Flexible monocopter drone can be completely rolled up | Hacker News Reader

Flexible monocopter drone can be completely rolled up

spectrum.ieee.org

I tried to do a similar thing with a single actuator frisbee. The goal would be a frisbee with a single fin could fly anywhere when launched out of some high power launcher.

The goal was to make a single-use aircraft which could go anywhere in a city for under $1. By making the aircraft unpowered and under 250g, lots of regulations no longer apply, and it becomes practical to use them to deliver small items.

The fin was actuated by a 10 cent microcontroller using a ultra low cost GPS and light sensor (for sun position), powered for up to 2 minutes by a coin cell.

The challenge was that it seems gyroscopic effects means that if a frisbee has lift and moves forward in the air, it will always see a (very slow) turning moment that requires a massive fin to counteract. The end result is the frisbee eventually turns to vertical and falls out of the air.

My prototype could still land gently anywhere in a football field when thrown by hand, which I was pretty pleased with, and had a BOM under $1!

If anyone has any ideas to make this not happen, I'd like to hear them.

Sounds like something the military would like. Cheap way to see if anyone is behind a barrier in front of you.

If by vertical you mean that the frisbee has rotated so the disc is perpendicular to the ground, that effect is actually aerodynamic, not gyroscopic. It comes from imbalance of lift due to two halves of the disc see different relative airspeeds.

In competitive disc sports this is usually mitigated by adding more spin (to add more moment of inertia in the spin axis to gyroscopically resist the turn), and by releasing the disc with the advancing edge pointed down to extend the period before it over rotates.

If this is what you were talking about, then you should be able to just make the disc more edge heavy to try to counter act that. Or maybe you can try to someone add more drag to the forward spinning edge. Presumably if you have a single fin, that fin will usually be at the 6 o'clock position (from top view). Assuming we have a forward moving disc rotating clock wise (from top view), and the fin is at 6 oclock, then the forward rotating edge would be at 9 oclock. If you had a second fin (or some other type of feature) basically mounted at 9 oclock (so at 90 degrees relative to your main steering fin), then it should always spoil that air... might be able to use that to somewhat counter act the rotation.

However, if you were actually talking about the gyroscopic effect (trying to rotate a spinning disc along it's axis pointing in the forward direction should cause a torque in its pitch axis)... I don't really know how to deal with that.

I was talking about the fact that the center of lift is ahead of the center of mass of a circularly symmetric wing with positive lift. [1]

That force would normally cause an aircraft to 'pitch up' (and hence planes have tail's and carefully balanced weight to stop this happening). In a frisbee, the gyroscopic effect means that this 'pitch up' force actually translates into a 'roll right' motion. By making the rim heavier, you can make this happen slower, but as far as I can see, you can never eliminate it entirely.

My fin is on the outer edge of the disc, spinning with it, so it can oscillate at ~50Hz to get independant roll and pitch control, while also being able to trade off loss of angular momentum vs lift. It also has the benefit that it can be made from cardboard in a single pressing with the disc with no glue or assembly labor.

[1]: http://ffden-2.phys.uaf.edu/645fall2003_web.dir/Mike_Abrams/...

Ah hm. I wonder which one of our two effects is dominant. I can see that your explanation must plainly exist - it's just not part of the standard "intuitive" model taught to frisbee players.

That's a really clever fin design.

If the dominant component of rotation was due to the center of pressure being ahead of the center of mass, could you to use timing of the fin being in a "high drag" configuration to pull the time averaged center of pressure back towards to the center of mass?

> could you to use timing [...] back towards to the center of mass?

Yes, but sadly the effect isn't large enough until you end up with a very large fin and a very powerful actuator to move it. That difficulty is what made me put the project on the backburner for now.

I'd love to see your setup and how you attached the fin. I've thought of doing something similar, but instead of a fin, I figured the easiest way to steer the frisbee would be to bend the disc either up or down. Depending on spin direction this should, at least according to my theory, gently steer it left or right. A single actuator mounted to the center of the disc, pushing or pulling on two points on the bottom lip would be all it should take. Intuitively I think this might also eliminate the effects of the asymmetric lift torque since it could continually bias which side of the disc has more lift based on speed, rotation frequency, and angle to keep it balanced.

To counteract the gyroscopic effect you'd need 2 frisbees with attached axes spinning in opposite directions. Lots of choices for which way to arrange them - side-by-side, forward-back, vertically, some staggered formation...

If you added a vertically oriented 1-dimensional camera, i.e. a single column of pixel sensors, then it could automatically extend that into 360 degree video.

It spins at ~5 Hz. 5 fps isn't great, but this is a cool idea.

Use 6 of them, each facing at 60 degree angles. That would give you 6 updates per spin, for ~30hz video.

that would be cool, then you could do your visual inertial odometry stuff with that data, even if low quality might be like working with IR

The ones used in barcode scanners have quite a high refresh rate (100kHz IIRC) so that could be doable.

Mounted on a hinge in a kind of ¬ shape with a weight at the end (hinge at left, sensor then weight at bottom right), it could perhaps capture at an angle in self-stabilizing manner, ie independent of the tilt of the main body.

Don't these cameras require to be stabilized? You only have one column of pixels so you can not correctly align the columns through stitching (as in panorama mode on phones).

As impressive as the stabilization used is that these dromes manage to hover in roughy the same spot (instead of drifting everywhere), I do not believe it will be sufficient for a rotating line camera.

I don't know how well it would work, but I think you could still perform alignment using a single column assuming that you would expect to see patches with the same color connected together more often than not. Especially if you're trying to do video since then you can also exploit temporal coherency (meaning that you expect many objects to be stable over time). I'm sure in my head it's much simpler than practice, but it seems like it is achievable at least in theory.

Sounds like a very fun problem for computer vision researcher to work on. Probably not terribly difficult to generate a bunch of test data from existing 360 cameras by downsampling.

The column images can be tagged with samples from an onboard gyro and compass. The stitcher does not need to rely only on the pixels themselves. Normally the next pass will be pointing a little higher or lower, so such annotations will be essential.

Gyro? Relying on just angular velocities doesn't provide any good tags. You probably mean accelerometer or a combination of these referred to as IMU.

And even then, the raw data of these sensors are extremely noisy, so you need something like an EKF to get a more reliable attitude of the drone. THIS can actually be used to tag photos.

If "onboard gyro and compass" means something to you substantially different from an IMU+EKF system, it makes me wonder what century you are talking about.

My comment was about tagging pixel column data with non-optical measurements. Spewing alphabet soup fails to add clarity, and does not even make you look smarter.

Given adequate bandwidth, EKF can be better applied at the receiver, where more computational power is available.

Alphabet soup? Sorry, but nothing that you say makes any sense and I politely corrected you regarding the extremely stupid idea to tag any raw sensor informations. Your new proposal of streaming this kind of data to an offboard processor just proves that you have zero dot zero zero experience about what you are talking about.

You have contributed exactly nothing with your grandstanding. You should be ashamed.

The first thing I thought of was a single-'wing' maple seed [0].

https://www.researchgate.net/figure/The-Samarai-MAV-design-i...

The first maple seed inspired 100+ years ago https://en.wikipedia.org/wiki/Monocopter

Interesting and thanks for sharing, I also ran into another design that is maple-seed like in terms of aerodynamics. Sadly this is one of the cases where human ingenuity is misdirected and is still causing harm today:

The BLU-43 Dragontooth mine [0][1] and the Soviet PFM-1 [2] which still are responsible for children losing limbs in Vietnam, Laos, Cambodia and Afghanistan.

The BLU-43 came with a chemical deactivation mechanism but of course that failed in a subset of dispersed mines. The PFM-1 in Afghanistan came without a deactivation mechanism, the later PFM-1S had one.

[0] https://en.wikipedia.org/wiki/BLU-43_Dragontooth

[1] https://johnryle.com/?article=the-invisible-enemy

[2] http://www.one-step-beyond.de/en/countries/afghanistan/mines...

Edit: formatting

https://youtu.be/n_q_DD_4LNg

In action, neat flying mechanism!

> The total weight of F-SAM is just 69g, of which nearly 40% is battery, yielding a flight time of about 16 minutes.

Nice to see mechanical performance qualities (battery mass percentage), and battery/aerodynamic performance qualities (minutes per mass percentage) condensed to such a small set of numbers. Unfortunately battery and aerodynamic qualities aren't really separable, but if we assume batteries to be a constant between designs (they are all generic off-the-shelf parts after all) "minutes per mass percentage" becomes a quite interesting metric.

That does not fly in the way I initially presumed it does.

I would never have thought of something flying like that. It's a nice reminder for me that we don't know what we don't know.

I also thought the small propeller would point upwards (rather than to the side).

Now afterwards it feels so obvious.

First i thought, but the propeller, it's so small

Turns out the whole thing was the propeller :-)

The real propeller was the friends we made along the way

It folds into a small storage tube, is lightweight, has good autonomy, makes little noise, and safely autogyrates to the ground on failure. If you could somehow add a camera and stabilise it, it could make an excellent search drone for search / rescue / military ops.

What would be the advantage over existing drone technologies? Look at these autonomous drones [0]that can fly through the forest at high speeds. High speed, highly maneuverable and can hold other equipment like a camera. [0] https://electronics360.globalspec.com/article/17357/video-fl...

Cheap. Not every mission requires a stable camera.

If you don’t need stable hovering or tricky maneuvering, a traditional airplane design is almost certainly going to be even cheaper.

And there are already relatively cheap commercial offerings with very impressive collision avoidance and path planning (www.skydio.com)

Since it folds into a cylindrical shape, it's easier to transport and could potentially be fired off from something like a grenade launcher. It's the computer vision and shaped charge that I'd worry about.

what a beautiful piece of engineering, its flight seems so...biological.

Amazing tech. Combined with a fast fixed cam it would maybe be possible to get a 360 degree FOV.

A line scanner, like the proposed LIDAR implemention. We use line scanners for paper. You could certainly put a line scanner behind a conventional lens but it would feel kind of weird to "waste" an entire dimension of the lenses projection qualities. Would there be opportunities for design compromises favoring one direction over the other? (I'm thinking of the weird eyes we find in flight animals like sheep that need full horizon perception more than point of interest focus)

I guess it's using thrust to control cyclic? Fast pulse to pitch up/turn opposite way and then minimum throttle to stay airborn keeps it level?

I wonder if you attached a 360 camera whether the footage could be stabilised or 'fixed'

Not all images must be video. With a rotation of only 5Hz, fast shutter speeds should produce clear snapshots.

I wonder if it could be even better and lighter with carbon fiber instead of basla wood.

Carbon fiber is much more difficult to iterate on though. Using accessible and cheap materials is beneficial for adoption by hobbyists or other researcher.

Probably not by much. Balsa wood is extremely light, cheap, and easy to form into a wing. It is also easily repairable.

The real weight savings would probably come from the motor - and a carbon fiber casing might work here instead of the presumably steel casing currently in use. And though shaving weight off the wing would almost certainly move the center of mass further outwards, shaving the mass off the motor would move it further in. That might be detrimental, though, as the motor needs some leverage on the COM for this design.

This is a terrific project that invalidates many traditional ideas that we have about flying machines.

Great idea. Wouldn't it be more stable with two rotor-wings?

Symmetry would ruin the self-stabilizing autogyro descent on unexpected power down.

Would it? Aren’t most autogyros 2-blade? Multi-blade helicopters also use autorotation for emergency landings.

Helicopter autogyro requires deliberate control input. This one enters self-stabilizing autogyro when control input stops at any given operational state.

It looks dangerous, but awesome.

The whole thing weighs 69g, meaning that the kinetic energy even when rotating at full speed is quite low. You'd still want to avoid getting your eyes in its way tho.

Reply on news.ycombinator.com