool.
- Can I jump back in?
- Go for it, Abby.
- I just want to mention there are a couple of
comments about "It's Jason Bourne! and "Matt Damon!"
- I get the Matt Damon thing. You should see my
older brother. Literally looks like Matt Damon.
- We are honored to have you with us.
- There are a couple of questions that people ask
twice. I want to guess to those. [surbazmeister] really wants to know, how is the research
in robotics advancing in the field of asteroid
mining? Do you know anything about that?
- I will admit I know nothing about asteroid mining. We do have a lot of work here at NASA
in terms of how do you make use of resources
on different planets. One of the biggest areas
that we are interested in maybe they talked about
this on the last episode, was in terms of things
that you can find on the moon. We are very interested
for example in going to the moon and locating pockets of underground water ice, because
we care about the hydrogen that's there. So that's
a resource and we identify a place to go mine
that. Then we will go mine it, but it's probably
not going to be like the mining we see here on
Earth. We will not have giant bulldozers and big
giant trucks hauling away tons and tons of material
but we will find ways of drilling down and excavating
quantities of say water ice, just so we can get
out the cool thing we care about, which is hydrogen.
- Yeah, cool
- And, of course, we will do it all robotically.
- And why do we want the hydrogen?
- We want the hydrogen, because we care about making fuel and we also care about water,
just to keep people alive. And water tends to be a
good thing to have.
- Yep.
- Nice.
- [YoungReefer] asked a couple of times, moving on
to spaceships far from Earth, we have built them
to be autonomous, but something could happen. Something goes wrong and they are too far
away from us to fix it. What can do you? - Is there
a solution?
- Well, it demands on what breaks, of course. NASA
is doing a fair amount of testing these days on
the space station of 3D printing. And so, we are
trying to look how can you 3 D print replacement parts, if a part breaks. If you are in deep
space and it's a spacecraft that has humans on board,
and someone falls sick, another thing we are trying to do, is try to figure out well, what
information does the chief medical officer in real
life, so not just in science fiction, have to do?
And it is interesting if you think about this from
a medical point of view. Here on Earth, especially in the United States, when you
get sick you go see a doctor, but it's not like you
see just one person. That doctor is tied into
the whole community. He's got all the support
around him. Labs and tests, and specialists and you
can get all the referrals. And that's great in
a connected world, not so great if you are out
on a spacecraft that's in deep space and there
is nobody else there. And if you want to pick
up the phone and call home, it's like, well, I will
call home and they will get the call like five
or six hours later. So, part of what we are trying
to do also is figure out what sort of on board
knowledge, maybe it's a computer system that can
help out diagnose things or treat things that will
give support to, say a chief medical officer, on
board a spacecraft.
- We are also, when it is non-human, when is a
spacecraft far away, we will put the spacecraft in
safety mode. If it detects that there are something wrong, we do have system checks
that are checking for the health of the vehicle and
that that will go to safe mode and wait for that
call from humans to tell it what to do next and
how we can continue on with the mission or do we
have to abort. Then we are also looking at bringing
some of that intelligence to the spacecraft itself.
So it can figure out what is wrong, what are
my capabilities and make that decision without
waiting from the call from home.
- Cool. - Yeah. Done a little bit of that on the
aircraft too.
- Interesting.
- Lots of parallels between the different domains
- One question, this is related to autonomy. Obviously, nowadays you think people think
about like the drones that go in swarms or self-driving
cars, and you know, like my favourite sub-reddits are like R slash self-driving cars and everybody
is looking forward to this future where I can call
my car and it will come. Of all the talk that happens with self-driving and autonomous systems,
what is your guys' opinion of the gap probably between what's really going to happen and
then what's like more science fiction and hopeful
thinking. So kind of like, what is the stuff that
you can see or that excites you?
- Yeah. You want to take that one?
- You go first.
- I don't know.
- You want to go hopeful or buzz kill?
- Yeah. It can go either way. I'm really excited. This is a really exciting time, especially
given the career that I have chosen. People see
all these changes and the investments and the
thousands of people who are working these really
challenging problems. And I honestly don't know
where it's going. I am eager to find out. I'm very
fortunate and proud to be part of it. And it's
going to be a fun lifetime.
- Good time to be alive.
- Yeah.
- I will tell you I believe the future is full of
robots and self-driving cars and self-flying planes but I guarantee you that those robots
and cars and planes will be doing things that
we just don't imagine today. I mean, you roll the
clock back even 10, 15 years ago you asked people,
why do you have a cell phone? And people say "Oh,
it's make phone calls." And you ask the same thing
to people today, they are like "Phone calls? I
don't make phone calls. I send text messages. I
watch videos." So the things you carry around, they are not phones. They can be used as phones
but they do different things. And I think if we
think "Oh, I'm making a self-driving car because I
want to be in a vehicle where I don't have to
drive," it may do that but I think it might do
something else in the future. Maybe it will bring
us groceries. Maybe it will entertain us. You get
in the car and it's a place you go to - instead of
going to the movies you get in the car and have
fun for some reason. But the point is, that I
think that we don't yet see what's going to happen
with all the robots and cars and planes in our
lives but I'm sure they're going to be there.
- Awesome. Cool. Can I divert us?
- Yeah. Let's go into the chat.
- Quick one for Matt. From [Realtoring] Where can
I get that cool NASA Silicon Valley shirt?
- This was a special order one but we do have a
little store over here at NASA Ames, over at the
little visitor center and I've been talking to
Kenny, the guy who runs it. Hopefully we will get
more copies of the shirt to come out. This is
where I dance that fancy line between a federal government entity and like endorsing stores.
- Well, maybe some day they will visit us and -
- Come and visit… Come to NASA Ames in Silicon Valley. There is a visitor center tent. There's
a bunch of… there's a little store you can
buy stuff there.
- Now, here is a question about hacking. This is
from, sorry, [Markusalaya]. How safe are the self-driving cars and robots now from hacking?
Or are they at a really early stage of security
levels?
- Well, I used to think my computer was safe from
hacking until just a few weeks ago and …
- [Laughter} Oh, no…
- We learned that every computing device in the
world is susceptible to things. I think it's a
real good question of how you have confidence that
your car, your robot is not hackable. I think it
will go beyond that. Anything you hook up to the
internet these days, you might worry about. I
look at all the people who all the different devices that control your lights and thermostats
or your often ovens remotely, and you think "Well,
are they safe or can somebody going to be able to
tap into them?" All the people that now have all
these home speakers that you can talk to and ask
them to do things and not just tell jokes and
stuff. And all that is really then tied into the
question of how do you make them secure. And I
know it's a really very important area, if we want
to trust the systems and rely upon them. Trust is
something that I think as humans, has to be earned. When we work together as humans, when
I first meet somebody I'm like yeah, can I really
trust them with my life? Maybe not so much. But
over time, as you work with them, and you understand what they can do, and especially
that they show you over and over again they can
be relied upon, then you trust them more. And
I think to some, extent robots and cars are
going to have - self-driving cars and that sort of
same category.
- Fair enough.
- As we go forward with our space industry, it is
not going to be just government out there. And
just a few communication companies. We're going
to find that industry can really utilise space in
ways that we cannot imagine right now. And as many
people become spacefaring, they will have the
capabilities to reach out to different spacecraft. Right now, we sort of rely on the fact that
not many people are able to communicate with our
spacecraft and our robots and if they could even
communicate to them, that they wouldn't know how
to be understandable and be able to give them something that would be - would hack the system.
But as we go forward, it's going to be something we have to address. Because cyber security,
for all the spacefaring industry, is going to
be real and we will have to follow suit, other industries
have done this before and we will do the same.
- It's not just space too, obviously aviation in
general too as that becomes more connected.
- Right.
- All the aircraft that are flown, whether large
transport aircraft, airline companies, or general
aviation pilots, people want to rely on connected services just like we all do when we are just
walking around our neighbourhood or at home. And
so, they need networking but that means is that
networking safe, is it secure? Is it reliable?
- Right.
- That's a real big I think challenge for everybody.
- It is true. We are moving away from voice and in
our cockpits to data and that's that will be a
problem if we don't address it soon.
- So folks, if you are just joining us now, you
are watching NASA in Silicon Valley Live. The new
conversational show we are Trying out on Twitch dot TV slash NASA. We are chat being self-driving
planes and robots and automobiles with Terry Fong
and Diana Acosta. We are heading into the last 15
minutes, give or take, we're going to jump into
rapid fire questions. I think we have got a ton of
them and we want to get as many questions from the
chat as humanly possible. So yeah, let's do this.
Let's go on through.
- The challenge before you is to answer in one
liners or as concisely as you feel you can manage.
- [RainarTheConqueror] asking, will AI ever power
any kind of functionality on the international space station
- Yes.
- Is it already?
- No.
- In research, yes.
- In research, yes.
- Yeah? Yeah…
- Do you want to take a little line to talk about
that?
- So, we have done research on the different systems for the ISS. Being able specifically
to monitor the life support systems: water, the
cleanliness of the water and how well that's working. So, we have utilized AI in that.
For research purposes.
- Congratulations. OK. Let me find a good one.
This is vast. What do you think the next significant breakthrough in space exploration
will be?
- Boy, the next breakthrough in space exploration?
- Yeah. That's tough.
- Sending humans back to the moon.
- That will be exciting.
- Yeah.
- No less vast. [Robbie1896] are we alone in the
universe?
- [Humming musically] Dun, dun, dunnnn…
- I would say watch the podcast or the live video
from last time.
- Oh yeah.
- With Jim Green and Greg Schmidt.
- Those are the science experts and they addressed that question a little bit.
- Yeah, they talked about that in detail. You can
check it out our YouTube.com/NASAAmes or I think
it's actually on demand on Twitch.TV as well.
- Now seriously, I have several questions asking
about their careers, their future careers and what
you guys might advise. So [montrealchrislee] "What
the hell should I do to get an internship at NASA
as a software engineer?" Or more broadly, what
can anybody do to work with you guys or with other
teams here?
- I think it is really important for people to
understand that NASA is a place that's very open
and very welcoming to people who want to get experience at NASA, to do internships here,
to work with NASA people. One way you can get
involved is that NASA releases a fair amount of
software open source. So you can actually download
code from NASA. And we actually do take back contributions in various projects. Another
way, is that every single NASA center across the United
States has a very strong program for summer internships. Here at NASA Ames, we typically
get like 800, 900 students every summer, which
is a large number. And there are lots of ways to
get involved from the high school level all the
way through to grad school level. So just contact
NASA centers and get an internship.
- That's intern.nasa.gov. That's the one stop shop. Go in there to do all the applications
and stuff.
- Alright. Moving on. [Gralic] asks how long does
it take for a program like Superball to go from
idea to actual prototype? How much does it cost?
Also, Matt how does your hair look so darn good?
- Maybe talk to us about Superball first.
- I want to hear about Matt's hair. I mean, no,
but -
- I do I what my wife tells me to. [Laughter]
- Superball is something we have been working on
for the past maybe four years or so. But that's because we have been continually coming up
with new designs for it. The first concept, from
sort of like paper… "paper." All right, computer.
From computer sketch to actual hardware, just a
couple of months. But once you build it, there's
a lot more to actually make it work in a really,
you know, high performance, reliable way. That
takes a lot of time ,in terms of controls and modelling
and simulation and testing. And the testing of
course, is the most fun part. But you have to
build it.
- Yeah first of all. That's shorter than I might
have thought to get started. Tell me if you guys
know enough about the field of physics to answer
this one. [VonetarWolf] wanted to know how successful is a career in physics, like
theoretical or astrophysics? Assuming you have a
graduate degree, versus a career in aeronautics..
[crosstalk]
- I feel strangely qualified to answer that because my bachelor's degree is actually in
aeronautics, not robots and my wife has a PhD in
particle physics.
- Cool. [Laughter]
- So, I will tell you that of all the fields in
science of course, now I'm going to get all the
strange comments from people who aren't physicists. I think physics is the most universal
field. Just about every single physicist that I
know can get a job in almost anything they want.
Partially because they have more math background than anybody else and everything is driven
by math. So, I think that, for people who are
studying physics, really the sky is the limit. For
me, I was trained as somebody in aeronautics but
then, when I got to grad school, I got interested in robots and computers and that sort of took
me down a different path. You know, and aeronautics
too, is actually a good starting point, because it's very interdisciplinary. You learn about
lots of different things. It's sort of a classic
systems engineering discipline. And it is not
just, oh, you are going to make planes. You going
to make planes. You're going to make spacecraft, you will do math and physics and all kinds
of stuff.
- Is that how you see it too, Diana?
- I see it that way as well.
- That's nice. Interdisciplinary. Here is a quick
one. [Shocarscon] is asking "Which programming languages do you use?"
- So, in robotic, we rely a lot on C, C plus plus
and Java. I would say for some of the systems, there is a lot of work that's been done in
Python, as well.
- Same for you?
- C, Python.
- Alright. Things I myself have never learned. Do
you know any programming.
- I - I took - my - the intro to my computer sciences, this is like in 1999, like first
year of college and fiddled around making little like
windows and programs and stuff. And the thing that
struck me was how similar it felt to like learning
another language. It is like learning Spanish and
French and you quickly realise, that I could follow instructions, but then it quickly went
to the point where I'm like, I don't even have
the vocabulary for this. And so yeah, that was
the end of my computer science career.
- That's OK, Matt. Back to self-driving cars. [NickGares] asks, what do you think the adoption
rate for self-driving cars will bein the next decade and will they ever reach the price
point of manual cars?
- Wow, so two part question there. So, first part
here is adoption rate, I think it depends on where
you live in the US. I think that if you're in an
area where you have the "good weather," like Silicon Valley, it is a lot easier because
there is a lot, you know, fewer things that we have
to deal with, like thunderstorms and snow and
stuff like that. So, I think it will be easier to
have driving cars out in places like California,
Arizona, New Mexico first. Will they get down to
the price point of normal cars? Well, it depends. A lot of cars these days are just really becoming
software platforms. You know, all the Teslas out
there these days. I mean, they get software upgrades. You flip a switch and they go from
manual driving to self-driving. It is not like, so
much a question of an add on. I mean, the cars
themselves will be ready to be self-driving --.
- That's interesting. Cool. Diana quick. [Wordsworth] asks "What is your favourite
science fact or theory?"
- [Laughter] The pressure…
- No pressure. I'm stumped.
- OK. That's good.
- Yeah.
- Ponder that and if something comes to you, you
just jump in and let us know.
- Terry, [JCBaby] asks "Star Wars or Star Trek?"
- Star Trek.
- He didn't hesitate. He did not hesitate! You
guys?
- I know way more about Star Wars than any like -
than I should. But way more. I know way more about
Star Wars and read more stuff than necessarily Star Trek. So, that's probably my world.
- Alright. Firefly.
- Firefly. [Laughter]
- Nicely done.
- A shout out to R slash Prequel memes. They know
what I'm talking about.
- Terry [keekz] is asking who came up with the
name Superball?
- Superball! Actually, that was a former researcher here at NASA Ames with a really
great name, Vytas Sunspiral. You can find him working
at a start up company in Emmeryville these day.
- Let me see. We had a question about robotic surgeons. Do you guys feel you could handle
that?
[mellowcanuk] was one asking about robotic
surgeons will be used in space or the moon. Is
that something you hear about? Are you involved in something like that?
- We hear about those things and I think that goes
back earlier to what I was saying about, you know,
what's needed for a chief medical officer or
somebody who has to do medicine in space. And I
think that robot surgery is certainly going to be
sort of like one of the tools in that person's toolbox.
- OK. Because actually do we have robot surgery on
earth right now?
- We do. Yeah.
- That sounded so far out to me, but no, actually that exists.
- It's used very widely.
- Interesting.
- Yeah, the idea of surgery in space though makes
me a little squeamish. We have gravity here, to
make sure our blood goes in certain ways when we're… we have an injury. But in space,
not so much. So -
- Yep. Well.
- It's a little a little more complicated in space
than it would be on a different continent.
- A lot of things are, aren't they?
- We going to get close to wrapping up. What do
you think? Have we got one more?
- Oh my goodness.
- Let's get a ready good one.
- There's kind of too many.
- Well we didn't get anybody's favourite science fact or theory, which I'm disappointed about.
- What's yours favorite fact or theory? Yeah, Abby!
- I don't know.
- What is your favourite space fact or theory?
- I ask the questions around here, OK.
- Or why don't people in their chat put down their
favourite theory? I would love to see that.
- If Terry or Diana think of one we will have them
add it to the chat towards the end.
- This is a question that's quite different from
the others in terms of uses for AI. [BearskinRug]. How do you think AI could help us manage our
natural resources? Is that an area you guys know
about? Is that something AI is busy working on?
- There already is an AI or big data techniques, machine learning, to look at our natural
resources, especially that the big data we are
collecting from satellites, of the different rainforests and such. We are watching how
those are changing, not just in terms of the green
also the temperature and other scientific data.
And the oceans, as well. So, we are collecting a lot
of information and processing it with AI and
then that informs the humans to make decisions
that are can be regulatory in nature or other.
- Neat. AI is kind of everywhere, isn't it?
- Robots are everywhere.
- Algorithms, math.
- Math.
- Awesome.
- Excellent. So, well thank so you much guys. For
folks, this has been NASA in Silicon Valley Live.
Huge thanks to Terry Fong and Diana Acosta for
joining us today. For folks listening or watching on demand, we are on all the major social
media platforms under NASAAmes and using the hashtag
#NASASiliconValley. We even have a phone number, analog, where you can call in and leave comments
and feedback. Don't ask for a call back because it
will not happen. But we can add your comments into the chat or into future episodes. That
number is 650-604-1400. We will be back next week.
This is Friday, February 2. At 2 p.m., Pacific
time. And that's where we will be doing a special
"Let's play space video games" episode that we have
been working on. So, get ready for that. That is
going to be a lot of fun. But if you haven't already
go ahead click like, share, subscribe, whatever
it is that you need to do on the screen or the podcast
app, so you can check out out next week. Don't bother sending us tips or subs because this
Twitch show is brought to you by your taxpayer dollars.
For all the of our international fans, you are
welcome. In a shout out to the international science community that provides so much as
well. But we are a NASA podcast, but we are not
the only NASA podcast. So, shout out to our friends
who do "This Week at NASA," "Houston, We Have a Podcast."
and "Gravity Assist." Thank you so much and we
will see you guys next week.
(Music)
Không có nhận xét nào:
Đăng nhận xét