You’ve got to hand it to [Tom Stanton] – he really thinks outside the box. And potentially outside the atmosphere, to wit: we present his reaction control gas thruster-controlled drone.
Before anyone gets too excited, [Tom] isn’t building drones for use in a vacuum, although we can certainly see a use case for such devices. This is more of a hybrid affair, with counter-rotating props mounted in a centrally located duct providing the lift and the yaw control. Flanking that is a triangular frame supporting three two-liter soda bottle air reservoirs, each of which supplies a down-firing nozzle at each apex of the triangle. Solenoid valves control the flow of compressed air from the bottles to the nozzles, providing thrust to stabilize the roll and pitch axes. As there aren’t many off-the-shelf flight control systems set up for reaction control, [Tom] had to improvise thruster control; an Arduino watches the throttle signals normally sent to a drone’s motors and fires the solenoids when they get to a preset threshold. It took some tuning, but [Tom] was eventually able to get a stable, untethered hover. And he’s right – the RCS jets do sound amazing when they’re firing, as long as the main motors are off.
This looks as though it has a lot of potential, and we’d love to see it developed more. It reminds us a bit of this ducted-prop drone, another great example of stretching conventional drone control concepts to the limit.
Pretty sweet. I wonder if you could use a single small CO2 cartridge instead of the three 2 liter bottles. The total weight should be about the same, and you could get longer run time. And it would be more aerodynamic.
People forget, in addition to the standard 12g cartridges 40g ones exist, for not much more of a size/aero penalty. Also copper pipe can hold compressed air quite well.
Valving to control high pressure CO2 would likely have to deal with 2k PSI (140 bar) and corresponding plumbing. That might be a worthwhile trade. Consider nitrous oxide (for whipped cream) for lower pressure.
Fun and brilliant!! Also reminiscent of the “flying bedstead”
https://www.nasa.gov/feature/50-years-ago-the-lunar-landing-training-vehicle
I was going to say that myself, and now that he has worked (is working out) the control for a system like this this could open up a new field of reaction (rocket) drones!
It’s an interesting concept but if you watch the video the thrusters do some odd things, triggering at weird times or not triggering at all. On top of that, the on/off control seems prone to instability.
I’m interested in how difficult it would be to produce a more complicated control model that runs the thrusters for a period appropriate to the goal orientation. It should be easy to get the current angular velocity from the accelerometers, and the moment of inertia could be calculated the roll in response to a single thruster.
just pwm the solenoid valves in proportion to the servo timing
Remember solenoids are rarely rated to run contentiously (which this might well be near enough to) and have a pretty slow switching time (in PWM terms at least). I think crushing a tube with a high power servo and lever arrangement as a primitive regulator might be a less problematic but still simple solution, might even be able to do that without the solenoid for on off.
It would allow a simple offset curve in the program to account for the dropping pressure.
PWM would have to run so slow that the drone would be uncontrollable. On the other hand Pulse Position Modulation would be perfect. I.e. measure the shortest good pulse the valves can make, and fire that pulse off every time the error gets too high. For an add on to the controller, you’d need to integrate the servo PWM vs time then fire a pulse and clear the integrator after a some threshold was reached.
The flight controller is likely outputting R/C pulses to electronic speed controllers (ESC). R/C pulses are 1 to 2 ms long, repeating about 20ms. ESCs convert that to 20kHz+ 3 phase BLDC PWM.
.
Capture an RC pulse frame (20ms) and expand it to about 100ms. Ignore the first 60ms of the output pulse as the flight controller is likely outputting drive for a lift motor rather than a control thruster. Apply the remaining pulse to the valve which now fires every 100ms for some time between 0 to 30ms. Maybe the flight controller can be programmed to do this for you.
.
I suspect 4 thrusters would be easier to control than 3. For 3 thrusters, tipping X one way would fire 1 thruster, but tipping X the other way fires 2 thrusters and likely introduces an uneven Y impulse.
.
Tricopters also have torque issues to compensate, but these would be absent when using 4 fans / thrusters.
Bang-bang control should not necessarily be a problem. Combined with a minimum pulse length it puts a lower bound on the maximum deviation you have to accept.
He could use some pressure regulator to maintain a constant pressure.
It would be interesting to see if air from the main motors could be steered and redirected to the thrusters for control. The air pressure is much less, but you get a higher volume and some flow control with servos.
Being pedanticpedantic for fun: most drones I’ve seen use gas thrusters for control.
I was wondering who’d go there first.
Is there a reason not to use pressure regulators in each tank (sounds better than bottle)?
That way you could put a 100PSI in each, regulate at the minimum you can get control, and have control thru the whole range from filled to depleted tanks.
I’d guess weight/cost. Assuming you aren’t a competant machinist with a tuned lathe, regulators are expensive.
The sounds of the first test reminded me of the launch sequence from the “Fireball XL5” show. Ah ,memories …
Ah, the humble cable tie (zip tie).
In my opinion truly one of the great inventions of the 20th century, and also one of the greatest fastener concepts of all time, right up there with nuts and bolts, split pins, circlips, jubilee clips and other indespensible ways of holding things together.
Totally agree: https://hackaday.com/2018/03/22/mechanisms-cable-ties/
He should lookup Venturi nozzle blow guns. Lots more thrust for the same gas supply.
I was thinking the same thing. Air expelled out the nozzle enters a ventri, which entrains more air, increasing thrust. Search for “eductor nozzle”.
OK, read that as “educator nozzle.” Feeling better now.
fun project, lot’s of challenges, thanks for posting
Nope, not practical. Neither relays nor solenoid valves like PWM. They don’t respond quickly enough, and they wear out after relatively few cycles.
A 3D printed fan duct is not smooth. Lots of drag happens at those layer slice steps.
You might consider looking for ducts (and fans) in the hobby R/C shops, as used for model jets. (I know you have a printer, but smoothness of the duct and small clearances with the blade tips are critical for efficiency.)
Or maybe acetone-smoothed ABS? It’ll still have some texture, though, I suppose. He could do a body-filler job on the inside of the duct too, and sand that down nice and smooth. Doubt it would add much weight. Or maybe even a filling primer.
I wonder how much can be gained by using something like SodaStream bottles. They are rated to very high pressures (i believe 1000’s of PSI / tens to hundreds bar) compared to a regular soda bottle. They are a bit heavier, but the increase in gas “payload” could make more than up for it?
Well this is a few years late, but here’s a similar idea on steroids :-)
https://www.youtube.com/watch?v=UCwjZP9NgrA&t=59s