Gas Thruster Controlled Drone
Articles Blog

Gas Thruster Controlled Drone

September 8, 2019


A conventional drone is controlled by
varying thrust on each one of its motors and propellers and by increasing the
thrust on one side of the drone it will cause the drone to tilt, now this works nice and well down here on earth where the atmosphere is nice and thick, but up
in space, things like rockets and large satellites
require a different form of control now probably one of the most common forms of
control for the larger rockets and satellites is gas thrusters, these work
by expelling high-pressure gas out of very small nozzles mounted at various
points on the spacecraft now what I want to do is try and apply this gas thruster
reaction control system to a drone but first let’s talk to someone who might
know a bit more about this subject than I do Joe: hello Tom: Hello Joe: Hey Tom, oh wow that was really loud. How’s it going man? Tom: Good thanks and you? This is Joe Barnard, he has a YouTube channel called BPS.Space where he builds model rockets. but the rockets that he builds aren’t your average model rockets yep this is a
replica of the SpaceX Falcon Heavy and it uses thrust vectoring for control Joe
has been developing this thrust vectoring system to eventually land a model
rocket propulsively that’s right, this video isn’t being played in reverse more
recently he’s been experimenting with a gas thruster reactor control system so I had to ask him a few questions if you could split up the hole of an RCS project into three components, what would
they be? you wanna start with the plumbing umm that actually is pretty
accurate you you probably want to start with the plumbing, so you can figure out
which valves you need to actually well okay the problem is it’s all
interdependent, if you start with the valves you also want to consider the
actual moment that you want to impart on the vehicle if you want to do the math about it, you’ll have to consider the flow rate in your valves but mostly you’ll
start with the plumbing then I would focus on how are you gonna control this
so like what is your flight controller going to be your power source, your
wiring and maybe a little bit of layout stuff and then the last part of the
project which in my mind or at least in my project has taken the longest is the
control side of things what does the control logic look like
when does the computer decide to fire a valve and how do you keep your vehicle
from spinning out of control or not controlling itself at all and crashing and crashing right yeah like how do you minimize your your thump factor so as Joe mentioned
it’s best to get the hardware sorted first this will include things like the
pressure vessel the valves for controlling the airflow and also the
nozzles for exhausting the high-pressure gas now with Joe’s system, he has to fit his
into a model rocket which means it has to be quite compact, whereas on a drone I
have a bit more space to play with so for my pressure vessel I’m going to be
using plastic bottles this is a 2 litre plastic bottle and it can handle upwards
of 150 psi, it’s really lightweight and they’re really cheap, also my compressor
only goes up to about 116 psi so these bottles are pretty ideal in order to turn this plastic bottle from a drinks container into a pressure tank I need to connect a sealed tube to the cap this is done by 3d printing a heavy-duty
threaded cap then an o-ring is inserted to keep a
tight seal with the throat of the bottle and after a quick pressure test it seems
to hold fine because the control of the drone is somewhat limited by how much air can be stored in the bottles it’s important to optimize the nozzles for
the maximum thrust possible without wasting too much air so I’ve built this
thrust test stand it’s comprised of a load cell to measure the thrust the thruster nozzle, a set length of tubing and the solenoid valve that’ll be used on the drone the data from the load cell is then logged to an SD card using this
Arduino board so that I can plot a graph of the thrust later the air is supplied from a single two liter plastic bottle so I can get a rough idea in
terms of thrust duration it’s then filled using this larger tank
via a pressure regulator so that all the tests are as consistent as possible, I
chose to run the tests at 60 psi pressure as it would allow multiple tests to be
run without filling the tank every time plus I’m more interested in how it will
perform at lower pressures as that’s when I’ll lose control to find the best
nozzle design I chose several options none of which had any theoretical
calculations behind their design but that’s the advantage of having a test
rig, the first nozzle was a straight-through nozzle which would act
as a control the second was a straight cone nozzle the third was what I’d call
a conventional rocket nozzle with its distinct bell shape and the fourth is a
reverse or a horn style nozzle each nozzle has the same throat diameter of four millimeters as well as the same exit diameter of 10 millimeters and a
length of 15 millimeters take a guess which nozzle you think will perform the
best in the comments down below I ran each thrust test six times per nozzle to get an average and imported the data onto the following graph. Now before I tell
you which nozzle performed the best let’s just admire how little thrust this
system actually produces yep that’s right the best nozzle produced a peak thrust of 0.62 Newtons
or the equivalent of lifting about 63g so here’s the lineup, did you guess
that the straight nozzle would perform the best? also the bell shaped nozzle
performed the worst out of all of them now because these solenoid valves have a
small orifice of about 2-3mm I thought I should probably
try a smaller diameter nozzle so I made this two millimetre diameter straight
through nozzle and it outperformed all the previous
contenders in both peak thrust and duration so I’m gonna go with this, let’s
talk about the drone design so the plan is to use two counter-rotating
propellers mounted inside of a single duct this will be mounted in the center
of the drone and won’t provide any pitch or roll control, it will be able to
provide yaw control by adjusting the torque from each of the motors, then for
pitch and roll control I’m going to have three nozzles mounted on three long arms
extending out from the center of mass the reason for choosing three nozzles is
because it’s the minimum amount of point of thrust
you need to have stable control similar to how you can’t have a stall with less
than three legs, now the reason why a regular drone will have four arms or
more is because they need equal amounts of propellers to counteract the torque, but
because these nozzles don’t produce a torque in the yaw direction, I should be
able to get away with three the frame parts consisted of a few 3d printed
parts using PLA filament supplied from my filament sponsor 3d printz UK, as well
as some aluminium spars when assembled, it created a triangular frame around the center duct, then I mounted the electronics starting with the relays
that would apply power to the valves then the valves themselves, which one of
them weirdly had a hole at the top which would leak air from but the leak was
very slow so it shouldn’t be an issue then came the flight controller, Arduino
board and radio controlled receiver which allowed me to actuate the valves
from my controller I then mounted the arms which the
nozzles would attach to and temporarily mounted the nozzles using tape in case
they need to be removed later for modification so now most of the hardware
and electronics are complete apart from mounting the bottles but
before I give it a test we need to consider how the flight controller will
control the valves with the regular drone the throttle can be increased or
decreased at different rates depending on the required control output but these
valves can only be fully open or fully closed therefore we need to configure
the system to know exactly when to trigger the valves, which is where once
again we need to seek Joe’s advice so this is like uhh… this has been the constant theme of like all of the rocketry stuff I’ve done it’s it’s
actually comparatively easy to develop all the hardware and build everything
but the devil is in the software umm so right now I have, like if you were an
actual controlled engineer you wouldn’t want to hear this but I it’s just a
gated and saturated, saturation limited PID it’s fairly primitive
if you do enough simulations so I have a simulation built up for how the vehicle
should actually be able to control itself in simulink that helps you dial
it in but it’s there are much better ways to do it if you’re willing to get
into like full state space control or something that actually considers the
torque that each thruster can put out but that’s also kind of a nightmare so
it’s easier to just set it up on a test stand I don’t know if we’ve covered this yet but that’s
kind of how I’ve been doing it you have basically uhh, you can see actually two holes
in the vehicle right around the centre of mass here and so it’s actually slightly
below the center of mass so the vehicle wants to fall over and what I’ve been
doing is just setting it up in this mount where it can rotate on this axis
and this axis and just you just start loading PID values and giving it a shot
Tom: just trial and error yeah that’s not how it really works in
the aerospace industry but these are model rockets so Tom: gets the job done
yeah yeah haha all right well I think that’s enough
information for now, hopefully I’ll keep you posted on how this thing goes good luck man control systems are hard but it seems promising well yeah you’re far better at control system than me well thanks a bunch Tom good luck Tom: yeah you to yeah all right
I’ll take you soon bye now I’m going to take a slightly easier
route to Joe as I want to use a conventional drone fly controller this
will save a lot of time as the code in these things is far beyond my knowledge
level, so to convert this fly controller into the similar gated PID system that
Joe mentioned, I’m going to need the Arduino to run a small bit of code,
essentially the fly controller will output a throttle value as if it were
communicating with a drone motor the Arduino will read this signal and
convert it to an on and off signal by gating it with a set threshold, when the
output signal from the fly controller increases above this threshold the
Arduino will open the valve now this threshold will depend on many different
factors so let’s program it into the Arduino and give it a test so I’ve hung the drone from the ceiling so that I can sort of simulate it in
midair so it can now swing forward and backwards and as you can hear the valves
do actuate, now I can’t test the roll in this configuration I’ve just hung it
roughly about the balance point for the pitch axis and you can hear the valves
actuating when I move the drone so what I’m going to do is I’m going to
pressurize the whole system using my large tank down here on the ground and see if this, see if this works, it should
work all the valves are actuating I just haven’t tested it with the air yet so I
should be able to push down on here and this nozzle will fire that’s actually
really well tuned already oh I don’t what happened there it’s working! Let’s try it the other way so if I push up, the two jets
at the front should fire So it will try and hold a specific angle okay it slowly gave up then what I can do is I can also flick this switch and it will always try and self level it’s trying to self level so because it’s
hanging slightly nose forwards as you can probably see apologies my remote
keeps talking to me yeah when it hangs nose forward it will try and lift the
nose to bring it level if you watch 3.. 2.. 1 that is working way better
than I expected Wow err right that is so cool
it sounds absolutely amazing aaaaand level right cool, now I think we need to install the
bottles for its own air supply and add some motors the motors were mounted
bottom to bottom and soldered in a way that they would rotate in opposite
directions so they provide a net zero torque the bottles were then mounted to
the frame using large zip ties and then it was a case of plumbing all the
bottles and valves together this was done using flexible PVC hose that would push fit into various connectors which makes the plumbing
super easy and also reversible if something needs changing,
I then hooked up to the air tank for a pressure test and to check for leaks
which was successful okay so the drone is all ready to test,
I haven’t pressurized it yet but I’ve got all the electronics turned on, I’m
actually going to run it on a tether that way I can use the full volume of
the large tank and I don’t just run out of air really quickly because for his
first test I wanted to sort of gauge how it feels on the ground so, let’s hook up the pressure tank and give it its first test I’m slightly nervous about
this but you know, it’s got to be tested so let’s armed the motors let’s check the valves okay what the valves are working increase the thrust let’s try that again the motor and propeller
I chose should be capable of 1.5 kg of thrust per motor and this
should be plenty to lift this drone but for some reason, it didn’t want to lift
off the ground, so I tried raising the duct off the ground a bit to see if it
improved the airflow not far off that worked it definitely needs the bit off the
ground just to get the airflow yeah I had control though not far off, yeah it’s controllable these initial tests were run at about 60-70
psi, the reason for this is when I fill a 24 litre tank to 100 psi and then fill
the 6 litres of bottles from that the pressure drops to around 80 psi, then
after messing around on the ground for a bit the pressure drops some more okay that seemed an issue with the
tether, it seemed like it pulled on drone so it was time for the first untethered
hover for most of the following test flights I
was adjusting the values inside of the flight controller to make the drone react
quicker and hopefully achieve a more stable hover much better, that was a lot better, much better that was fantastic yeah the battery dies like instantly I
was full throttle there you can charge the battery yeah I think I need to get a better battery really with the new battery and the bottles filled to around 80 psi it started to
look a little more promising so I think I need to tune the gyro
values up a little bit more because at the lower pressures it does start to
lose control, let’s just try that again with the tanks relatively empty yeah it just has no control and the lower pressures pressures bottles filled to 100 psi was to connect it straight to the compressor okay this is the first test at 100 psi
bottle pressure YEEEEEESSS!! Psshhh Psstt Pshhh Psssshh Psstt Pshhhh Pssssssh (yes I’m typing this by hand) Psshhh Psssst (repeat a few more times) now I know it’s a bit disappointing that
the noise from the main lift propellers are louder than the gas thrusters but
there’s not much I can do about that so just imagine it sounds like this Psshhhhh Psst Pop Psh Pop Pssst (ah please no more subtitles) I don’t know about you but I’d call that pretty good success the motors are a bit toasty and so they don’t melt through these 3d
printed mount I should probably end the testing there for now I’m honestly so happy with that if you’d like to learn more about how
high-pressure gas flowing through a nozzle nozzle causes this drone to move through
the air or how a rocket accelerates into space the most effective way to learn is
by doing, which is the reasoning behind carrying out this project but a far
simpler method of learning is to challenge yourself with interactive
content, brilliant challenges you to solve problems to broaden your
understanding of various topics they have a wide range of interactive courses
from basic forces and acceleration to real-world challenges, such as finding
out the correct sized battery for your drone each course contains clear
diagrams and animations to help you learn your way to becoming an engineer,
to support these projects and learn more go to brilliant.org/Tom Stanton
and sign up for free and also the first 200 people that go to that link will get 20% off the annual premium subscription so there we have it a drone controlled
by gas thrusters I’m really happy with the way it performed in the end, I honestly didn’t think it would even fly so the fact we’ve got a good 10-second
hover or so was really good you may wonder why there’s these huge tubes on
it now and the massive arms and that’s because I actually wanted to test whether
the longer arms would give the thrusters a bit more authority to keep it under
control and this worked well at the lower pressures but when I put up to 100
psi it then didn’t fly too well this resulted in it crashing and
breaking one of the arm mounts so the drone needs some repairs doing to it, I’m
not sure exactly how we could tackle the issue of the pressure changing so much
throughout the flight because I guess theoretically when the pressure halves
so does a thrust which then means you need to double the gyro sensitivities
or maybe even at those pressures it doesn’t have enough thrust to control
the drone, I’m sure it can be fixed on a newer version but I think this is the
best that this drone will fly I would like to thank you very much for watching this
video I hope you enjoyed this project and if you did then it’d be great if you
can leave a thumbs up on this video if you’re new to my channel and want to see
more videos such as this, then please click subscribe down below and a huge
huge thanks to all of my patrons for supporting my channel and letting me try
crazy projects like this, now I believe the sun is just about to set, so good
timing to wrap up this video thanks once again for watching and I’ll see you in
the next video goodbye

Only registered users can comment.

  1. 23:18 more plumbing (& weight).
    If the bottles have a common line, wouldn't a regulator extend useful time?

  2. Hello there, I am Paul. Please take two videos I attached as your reference. I designed a simple structure air motor many years ago. Perhaps I can be helpful.
    https://www.youtube.com/watch?v=8ifVmHx5ZWc
    https://www.youtube.com/watch?v=UjQyUTKmqcE

  3. So, you need to build into the Arduino software a timer. Amount of time spent above the thrust threshold, reduces the thrust threshold for next time. Base the change in threshold on your graphs of nozzle performance.
    Make a simple reset for the arduino (resets to baseline on power?) and just reboot the arduinos before every flight.

  4. As usual, you don't disappoint. Reminds me of the rig Neil Armstrong practiced on the nearly killed him. Great Video, Great Content, Great Creativity.

  5. you want a nozzle that starts wide and ends thin at the exit, because this forces the flowing air to accelerate, thus increasing your max peak thrust

  6. Good idea – did you try building the valves and nozzles and into the bottles and mount the bottles vertically – you could have less pipework and the bottles could give that clearance of the ground the wood was giving.

  7. Oh shit now we are really really screwed because now exist a terminator of the future 3d print and now drones

  8. Seems like you could have almost got that result without any flight controller at all, if things were well balanced enough

  9. Yo Tom, why not a nozzle that has holes in it, the same logic as in those inflatable rafts on airplanes, their oxygen tanks could never fill them up fully and fast, so they have vacuum holes that suck and push more air inside them all due to the fast air, i am explaining this badly but you need to look it up

  10. Hey, regarding your relay HIGH, LOW output thingy, I wonder what if we can do this way – We can send continuous HIGH-LOW pulses at a rate based on PWM input from the FCB… But I wonder would a relay be that fast enough to react to rapidly changing signals from FCB and overall, be a reliable way… Or try something like MOSFETs… Eager to know whether that can work… Yes, it can be complicated over simply putting a threshold to decide HIGH or LOW

  11. Very cool.
    How about changing the propulsion and making it with a centrifugal compressor instead of that counter rotating propellers. The pressure the compressor creates is then transmitted via 3-4 or more ducts (3 cheapest) to extended arms where thrusters are used not only for direction/stability control but also for actual thrust… much like a Harrier Jet but just with compressor (centrifugal would be easiest imho) and the nozzles much much more flexible since flight control is done through them.

  12. In a rocket the bell is where fuel burns. When you use pressure air, that shape just drops the gas speed. You want as much speed as possible while having sufficient flow volume for needed force. So it's the opposite, you want the nozzle hole to be smaller than the pipe leading to it to minimize pressure loss.

  13. So the issue with the drone is that with its single thrust point straight down it is dynamically unstable. As soon as you get anything off perfectly vertical thrust vector that exists through the center of gravity the force will push it further off balance. Moving the center of gravity below the center of thrust, increasing the number of vertical thrusters to three canted ones, or having a baffle that splits the thrust into a circular or triangular flows will help increase stability and controllability.

  14. I don’t know about those solenoids, but I know automotive boost controller solenoids use a low frequency PWM signal. I began building an arduino-based boost controller, and was going to attempt using a PWM frequency of about 16hz. Obviously your drone needs to respond pretty damn quick, (unlike the boost controller system) but I wonder if it’s worth a try because incorporating that into your arduino solenoid controller would be comparatively simple. You could even use a pressure sensor to adjust the PWM to help even out your thrust due to pressure drop. You are far better qualified than I to know if that’s a tenable solution, but I thought I’d throw it out there in case it was something useful that hadn’t occurred to you!

    I know nothing of aerospace engineering, and aren’t generally interested in drones and aeroplanes, but I keep finding myself watching your videos because I enjoy your approach to problem-solving. Keep up the good work!

  15. Looks like you had a bit of saturation on the PID around the 12' mark. Either caused by delay, deadtime or windup. In any case the control isn't robust. Judging from my control theory classes a couple years back. You'd probs want the PID to have some pole compensation. But that's a bit complex. https://fei.edu.br/~psantos/PID2737725.pdf https://www.ijareeie.com/upload/2013/november/17_COMPARATIVE.pdf

  16. To get recording of 'real time' audio performance of the valves without the 'prop/motor noise' perhaps installing a 'noise cancelling mic' on the 'prop/motors' would be of benefit ?

  17. 🚀👨‍🚀Spacex Rocket toy 🤷‍♂️😳 go figure 👍in stores soon or d/load the data free & build it 😳🦘💉
    your test flight keeps falling left like when your Drunk 😳

  18. For your self balance/level test, a drop of hot glue on 1 string at centre of top may have allowed the system to show results of how it's doing in both axis

  19. Here's to hoping I'm wrong and I'll keep on watching, but this looked like a total waste of time right from the concept stage with weight being such the critical factor it is with drones.
    Q?: Why add weight for some wrinkled-inky theoretical thrust system when at the best you might be to lucky to just barely fly. EDIT: 12:25 Seems you too were skeptical. ;]
    A!: Because it's cool and for just that reason; "_some wrinkled-inky theoretical thrust system"!!_ Cheers !! Just watched and you had more success the I'd have or ever imagined… ;]

  20. Interesting point being made about trial and error vs calculations.
    The Russian space effort was based heavily on trial and error, especially the closed system breakthrough innovation. It's hard to imagine any other approach before computers of todays calibre.

  21. I know for sure the straight would work best and the bell the worst. It is not a rocket engine. Jets would have bells on the air out side if this was the case

  22. Can dry ice sublimate fast enough to increase duration? Would the weight of the dry ice be better spent on tank capacity sans dry ice? The overpressure release valve would add complexity and additional weight.
    There might be something useful from these water rocket guys. They are gluing bottles together and reinforcing them with fiberglass or carbon fiber.
    http://www.raketfuedrockets.com/en/pressurevessels.htm

  23. reaction thrusters with hydrogen peroxide monopropellant= consistent pressure. or you have to store moar gas at a higher Psi and regulate it before the valve.. like maybe the CO2 cartridges and a few torn up air-soft gun

  24. wholley shit, that guy is Elon musk before he grew up then came back in time. Don't let them meet or the world will implode

  25. Treat the signals like edge triggering, not level triggering. Trigger on sooner on a up signal, and off sooner on the down. You're not trying to fire until level, you're firing to start leveling and let momentum do the rest. The adjustment of the how soon/late is a measure of difference between current and desired acceleration. Air thrusters can also be PWM, not just an all or nothing.

  26. how does old saturn rockets burn in outer space without the oxygen? they we cant go to the moon today, but could in 69? how they overcome the vaccume issues when ppl dies and cannot even be in the biggest/strongest one on earth? just curious, hence ive been fed some bs from ppl like you. liking the $

  27. So, since the drone is going to fly in an atmosphere, why not use aspirator nozzles for thrust with pressure regulator to maintain relatively constant pessure through the nozzles and the flight. If you make the nozzles with small enough diameter, the aspirator effect will compensate for the small volume of air going through them by accelerating the ambient air to create thrust. And the pressure regulator will slow down the usage of the air by obviously regulating it to a lower pressure, that will be enough for the nozzles to operate. And then a little more adjustment on the gyro controler and you should make a decent flight with it!

  28. I was wondering, what is the speed of the valves?
    Couldn't you switch them off an on again really fast at different intervals to simulate variable thrust?
    Surely that would make control easier.

  29. Reverse funnel thruster nozzle is the best choice but, nozzle opening is perpendicular to a large flat surface causing too much turbulence, I suggest modified external shape to be more like the interior shape.
    This will boost the thrust while consuming same amount of propellent.
    Too bad a sketch is impossible via the comment section.
    I think you are bright enough to get the jist of what I see.

  30. In zero G the speed of falling doesn't have to be counteracted by the speed of the control system. Gas thrusters are just too slow for counteracting gravity.

  31. space may be the final frontier but its made in a hollywood basement.. sorry… the ground is fixed and not moving.. a bottle of water proves it… TEST IT before you go on any indoctrinated rant… hold a half full bottle of water on its side.. can you cause the water to be 'motionless' ??? your kitchen table has NO problem with this challenge… now, compare the two results…

  32. Have you considered using a bit of dry ice in your gas bottles as it converts from a solid back to a gas it will keep your pressures up for a longer period, If you had stronger bottles you could use liquid CO2, actually that's a great idea try and use the small CO2 cartrages that are used in air rifles or pellet pistols they are small and lite weight, and would last much longer than compressed air as they are liquid CO2 compressed into very small aluminum cartrages, but you could try some pieces of dry ice with a must of water on it if need be to encourage conversion!!!! But I'd definitely look into the CO2 cartrages, I think a box of 12 is around $10.US, also just had another brilliant idea you can put a few CC's of liquid nitrogen in your bottles as it expands twelve times or more when it goes from liquid to gas and it does it much quicker than dry ice will since it's already in liquid from the start. Anyway sorry for the ramble but I think you done an awesome job either way, and I found it to be very interesting, exciting, & educational all at once, well done!!!! And I hope my little brain fart helps you progress, good luck & Cheers

  33. I do internal combustion engine design, and I knew the straight one would be the best. In fact if you ran the nozzles backward, ie: converging, they would be better. As you found out with the 2mm nozzle.

  34. perhaps incopirate a presure regulator into the design that way the presure will remain constant untill the main tank presure drops to your reg presure, then the more air in the main tanks you can get will mean longer flight time

  35. Your valve with a hole in the top is likely a three way valve. Flow can occur through one path when the valve is not energized, and a different path when energized.

  36. The reason your thrust don't add up are due to the lower propeller working in disturbed air, therefore not adding the thrust you expect. So even if they are counter rotating the turbulence is a large effect. A few tips. Try switching one propeller to a different grade, also minimize the gap in between the duct and the fans. Also more blades equals less thrust per blade but using a high speed fan made for ducted fan projects may help, but then you need a higher KV motor. If your current system can handle more voltage, try going for more cells on your lipo. Cheers 🙂

  37. What about using one of those small 12v compressors to supply continuous pressure? I found one about the size of a tennis ball inside one of those glove box compressors for your car. Of course you’d have to use larger lift motors but I think it could work.

  38. I love how you Science the heck out of your projects! Was going to suggest a nozzle like the one you eventually went with. Good call mate!

  39. 2:00 so funny to see this interviewed guy even looks a bit like Elon Musk… same style of eyes, or is that because of the lack of sleep? Lol

Leave a Reply

Your email address will not be published. Required fields are marked *