Hacker News new | past | comments | ask | show | jobs | submit login
Ikarus electric “rocket” – Thrust-vectored flying ducted fan [video] (youtube.com)
357 points by inetsee 6 months ago | hide | past | web | favorite | 115 comments

Great project. Though as a former pilot, I was wincing at all the foreign object ingestion happening in the opening sequence! :)

I was wincing at that too, but then I realised its probably just as well it was tested in such an environment. ;)

Yeah it could have used a FOD trigger warning ;)

FOD: Fear of Destruction?

Should have flown a MIG-29.

I had a similar project on Kickstarter. https://youtu.be/xRviasOMCvk

We made 300k$ but then dropped the project and gave money back to backers. Basically beyond the "damn cool" effect there is no edge over quadcopter. It is less efficient. You gain something by the duct effect but the duct is heavy. It is very noisy. More sensitive to wind, especially when you try to reduce weight to make something affordable to build and sell. The only benefit was safety due to ducted fan but the point vanishes against mini drone the size of your hand like latest dji.

But for sure it s dann cool and fun tech to master. Kudos!

This is really cool. I find it very strange that it's less efficient though.

I believe it is because propellers get lift from air pressure, which is more efficient than just pushing air for thrust. If you are sucking air through an engine you don't get that bernoulli effect.

I could be wrong. I am not an aerospace engineer

The efficiency of a propeller depends on the ratio thrust/area, the smaller this ratio the higher the efficiency. (See 11.7.3 at https://web.mit.edu/16.unified/www/FALL/thermodynamics/notes...) The issue with the ducted fan is that the propeller size is constrained and the weight is higher due to the mechanical structure. A small quadcopter has twice or more propeller area for a portion of the weight.

Thanks! I learned something from this.

Cool, add a camera and some AI and you've got an autonomous leaf blower.

Oh wow. Making those way more annoying previously seemed impossible.

I've used an off the shelf drone to blow leaves off an inaccessible roof. It can work fairly well, but you waste most of the blowing power since the air blows in all directions.

Wow, it seemed straight forward for the first 10 minutes or so, when he introduced the biggest challenge: gyroscopic motion due to the high speed rotation. this is a gripping video.

I was expecting some sort of dynamic feedback system based on a gyro sensor, but apparently the compensation is just hard-coded? I'm amazed that can work.

Ducted-fan VTOL aircraft are surprisingly stable. Here's the Hiller Flying Platform of the 1950s.[1] It's very stable, even though top-heavy. It's mostly steered by leaning, but there are thrust-vectoring vanes to control yaw.

Many variations on this theme were tried in the 1950s, and many of them flew. But it never caught on. There's been recent interest in larger ducted fan drones. This apparently scales better than quadrotors. But nobody seems to really need large ducted fan drones.

If you like early small VTOL craft, many of them ended up in the Hiller Aviation Museum in San Carlos, CA.

[1] https://www.youtube.com/watch?v=W3FS3D1rCos

The gyro effect is pretty predictable. Anything where you're pretty confident in your model, it's better to compensate in the feedforward part of the controller than to push it into the feedback part. Think about gravity compensation - adding a constant amount of thrust to cancel out your weight. It's not going to match the model perfectly due to error in your attitude estimate, but it's still better than treating gravity force as an unmodeled disturbance. Gyro compensation is similar.

You have to pre-program in the compensation because it basically rotates your controls 90 degrees which isn't something a simple feedback loop can deal with. No matter how fancy the feedback system is if you swap 2 controls its not going to be stable, it'll spend the whole time chasing the problem.

Yes, it is impressive that he got it going with PID controllers plus whatever else is in the controller firmware.

This sort of inverted pendulum type system would typically require at least 1/s^2 control. It also has the additional complexity of gyroscopic torques, center of thrust not through center of gravity and servo control lag.

This is similar to controlling a rocket, something the Germans took a long time to perfect with the WW2 V2 missile.

I'm glad I'm not the only one that found this project fascinating (see my http://weeklyrobotics.com/weekly-robotics-4).

If you are interested in drones/robotics then it's very well worth it to watch this video. You can learn a lot from the author.

I wonder how much more power you can provide it to lift a human for a say half an hour.

I've always wanted to build a jetpack like that from xDubai [0]. This is definitely easier to build for an amateur compared to the one professionally built in the video.

[0] - https://www.youtube.com/watch?v=_VPvKl6ezyc

Richard Browning's suit is even more impressive, but presumably infeasible to do with electricity with current technology https://www.youtube.com/watch?v=DafoDiRvGLM

It would probably look something like this:


I saw that and while that is super cool - (it's being sold for ~$100K btw), that's not what I was looking for. (think more human plane / human with wings rather than an human + machine if that makes sense)

(more horizontal than vertical if you will - I'm getting my solo skydive certification soon and my plan is to move on to wing gliding - which I want to follow up with something similar to xDubai's suit).

(earlier non electric)

Not quite, as they use contra rotating fans to counter the gyroscopic forces the creator had to tackle. The RS-16 might have a single but can't find any info on it.

Somewhere in the youtube comments it says it uses fans below the propeller to counteract the torque.

EDIT: Nevermind, it appears you're right!

Total noob here, what's the difference between wrapping a quadcopter with a duct and this?

Is it because a single rotor is much harder?

Well, in simple terms with four rotors you vary their speeds to tilt the copter whichever way and thus steer it.

Here you have a single rotor, and just try to imagine what happens when you vary its speed. It might bob around, fly up and then crash into something, but you won't be able to point it anywhere. Imagine trying to fly a ceiling fan.

This guy uses something called "thrust vectoring", i.e. "pointing the thrust of the single rotor in some specific direction" to steer it around, but for that you need something to redirect the flow of air, which he designed and custom 3D printed.

Which is pretty damn cool.

But what's the benefit of taking this approach instead of the classic open propeller design or to ducted multi propeller designs?

The advantages of a single blade over the equivalent 4 smaller blades is efficiency. A tiny part of that is losses from extra motors but mainly its from blade length which dramatically affects power efficiency, up until the point where the tips go over the speed of sound, helicopters heavily optimise for this.

The duct itself offers some advantage but it's used here to vector thrust for control rather than just go up and down.

Battery power is quite constraining already for aircraft, ducted fan offers some benefit. The US military use a petrol powered ducted fan drone operationally, though i believe that's dual rotor, this is single rotor that offsets counter rotation with vectoring.

So, is this thing guaranteed to go faster in a straight line even without wings to provide lift? (Slight downward vector on thrust?)

What would happen if rotors were added, ala Kamov helicopters.

Fun building it

As the parent said, it's pretty damn cool :)

Mainly a completely different control mechanism and single vs multiple power-sources.

On a multirotor, there are obviously 4/6/8 motor/rotor sets, typically positioned far from the mass centroid, so control is effected by 1) varying the thrust from each rotor (primarily pitch & roll), and 2) varying the speed of each rotor for inertial effects (primarily yaw). This uses a single rotor and thrust vectoring fins to influence the direction and reaction to the thrust.

I'm more interested in how the efficiency numbers work out.

Typically, a larger rotor produces more thrust per unit of energy. However, a ducted fan is also typically significantly less efficient than an un-enclosed rotor. Since I don't have numbers, my question is which effect dominates -- is it overall more or less thrust efficient than a multirotor?

Either way, it is a very cool design & build -- nice work Ikarus!

I believe a quadcopter has variable-pitch wings that spin to create lift. This is a fan that pushes a column of air down to create thrust.

Helicopter blades pull the vehicle up from the top, rocket thrust pushes the vehicle up from the bottom.

Quadcopters never have variable pitch props, they just vary the speed of them to vary thrust. Moreover, thrust from the bottom or top doesn't really matter (look up the Pendulum Rocket Fallacy)

Well, they usually don't have variable pitch. But there are a few rare examples of variable pitch quadcopters, for instance this monster from Curtis Youngblood:


For most applications you don’t need variable pitch blades on a quadcopter, they’re also very delicate and break easily on drones.

I used to fly r/c helicopters before drones became popular. The pitch mechanism was routinely a source of problems for me. It’s pretty amazing to watch it work though.

neat! but (speaking of the pendulum fallacy) if the fuel tank was attached with cables instead of a fixed structure, wouldn't that actually work?

No, not at all.

The technical difference is how the overall thrust is steered:

* Quadcopter: the speed of four (fixed pitch[1]) propellers is managed independently. Spinning a single propeller a bit more tilts the copter away from that propeller[2]. This is very unstable, so speed needs to be adjusted at kHz rates[3][4].

* "Electric rocket": the thrust of the single propeller is steered using the control surfaces at the bottom to deflect the thrust. This is called thrust vectoring. Since the propeller is not directly used to steer, the speed doesn't have to be adjusted that often. However, the position of the control surfaces does.

* Helicopter: tilts the rotor plane[5] to tilt the aircraft, and inclines the individual rotors to change the magnitude of thrust. The rotor speed typically remains constant.[6] Helicopters are aerodynamically stable and don't require an automatic control loop [edit: no, see follow-up post!].

In theory the "electric rocket design" with a single propeller might allow higher efficiency: larger propellers are generally more efficient (thrust per watt). (However, the model in the video won't be very efficient, the props are quite small. And there'll be other tradeoffs too.)

Also, this is a refreshing take on RC aircraft that hasn't been done as much before.

Lift is generated on both sides of the propellers in both cases.[7]

Optional reading:

[1] Generally - variable pitch quadcopters exist. However that gives up the mechanical simplicity that's one of the main advantages of quadcopters over helicopters. Their advantage is that they can actively decelerate a lot more efficiently and also fly upside down easily.

[2] And turns the copter (yaw). Quadcopters use two pairs of counter-rotating propellers to be able to cancel the yaw.

[3] Quadcopters are unstable and need to measure linear and angular acceleration and correct the propeller speeds very quickly. Control frequency of 100Hz flies, 1kHz is somewhat standard, 32kHz is state of the art. (Note: Propellers have significant inertia. To give an idea, reversing prop turn direction in-flight takes ~1sec (and requires special symmetric props if it's supposed to work efficiently).

[4] This is feasible since the propeller is smaller (less mass, less inertia) than for helicopters.

[5] Due to gyroscopic effect, the tilt of the aircraft is 90 degrees out of phase.

[6] Assuming the typical variable-pitch helicopter. There are other options both for real life helis as well as for models (fixed pitch, coaxial, ...) but they're rare (or very cheap toys).

[7] Side remark: lift is generated due to flow turning, not due to different pressure differences caused by different airspeeds over top/bottom of wings as is taught in school. Search for "NASA incorrect lift theory" to learn more. https://www.grc.nasa.gov/www/k-12/airplane/wrong1.html

> Helicopters are aerodynamically stable

That’s not my experience as a helicopter pilot.

I believe the textbooks say they are “statically stable and dynamically unstable”, but the bottom line is you surely do need a control loop.

Thanks for correcting that part then.

I've never even been in a real helicopter unfortunately. I really wish I could touch the stick of one once to see how it compares to models in terms of touchiness... it's (un)fortunate that there's probably no realistic way to try that.

If you can afford $150 some helicopter flight schools offer prospective students a brief introductory lesson where you can touch the stick.


I really enjoyed Smarter Every Day's video about learning to hover: https://www.youtube.com/watch?v=eXR1olg_I0w

I think the main advantage of variable pitch quadcopters is the ability to change the thrust of each motor more quickly. You don’t need to wait for a motor to speed up, you just slam on the pitch of the blades and the rotational momentum in the blades gives you zero-latency thrust (meanwhile the motor is spooling up to prevent the rotation speed from dropping too much).

Of course, you’re also right that the aircraft can decelerate faster, since each rotor can thrust in either direction (the same effect that allows for flying upside down and accelerating toward the ground faster than 1g).

Lastly, I assume that variable pitch quadcopters are able to autorotate in the event of a motor failure (not that that’s super important for relatively inexpensive unmanned aircraft).

With regard to footnote 7: As someone who has played a very small part in correcting some misconceptions about how airfoils generate lift, I find that I now have to apply some corrections the other way: there is necessarily a pressure difference across an airfoil if it is generating lift, and Bernoulli's equation does correctly describe the relationship between that difference in pressure and the equally real difference in local airflow speeds (up to airspeeds where compressibility is an issue, as Bernoulli's equation is only strictly correct for incompressible flows.) What was most wrong with the old explanations was the use of a fallacious 'equal transit time' argument in an attempt to explain the difference in speeds.

> Also, this is a refreshing take on RC aircraft that hasn't been done as much before.

Maybe the reason it hasn't been done much before is because one doesn't have to actually go and build this to know that it will be extremely inefficient and painful to fly...

Surely the only difference between push and pull is near the ground right?

Another example would be rapid descents: the aircraft can become quite unstable when it gets into its own downwash.

(A high control frequency makes all the difference here for quadcopters. Compare a toy and a hobby quadcopter, it's easy to spot the difference.)

Also, lift generated due to the top and bottom of airfoils can be quite different.

The YouTube codec did some wild stuff with that slow-mo flying mulch at the beginning... any video experts around that can explain?

Tom Scott did a video on this very subject with a decent amount of depth: https://www.youtube.com/watch?v=r6Rp-uo6HmI

In the future I have a feeling that's the sound you hear before death is approaching in a battlefield. Much the same way as the sound of jets overhead or the squeal of dive bombers in WW2 or the whistling of artillery in WW1.

Interestingly enough, in two of those cases, if you can hear them, it means you survived.

Artillery travels faster than the speed of sound, you hear the impacts before you hear the guns firing.

Similar sort of story with fighter jets. The loud end is behind the jet, you just hear a faint whining as it comes towards you, and an almighty woosh as it flies overhead.

In both those cases, if you can hear the source of death, it means you've escaped it, at least momentarily.

Surely, by definition, an approaching supersonic jet would be completely silent until shortly after it's passed the observer?

Yes, but usually jets aren't flying supersonic, especially for ground attack missions (if they're even capable).

It uses a lot of fuel, you can't maneuver easily, and you fly over your target too fast.

But how long does it take for the munitions to detonate over the ground to when the aircraft is the target? Particularly the modern-era "dumb" bombs used over syria.

I'd imagine that takes a bit of time to hit the ground, unless it's rocket propelled. So you will probably hear the jet just as it's about to detonate, no?

I think it depends on the trajectory. You could hear it if it's not flying directly towards you. But if I'm reasoning correctly, as long as it's following a straight line path then even if it's not going directly at you, you could only hear it after it passes the point that is closest to you. So it would probably had to be turning.

If the flying death machines are going to make that much noise, it should be fairly straight forward to use a few microphones and CNC machine guns to locate and shoot them down.

The microphone part: https://youtu.be/I6U-NkOQTe0

The military’s Phalanx CIWS is probably capable of shooting down small RC / autonomous aircraft?

Phalanx was successfully adapted to shoot down mortar shells in Iraq (C-RAM.) If it has a radar return Phalanx can probably kill it. This sort of weapon could be adapted with infrared or other anti-radiation sensors as well.

I think the answer to defending against offensive UAVs is defensive UAVs. Just today a story[1] appeared about a USAF Reaper shoot down of another UAV. Supposedly a first.

[1] https://www.popularmechanics.com/military/aviation/a23320374...

If it can stop missiles, I don't see why it can't stop a slow UAV.

I could see it being hard to aim at them accurately enough to hit, iirc the CIWS systems are radar aimed for missle defense. Depends what kind of UAV we're talking about an antipersonnel UAV could probably present a small enough radar signature.

Car Bombs + self driving cars is going to be an utter shit show.

You joke, but IS and JaN/JFS in Syria used the piloted versions of these: SVBIEDs aka suicide car bombs. Soviet APCs were abundant enough that they could even deploy armored variants.

They were extremely effective in combat, and called the "poor man's cruise missile" by analysts. Here's a video of a strike; note that they were often used in as the opening salvo of combined arms attacks. https://www.youtube.com/watch?v=AZBAU_TTZ-4

I agree with everything you mention except the I'm joking part.

Why do they need to be self driving cars? They’ve already proven effective using suicide drivers. They’ve also proven RC quadcopters are an effective delivery mechanism as well. Combine the RC aspect with the car, no fancy AI needed.

Of course there’s an aspect of their ideals that glorifies suicide so maybe they’ve already thought of this but just prefer to keep the suicide route.

Because a quadcopter is meant to be controlled remotely, while a traditional car isn't, so retrofitting it for remote control is likely time consuming and error prone (ever watch one of the mythbusters episodes where they rig up a car for RC?).

Self driving cars already have everything required to control the car built in. They just need to hijack the existing controls and hook it up to a RC system (or just tell it to drive to location using internal controls).

Don’t even need the bomb part, just command it to drive through a crowd on a national holiday / event. No driver to shoot dead.

I’m surprised someone hasn’t hooked up a remote control interface to an existing car.

Or hack the network they're all connected to and direct hoards of cars traveling at maximum speed into your target.

This kind of stuff was done in the Daemon/Freedom sci-fi books, which are entertaining and becoming concerningly less fiction with every passing year.

If they’re capable of acting on such a command, somebody has messed up.

That’s not to say it won’t happen (Uber) but it shouldn’t.

So you order the $30 mod chip that will definitely exist.

That's a good point. A stock standard autonomous vehicle ought not be able to. One wonders at how much modification would be necessary to override or bypass any software / hardware lockouts.

Not when it's flying at roughly head height for its entire trajectory in a civilian environment.

That’s a good point. Simple netting might be more effective in that scenario?

Edit to add: of course, it’s a never ending cycle of attack vector vs countermeasures I guess.

This is our future https://youtu.be/TlO2gcs1YvM

I’m not really sure how we counter this type of threat.

Wow, mind blown, your video link scares the shit out of me.

Yeah, wow; it's just an application of existing tech. very scary.

There's a couple sizable leaps in there in particular the combination of size, payload, and autonomy for the main mini murder drones. Particularly quads/drones of that small size just don't have the payload/flight time combination to really do that, they're extremely sensitive to weight and have really short flight times that are hard to extend because more weight requires more power in kind of a mini version of the rocket equation.

Well, if you deploy them close to the targets, they don't need much flight time to deliver the charges.

The idea is that instead of packing a truck with explosives (which create a large but very localised detonation) you just deploy a swarm of these and let them take out whatever group of people you want to target.

Unlike in Black Mirror's Metalhead this isn't a scenario that requires long-lived machines. You just need them to be readily available at scale and to use facial recognition (or whatever -- the US already targets drone strikes based on cell phone signals I hear) to determine their targets.

Or hey, imagine a white supremacist group (not like the US has any of those) using them to just kill people of color. Terrorism doesn't need anything approximating a 100% success rate to achieve its goal.

The video looks a bit in the future, but even a crude attempt at this kind of attack with what's available right now, would be terrifying.

Mosquito net is the first thing that comes to mind.

Then EM pulse weapons, lasers etc.

> Mosquito net

That's addressed in the film.

> EM pulse weapons

That's not actually a thing.

> lasers

Aside from being impractical, this is bringing a knife to a carpet bombing. The video is about the potential of autonomous drones being used against civilian targets. Unless you put all civilians in bunkers guarded with lazors, they are irrelevant (plus of course you still have created a nightmare scenario where humans need to live in bunkers to avoid random electronic micro-terrorists).

By caring about privacy and voting to prevent this.

I believe this is also the biggest intent of Black Mirror (where I believe the video is from).

For example Facebook and Google have so much data about you that a drone can track you down by using this data. Geo data, face data, habits, family and friends, you name it.

Let's hope the video will never happen.

The video's not actually from Black Mirror, it was an independent short film.

Black Mirror does have a very similar episode though: Hated in the Nation (S03E06).

EMP? Cyber Intrusion?

A much more practical/affordable application today is to use them as forward spotters and walk mortar fire into your enemy's position.

Amazingly cool. If he can find some lighter batteries with the same output it would make a huge improvement in stability I am assuming.

If it's heavier it's actually more stable because it's harder to move a heavier object. But maybe you mean that a heavier craft is slower to stabilise?

I think it's because the batteries are teetering above the centre of gravity?

you want center of gravity above the propeller, its the classic balancing a broom on the tip of your fingertip trick

So glad birds don't make that kinda noise.

As I understand, much of the difficulty controlling this comes from the gyroscopic effect. So wouldn't it be feasable to add a gearbox and a secondary rotary mass to cancel out the angular momentum? If you spin the secondary mass fast enough it also wouldn't add that much weight.

Adding weight to solve this problem doesn't sound like a good idea to me.

The video covers this...

This almost looks like the original App Engine logo.

Pardon my ignorance. Is this that different from a helicopter besides having walls? Does this mechanism serve a different purpose?

A helicopter uses a rear rotor to counteract the main rotor rotation. Here he is using thrust vectoring to perform that task.

Ah, thanks! Some helecoptors also just use a second main rotor spinning in the opposite direction instead right?

Would that not be sufficient? Sorry if it's covered in the video I've only seen the beginning so far.

Yes but he only had one fan. Just a rotating ring mounted somewhere would help too. But not as much fun I guess :-)

This is visible in the very last scene where it automatically tilts its blades to counteract the rotation induced by the accelerating propeller.

Nice! I wonder if it would be possible to add wings and build a VTOL plane with thrust-vectoring?

Fixed wing and let the thrust-vectoring pitch it over for forward-flight. Pitch up for SpaceX style landing.

It's been done with two electric motors. This is the E-flite X-VERT:


You can fly it like a regular plane, or hover like a quad. It doesn't do that well hovering in wind though.

But no trust vectoring :( still looks neat.

I was wondering the same. Or would external fins/wings defeat the purpose?

What's the purpose of this? Seems like "I want to build that" to me. So wings would defeat the primary purpose - but once the initial goal is reached, I personally would give it a shot (for the pure fun of doing it and see if that works, too).

Disclaimer: I believe building good wings is not a trivial task; even compared to this awesome rocket.

the future is loud

Very cool, some nice hands-on engineering going on here.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact