Tip:
Highlight text to annotate it
X
[Transcription is missing for 3 mins, apologies for the delay]
So, you know, how do you design an underwater glider? You have to learn how to
design an underwater glider so it was really the start of an education, you know, self‑taught
in design and manufacture of underwater vehicles. So was studying machine tool technology at
the time, manufacturing, engineering type stuff, eventually got a transfer degree so,
yeah, it was a great vehicle to actually learn and gave a really good incentive to study
stem subject matter and ultimately transitioned into an interest in an engineering transfer
degree. What is an underwater glider. Normal vehicles
glide themselves through the water using a propeller, pretty straightforward. There's
no propeller in an underwater glider generally speaking. They use something similar to a
fish's swim bladder in the water and as they float to the service they have wings on them.
Does this thing work? >> Apparently not.
>> So wings? And those actually help transfer the change in altitude into forward movement
and so the efficiency of the vehicle is going to be determined by a couple things. First
of all, the efficiency of your fish bladder the engine, the big factor is the drag, and
also kind of related to drag is your glider ratio, how many feet forward you move through
the water for every foot you lose or gain in altitude. One interesting thing in underwater
vehicles is they spend half the time flying upside down because your gravity vector flips
its head when you are going against the water column or going down with it so, yeah, they're
autonomous underwater vehicles that can travel long distances on battery power.
So history, you know, it all started back in, you know, '80s or something, I wasn't
really born yet so it don't really matter, very ego sent trick view of the universe so
they had these things called argue goes floats, they were essentially a coffee can with a
linear motor powered syringe and go through the water column and collect sensory data
and when they had enough data they'd float back to the surface, satellite modem, phone
home and do it all over again and based on my understanding of how things went down,
one person was like, well, these things are really cool, you know, we're getting really
data, we're not having to send people out in boats at $40,000 a day but the problem
is we can't actually control where they go, they're like hot air balloons. So it was like
what if we strap wings on them and underwater gliders were born. Henry Stommel was one the
people who wrote some kind of pioneering articles on the potential applications of these things
and what you see before you today between the Scarlet Night, the Slocum, the Spray,
it's become a very popular vehicle class simple because the range you can cover for the amount
of energy you have to store. The Scarlet Night was one of the record holders, it was put
together by Rutgers university, and it was essentially a Slocum glider that had been
loaded down with extra battery mass and stretched out a little bit. It uses lithium CSC cells,
is my understanding, which is kind of impressive because the high test peroxide rocket fuel
you see in the movie Moonraker as far as I can tell has lower energy, so are pretty impressive
batteries although also fairly expensive. That's not what I ended up using. So I don't
know if this is the correct way to design an underwater glider but after going through
10, 20 designs in my head before trying to build the first draft this is kind of what
I settled on as the proper procedure. Maybe somebody can correct me if I'm wrong but you
don't really know how you can your package your components until you know what your components
are, and buoyancy engine being one of the key elements, it's going to determine the
efficiency, you really have to start with that. Energy storage system is going to be
dictated by the energy requirements of your buoyancy and then whole design, it's a pretty
straightforward process. Very few moving parts. So eventually I did, you know, I'm not a very
decisive person but eventually I did come to a conclusion on how I wanted to build one
of these things and that's with a phase change material. There's another vehicle out there
that uses PCMs to propel itself through the water, it's called the Slocum Thermal and
I'm not doing that because it looked very expensive. I read NASA tech briefs kind of
talking about that aspect, and I think it was like (inaudible) decking was the alkane
they used and really expensive per ounce, wasn't really on the table because my goal
going into this ‑‑ because, remember, I said I could do this for $100 if I'm spending
$200 for my face change material I'm already blown out. So these were the design requirements
I went into it with. I wanted the barriers to entry to be super slow because even though
I had access to a milling machine and a lathe, not everyone does so I wanted all of you to
be able to take the DEF CON CD, take the solid models that on there and go home and build
one of these yourself. And then range of efficiency, you're not going to take a two order of magnitude
reduction in costs without taking some hits on range, efficiency, performance. So it was
kind of a best‑effort basis. I'm pretty optimistic. I've got a fairly large battery
pack compared to the currency, or compared to the current consumption I'm looking at,
so we'll see. Here's one of the earlier efforts I made into
actually trying to design for the manufacturer. That's a Harbor Freight air compressor gear
being used to index several syringes and this design actually served as a little bit of
inspiration for a hack I had to make towards the end of this project. Overall the check
valves involved were kind of a prohibitive feature that just disillusioned me on this
concept. So considering a lot of thing, I struck out early on in the process because
at the time a lathe and milling machine were pretty much, you know, a requirement to make
those happen. The weight of using off‑the‑shelf linear actuator was pretty massive, you'd
end up with a scuba tank pressure housing and that's no fun. So wave and solar power
interesting concepts, I'd like to explore them a little bit especially for one concept
I might mention at the end, but overall hydraulic pumps and face change materials kind of showed
out to me as the best choices available. So paraffin wax expands 10 percent when it
melts, ish. It's got a high specific heat when it does melt it will stay liquid for
a very long time. I was looking at using soldering irons as a (unintelligible) to melt down the
wax. And this was just kind of a breakdown at the various energy densities I was looking
at for ‑‑ kind of guiding my decision on how to build the buoyancy engine. And when
I first started on this project, there really wasn't very much available of low‑cost inertial
navigation. You had the multi‑weave project which I thought was pretty ingenious, they
took a Numchuk and a motion plus from a Wii game console and they actually made an IMU
something out of it. Welcome to the future. This is really cool, a $33 board available
from Hobby King, I can't imagine they have more than $5 profit margin built into that.
You get a really cutting‑edge chip on there called an MPU 6050, or 60,000, on of the two
depending on if it's SPI or ITC. But the point is it's got a built‑in magnetometer, and
so it can do complete 90‑degree freedom sensor fusion which is the ability to distinguish
between gravity and linear movement through space. And when you're trying to actually
integrate your inertial data, you'll run into problems unless you have some pretty good
math background or a fantastic product like this behind you.
That's a 3‑D R robotics GPS chip, it was available and it's buried in the nose cone
of my vehicle right now, coated in a nice, thick goop of RGV compound.
The whole design was kind of fun. I spent most of the past year working on that specific
aspect because it ultimately guided the production of the rest of the vehicle, once I knew what
I was packaging, I had to actually wrap it up into some sort of fluid dynamic whole form
and so based on the research I had, I'd heard some interesting things about 30‑1 glide
ratios with something called the McMasters air foil which is pretty much well which is
pretty much a NACA four‑digit series 0030, but based on the simulations I did using a
program called Profili 2 marketed at RC Plane Design, its main claim to fame is it didn't
stall out at high angles of attack whereas the black line has one of the sharpest peaks
on there so NACA 009 or 0009 was kind of out of the question, too, so kind of middle of
the road choice of this metric air foils who were well understood at the time. Pretty common
options are NACA 0015ish, so I think my wing route is NACA 0015 and wing tip is NACA 002,
and what that does is hopefully like the Rutan Long EZ, if you end up stalling out, the wing
route stalls before the wing tips and that will cause the pitch to self‑correct. Fingers
crossed on that. Not a lot of testing in that regard.
So first concept, I took a polycarbonate tube, a wax motor used to control a dishwasher latch,
funny enough, high‑force, high late answer, low efficiency unfortunately as I found out
later. And, oh, no, I'm not sure if I'd submitted my CFIP at this point but I was kind of sweating
at that point because I'm apparently really, really bad at fiberglass. Who knew? The original
idea was is that I would be using hot wire foam techniques to actually generate the whole
form and lay fiberglass over it and as you can see from this picture that wasn't going
so hot so not as easy as it looked on YouTube. Was not going to hit my deadline at that rate.
So throw money at the problem. I've generally found that when time is of the essence money
can generally buy you time back so I bought a 3‑D printer. I really am happy I made
that decision. 3‑D printers are really cool. One of the design requirements being able
to build this thing in my underwear ‑‑ check.
So similar design, same whole form but now broken into smaller chunks to help with the
print volume and also, you know, trying to minimize the overhang of the printed parts.
Support material is an option and one of the large parts actually was using support material.
One month turnaround time, you really can't argue with that. And I ‑‑ I knew 3‑D
printing could really help expedite the design process in getting you through multiple iterations
but until I'd experienced it firsthand kind of saving my bacon of getting a prove concept
throughout the door and second, third revision, whatever I was incredibly impressed with the
value. The 3‑D printer was a road stock max, it's a delta style design and they're
one of the coolest when they're running. Cartesian bots a little bit lame, in my opinion. And
that's the robot you see in front of you so $100 price target. $31 a kilogram for my plastic.
That could be cheaper but let's, you know, try and be generous just to make sure I don't
underestimate. $21 in plastic. So if you throw a significant amount of capital investment
at the prospect of building injection molding tooling, then, that could come down significantly.
Remaining bill and ‑‑ building material was actually published to the DVD if I remember
correctly and I mentioned there may have been some hiccups along the road, this building
material is definitely not accurate to the 30‑cent mark because the face change material
concept ended up working out to not just moderately inefficient to virtually unworkable, at least
with the energy storage system I selected. If you need your robot to go to the bottom
of the Marianas trench, the fact that it's a solid‑to‑liquid phase change as opposed
to expansion and contraction of gas really work to your advantage, but if you're going
for range and you don't particularly want to be dragging bottom with seaweed, trying
to get 3,000 meters of depth with that process probably not the most efficient use of battery
storage. So I think it was early May when I realized, wow, I've got, you know, a 200‑volt
pack or whatever and it's browning out at 100 milliamps, .1‑amp, I knew coin cells
weren't exactly like lithium palm or RC plane batteries but I was a little bit blown away
so I needed a plan B, or plan C, I guess, so what I came up with was something a little
more conventional, you know, trying to avoid too many experimental, you know, wild concepts
in one vehicle build can be a life save her so I went to the tried and true liquid motor
linear actuator because if you remember those aren't really an option because people wouldn't
have access with a lathe and mill but with a 3‑D printer that potentially changes so
I ended up using commercial off‑the‑shelf McMaster car parts to build the assembly which
you can see white on the actual robot now. So the only change between what's on your
DVD and what actually is here in front of you is the white assembly. Everything else
I was able to recycle between designs. Modular design is a good thing, I guess. The gray
pipe was a really good choice it will ‑‑ sorry about that ‑‑ the gray pipe was
a really good choice because it allowed for routing of the wiring and ultimately acted
as kind of a skeleton or a backbone for the entire vehicle. So what I learned from that
process is 3‑D printing's awesome and the great thing is when you're actually done with
it you don't have to go through an entire twisted any process to design injection molding
parts you're practically there when you have a working prototype.
What I learned through the process, I feel like I actually hit a pretty good balance
between fail early, fail often and not making stupid mistakes for lack of planning. I didn't
build anything before I had complete solid model data. And that's really valuable when
you're trying to avoid these last‑minute crazed runs to Home Depot, West Marine, Auto
Zone, whatever it is, if you have the complete building materials before you make the first
part, you know exactly what's going into the thing and you don't find yourself running
into these issues where parts are colliding with each other so ‑‑ and then it was
also extremely difficult to quantify whether or not a design decision was a good or a bad
one such as the wax motor until actually trying it. So it was a design decision trade‑off
to not use so much simulation and I was pretty happy with how that went because you can simulate
the daylights out of a bad idea and then find out you wasted the last year doing, you know,
FEA or CFD on a broken design and then it's all wasted so prototypes can little nature
things that you otherwise wouldn't be aware of.
So I did this entire thing out‑of‑pocket. No ‑‑ no grants or funding agencies and
I actually really like that because I'm accountable to myself and I don't have anybody breathing
down my neck, forcing me to, you know, chase some costs or make bad decisions because,
oh, I need to save face over that, you know, thousand dollars I wasted on, you know, wax
motors. On the other hand, I gave pretty much the profit of the design away for free on
the DVD this year. I probably spent around 15 grand on the project over the past decade.
And so I would consider it a dubious appropriation of my retirement savings. But, you know, that
was the mission from the start. I wanted people to be able to build these themselves and kind
of empower themselves to deliver sensors, communications devices and payloads to remote
destinations where traditionally it would be prohibitively expensive at, you know, tens
of thousands of dollars per vehicle. And how that's possible is if your price is low enough
you can consider it disposable and you don't have to pay somebody 40 grand to go out and
water and change the batteries. So we'll see. Ultimately if you build it they
will come and if it's a good idea hopefully people will say, hey, maybe we should build
some of these. So where to go from here? I didn't have an
opportunity to test max depth because I only had one prototype as of two weeks ago and
I didn't really want to lose it out in the ocean scuba diving like, I guess, I could
have tried to put a dog collar on it and walked it like with a leash but overall it's disposable
after my talk is over but the thing took over 100 hours to print and, actually, that guy
right there put in a significant amount of labor helping getting ready for your guyses'
eyes so small applause for him, if you're willing.
[APPLAUSE] >> Thanks, guys. So in terms of trimming vehicle,
I can say with high degree of certainty this thing is very positively bouyant. Traditional
buoyancy foam, polyurethane, or epoxy and glass micro balloons. High‑crush depth kind
of unnecessary seeing as how I designed the vehicle with syringes, but overall a good
choice. So generally speaking, epoxy or urethane, I chose paraffin wax, it's lost cost and over
half the dollar per pound of buoyancy I was seeing was either epoxy or urethane, so paraffin
wax is like $4 a pound. So very positively bouyant, every empty cavity in that entire
thing has a specific gravity about .5 so it's really just a matter of how you distribute
your lead and batteries at this point to make sure that when the buoyancy engine's in neutral
position it's sending about horizontal because when the vehicle goes buoyancy the center
of buoyancy moves the vehicle and it pitches toward the suffered and the GPS chip rises
up towards daylight and when it goes negatively bouyant the opposite happens. So if you're
in neutral buoyancy you should be sitting around level in the water and it's really
just a matter of adding weight and removing buoyancy foam as necessary to achieve that,
I'm looking at around 40 to 50 milliliters of displacement change. So it's unknown at
this time what the vehicle's velocity through the water is, the more bouyant it is the fast
e it's going to pop to the surface like a cork and your speed through the water is going
to determine what the Reynolds number is and that guides, you know, things like what air
flow you select. But first draft, I'd call it an alpha stage design and openglider.com
is where I'm going to be adding future revisions. The one on your CD is rev0.1, and rev 0.2
will be on openglider.com tonight and that will be have 9 new white syringe‑based assembly
so yeah. In terms of research that went into this, this guy named Bruce Carmichael was
one of the pioneers of the airfoil you see before you. It's called the X‑35 and originally
it was known as the Dolphin but if I understand correctly they ran the Dolphin through a genetic
algorithm and ultimately the citation shaping of axonometric bodies from minimum drag, that
one was I think was where I was able to find the coordinates for this design. It does bear
some resemblance to some other vehicles on the market and that shouldn't be too surprising
seeing as as far as I can tell we did use the same curve rotate around its axis to generate
the main body. Otherwise, the NACA four‑digit series, pretty common equation. And I threw
something in here that I found really fascinating. It's the geometry of a blended wing body morphing
wing and that was the idea that you could actually design an entire vehicle around variables
and equations so you can change something like a static wing sweep from 30 degrees
to 15 to 25 degrees and rather than having a lot of hard engineering labor going into
remaking the entire design it just regenerates itself so I very early on in the process I
was trying to stick to that concept and I was pretty pleased. I would change a variable
from like a NACA of four digit 002 to NACA at 005, you know, kind of crazy, and hit rebuild
and the model would regenerate itself with very little labor on my part. One thing I'd
like to explore is potentially even doing a genetic algorithm around that concept to
optimize for parameters like speed through the water column and lift‑to‑drag ratio.
So that's pretty much how I got here. The X‑35 took me like six months to a year to
even find. I had no idea where people were getting this curve from but I saw it in several
places. Ultimately it was just a mad Google foo that was able to illuminate the coordinates
to build this thing. And in terms of other information on how to build an underwater
vehicle, I was going to joke, you know, every story's hero has a favorite weapon. Rock Sampson's
Bowie knife, Will Smith's little cricket and Flipper's hot glue gun. When you're try to
waterproof stuff, I've tried a lot of different things, (unintelligible) coat, epoxy potting
and pressure housings and overall, even paraffin wax, actually ‑‑ it's not going to do
very good in the Caribbean where the water gets down there 80 degrees Celsius. And so ‑‑
but it does a pretty good job because when you pot your electronics in epoxy it's difficult
to recover them or repair them if something blows up. But when you're using paraffin wax
the 80‑degree C melting temperature is actually below the rated temperature for most electronics
so if you decide you want to change something about them, and I've had to, you can just
use a hot air gun or boiling water to rescue your electronics. That said, from a durability
standpoint, paraffin wax, even if you anneal it with a little bit of mineral oil, not the
most durable thing ever. So ultimately hot glue has been my favorite approach to waterproofing
and if you need something a little bit lower viscosity, get down in the English muffin
nooks and crannies then they make low‑viscosity silicone RTV. The acetic acid can potentially
corrode your electrical contacts, but we're talking about disposable vehicles here. It
doesn't have to last three years. I've got electronic speed controllers I potted for
a competition two years ago that still work and that was just plain Jane Auto Zone RTV
compound. And so pretty much every piece of electronics on that thing are just either
coated in hot glue or RTV compound and it's a pretty easy low‑cost way of waterproofing
electronics. So I guess at this time it would be a good way to go for questions.
>> (Off mic). >> Yeah, that's a good yes so one of the problems
with the underwater vehicles that is interesting to me, unlike mars rovers, you know, easy
environment, you know, aerial vehicles, bunch of light weights, you know, these people,
they're so spoiled, with underwater vehicles you don't get to talk to your robot after
you let it go until it comes back to the surface at which point it's kind of a surface vehicle
until it goes back under water snorkel depth style. So no radio frequency communication
so it has to be pretty much ought autonomous and the source code I used with the IMU to
actually ‑‑ it's called a Haversine formula, it finds your current GPS coordinates and
your destination GPS coordinates and gives you a bearing and the distance you need to
go to get there. So it's pretty much a line‑follower robot, that's what makes underwater gliders
potentially so successful unlike mine‑detecting robots or antifrogman stuff. They have a pretty
basic mission which is to go from point A to point B. So object avoidance potentially
with sonar, but suddenly when your payload starts getting into the high‑dollar sonar
stuff it stops being so disposable. At that point my 3‑D printed open glider may not
be the best choice of vehicles. >> (Off mic).
>> Um, so testing has been kind of an iterative process. I've learned every time I stuck it
in the bathtub, early on it was pretty much with a small car battery and some red and
green wire hanging out of the thing. Underwater vehicles can actually be communicated with
remotely using something called a tether at which point they're no longer AAVs or UUVs
they're ROVs, they're remotely operated vehicles. And I had two years of tethered vehicle competitions
that kind of prepared me for working with the unmanned side of things. And it's a pretty
good way to go. So for testing, bathtubs or surprisingly effective. You have them in hotel
rooms. And, yeah, generally speaking, the main thing you're going to be testing is either
what it decides to do when it gets in the water or how it sits in the water when you
put it in there and both of those things don't require a big swimming pool. Also a lot less
terrified sleeking swimmers running away from the evil robot.
>> (Off mic). >> It's the second one. So I don't really
rely on pressurizing as a rule. I had to depend on some amount of air cavity for the design
fix and it's the most conventional thing about this design was you've got one of your actuators
driven by three‑volt gear motors, actually, I think I have one, anybody want one? I know
audiences love things being thrown at them. Not you, Mark. Ooh, hey, speaking of that,
anybody want a flight controller, $33 gimme? Remember two is one, one is none so bring
spare parts. Oops. Sorry. I throw like a roboticist. More questions?
>> (Off mic). >> Oh, good question. So when you have like
a tank light when you're scuba diving you can make them light not wasting battery life
and trying not to lose your dive buddy. In my case I don't have anything smart like that
on my robot, those wires allow me to disconnect the power supply from the micro controller.
Once I plug it in it goes and keep going until I unplug it.
>> (Off mic). >> Good question. I was on a ROV conference,
and this guy was like super salty, had been out in the field for a long time and told
me a really cool trick. You go to like Harbor Freight and they'll sell these heat shrink
packages, marine heat shrink, and it's just heat shrink that's been lined with hot glue
and you don't need to use special heat shrinker wire crimps, you can just take your wire splice,
coat it in hot glue and then run heat shrink over it and then when you heat the heat shrink
it will melt the hot glue again and pull all the air bubbles out. Great question.
>> (Off mic). >> How does it what?
>> (Off mic). >> Oh, yeah, no, I totally glossed over on
that. Really good point. I don't actually have one buoyancy engine, I have four. And
generally speaking, I might have them towards the top of the vehicle. I have them low on
the center of gravity because I had two places I could put them top or bottom and the top
was all being used by syntactic foam or syntactic wax, buoyancy sandy and since that was more
bouyant than the buoyancy engines themselves I had them down low and so I have broken the
cross‑section of my vehicle into eight little slices of pie. The top four are pull of buoyancy
sand and the bottom four, there are two pairs. So when you trigger the two on the left, your
left side is going to be more bouyant than your right side so one wing will lift up and
likewise if you use the two on the right it's going to change the attitude the other way.
So this is one unique thing about my vehicle design, traditionally you actually use your
battery mass whereas I'm just having four times the buoyancy engines that are typical.
And finally, you know, if you ‑‑ you change buoyancy on the bottom two then you'll
just go straight. >> (Off mic).
>> Good question. I can say with before is pretty good confidence my inertial confidence
is not high enough quality to be U.S. munitions grid. That's total relief. These are hobby
components. Video game grade. In terms of actual like integration of my, you know, velocity
or whatever to get my position I don't really do any of that. I've been working a little
bit in my free time with trying to use fetal Doppler monitors which measure the velocity
of fluid through babies' blood to, you know, let you listen to the kid's heartbeat, kind
of cool and they're ultrasonic *** dines and I've been potentially getting velocity
information from that but pretty much I figure out which direction I want to go and it's
a line fall robot it says I know I want to go left, right, or straight and it comes to
the surface, oh, wow I missed my target but OK, it doesn't look backwards, only forward.
Good question, thanks. >> (Off mic).
>> Oh, bathtub depth? >> (Off mic).
>> Originally I had a washing machine wax motor, like I think it was a whirlpool, Neptune
or something. And that one didn't get a lot of mileage simply because the fiberglass thing
went so terrible, just didn't really work into the 3‑D printed picture. So then I
worked to a soldering iron based design. I was using silicone high‑temperature rubber
hose and soldering irons and filling them with wax and I was browning out my power supply.
Even though I only needed about 100 milliamps, it was still too much, and it would have been
a pretty big design tearup to change my batteries so I just changed my buoyancy engine. Questions?
>> (Off mic). >> Really good question. So and the original
solid model I designed around, I actually made a solid model of the PX‑4 auto pilot
from the drones or PX ‑‑ I don't really know what the relationship is but it's a really
good board it uses a publisher‑subscriber system so right now I'm only using the IMU,
the micro Wii board for everything and one of the issues with that decision is it is
so flash memory constrained that you can only just barely compile the source code included
on your DVD. In fact, if you try and use Arduino 1.5.2 to do so, it's not going to work for
you, you have to use Arduino 1.0.4, if I remember correctly and that will work, and I didn't
remember to put that in the slides and I'm glad you brought it up. So there is a reasonable
amount of space, about one‑quarter of the size of an Altoid tin, maybe half the size
of an Altoid tin for a supervisory controller to supplement the IMU. The IMU is broadcasting
yaw, pitch, roll information, that's already been fused so it's pretty much relative to
your world coordinate system and it's really nice because it's publishing it on the serial
port and it's also telling you ‑‑ oops, time's up?
All right, any further questions or wanting to see the robot in person, you know, touch
and feel? I'll meet you guys out in the hallway, I guess.
[APPLAUSE]