- Good afternoon, everyone.
Thanks for coming. Appreciate the time and looks like we have a pretty good
house full of people here, so we're very excited to be here. My name's David Bader. I lead the go-to-market strategy for robotics and factory
and warehouse automation for AWS worldwide. Super happy to be here and
to work with Lockheed Martin. So today we're gonna talk a
little bit about robotics, but we're also gonna
talk a little bit about how we were able to, you
know, kind of tie together t
he IT aspect of a manufacturing facility with the operational
technology, the OT side. So I think as we go
through this presentation, we'll talk a little bit about that, and that'll be the overwhelming theme across the entire presentation. We will leave a little
bit of time, hopefully, at the end of the
presentation, for questions. Hopefully they're very easy ones. Usually what we do when we
have questions at the end is to volunteer for a question, you also volunteer to buy us
a cocktail afterwa
rds, so... It's just an incentive. That's just something
that I thought we'd go in. So we'll go through some introductions. Guys, wanna introduce yourself? - How we doing? Good afternoon, evening. What time is this? The
afternoon, Vegas time. - It's evening in the east coast. - Evening for most people, maybe? Okay. My name is George Kaniamos. I'm the Director of Advanced
Manufacturing Technologies and Operations Engineering
with Lockheed Martin, Missiles and Fire Control business area. I oversee
the advanced manufacturing, the manufacturing engineers,
the industrial engineers, and a lot of what goes into
building all of our products, enterprise-wide, within MFC. We build a lot of really neat things, and we solve a lot of
really hard problems. Some of the products that
you may be familiar with, right off the bat, Javelin,
HIMARS, F-35, Hellfire missile. Yeah. So that's us. Happy to be here, proud to serve. Look forward to chatting
with you guys today. - Hi, I'm Gregg Doppelheuer. I am t
he LM Fellow and Chief Architect for Missiles and Fire Control. So as you can imagine, I have a pretty close working
relationship with George and a couple of our other
team members out here. And part of what I'm
supposed to do for Lockheed, and let's hope I'm doing an all right job, is bringing technologies, like with Dave, into the factories, into
our manufacturing processes, to try and improve our
yields, our outputs. You know, "How can I build
10,000 Hellfire missiles for the cost of 5,000? H
ow do we improve all of our efficiencies across all of our manufacturing lines?" And then when we do
solve a problem for MFC, we contribute it back to the
greater good, Lockheed Martin. So we build solutions that
can scale not just MFC, but to the size of our enterprise. So hopefully we'll nerd out
here a little bit more later. (Dave laughs) - So a quick agenda.
We're gonna go through... Anybody ever heard of the AWS
Working Backwards process? Is that something that
anybody's familiar with? Okay
, so we'll talk a
little bit about that. Doesn't seem like too many
people are familiar with that. We'll talk a little bit
about how we use that for this project in particular,
then we'll go through the actual manufacturing
technology landscape and talk about the digital shop floor and how we kind of
solved a lot of problems. So the Working Backwards process is really something pretty simple, right? But we live, breathe, and eat it at AWS, and it's about looking at a problem from a customer's pe
rspective, and then working backwards
to solve a problem. So we never enter into a solution in the beginning of a customer engagement. We try to talk with the customer, we try to learn as much
information as we can about the opportunity, and we work backwards
from that opportunity to be able to come up with either a solution or a group of solutions that fit the process. And it's really not about
solving the problem initially, it's about getting clarity on what the overall
problem set looks like.
And I went backwards. I had one job to do. (George and Gregg laugh) So in this particular case- - [Gregg] You worked backwards. - Working backwards, that's right. So in this particular case,
we were really fortunate to be able to work with
a digital artist, right? And we were able to take
our learning that we had from a series of discussions
with George and with Gregg and with their entire team, and we worked with a digital
artist to actually visualize this Working Backwards process. Now, we do
n't do this every time, right? But in this particular case we did. And this is kind of what it looked like. And again, keep in mind, we're not trying to solve
the problem at this point. We're really focusing
on the what and the why, and worrying about the how down the road. We're not solutioning anything. So the idea is to be
able to do, in real time, is to be able to get these key metrics that are important to the operation, and important to the IT team. All kind of in one sitting,
in this part
icular case. And this is kind of what it looked like. So it was really pretty exciting, to be able to do this process
and to be able to say, "Hey, these are these important things, and these are the North Stars, and this is how we're going
through this as a team to be able to build this out." Ah, I keep doing it. Go ahead, George. - I don't know, you rushed through that. You hungry? - [Dave] I am. Very hungry. - Gotta get out of here early. All right, so Dave
talked a little bit about the what a
nd the why, and
how we work backwards, and how we were going through
this whole process with AWS. Well, now we're gonna talk
a little bit about the how. Because at Lockheed we
have a lot of engineers, we like to solve hard problems, it's just usually what we do. He presents the problem, and we end up spending
a lot of time and effort trying to solve it. But in the process of doing this how, we encountered that there
were several different kinds of how. There's the technical how, which was, "Hey,
how do we get this data from one database into another? How do we get... How do we extract this
information out of our... Out of this machine, and
do something with it? How do we get the information
to our technicians in a timely manner, get
them the right information in a way or method that they
can actually interpret it and see it? How do we get some of our
systems to talk to each other?" Those are all the technical hows, and they were good discussions to have. We had them, you know, we
devel
oped some roadmaps, we took some actions, we were
able to do a lot of that. But as we moved forward in this process, we discovered that there were other equally as important hows
that really drove challenges for us, that at times
could become roadblocks. Things that have to do
with change management. Things that have to do with how do we upscale our organization to be able to deal with the tools that we're putting in place with them? I mean, yeah, we've got a
room full of engineers here, likely
leaders in an organization, and what are we talking about? Well, we just digitized something, we made things easier, right? We put buttons in places
for people to touch. That's easier, right? Well, not for everybody, let me tell you. So the hows, as we worked
through some of these hows, and we looked at the
organization that we have, the organization that we need, we have a lot of commitments
to our customers. If you look at the
geopolitical landscape today, we have a lot of commissars, a lot of
commitments to our war fighters, our allies around the world. We can't stop production. And what are we talking about here? We want to put Spot in production
at one of our facilities to be able to do work. So how do we start addressing
some of those hows? How do we do that? So as part of that, next slide. We went through and we
looked at our organization, how are we gonna do this
within our organization? And so we tried to
leverage some of the roles that we currently had in
place as part of our
org. We had an organization already called the Advanced Manufacturing
Technologies Organization. That organization is
predominantly focused on research and development
of new capabilities, manufacturing capabilities,
robotics and automation, the materials and additive manufacturing, developing new manufacturing processes. And it's really where a lot of our operations fellows reside. In addition, we created an organization called the Intelligent Factory Framework. And I'll let Gregg geek out lat
er on what the Intelligent Factory
Framework is at Lockheed. But ultimately, the intent of this group was to help find those
strategic insertion points, and to help be that
closed-loop feedback point to the AMT, our advanced
manufacturing organization, so that we're developing valuable
solutions to the business that will actually drive change, that are what our technicians,
what our workers want, what the business needs
to drive the efficiency, the span time reductions, the quality that we need
moving forward. We also created another organization that kind of identifies, is the Replication and
Proliferation Group. So this is a group that is
predicated on taking information from this Intelligent
Factory integration team and from AMT, and how do you
take ready-now technologies, and roll them out across an enterprise so you can compound your return? A lot of times what we
find is no one technology or no one solution will
actually pay for itself. You have to combine it with
multiple other
solutions and change your business process to actually make something
matter, and actually make change. An example, you know, we do
lots of inspections, okay? That's great. Well, I don't need to
automate my inspections, I don't need to set up
cameras to do inspections. Why don't I just use the IoT
data that's on my devices to verify that I actually
performed a task? Things like that. Needless to say, spoiler
alert, that's part of this. I've gone through now
these last couple slides and I never b
rought up
our IT organization. That's just a spoiler alert,
we'll talk about that next. Now as we're going down all this, developing these organizations, trying to figure out
how we do this stuff... Our information group, our
IT group, the AWS team, our partners, and part of my organization, as well as myself, went through and we started
mapping out everything. We started digging deep into the details of all kinds of things. We were mapping out processes, we were developing value-stream maps, we
were creating systems diagrams, we were creating spaghetti charts. I mean, it was interesting
and very insightful for me, because as an organization
like Lockheed Martin, we have decades upon decades
of tools developed over tools, over infrastructure, over processes that were created
because we got a complaint, or we found a defect, and we
didn't want this to happen, so we put a corrective action in place, and that corrective action
wasn't very efficient, or whatever it is. It turned into a bit
of a mess. So lots of disparity, lots of
complexity across the board. And what we ended up finding
was 13 functional personas, at the end of the day. And when I say personas, we
mean roles and responsibilities, roles within the organization. There were well over 15 process workflows, well over 16 software platforms
that we're dealing with. And the reason we limited
15 and 16 is we got feedback when we did these charts
that if we added more, it was gonna be too small, so
that you guys couldn't r
ead it so we ended up just
truncating what that was. But this is what we were dealing with. And when we ultimately found this, we started having discussions, and it was more of
spirited arguments over, "Why do you need access to my data? Why do I need to have an ERP integration when all this stuff is
happening in my MES? Why do I need to integrate
quality, why do I need my CARD, my corrective action request database, have access to anything that
we have going on with Spot, for crying out loud? O
r warehouse management? Why
does that have to be in here? On the flip side, our IT partners
were sitting here saying, "Dude, you don't know how this operates. We need to be able to combine our data, we need to be able to
have the data format, we need to be able to clean our data, we can't apply any of the models, we can't apply any of the
things that we want to do to our data unless we
clean all this stuff up." And then it hit me at some point, and it kind of brought me back
to a conversation th
at I had with our IT leaders and our AWS partners, which was really surrounded
around core competencies and silos. And this was exactly what was happening. And exactly what we vowed we wouldn't do, we were actually doing. And that was creating
massive problems for us. And one of the reasons I'm
bringing this up is because I know a lot of you are engineers, and we want to talk about,
we'll geek out at some point, you know, Gregg will geek out on you. But I also know that there
are many leaders in
this group who might be running organizations. And this can really screw up your project if you don't pay attention to
the other side of the hows. Okay? Okay, so, with that, we went
back to the drawing board, we had to get some of our leaders together and make sure that we took
a more conscious effort at breaking down these silos,
breaking down these walls. We started interviewing technicians, we started interviewing
engineers together. We started mapping out processes and looking at solutions
together. We started educating people
about our individual roles and responsibilities together. And in doing that, we were able to break
down all of these silos, and we were able to take this
cluster-blank of results, and... And we were able to... We took a portion, I'm just showing you guys
a portion of this, right? I mean, this was a much larger integration than we're looking at. But, you know, even something
as relatively simple as views. What information does what person get at what domain,
or in what
level of the organization, to ensure that they can
do their job effectively? And we were able to take
all of those personas, all of those processes,
all of those systems, and neck it down to six. And, look, I'm telling
you, that was quite a feat. And we were really proud
of being able to do that, because it really brought
us to a point where we had a level of expertise,
and of situational awareness of our individual roles, and we were able to go back and look at, "What information do
people
really need to do their job? Do I really need a
technician having our... Getting reports on quarterly
financials for the company or for the center? Is that something that they really need, or do they need more information related to the work that they have at hand, and the work that they're
trying to do day-to-day?" And that's ultimately where we ended up. So that was a positive note. Go to next slide. Okay, so, our objective as
part of this whole thing was to remove the data silos, we ta
lked a little bit about that. Fill the data caps and remove blind spots. Now you'll know from
the title of this chat, or this discussion, you know,
we're talking about Spot. Right? And so we were really
trying to figure out, "How do we integrate Spot?" I mean, just take Spot out
of it and put in X robot. You know, how do we integrate
this into our factory, and really trying to figure out, how do we fill in some of the
data gaps that we may not have or may not have a solution for, and how do we d
o that autonomously? And then of course, improve the quality and the overall productivity
of our workforce, and of our technicians. And for us, at least at some of our sites, we have a unionized workforce. So that becomes a challenge. And so really trying to
help people understand that this is not about removing jobs, this is not about
people's jobs going away. And it's about augmenting a person's role, it is about helping support
and perform activities that will make somebody more efficient. An
d it took a little
while for us to get there, but when we were able to
demonstrate that to our workforce, we started to get buy-in
from them and from leadership within the union, so that they could understand
that this is not about, "This robot is gonna go
off and do their task." That wasn't what it was. We just discovered that
there are a lot of tasks that people do day-in, day-out
that take a lot of time when you start aggregating
it over a period of time. And if we can start reducing some of
that, we can start impacting spans, we can start impacting takt times, we can start impacting the quality, we can start impacting how many Hellfires we can get out the backdoor
when the customer needs it. So that's that in a nutshell. Okay. - So part of the thing
that Lockheed has done, very progressively, is to
connect their machines. So this process, and Gregg will get into this
a little bit more detailed, but being able to connect
and to have visualization into all of this data
really made th
is process a lot more coherent. We were able to see
things a lot more clear. So in this particular case, as
we kept boiling things down, we looked at the really
big picture and we said, "At AWS, we start big and
we take small chunks." And that's how we approach
most projects, right? So we boiled it down and boiled
it down and boiled it down, and what we all came up
with as a group was that the CNC machine optimization was kind of that low-hanging fruit. At this particular
facility outside of Dal
las, there's 70 CNC machines
all in one very, very large manufacturing area. And the idea was, how do
we help that process flow? So if anybody has CNC
machines in their facilities, they may have five or
ten or one, or whatever. And being able to move material in and move the finished material
out is not that big a deal. But when you're talking about
70 machines in one facility, being able to get that material
flow and automate that, without conveyors or without
augmenting the facility, was a cha
llenge. But we decided, "Hey, that's going to be
that low-hanging fruit. That's what we're going to go after." And that's where we came up
with the concept of Spot. The idea was Spot can overcome some of the facility difficulties,
like the uneven floors and there was actually wiring
trough and plumbing trough in the middle of the facility. Could we have built a large
IoT structure in the building? We had 120-foot ceilings in many cases. So being able to mount
cameras or do other things became a
real challenge. So we decided that we were going to augment the material flow. And oh, by the way, we identified
that, in these 70 machines there were upwards of 15,
per day, tooling breaks. That's where somebody had
to leave their workstation, walk God knows how long to
a tool replacement location, wait there until the guy reset
the tool, and then walk back. That machine was down, that operator wasn't doing
anything at that time. So we augmented their process
by implementing robotics in the CNC
process. - [Gregg] Good, finally me. - Yeah, it's you. - So a couple things that
I wanna piggyback on George before we get to some of my stuff. Most of you probably think of
your manufacturing facilities as, "We build one thing." Trouble with us is our
manufacturing facilities, some of them are 100 years old, and some of them build 10,
15, 20 different products, or pieces of the product. So that adds a layer of complexity to what we were trying to do as well. The disparate nature of
each of the
se facilities and their ability to provide
product where we needed it, when we needed it, at the right time. So in this chart, we're
gonna start talking about how we started our journey. And it's really interesting. Anybody here from 2018? re:Invent 2018, saw me talk about the
Intelligent Factory Framework that we built at Lockheed Martin? Well, Adam, you were there. So back in 2017, we undertook a task to build a network fabric and a framework that we call the IFF. That allows us to connect
our
machines and protect not just the IT infrastructure,
but the OT infrastructure. The best way I can describe it is it's a very secure private cloud that runs inside of Lockheed Martin, that all our machines are connected to. And by presidential mandate,
Steve, if I'm not mistaken? If you work in the DIB, you have to provide this level
of protection and complexity to the tools that you're
running in your facilities to build... - [George] Defense Industrial Base. - Yeah. Okay. I'm sorry, I'm... If
you work at Lockheed Martin,
or probably most of the DIB, you're gonna be dealt
with a ton of acronyms, so if I hit an acronym you
don't recognize, just shoot me. So reality, if you look at this chart, I mean, we may sound like we're mature. We're just starting our journey. And it's a long journey. And it's one that we've had
to build upon our successes, keep going back to leadership, saying, "Here's the next step.
Here's the next step. Here's how we take it to the next level." So the funny thi
ng is,
we're bringing in Spot, we're bringing in a couple other
different types of robotics and we're relaying it right into our Intelligent Factory Framework. We have no infrastructure
work to do whatsoever. So we've laid the base, and we're moving towards full automation. Let's hit the next one. Yep, it's you. - So if you've ever
implemented a mobile robot, and probably some people
in this room have, and if you haven't, we kind
of start by identifying the area in which the
robot is going to tr
avel. So in this particular case
we had a very large area. We had a very uneven environment. So we took an overall view of the space which you see here in the picture, and then we started to
drop in the tool drop-offs, the tool pick-ups, and
those kinds of things. So we dropped those in to be
able to kind of get a picture, and to communicate that
picture to the team. So George talked really well
about the idea of talking with the unions and making
sure everybody's on board. Any time you implemen
t a robot solution, especially a mobile robot solution, it's really important to be
able to communicate to the teams what to expect. And so by being able to
do something like this in a kind of a graphical format, although it doesn't look super
sophisticated, it's probably not the best 3D CAD drawing
anybody's ever seen. But being able to take this
and have an internal meeting with the operators and say, "Hey, this is where you're going to have a tool dropped off for your machine," or "This is wh
ere a tool
is gonna get picked up," was a really big deal. And it really made an
impact on the operators. They all of the sudden
became part of the process, rather than like a cog in that wheel. - [Gregg] All right, so now
we're going to talk about how we brought him in, right? So this is really interesting. We have the Intelligent
Factory Framework established. Now we're gonna introduce this new device. So now we're partnering with George, we had to go to George to get
the money to buy it first
. (George laughs) "Hey, George. Got a few bucks?" And it's interesting, right? I think we developed the relationship that I think we've established here at MFC is unlike any other in a corporation. We have a pretty good relationship
that allows George and I and Steve and Matt and Jonathan
out here in the audience, the rest of my team, to really
have hard conversations, have some interesting
conversations about security. You know, George's team, once
in awhile, will go out and say "Hey, look at t
his new thing I bought." Without any concept of, "How
do I put it on the framework? How do I introduce it to
the manufacturing facility? How do I securely bring it in?" And when I say secure,
the biggest concern for us is not just north-south
movement, east-west movement. Many of you probably
heard the Honda use-case, and a couple others. Lockheed wouldn't stand this
up until we could guarantee you that we would not have an
incident that happened at Honda and some other manufacturing facilities.
So taking this a little
further, we said, "Okay." Now we're talking to Dave, right? "Dave, how do we take..." And if you've been in a
manufacturing facility with MES instructions and all the things that go
into building something, you've got work steps, you've
got long-winded standards, you've got all this other
kind of stuff going on. And in the meantime, I'll take an example I think we used as part of JASSM, or... Yeah, it was JASSM. We have a welding operation.
Takes the welder two hours. Cl
ocks onto that job, performs the job, he or she then has to go find, not notify, go find an inspector. After that inspector comes
in, they log my user out of their manufacturing execution system. They log in, find the work step, and then they execute the inspection. When they're done, my
welder has to go log in, and begin that same process
over and over again. You can see we've got a lot
of time-consuming steps. So working with AWS and one of their
preferred partners, Tulip, we started looking a
t
process automation software. How can we take steps out of the process that are redundant, unnecessary, and still maintain the level
of security that we need to make sure that it's Gregg
doing the welding operation, or, you know, George doing the inspection? So we're looking at Tulip
to come in and be that glue that stitches together
our process automation with our backend MES,
with our backend WMS. You know, there's a whole... You know, it's not just your MES, it's your ERP, it's your WMS, it'
s... And if you look at the
way we build things, it required us to think
about the integrations across major enterprise systems. So working with Tulip, we're developing not just
these automation screens that are just touch-button or touchscreen, and red lights for notification, or even SMS text messages for
notification to inspectors to come do a job. I might not even have my
inspector log in anymore, they just come, hit the inspection button, and we move forward. So there's a lot of things
to l
ook at outside of just the robotics going into the factory. We were looking at process
automation as well. Because if you look... I think I counted probably two hours worth of wasted time a week just
logging in and logging out. So how do we take and
minimize that activity? So we're working that
with Tulip, which is... Would you call it SaaS? It's something that's available
on the AWS marketplace. - [Dave] So I think the
key there with Tulip, or with many of the AWS partners, is that using a serv
ice like this, we were able to take all of
those disparate software systems that you saw in the earlier slide and kind of connect all of those, and then to be able to bring
that out to make sure that we're saving time, but to also be able to collect
those operational things that are going on, so that we can then continue to change and continue to iterate. So being able to tie all
of those systems together in one format I think was a real key to this overall solution, here. - Yep. Somebody get th
e next slide? - Yep. So this is you. - Yeah, so Gregg already
covered most of... Most of what we want to do here. But the... So Tulip... You know, "Hey Gregg, look
at my new shiny penny. I'm gonna integrate Tulip. (Gregg laughs) So Tulip's been a great
tool for us thus far. We've done several
demonstrations with Tulip, we're working to try to figure out how to get that integrated. We've been working with our corporate information security team, who's going through and
doing their analysis. They'
re super outgoing
and flexible people, which has been super helpful. That was total sarcasm. - [Dave] On video. - And it... (laughs) Anyway. Tulip brought forth a
low-code, no-code solution for our manufacturing engineers, because as we talked about earlier about bringing in new technologies, and are people familiar with them, how do you upscale
people in various domains so that they can use the tools that you're putting in front of them. Tulip brought to the table
a way for our engineers to bri
ng their creativity,
citizen development if you will, and help them work to manage certain processes in the factory. So by giving them the tools
to be able to take sensors, find sensors, go buy some sensors, integrate a smart workstation together and put the controls in place
for the manufacturing process, is that you just wrote work instructions for a technician to perform. And allowing them to be able
to do that was very helpful. - [Gregg] I think one of the
things we're working on, too, Georg
e, is that your team can
do those low-code, no-code modifications to the process and instantly make an impact
in the production area. It doesn't go... You know, we still follow
a DevSecOps procedure and a deployment, but his team can now
recognize an efficiency and implement it in Tulip
in a matter of hours, as opposed to days, weeks,
or months it previously took. - Without tying up a developer, which... - It's come at a premium. - Yep. - All right, I think- - When the heck are we
gonna talk abo
ut robots? - Yeah. (George laughs)
- You know? - I think this is where you
guys really wanted to get. I mentioned already that in 2018... I'm trying to remember when
the world shut down, right? 2018 or 2019...
- 2020. 2020. - Came out here, spoke
two years in a row about the work Lockheed's doing
with IFF and IoT Core. I think the biggest benefit
for us was AWS IoT Greengrass. The ability to push edge
compute onto our network fabric of the IFF, and then begin to
traverse the network fabric to ou
r data lake in AWS. I think we're hitting somewhere around, correct me if I'm wrong Steve,
75 petabytes of data today, coming off our machines? We're working now with some
of the citizen data scientists and some of our other data
scientists to figure out, of that 75 petabytes, if I've
got 200 tags on a machine, do I really need all 200? What are the top 20 that
really give me value? So we're really trying to narrow that down and get to a point where we can say, "Okay, I've got 20 tags on
every m
achine that drive value. And then we're gonna build
models, and we're gonna build dashboards and displays for leadership, so that they can understand
what we're doing. I talked about the extensibility
and the strong security, that's, you know, successful
machine connections... I think one of the other
interesting things, all you guys have here
heard Greengrass today, and they were around here today
in a couple things I was in. You know, really great job at OPC UA, really great job with MQTT. Wel
l, the higher-ups in Lockheed Martin decided on a format called MTConnect. It's a whole lexicon around observations. So we decided on that, so
the first thing I had to do was write a normalization
that takes all these tags that had come off the machines and turn them into MTConnect format. So as many of you know,
Greengrass runs at the edge, I run a LAN at the edge that takes and turns that JSON object that we get, 'cause we can't send straight
OPC UA across the network. Lockheed won't allow, it
's
'cause you can't encrypt it. So what we do is we put
it in an HDPS packet and a JSON object, and we send it up. So what I needed to do
was normalize that data, or figure out how to normalize that data. So a little Python program
I wrote 2 or 3 years ago, still running today I think. And then they turned it into... I think we have a Java
one running as well. So there's a lot of really
interesting things that we do at the edge to get the
data in the format we need. All right, so, how does our I
FF work? And then you can see our
IFF, along with Spot, we've added Spot down at the bottom, but most of the IFF today is CNC machines, even machines that don't
have any sensors on 'em. Think about that. We had metal shearing machines that we needed to know what the usage was. Strictly an analog machine. You push a button, drops a
shear, cuts a piece of metal. Worked with a couple vendors, were able to determine when
that thing was in operation and when it wasn't. So if you look at the stack
her
e, we come up from the edge and everything is based
in IoT Core services. There's some... I think it's funny if you... This may not be known to
everybody, but I think we're one of the largest
Kinesis users in the world, because every Lockheed
PC has forensics on it, and it streams through Kinesis. So we not only stream our IoT data, we stream all of our information
about our corporate assets. So as you can see... - [Dave] Everybody's gonna
know that now, by the way. - I'm sure they knew it anywa
y. - Yeah. (laughs) - As you can see, we execute
a lot of services, right? We've got IoT, SiteWise we
started implementing this year. That's been a little bit of a challenge, but we're working through it. SiteWise is a very complex
product for visualization of what your factory looks like. Works as a graph database on the backend. You can develop functions
and real-time visualizations as data's coming in. So it's been a little bit of a challenge. I think our team has struggled
to skill up in tha
t area. But I believe there's a
lot of power in the tool. So we're gonna continue down
that path and see if we can get to a really steady state that allows... You moving me, trying
to rush me along, here? - No.
- Yes. - Really gets to a steady state. What do you say, how
much time we got left? - [George] We've got 27 minutes. - So yeah, we're good, we're good. - [George] He's hungry, that's why he's... - But I mean, you could look
at all the services, right? There's a set of streaming services,
there's a set of data services, there's a set of analytic services, that all are being put into use, all based on the Intelligent
Factory Framework. And now we're just adding
Spot and robotics into that. So we're gonna communicate with Spot via some of these tools here, you're gonna see in the next slide or two. We're also going to use some
of the tool APIs and data, data functionality and connectors,
to be able to direct Spot along its merry journey in our plant. So if you go to the next one, D
ave, we can talk about the deep dive here. - No this is...
- Oh, this is George. - Deep dive's on the next slide, I think. - Oh, okay. - I guess. Okay. North Stars, technical KPIs, OEE. I'm not gonna belabor
this one, but ultimately, at the end of the day is
what makes this real for us? It was really machine
availability insights, simplification of the process, shop floor quality, shop floor management, ensuring the factory has the
information that it needs to be able to perform work, and that i
t gets that
information in a timely manner, in a format that can be consumed, so that our technicians and
that our workforce knows what to do, and they have the
information that they need when they need it, in a
way they want to see it. - [Dave] So I think my
takeaway on this one is it's very easy to say, "Hey,
OEE, let's do some OEE." But when we start to talk about what Gregg and his team built,
and what George wants to see, we have to start at that beginning piece and build that whole
infrast
ructure so that we can then be able to do these kinds of things, and get to those top priority KPIs, and ultimately get to OEE, right? - Yep. - [Gregg] I think it goes
back to too, our MES, right? We've gotta be able to
communicate with our MES to provide that sensitive
data that we need. The other issue that I don't
think George talked about, that he always likes to say, and I'm gonna steal his thunder here... - Go for it. - Is one of our... You know,
think about our products. Okay, everybody k
nows the
F-35. It flies and comes home. 90% of the products, 99%
of the products we build don't come home. We don't get a chance, we
don't get a second chance to make sure they work. So you ask, "Why so much
testing? Why so much test data? Why so much rigor?" We only get one chance for
these products to work, and work where they need to work, not in the hands of our
men and women of the US, or our allies' war fighters, so it's an extremely sensitive... - And important. - And important thing that
we undertake, and, you know, I like to brag and say that we do a pretty damn
good job of making sure that things go boom where
they're supposed to go boom. - Yeah, yeah, yeah. Not here. Just saying. So by pulling together, how
do we do all of this, right? Where does AWS fit into this? We talk about some of the services, but what we did was we
have an advisory team, we have a manufacturing advisory team that was part of the ProServe team. They came in and they worked
with the Lockheed teams to c
ome up with those high-end systems. That's how we discovered the 15-plus or 16-plus
disparate software systems and the numbers of personas. So we brought in an AWS advisory team. We also had our ProServe team,
so our ProServe team comes in and says, "Hey, we can
build the data lake with you to make sure that all of the data is going to the right place,
and it's structured properly." And then also we talked about partners and we talked about Tulip. So the idea is, in this particular case, it was
a community of efforts to get us to be able to build something
as simple as deploying a robot for delivering parts to CNC machines. So we could talk about the robot part, and we will a little bit. But the idea is all of the infrastructure and all of that capability to
be able to build that securely and robustly, and then
be able to replicate that for the next robot, or the next process. So I think it's really
important and, you know, we have this slide here from
Tulip to show that there's dashbo
ards and there's
interfaces and all of this stuff that we are adding into the
overall system, but again, keep in mind that we're tying
all of those disparate systems together, so that now we don't
have to worry about people drilling into each one of
those systems separately to do their work. - And I think the other thing we did is we called you guys in and made you a partner in our strategy. Early on, we had some meetings. I think after the world
started opening back up, I called Robin and proba
bly
half her team down to Orlando and half of my team down to
Orlando, brought in George, brought in executive leaders
from the business and said, "What moves the needle for you?" And I think a lot of those discussions spurred a lot of projects. I mean, we're working on machine learning, we're working edge compute with Spot, we're working a whole ton of
solutions to improve our yield across our product line. I mean, I don't think we've not looked at a probable solution for some
of the things tha
t ail us, and I think it's been a really
successful relationship now. We haven't gotten to half
the things we want to get to, so look for us to be here next year to talk about something we solve next. - I remember that meeting because George started
that meeting by saying, "Damn it, I want robots in two weeks. I want a whole mess of
robots in my facility." And it was the first time,
I think, in George's career that an older guy said to him, "No." I said, "No, we're not
gonna put robots in here u
ntil we understand what
the business value is." And then he called Robin and
said, "Hey, who's this guy, this robot guy that you
brought in here telling me no? I want robots." And we worked through it, and
we developed this process, and I think that now we can start doing... Adding more robots, and adding more things in a kind of a sprint fashion, the way you would do software. - Yeah, I think the thing
that stuck with me the most out of your conversation,
and then I think it really shed a light
for you guys,
was you came in and said, "We do the hard things really well. We do the easy things really, really hard. So we need to fix that problem." And this is part of fixing that problem. - [George] So where does Spot fit in? That was my only line
on that last, to you. - [Gregg] So, okay. Now here we get to, how
do we get Spot in here? And I'm gonna hurry it up, 'cause we've only got 17 minutes left. But again, we talked about IoT Greengrass. You guys know that we're gonna connect through
IoT Greengrass, we're
gonna put it on that fabric, all the edge data, we've
selected a few different payloads for the back of Spot, and payload on Spot is a
different functionality. We've got an arm, we've got a camera, a couple other things, some edge compute. I'd really be curious to
see if we can run Greengrass actually on Spot. That's one of my next, you know... - [Dave] And we've done it, it exists. - So I think if you
look at the stack here, and the capabilities here, everything we wanna d
o is possible. It's just stitching it all together, and getting Spot to
fit in to the scenario. So let's go to the next one. I think there's another
one that talks about... As we build this up... I mean, this slide's gonna build, it's gonna get geeky on you, right? We're gonna use IoT Greengrass, we're gonna use the developer kit, the Greengrass developer kit at the edge for some API management at the edge, and the communications with Spot. We're gonna be telling
Spot right at the edge what to d
o, when to do it, sending instructions to
the APIs that exist on Spot as they come from the
manufacturing systems. And then you're gonna
see the data traverse up through Greengrass, coming into, we're gonna be sending... We send recipes, now, through Greengrass, and some other ways. But you're gonna be
seeing data and artifacts come up from Greengrass as well. It's bidirectional. And then we're gonna push inference. How can we start to use
Spot for some other things, inspection, machine
learning
, visual inspection? The capabilities that we get
in that upper-right-hand box become limitless when you
look at your business process and where you can improve. We have a business, I won't
mention where they are and what they do, it's very sensitive what they do. It's a lot like chip
manufacturing, your yields are 25%. If I can increase that by
20%, that gets me into 30%. I add another B to the
end of that business. So that's one of the things
we're going after as well. How do I take machine le
arning and learn earlier, earlier,
earlier in the process when I'm gonna fail? How do I relate that
failure back to the process, and then take that failure
out of the process? - [Dave] And I mentioned earlier that the buildings that we're
talking about are unlike many of a standard manufacturing site. So think about a 120-foot ceiling hangar, an airplane hangar, something like that, where it's not easy to put
in cameras and infrastructure with wiring and all in
that kind of environment. So being
able to put this on a robot and do the inspection in a mobile fashion is really kind of innovative
for this business. - Right, and I mean, we're even talking... We're sharing some information
with our Aero partners. Adam's sitting right here from Aero, Digital Transformation Senior Manager. We're sharing some information with them, everything we're doing with Spot. So how do we move this
work to the F-35 line? How can we improve processes on the F-35? Then we'll go from the
F-35 maybe to Sikors
ky. How do we go help Sikorsky? Then how do we help space, right? It's not about MFC, it's
about Lockheed as a whole. So everything we build, the IFF is a Lockheed
Martin enterprise system. Everybody can build on the IFF. We're just stepping out first. So we're leading the corporation
in developing some... I mean, Aero does some
robotic assistance inspection, I'm not gonna say they're not. But they're dealing with some
very proprietary technologies. Here, we keep it as agnostic as possible. I co
uld plug into... If Spot comes out with Spot 2.0 tomorrow, I could plug it in much easier than if I were working with
a proprietary technology. - And I think you said it earlier. It's not... Spot isn't
the message here, right? Spot is a great partner, Boston Dynamics is a
great partner for AWS, a great customer for AWS. But it can be any kind of
robot, any kind of mobile robot. The same kind of process holds true. In this particular case, Boston Dynamics is the right solution, because as I menti
oned, the floors, and all of these other things. It's pretty exciting, and
everything that Gregg said, we're super proud, from AWS' perspective, to be a part of the IFF, and to be a part of all
of this data gathering. So it really makes us a true team, because it's all built
from that ground up. - I think George talked about
the geopolitical climate. You know, those products
are performing well. So we had to make more. But we can't make 'em... There's no additional cost to make 'em. So my custom
er not only
wants their things faster, they want more of 'em faster. And the only way we're gonna get there, process improvement, process automation, nothing is off the table. So we have to start using the technology to help us get there. And again, this chart's just
to show, everything we built, we're building to scale. Everything built with DevSecOps in mind, everything we do from
a robotics perspective, we can deploy multiple places. Everything we build
from a tool perspective, we can deploy
multiple places. Even where we're getting, it'll
force the teams at the edge, George's functional folks,
Aero's functional folks, they'll have the ability
to do these things and make the changes on the fly. We'll build the automation for 'em. Just make your changes
and let's process it. So I think that's the key message
I want to get across here, is it's about building
a solution that scales, not just to your business,
but to other areas in your enterprise, and being
able to build it on tools th
at can support the scale. - [Dave] Did I... Oh. - All right. You gonna introduce? - Yeah, so we brought Spot
out for the first time, and we kind of mapped an area. We did a demo where we used
machine learning initially. So this is the environment
in which Spot will work, in some cases. It's not particularly the CNC area. But this video comes
kind of self-explanatory. It's available everywhere. Don't pay attention to that
guy in the back on the phone, in that particular picture. - Nobody saw that
until you... - Yeah, that's what I'm saying... - I think this was released... We released this on National
Dog Day on LinkedIn. - Yep. So there was National Dog
Day, so Lockheed put this out as a public support to pets. So we'll just let this run, kind of understand what's going on. - Lockheed Martin Missiles and
Fire Control, Operations team collaborated with Amazon Web Services to reimagine how we work, and demonstrate how the Boston
Dynamics' Spot the robot dog can help us transform our
Gran
d Prairie Operations to generate products faster and easier. During the demonstration,
Spot showed the team various interactions and scenarios to explore automation
on the operations floor, like material delivery. Spot will benefit manufacturing operations by augmenting our existing workforce and supporting people's
day-to-day monotonous tasking, allowing our employees to
focus on the critical aspects of their roles, ensuring we
are meeting our customer needs and manufacturing quality products.
AWS tools provide machine learning and Internet of things capabilities that enable predictive ordering and direct Spot to inspection tasks to keep the facility running smoothly. Together with AWS, Lockheed
Martin is revolutionizing our factories, tools, and
processes to enhance production, answering our customers' needs, and delivering mission success. - And all those black boxes
you saw on the back of Spot, there's a ton of 'em,
there's a ton of options. But Dave can go into
more of their paylo
ads. - So we have a little bit of time left. We'll be happy to take some questions. And then if we run over
from a question perspective, or what have you, we can meet
you out in the back as well, and take some more questions, or if you don't want to stand
up in front of everybody. But we'd love to hear if
anybody's got any thoughts or comments or anything
about the overall discussion. Hopefully it was helpful.
Comments