I'm not surprised you couldn't feel much ground effect with this glider, the Fox is more of a glider shaped sports plane than a real floater. No 1200g 2.3 meter plane is going to be much of a glider, my X-Dream 2m is 474g RTF and that can float on ground effect for the length of our runway - you need spoilers to land it
What if you add a shock from an rc car as a dampener on the rods. Ideally I would thing you would want the rods touching the ground constantly. Maybe add a really weak spring in line so that the rods would be more progressive in the authority adjustments. Maybe have a gear reduction on the arms so you can still have large surfaces but less aggressive inputs.
A problem could be the sonar bouncing unevenly off the water surface and it’s getting a little bit of mixed feedback from the mix I’ve heard it happens when water is choppy so idk just my thoughts love your videos and can’t wait for more projects
Does your radar buffer the altitude before it outputs the control? I would think (if I understand what you are doing now) if you just slowed the response of the output, the software would be catching peaks and troughs of the waves, and the delay between what is sees, and what it outputs, would be problematic. So some outputs would be greater than others, and some would be negative vs positive. I.E. You are compensating for problems you already passed. I would assume this action would lead to porpoiseing. If you forward angle the sensor a bit, and did some math to compensate for distance and time of the radar, you could avg the wave height, before the aircraft is over it. In this sense, the aircraft would fly as if it was over a level surface. It would be looking at waves ahead of it, thinking it over, and acting at the appropriate time. Then you could leave the sensitivity up, but create a digital travel limit of the servo outputs, and set a fail safe climb altitude if those limits are exceeded. If you really wanted to get fancy, you could do the fwd looking sensor, and a direct down sensor, then have your software crosscheck the data of what is trying to accomplish, vs the actual result. An algorithm of altitude results from the intended vs actual altitude could also be running to fine tune on the fly. The only problem being the occasional freak taller wave, but I lack the understanding of how a wave peak would physically disturb ground effect, or if the plane would mechanically avg that out.
I had this Idea after your last video, What if you where to mount Rotating Magnus effect wings with the added benefit that they might bounce out of the water like the Dambuster Bombs. wuld love to see you try this.
As usual, fantastic film. Re your vortex shedding imagery, I used to be a scientific photographer at the UK’s MoD Royal Aerospace Establishment, not really what you were trying to do, but in wind tunnels we used to use UV mini tufts on models to view localised flow patterns on the surface of the fuselage. Short lengths of thin string, with a UV fluorescent dye on it, at night with a good UV source and a camera with a high frame rate might give you some lovely images. Keep up the great stuff.
@@FaustoTheBoozehoundRAE establishment, UK. We're in the presence of aerodynamic royalty! Thanks for your input @1967AJB! These Americans have trouble with places outside their country...
It seems like the feedback loop needs to account for both height above the ground and pitch. A pitched down plane nearing the water needs to pull up, but a pitched up plane near the water will soon be high enough above the water without further pulling up. In other words, the altitude adjustment loop needs to know if it is too low and pitched to go lower, or too low and pitched to go higher, with a weaker or even inverse adjustment in the latter case. Likewise for being too high. This was very apparent with the stick based design, but I suspect would be much easier to implement with the flight controller and ultrasonic sensor design.
Similar to my theoretical solution, which was to have the servos controlled directly by the radar output’s RATE (i.e. how FAST the plane is climbing or falling) that way as soon as the plane starts to level out, the servos will be at zero. You could have the plane SLOWLY approach the ideal cruise height, then engage the subroutine that controls the servos. I obviously agree this would be much easier with flight controller than with “mechanical radar stick”.
You should measure the lag of the sonar sensor. Any lag in the PID "plant" moves the zeros and poles of the transfer function, which affects oscillations and stability
@@calloutman Yeah, control loops with significant delay can be a right pain to get fast response out of. It might be an idea to move to a predictor type controller (e.g. Kalman filter) rather than a PID type. However, first job is probably trying to figure out what the transfer function of the actuator->control surface->attitude/altitude->sensor combination is. I wouldn't even like to even hazard a guess what order of response it has.
Wingtip vortices are fascinating things. At Oshkosh one year they had a Sea Fury with wingtip smoke being followed by a Mustang down show center. At the end of the runway the Sea Fury would pull up hard and bank to the right...which turned the smoke to a swirling mess behind it that in a couple seconds all came together in a perfectly round smoke ring...that the Mustang would then fly through. I thought this was the neatest thing I'd ever seen.
@@NoNameAtAll2 Sure...why not? The air in the center of the smoke ring was obviously still and at speed with the high wing loading of the Mustang it blew through it with barely a bump. If it had been causing any issues I really doubt they'd have done it again and again for their entire demo time...but they did.
Great series! In my experiences with PID controllers, sometimes using x*abs(x) on the output can smooth out oscillations. In other words, square the output of the PID, but keep negative values as negatives. This results in exponentially lower sensitivity the closer you get to zero.
@rctestflight I've worked a bit with terrain following systems, and it might help you to aim your transducer more forward instead of straight down. That's generally the biggest flaw that most folks encounter with adverse pitch cycling is that the transducer is looking straight down, which causes the angle to point backwards when the aircraft is in a dive, and causes it to look forwards when it's climbing, and this creates a cyclic failure. However, if you mount the transducer pointing more forward as the aircraft climbs the transducer will not get the necessary feedback, and so it will slowly come back down in pitch which is much more controllable and creates a more expected less chaotic pitch control.
I'm flyin low checkin efficiency gains Testin wing shapes on RC planes Deep cords feel stronger ground effect buzz But thin ones might get better results This Fox glider got a long wingspan Should feel ground effect above a meter scan Installed sensors, art pilot too Try fly itself at the right altitude Tried followin the surface below But it kept bouncin, just wouldn't flow Tuned the PID but nothin worked right Still oscillatin out of sight Pontoons I added for water tests Flyin on the lake, flyin my best But the controller still had issues Keepin it steady, raisin fizzles Tried flaps next to make it rise higher Near the water, make the altitude wire But they couldn't stop the jumps The pitch and airspeed just too much pumps Wand concept next, canards it did flex Rotatin surfaces based on checks But responses had to be lessened Then it was too weak, flow now questionin Came up short on solutions galore But problems like these I want more Control systems is beyond me ya'll Hit me up if you can help it's not small Learned thick cords feel it most true But efficiency gain scales are tough to accrue Scale affects it heavy too it's clear Big planes don't feel it, only up close and near So ground effect tests will keep rollin on Till self stabilizin designs are fully gone Appreciate y'all watchin my vids Subscribe if you dig it, peace I bid
Speaking as a pilot, I can say that when I went from low-wing aircraft to flying a high-wing Cessna, I definitely made some rough landings at first, because of the reduced strength of the ground effect. When I was learning to fly, getting a good flare was one of the hardest things, exactly because of the ground effect. When the aircraft almost touches down, the efficiency of the wings increases dramatically, which means that even a slight movement of the elevators might pull the nose up too high ("ballooning"), exiting the ground effect again, losing a lot of airspeed and coming down hard.
Flight ontrol systems engineer here. Your videos are a delight, but with a bit of frustration around control design. I am sure that if you invested half as much time in beefing up your control systems engineering skills as you spend with 3d printing and whatnot, you'd be much less often puzzling over what's happening and you'd cut on the trial-and-error stage. The hard part is probably the aerodynamics though. Unless you make a wind tunnel, or you invest in CFD software, you'd have to characterize the plant's response in-situ, which is challenging (but since you seem to be able to always have something that at least flies, it might be sufficient). Anyway, hats off to your dedication.
Mechatronics engineer here! I'd love to see this project developed further. Im lead to believe that the oscillation problem may also be a result of the senor range. If the controller always has an accurate height measurement using a range of sensors (like IR and sonar fused with some form of Bayes filter), then the PID gains can be dynamically allocated. An airspeed measurement would also be useful to scale the gains, as the controll authority dramatically increases with airspeed. I like this approach because you don't need a perfect plant model, but you do need a half decent starting guess. A motion model is easy enough to find insitu for the bayes filter, though changing wind speeds/ direction might make for a new headache. So with just height and airspeed measurements, the controller should have enough information to fly without oscillation at any altitude. Pitch sensors may not be needed as sinkrate is arguably more important for stable flight at ANY airspeed. The only other useful information would be the terrain height AHEAD of the aircraft, similar to the terrain warning in real aircraft. this could just be a simple IR senor with a 5-10m range in this case. I think I still have the code to do this from a similar robotics assignment from Uni. Id be happy to dig it out for anyone interested. Loving the ground effect series! I look forwards to your next idea!
FluidX3d is an alright free for non commercial CFD program. Kind of a pain to use but it runs alright on consumer grade hardware if you have a decent discrete gpu.
EE with an emphasis on control theory and comms, its kind of difficult to just learn about control theory as there aren't many online resources. You'd need to take classes or buy some college level books and teach yourself, which isn't easy. I think if he really dumbed down the response rate he would have gotten better results.
@@Lost_Hwasal Either go through the hassle of teaching yourself, or better yet hire a graduate engineer to help with projects and teach you these kind of finicky concepts
On hydrofoil sailboats, the wand connects to the foil via two bell cranks and a push rod. The forward bell crank has a screw to adjust the ratio of angle change to pushrod movement (aka the gearing). In heavier seas we up the sensitivity and in flatter water we keep it as low as possible (because any flap movement is increased drag). It would be neat to see a mechanical tuning system for the canard’s instead of just changing their area. Keep up the good work!
Great video! As a control engineer, I suggest using MATLAB or similar tools to collect, analyze data and understand how the aircraft behaves. Testing altitude control ideas in the software before trying them out at the lake can help optimize the system. Also, be mindful of potential delay or lag from the sonar system, as ignoring them might cause issues with altitude control. Excited to see your progress!
While flying an IS-28B2 Lark sailplane (17 m / 55 feet 9 inch wingspan) on landing approach I was instructed to land long as the gliding operation was moving to the other end of the runway due to the wind direction changing. Fully retracting the air brakes while holding an altitude of about 10 feet and at a speed of 50 miles per hour the glider floated the length of the 2300 foot runway with very gradual speed loss, touching down at about 40 mph with an easy roll out to stop at the end of the runway. That, for me, was a very practical demonstration of ground effect. Thank you for all the analysis of the remarkable phenomenon of ground effect.
The text: My current project is building a plane which autonomously flies slowly and roughly low so I can run a waypoint mission and ride along with it on my bike, watching that footage of you flying your plane next to that boat got me hooked even more now. I got inspiration for this when watching your STOL multi-element wing plane flying a mission in that park. So yeah, shots of you chasing your builds are the coolest! 12:12: Shots like these are what makes your Videos special
Something to look at is the airspeed at which the highest efficiency occurs. A simple starting point is stall speed, as this is easily observable. (other measurements will require data collection) Typically ground effect helps lower stall speed, and lowering the speed across the performance curve. Effect feels like flying a lighter aircraft, makes if feel more floaty.
I'm a mechatronic engineer (control systems are part of that). A physical control loop, a digital one and an analog electronic circuit all follow the same rules and are modeled in the same way. In digital you have P, I and D. P reacts directly to the error, I reacts to the cumulative error over time and D reacts to the rate of change of the error. In a simple physical control loop you have a spring (which provides a proportional response to the error), a mass (which has inertia and doesn't want to change speeds), and a damper (which reacts to the rate of change). Your physical control loop has only a proportional response to the error, so it oscillates like any proportional-only control loop. If you want to add I and D to get a working control loop, you should isolate the probe from the control system with a rubber band, have a gear system so that a flywheel spins up in either direction as the error changes (to provide I) and a brake that rubs on the flywheel (to provide D). To providing tuning of I, you'll need a way to change the rotational moment of inertia of the flywheel, by either changing the mass or moving the mass closer to or further from the centre. To provide tuning of D, you'll just need a spring or rubber band that can have the tension increased to push harder on the flywheel. Probably the easiest way to achieve this would be by screwing weights into an electric scooter brake disc (about 75mm diameter) and mounting the disc to a drill's planetary gearbox (backlash in a 3D printed geartrain will not be kind to you). If you want to learn more about this, look into "spring mass damper" systems and state space modelling.
Regarding the sonar, you should really look at recorded data; measurement, control surface response etc. This way you can seen if your sensor is behaving, and you'll be able to better tune the PID, possibly with one of those tuning methods like ziegler nichols
I see one reason for the osillation in the linear actuation of the canards, the controller reaction should be weaker at low and stronger at high altitude. If you want to use a mecanical ground following controller, a possible solution would be to use a crankdisk and connect the controll surface at "top dead center" when the "ground following stick" is at its lowest point, with a second lever at the controllsurface right at 90°. But it would recuire some testing.
For the active control, my guess is, you would need to mix some gyro signal mixed i to the control, instead of just playing with the PID coefficients. Because what is most likely throwing the loop off is the temperamental variation of the control surfaces authority. Moreover if I'm not wrong, the system becomes at least 3'rd order (or even possibly 4'th order): Control surface displacement gets integrated into pitch rate, pitch rate gets integrated into attitude and attitude gets integrated into altitude (or first into vertical speed and that into the altitude). So to me nearly impossible to stabilize with just a simple PID (a single derivative is not enough phase margin recovery from 3 or 4 integral system). By using an attitude gyro signal giving you the attitude angle, you split the 3'th (when not relying onto that fast elevation changes) order response into two max 2 order loops, while the dynamics of the inner 2'nd order loop (surfaces -> attitude) is suppressed by the gyro signal (use the gain and D coefficients to stabilize), so then what remains is just the outer loop (attitude -> altitude) which is just one single integration, so very easy to stabilize. You first adjust the attitude control response (gyro -> servo PID response; better to be on the faster side, but still stable), then by varying the altitude sensor sensitivity adjust the final altitude stability.
As someone who formerly worked in a vape shop in the day of the classic mods. I definitely screwed several together and shoved up to 7 batteries in one stlet of copper tubes before it ignored several batteries and I had to throw it out of the garage door at the shop. It would probably create 1/2 the smoke of your machine only using 4 4680 batteries and a massive 8 coil atomizer with less than .1ohm guitar string colis. It was not in any way vapeble and required an air hose and a cut up 2 liter bottle to funnel air and a bottle of vg to prevent it from overheating immediately and going into flames. I installed the security cameras and my boss was there most of the time so he should have stopped us.
With the caveat that I only have seen the footage you posted here, I think you were closer to it working with the Canard design than you think. Instead removing so much of the surface, if you can trim it towards a more neutral attitude and change the gear ratio I think it would be much more effective. You are spot on that the response is very large, and by changing the gear ratio so the change in angle of the wand causes a smaller commensurate change in angle of the canard. In PID speak, think of it as changing the proportional term gain.
I think you should connect the giro to the sonar sensor, so you will be able to compensate for the angle change, and almost fully control the plane based on the giro, but adjust the altitude based on filtered and processed sonar data
The large tall T wing on the purpose built ground effect plane stabilizes it by being further away from the ground effect itself, and losing more lift as the plane climbs, which makes it pitch down slightly. Maybe the same effect would work with a negative stagger tandem wing, or with small fixed canards near the ground.
The ground effect on the lower aspect ratio ("deep chord") vehicle is more noticeable because it starts from a lower point in the efficiency scale, the glider is far more efficient with its high aspect ratio wings, so it seems less noticeable but is definitely there, glider pilots need to pay atention to it when landing. For the "terrain folowing canards" to work I would recomend using a thincker symmetrical airfoil (maybe an NACA 0018) and to limit its actuation to a incidence smaller then the stall angle, what is probably killing the idea is the stall od the surface and non linear response on the canards. For the flaps, it is true that lowering then increases lift, but it also increases pitching momment, so it would be necessary to compensate for that. If you want I'll be glad to help, I belive you come very close to getting it to work
Real neat wingtip vortex visualization! I live under the flight path for 12R at KMSP within a mile of the runway. We can often hear airliner vortices interacting with trees on the ground sometimes after a minute or more after the plane has passed. It's really damn cool to hear!
After watching this fly I have a suggestion. Go back to the canard; this was on the right track. The problem is not the canard, it's the surface following wand. The wand is so stiff it is acting almost like a 2 position switch. As soon as it contacts the water, it's all or nothing deflection of the canard. Rather than using a rigid pole, try using a thin piece of stainless steel fishing wire. The wire should be thin enough to deflect at least 90 Deg with minimal deflection of the canard when force is applied to the tip of the wand. You want the canard to deflect more and more as the wand goes deeper and deeper in the water. It should be at full deflection only when the wand is substantially submerged. To achieve this balance, you may also need to increase tension on the rubber bands you are using to return the canard to a neutral position. By making the wand much more flexible, you are in effect adding mechanical dampening to the system.
Use forward sweep to get better ground effect by thinner wing chord, this will also Improve the turning ability with less roll. Forward sweep has a similar flow characteristic as ground effect, therefore it will improve speed , lift and weight
This is the most visually pleasing video you or any rc video…er er er has ever made 👍 following your plane with the boat and reaching out and giving the wing a 👊 was tool cool. It would be very interesting to see more videos following an rc boat plane as it traverses through you experiments and local terrain. Keep the videos coming 👍👍👍
The ultra sonic sensor is picking up the wave height difference which is causing isolation. There should be a 2 second sensor delay once level on the water this will cause a less speratic data pickup. Then just adjust elevator sensitivity adjustments and should flow more evenly. The main issue is climate adjustment. To overcome this a larger heavier aircraft will be more efficient since even at a sensitive elevator rate the mass still requires more effort to elevate reducing the amount of oscillation.
wow, some very interesting insights. re: efficiency gains perportional etc, maybe a challenge, how many watts to transport 5 pounds one mile, or whatever. build a deep chord and shallow chord ground effect vehicle with same motor/prop? seems like adding some weight may make the ground effect the vehicle generated more apparent/easily measured? keep up the curiosity
Mechatronics engineer here! Click read more for some slightly nerdy control advice!
I'd love to see this project developed further. Im lead to believe that the oscillation problem may also be a result of the senor range. If the controller always has an accurate height measurement using a range of sensors (like IR and sonar fused with some form of Bayes filter), then the PID gains can be dynamically allocated. An airspeed measurement would also be useful to scale the gains, as the controll authority dramatically increases with airspeed. I like this approach because you don't need a perfect plant model, but you do need a half decent starting guess. A motion model is easy enough to find insitu for the bayes filter, though changing wind speeds/ direction might make for a new headache. So with just height and airspeed measurements, the controller should have enough information to fly without oscillation at any altitude. Pitch sensors may not be needed as sinkrate is arguably more important for stable flight at ANY airspeed. The only other useful information would be the terrain height AHEAD of the aircraft, similar to the terrain warning in real aircraft. this could just be a simple IR senor with a 5-10m range in this case.
I think I still have the code to do this from a similar robotics assignment from Uni. Id be happy to dig it out for anyone interested. Loving the ground effect series! I look forwards to your next idea! Thanks for listening to my TEDtalk :)
After seeing what the plane did with the stick/lever mechanism, I had this thought: apply the stick mechanism to a ‘normally’ shaped ground effect vehicle/plane but instead of thinking of control surfaces for an airplane consider it to be more like pressure control valves. If each side of the vehicle had a few sticks that linked to aileron like ‘vent’ flaps, as that side/area of the wing chord came up to a wave/high spot the stick/lever would open the vent in that area reducing the air pressure in that same area keeping the wing level. It shouldn’t take a very large vent flap to reduce the pressure. The same principle, of multiple small vents along the trailing edge of each side of the wing to reduce pressure, should be able to be applied to an electrical/radar/servo arm mechanism also. Great video as always.
I think maybe with the rolls on the water surface it might help to add dwell time to the surface indicated position as you would for temperature or water pressure. Those examples can take time to stabilizes as does the water level with a rolling surface. Up In Smoke has nothing on you and yes the license is on the back of the car.
My knowledge of the ground effect is that it's the wind that's normally kicked off of the wing reflecting back off of the ground then interacting with the wing underside and the flow of air, forming a standing wave. A high sweep flying wing should be able to catch some of that air bouncing back on the tips, and the stability can be modified with dihedral and wing twist nicely.
Oh my dear, i love your experiments, thoughts and your efforts. You should be a teacher, but not in school (is boring) but in a free class for all those children (and adults) who are willing to learn for their life and have fun! Thank you so much for sharing!! 🥰😘
One thing you could try is making the rods of more flexible material. Even more so you could have them of greatest flexibility at the point and gradually decrease flexibility down the length if the rod. This could be done in part by varying the width of a given material
In full size aircrafts that are aerodynamically stable one always talk about ”pitch changes speed and throttle changes altitude”, atleast throttle changes vertical speed. Have you tought about connecting the hight measurement to throttle more than to pitch? Might work or might completely fail, might be something to try out as well😊 and as always, great video! Much fun to watch and learn from
you cant tie the altitude error directly to the elevator. pitch controls vertical speed, which in turn controls altitude. so the altitude error must be used to generate a vertical speed command, which in turn generates an elevator command. i spent a fair bit of time myself dinking around with altitude control loops for uavs, getting a balance of stability and responsiveness is for sure challenging
Just to throw a curve ball, assuming fixed throttle, pitch also controls airspeed. As airspeed drops, so does climb rate. Among other things, this makes the physical response time vary based on the control output.
The mechanical control system on a foiling Moth (similar to the sailboat you showed), relies on what we call a "fast point", where you have maximum control input (rate) at the desired height, with diminishing sensitivity (rate of control surface change for a given unit of wand rotation) the further you get from the desired height. This is key to preventing the porpoising and oscillation you're seeing.
The issue is the input from your "sensors" and the output from your plane. Because the input from your "sensors" gets more intense the closer to the ground, and they seem to get way more intense before the plane has time to correct. So being able to adjust the sensitivity, or a delay or something along those lines would be the path forward.
I agree with this completely. Basic damping of the control surfaces should aid greatly. Maybe even something as simple as a friction damper attached to the control horns/surfaces. Maybe a rate slowing of the reaction speed in your duino? Or a steepening response rate based on altitude? Slow response with small altitude deviations, faster responses the further it deviates from its correct altitude.
I don't normally pay much attention to sponsors, but that battery looks tempting. I'm sure others can be charged from a car's 12V socket, but it's the first one I remember seeing a person specifically mention having that feature.
You may wish to notice that PID works well only when there are no other hidden integrators in the system. Integrating twice produces a nice oscillator. Hidden integrators may come in form of inertia of any kind. Things to do about it is either turning off the I component of a regulator, or shifting its frequency window a full order of magnitude away from the existing hidden integrator.
Daniels channel is the most interesting RUclips channel there is. I even watch his ads. LOL! Seriously, this channel always is interesting. Even when he was a kid.
In the sonar code try adding a smoothing code RNGFND1_WSP_MEDF: Moving Median Filter = 0-255 unless you’re using a custom code in a ardupilot. I think some of the oscillation is the waves Sets the window size for the real-time median filter. When MEDF is greater than 0 the median filter is active
11:39 Your canard issues might stem from the sprung mechanism's reaction force, causing a mix of aerodynamic and mechanical forces in your feedback loop. Test this by balancing the plane on your fingertips and lowering it onto a surface with the mechanism activated; if it pitches up, there's a mechanical force in your feedback loop.
I LOVE your videos!! Keep them coming! A Close Coupled Canard may be better for ground effect. IGE pressure under the wing increases which causes an upwash at the leading edge which will increase the lift of the close coupled canard, causing a pitch up away from the "ground". Also a swept wing pitches down less IGE than a straight wing because of less center of pressure shift. In fact 25⁰ to 45⁰ swept wings with a tall OGE T tail will also work!
Honestly I think using the skimmers on the flaps was the mistake, I think had you used that directly on the elevator with a switch to disable them for “normal flight” would’ve solved your problem. This is such an interesting “problem.” Awesome video, love your channel
You really need to take some classes in control engineering. Cutting down the canards didn't only reduce the p-gain, it also lowered the saturation limit. A better solution would have been to lower the gear ratio.
I would assume that the wands touching the ground actually reduce the lift required by taking up some of the weight. Aircraft is extremely light so imagine they would have an impact..
The wand/canard looked promising at first, don't know if anyone has already mentioned but it could be worth changng the feedback loop by adjusting the throw ratio rather than the flap size. You could do that by changing your gear ratio, but probably easier to switch to a coupled control arm set with different hole positions.
To apply the flaps to the controll loop is the right step. You create lift without loosing energy for pitch motion of the entire aircraft. The flaps need to be much bigger though, at least by factor 5! The canards are different, because of their leverage to the centre of gravity and it's coupled feadback loop with two different sets of time constants for the same prameters. The surest way to figure out a stable parameter loop is to keep the number of degres for freadom as los as posible. For instance, for ground effect mode - freeze the pich attitude and work only with flaps and rudder. The pich should be controlled by a gyro, but still keept fixed, when flying in the ground effect mode.
My wet grinder, besides using a similar non-fluid oil, has pads of felt to retain the lube along the shaft. Two of them form long tubes/rings around it. It withstands being submerged. It has a third trick too, gears on the ends with another parallel shaft to pick it up. Whole thing is in a bolt down cover (commercial manufacturer). It is legit waterproof. Something in there might help.
I really enjoy your experiments! On the topic of water surface skimming vehicles, I'd like to challenge your thinking here a little: perhaps the optimal foils for this application are hydrofoils rather than airfoils. Smaller foils, and the control loop being more naturally stable n simple to control... I suppose that makes it a watercraft not an aircraft, but those are just words. Keep being excellent!
Couch man here: The problem I see with the stick changingthe flaps is the stick: -I would try a longer and more nimble stick. My guess is, that would definitely work. Currently the stick is way too long. Just scale the length and size to the water craft, you showed. (Compared to that you are putting a tree into the water to adjust steering)
Great video, if you change the size of the gears at the front of the plane so that they are different size then you can reduce the amount that this will cause the plane oscilate as the distance moved by the lever could be reduced by having a larger gear control the pitch against a smaller one attached to the leader as the smaller gear will have to travel further to move the larger gear. A 2:1 gear ratio would halve the movement
Your deep cord tests are more effective because it takes more time to bleed the pressure off from under the aircraft. Think of it like a tarp gathering wind and billowing out. To stop the billowing you can put cuts in the tarp and the tarp will billow less with each cut. For a lesser cord there is a shorter path to release pressure despite more generated ground effect BUT the accumulated total air pressure under the wing surface would be less because of the amount that can escape relatively easily.
Double-check the barometer and right around where the barometer hole is. This behavior the fox glider had where it just moves up and down a couple feet randomly sounds like a problem I had where a little tiny piece of tape was being blown into and out of the little bay where the FC was, and it was changing something about how the pressure would change in there, and making the quadcopter bounce up and down a foot at random times.
I think that for a proper feedback loop you will need wands with progressive flex that match the weight of the aircraft. Think of it like a fishing pole. At the tip, the wand would deflect more allowing for less input. As the aircraft gets closer to the ground the wand provides more input because the wand gets progressively stiffer.
Changing the wands so that they are curved, may help with the oscillation. As the plane gets closer to the ground, the curve would cause the flaps to be turned less for the same unit of distance closer.
Dude! Another awesome video! A++ all the way around. Content, editing, narrative. You’re nailing bro! Thanks for inspiring my inner engineer to play. RC, micro controllers, 3D design & printing. PID. Checking all my favorite boxes. Love it! 👍
I can tell you as a Flexwing pilot our wing aoa is high on the take off roll, as soon as it lifts the wing aoa has to be reduced fast to increase air speed, or risk a stall. Your aircraft is oscillating because it is stalling.. the wand is producing too high an aoa regardless of control surface area. .Its a tricky one to solve dumb mechanically but you need to find a way to return to level flight and get your airspeed back at the top of the bounce. I would suggest use your sonar to activate a servo to bring the aircraft to level flights at anything ABOVE say 18 inch?. Wand is up control.. Sonar is level control. I'm rambling now 😅
The larger the model the more you can "feel" the ground effect, i have a 5 metre wingspan rc canadair waterbomber and you have to wait until it slowed enough for it to descend through ground efect and touch down nicely, if you force it it will get pio very easily, it has a deep cord flat bottom wing
This is fascinating, I believe that the efficiency gains of ground effect can influence the stability quite a bit which could be the cause of the oscillation, just my two pence.
Just a theory, for tuning the control loop you may make headway by considering both power and elevator commands. They say, "speed is life, altitude is life insurance" and while the saying goes for surviving (engine) failures, the same is true when landing. Landing aircraft must stabilize their approach, so that the aircraft's energy at the touchdown point is in a specific range. If they're going too fast, they can pull up and exchange the speed for altitude, and this exchange is done fairly quickly; a few knots for a few meeters is a balance between airspeed and altimeter readings. Their sum will calculate the aircraft's net kinetic potential energy, which you can add to by increasing power. You might have luck in your control loops by using altitude as input for the aileron loop, and altitude plus airspeed for the throttle. This would allow the plane to react to the situation where it starts too high, pitches down while adding power, and ends up low with too much speed. I would personally just use simple gains for both, but logic would say you'd need a gain for mass (speed and altitude multiplier), drag (speed multiplier) and should tune the rest with PID.
I fly a powered glider similar to his plane. I was surprised when he said he didn’t feel(?) the ground effect of the plane before he modified it. Well you can see the ground effect. The plane just won’t stop it keeps going and going flying along a few feet off the ground with out power. That can be a problem for new pilots trying to land. Here’s it comes great glide slop, almost down but it just floats right on by you slow and happy.
The Kasperwing U/L utilized a single surface airfoil between it's floats, it was very effective in shortening water takeoffs. Ground Effect was also noticeable with the wing between the floats, it might be worth a try on your R/C float plane.
The moment u started to throw the camera I scrolled down and tried to sub, yet I've been a sub for ages and u deserve more dude. Creatively your content is unique. Honestly lad keep up the good work, ya made us proud son.
This is where machine learning could really be interesting - with two ultrasound sensors (one facing down, one slightly forward) and a way to detect hitting water (capacitatively, or via an accelerometer), you could feasibly train a model to follow the water as close as possible, while predicting its future path (and increasing altitude for waves). It would also naturally learn to filter out noise from the sensors and water surface fluctuations without losing sensitivity to a sliding window average (like a regular control loop would).
Love your videos! I believe that the surface wand touch point with the water, has to be far ahead of the CG in order to have stability. This is evident in all of the foil boats that use this control method. When it's behind or close to the CG, the pitching motion may be causing a reversal in wand displacement angle. Put another way, think of the wand controlling aircraft pitch, not altitude. Suttle difference maybe, but it matters for stability.
I would say add a reduction gear between the wands and the control surface. Print some different ratios and continue testing. It will depend on the wind condition and airspeed which configuration will work best.
Heres the glider from this video: www.fmshobby.com/products/fms-2300mm-fox-v2-pnp/?ref=RCTestFlight
Use code $10RCTF or RCTestFlight to get $10 off
I'm not surprised you couldn't feel much ground effect with this glider, the Fox is more of a glider shaped sports plane than a real floater. No 1200g 2.3 meter plane is going to be much of a glider, my X-Dream 2m is 474g RTF and that can float on ground effect for the length of our runway - you need spoilers to land it
What if you add a shock from an rc car as a dampener on the rods.
Ideally I would thing you would want the rods touching the ground constantly. Maybe add a really weak spring in line so that the rods would be more progressive in the authority adjustments.
Maybe have a gear reduction on the arms so you can still have large surfaces but less aggressive inputs.
A problem could be the sonar bouncing unevenly off the water surface and it’s getting a little bit of mixed feedback from the mix I’ve heard it happens when water is choppy so idk just my thoughts love your videos and can’t wait for more projects
Does your radar buffer the altitude before it outputs the control? I would think (if I understand what you are doing now) if you just slowed the response of the output, the software would be catching peaks and troughs of the waves, and the delay between what is sees, and what it outputs, would be problematic. So some outputs would be greater than others, and some would be negative vs positive. I.E. You are compensating for problems you already passed.
I would assume this action would lead to porpoiseing. If you forward angle the sensor a bit, and did some math to compensate for distance and time of the radar, you could avg the wave height, before the aircraft is over it. In this sense, the aircraft would fly as if it was over a level surface. It would be looking at waves ahead of it, thinking it over, and acting at the appropriate time. Then you could leave the sensitivity up, but create a digital travel limit of the servo outputs, and set a fail safe climb altitude if those limits are exceeded. If you really wanted to get fancy, you could do the fwd looking sensor, and a direct down sensor, then have your software crosscheck the data of what is trying to accomplish, vs the actual result. An algorithm of altitude results from the intended vs actual altitude could also be running to fine tune on the fly.
The only problem being the occasional freak taller wave, but I lack the understanding of how a wave peak would physically disturb ground effect, or if the plane would mechanically avg that out.
I had this Idea after your last video, What if you where to mount Rotating Magnus effect wings with the added benefit that they might bounce out of the water like the Dambuster Bombs. wuld love to see you try this.
We all agree that you should continue the Solar Plane series!
YES
yee
Yes
yup
Yes fly it from one side of America to the other (within reason).
Those droneless camera birdview shots of the foam plane were top tier lmao
As usual, fantastic film. Re your vortex shedding imagery, I used to be a scientific photographer at the UK’s MoD Royal Aerospace Establishment, not really what you were trying to do, but in wind tunnels we used to use UV mini tufts on models to view localised flow patterns on the surface of the fuselage. Short lengths of thin string, with a UV fluorescent dye on it, at night with a good UV source and a camera with a high frame rate might give you some lovely images.
Keep up the great stuff.
That's a great idea.
Farnborough?
@@FaustoTheBoozehound
Correct.
@@FaustoTheBoozehoundRAE establishment, UK. We're in the presence of aerodynamic royalty!
Thanks for your input @1967AJB! These Americans have trouble with places outside their country...
It seems like the feedback loop needs to account for both height above the ground and pitch. A pitched down plane nearing the water needs to pull up, but a pitched up plane near the water will soon be high enough above the water without further pulling up. In other words, the altitude adjustment loop needs to know if it is too low and pitched to go lower, or too low and pitched to go higher, with a weaker or even inverse adjustment in the latter case. Likewise for being too high. This was very apparent with the stick based design, but I suspect would be much easier to implement with the flight controller and ultrasonic sensor design.
That's a good point. Maybe a plane that's designed for canards (like a Long EZ) would be better, as it would have a longer nose.
Similar to my theoretical solution, which was to have the servos controlled directly by the radar output’s RATE (i.e. how FAST the plane is climbing or falling) that way as soon as the plane starts to level out, the servos will be at zero. You could have the plane SLOWLY approach the ideal cruise height, then engage the subroutine that controls the servos. I obviously agree this would be much easier with flight controller than with “mechanical radar stick”.
Nice. Could be done with a forward and rear wand. I also think a damper on the wand/canard system could help.
@@ZenZooZoo That's already what's happening with the D part of the PDI controller, no?
@@snower13 that's a clever solution, would be interested to see if this sort of mechanical method could work
You should measure the lag of the sonar sensor. Any lag in the PID "plant" moves the zeros and poles of the transfer function, which affects oscillations and stability
The lag in the sonar sensor (miliseconds) will be far less than the response time of the aircraft (seconds)
@@calloutman Yeah, control loops with significant delay can be a right pain to get fast response out of. It might be an idea to move to a predictor type controller (e.g. Kalman filter) rather than a PID type. However, first job is probably trying to figure out what the transfer function of the actuator->control surface->attitude/altitude->sensor combination is. I wouldn't even like to even hazard a guess what order of response it has.
Wingtip vortices are fascinating things. At Oshkosh one year they had a Sea Fury with wingtip smoke being followed by a Mustang down show center. At the end of the runway the Sea Fury would pull up hard and bank to the right...which turned the smoke to a swirling mess behind it that in a couple seconds all came together in a perfectly round smoke ring...that the Mustang would then fly through. I thought this was the neatest thing I'd ever seen.
was that... safe?
@@NoNameAtAll2 Sure...why not? The air in the center of the smoke ring was obviously still and at speed with the high wing loading of the Mustang it blew through it with barely a bump. If it had been causing any issues I really doubt they'd have done it again and again for their entire demo time...but they did.
Great series!
In my experiences with PID controllers, sometimes using x*abs(x) on the output can smooth out oscillations. In other words, square the output of the PID, but keep negative values as negatives. This results in exponentially lower sensitivity the closer you get to zero.
@rctestflight
I've worked a bit with terrain following systems, and it might help you to aim your transducer more forward instead of straight down.
That's generally the biggest flaw that most folks encounter with adverse pitch cycling is that the transducer is looking straight down, which causes the angle to point backwards when the aircraft is in a dive, and causes it to look forwards when it's climbing, and this creates a cyclic failure.
However, if you mount the transducer pointing more forward as the aircraft climbs the transducer will not get the necessary feedback, and so it will slowly come back down in pitch which is much more controllable and creates a more expected less chaotic pitch control.
The most practical bit of advice!
I'm flyin low checkin efficiency gains
Testin wing shapes on RC planes
Deep cords feel stronger ground effect buzz
But thin ones might get better results
This Fox glider got a long wingspan
Should feel ground effect above a meter scan
Installed sensors, art pilot too
Try fly itself at the right altitude
Tried followin the surface below
But it kept bouncin, just wouldn't flow
Tuned the PID but nothin worked right
Still oscillatin out of sight
Pontoons I added for water tests
Flyin on the lake, flyin my best
But the controller still had issues
Keepin it steady, raisin fizzles
Tried flaps next to make it rise higher
Near the water, make the altitude wire
But they couldn't stop the jumps
The pitch and airspeed just too much pumps
Wand concept next, canards it did flex
Rotatin surfaces based on checks
But responses had to be lessened
Then it was too weak, flow now questionin
Came up short on solutions galore
But problems like these I want more
Control systems is beyond me ya'll
Hit me up if you can help it's not small
Learned thick cords feel it most true
But efficiency gain scales are tough to accrue
Scale affects it heavy too it's clear
Big planes don't feel it, only up close and near
So ground effect tests will keep rollin on
Till self stabilizin designs are fully gone
Appreciate y'all watchin my vids
Subscribe if you dig it, peace I bid
🔥🔥🔥
I wonder 🤔 how this would sound with a sick beat and Eminem performing 😎❤
These lyrics low-key fire tho...
Speaking as a pilot, I can say that when I went from low-wing aircraft to flying a high-wing Cessna, I definitely made some rough landings at first, because of the reduced strength of the ground effect. When I was learning to fly, getting a good flare was one of the hardest things, exactly because of the ground effect. When the aircraft almost touches down, the efficiency of the wings increases dramatically, which means that even a slight movement of the elevators might pull the nose up too high ("ballooning"), exiting the ground effect again, losing a lot of airspeed and coming down hard.
Flight ontrol systems engineer here. Your videos are a delight, but with a bit of frustration around control design. I am sure that if you invested half as much time in beefing up your control systems engineering skills as you spend with 3d printing and whatnot, you'd be much less often puzzling over what's happening and you'd cut on the trial-and-error stage.
The hard part is probably the aerodynamics though. Unless you make a wind tunnel, or you invest in CFD software, you'd have to characterize the plant's response in-situ, which is challenging (but since you seem to be able to always have something that at least flies, it might be sufficient).
Anyway, hats off to your dedication.
Mechatronics engineer here!
I'd love to see this project developed further. Im lead to believe that the oscillation problem may also be a result of the senor range. If the controller always has an accurate height measurement using a range of sensors (like IR and sonar fused with some form of Bayes filter), then the PID gains can be dynamically allocated. An airspeed measurement would also be useful to scale the gains, as the controll authority dramatically increases with airspeed.
I like this approach because you don't need a perfect plant model, but you do need a half decent starting guess. A motion model is easy enough to find insitu for the bayes filter, though changing wind speeds/ direction might make for a new headache.
So with just height and airspeed measurements, the controller should have enough information to fly without oscillation at any altitude. Pitch sensors may not be needed as sinkrate is arguably more important for stable flight at ANY airspeed. The only other useful information would be the terrain height AHEAD of the aircraft, similar to the terrain warning in real aircraft. this could just be a simple IR senor with a 5-10m range in this case.
I think I still have the code to do this from a similar robotics assignment from Uni. Id be happy to dig it out for anyone interested.
Loving the ground effect series! I look forwards to your next idea!
FluidX3d is an alright free for non commercial CFD program. Kind of a pain to use but it runs alright on consumer grade hardware if you have a decent discrete gpu.
EE with an emphasis on control theory and comms, its kind of difficult to just learn about control theory as there aren't many online resources. You'd need to take classes or buy some college level books and teach yourself, which isn't easy. I think if he really dumbed down the response rate he would have gotten better results.
@@Lost_Hwasal Either go through the hassle of teaching yourself, or better yet hire a graduate engineer to help with projects and teach you these kind of finicky concepts
OMG that vape car hotboxed scene was hilarious! Actually made me laugh so hard. Thanks for that.
Pull up to the vape store and tell them that my vape pipe has gone on thermal runaway.
Watched the whole ad just because of that part
JJKHNJJHNJJ?
The Rate in witch you pump out such high quality Videos is just insane!
That’s true.
which*
witches turn people into a newt :)
@@NoNameAtAll2 ops thanks for the correction.
On hydrofoil sailboats, the wand connects to the foil via two bell cranks and a push rod. The forward bell crank has a screw to adjust the ratio of angle change to pushrod movement (aka the gearing). In heavier seas we up the sensitivity and in flatter water we keep it as low as possible (because any flap movement is increased drag). It would be neat to see a mechanical tuning system for the canard’s instead of just changing their area. Keep up the good work!
Great video! As a control engineer, I suggest using MATLAB or similar tools to collect, analyze data and understand how the aircraft behaves. Testing altitude control ideas in the software before trying them out at the lake can help optimize the system. Also, be mindful of potential delay or lag from the sonar system, as ignoring them might cause issues with altitude control. Excited to see your progress!
While flying an IS-28B2 Lark sailplane (17 m / 55 feet 9 inch wingspan) on landing approach I was instructed to land long as the gliding operation was moving to the other end of the runway due to the wind direction changing. Fully retracting the air brakes while holding an altitude of about 10 feet and at a speed of 50 miles per hour the glider floated the length of the 2300 foot runway with very gradual speed loss, touching down at about 40 mph with an easy roll out to stop at the end of the runway. That, for me, was a very practical demonstration of ground effect. Thank you for all the analysis of the remarkable phenomenon of ground effect.
The text: My current project is building a plane which autonomously flies slowly and roughly low so I can run a waypoint mission and ride along with it on my bike, watching that footage of you flying your plane next to that boat got me hooked even more now. I got inspiration for this when watching your STOL multi-element wing plane flying a mission in that park. So yeah, shots of you chasing your builds are the coolest!
12:12: Shots like these are what makes your Videos special
Something to look at is the airspeed at which the highest efficiency occurs.
A simple starting point is stall speed, as this is easily observable. (other measurements will require data collection)
Typically ground effect helps lower stall speed, and lowering the speed across the performance curve. Effect feels like flying a lighter aircraft, makes if feel more floaty.
I'm a mechatronic engineer (control systems are part of that). A physical control loop, a digital one and an analog electronic circuit all follow the same rules and are modeled in the same way.
In digital you have P, I and D. P reacts directly to the error, I reacts to the cumulative error over time and D reacts to the rate of change of the error.
In a simple physical control loop you have a spring (which provides a proportional response to the error), a mass (which has inertia and doesn't want to change speeds), and a damper (which reacts to the rate of change).
Your physical control loop has only a proportional response to the error, so it oscillates like any proportional-only control loop. If you want to add I and D to get a working control loop, you should isolate the probe from the control system with a rubber band, have a gear system so that a flywheel spins up in either direction as the error changes (to provide I) and a brake that rubs on the flywheel (to provide D). To providing tuning of I, you'll need a way to change the rotational moment of inertia of the flywheel, by either changing the mass or moving the mass closer to or further from the centre. To provide tuning of D, you'll just need a spring or rubber band that can have the tension increased to push harder on the flywheel. Probably the easiest way to achieve this would be by screwing weights into an electric scooter brake disc (about 75mm diameter) and mounting the disc to a drill's planetary gearbox (backlash in a 3D printed geartrain will not be kind to you).
If you want to learn more about this, look into "spring mass damper" systems and state space modelling.
Regarding the sonar, you should really look at recorded data; measurement, control surface response etc. This way you can seen if your sensor is behaving, and you'll be able to better tune the PID, possibly with one of those tuning methods like ziegler nichols
I see one reason for the osillation in the linear actuation of the canards, the controller reaction should be weaker at low and stronger at high altitude. If you want to use a mecanical ground following controller, a possible solution would be to use a crankdisk and connect the controll surface at "top dead center" when the "ground following stick" is at its lowest point, with a second lever at the controllsurface right at 90°. But it would recuire some testing.
For the active control, my guess is, you would need to mix some gyro signal mixed i to the control, instead of just playing with the PID coefficients. Because what is most likely throwing the loop off is the temperamental variation of the control surfaces authority. Moreover if I'm not wrong, the system becomes at least 3'rd order (or even possibly 4'th order): Control surface displacement gets integrated into pitch rate, pitch rate gets integrated into attitude and attitude gets integrated into altitude (or first into vertical speed and that into the altitude). So to me nearly impossible to stabilize with just a simple PID (a single derivative is not enough phase margin recovery from 3 or 4 integral system).
By using an attitude gyro signal giving you the attitude angle, you split the 3'th (when not relying onto that fast elevation changes) order response into two max 2 order loops, while the dynamics of the inner 2'nd order loop (surfaces -> attitude) is suppressed by the gyro signal (use the gain and D coefficients to stabilize), so then what remains is just the outer loop (attitude -> altitude) which is just one single integration, so very easy to stabilize.
You first adjust the attitude control response (gyro -> servo PID response; better to be on the faster side, but still stable), then by varying the altitude sensor sensitivity adjust the final altitude stability.
As someone who formerly worked in a vape shop in the day of the classic mods. I definitely screwed several together and shoved up to 7 batteries in one stlet of copper tubes before it ignored several batteries and I had to throw it out of the garage door at the shop. It would probably create 1/2 the smoke of your machine only using 4 4680 batteries and a massive 8 coil atomizer with less than .1ohm guitar string colis. It was not in any way vapeble and required an air hose and a cut up 2 liter bottle to funnel air and a bottle of vg to prevent it from overheating immediately and going into flames. I installed the security cameras and my boss was there most of the time so he should have stopped us.
With the caveat that I only have seen the footage you posted here, I think you were closer to it working with the Canard design than you think. Instead removing so much of the surface, if you can trim it towards a more neutral attitude and change the gear ratio I think it would be much more effective. You are spot on that the response is very large, and by changing the gear ratio so the change in angle of the wand causes a smaller commensurate change in angle of the canard. In PID speak, think of it as changing the proportional term gain.
I think you should connect the giro to the sonar sensor, so you will be able to compensate for the angle change, and almost fully control the plane based on the giro, but adjust the altitude based on filtered and processed sonar data
The large tall T wing on the purpose built ground effect plane stabilizes it by being further away from the ground effect itself, and losing more lift as the plane climbs, which makes it pitch down slightly. Maybe the same effect would work with a negative stagger tandem wing, or with small fixed canards near the ground.
The ground effect on the lower aspect ratio ("deep chord") vehicle is more noticeable because it starts from a lower point in the efficiency scale, the glider is far more efficient with its high aspect ratio wings, so it seems less noticeable but is definitely there, glider pilots need to pay atention to it when landing. For the "terrain folowing canards" to work I would recomend using a thincker symmetrical airfoil (maybe an NACA 0018) and to limit its actuation to a incidence smaller then the stall angle, what is probably killing the idea is the stall od the surface and non linear response on the canards. For the flaps, it is true that lowering then increases lift, but it also increases pitching momment, so it would be necessary to compensate for that. If you want I'll be glad to help, I belive you come very close to getting it to work
Real neat wingtip vortex visualization! I live under the flight path for 12R at KMSP within a mile of the runway. We can often hear airliner vortices interacting with trees on the ground sometimes after a minute or more after the plane has passed. It's really damn cool to hear!
After watching this fly I have a suggestion. Go back to the canard; this was on the right track. The problem is not the canard, it's the surface following wand. The wand is so stiff it is acting almost like a 2 position switch. As soon as it contacts the water, it's all or nothing deflection of the canard. Rather than using a rigid pole, try using a thin piece of stainless steel fishing wire. The wire should be thin enough to deflect at least 90 Deg with minimal deflection of the canard when force is applied to the tip of the wand. You want the canard to deflect more and more as the wand goes deeper and deeper in the water. It should be at full deflection only when the wand is substantially submerged. To achieve this balance, you may also need to increase tension on the rubber bands you are using to return the canard to a neutral position. By making the wand much more flexible, you are in effect adding mechanical dampening to the system.
Use forward sweep to get better ground effect by thinner wing chord, this will also Improve the turning ability with less roll. Forward sweep has a similar flow characteristic as ground effect, therefore it will improve speed , lift and weight
This is the most visually pleasing video you or any rc video…er er er has ever made 👍 following your plane with the boat and reaching out and giving the wing a 👊 was tool cool.
It would be very interesting to see more videos following an rc boat plane as it traverses through you experiments and local terrain.
Keep the videos coming 👍👍👍
I haven't felt motion sickness from a video till now. Congrats
The ultra sonic sensor is picking up the wave height difference which is causing isolation. There should be a 2 second sensor delay once level on the water this will cause a less speratic data pickup. Then just adjust elevator sensitivity adjustments and should flow more evenly. The main issue is climate adjustment. To overcome this a larger heavier aircraft will be more efficient since even at a sensitive elevator rate the mass still requires more effort to elevate reducing the amount of oscillation.
wow, some very interesting insights. re: efficiency gains perportional etc, maybe a challenge, how many watts to transport 5 pounds one mile, or whatever. build a deep chord and shallow chord ground effect vehicle with same motor/prop? seems like adding some weight may make the ground effect the vehicle generated more apparent/easily measured? keep up the curiosity
Wow flying from your boat looked so much fun with that floater :D
The polite abruptness of "thanks for watching, bye" - is art. Only reason I'm not mad when these videos end.
Mechatronics engineer here! Click read more for some slightly nerdy control advice!
I'd love to see this project developed further. Im lead to believe that the oscillation problem may also be a result of the senor range. If the controller always has an accurate height measurement using a range of sensors (like IR and sonar fused with some form of Bayes filter), then the PID gains can be dynamically allocated. An airspeed measurement would also be useful to scale the gains, as the controll authority dramatically increases with airspeed.
I like this approach because you don't need a perfect plant model, but you do need a half decent starting guess. A motion model is easy enough to find insitu for the bayes filter, though changing wind speeds/ direction might make for a new headache.
So with just height and airspeed measurements, the controller should have enough information to fly without oscillation at any altitude. Pitch sensors may not be needed as sinkrate is arguably more important for stable flight at ANY airspeed. The only other useful information would be the terrain height AHEAD of the aircraft, similar to the terrain warning in real aircraft. this could just be a simple IR senor with a 5-10m range in this case.
I think I still have the code to do this from a similar robotics assignment from Uni. Id be happy to dig it out for anyone interested.
Loving the ground effect series! I look forwards to your next idea!
Thanks for listening to my TEDtalk :)
Always love your engineering mate! Out of all the RC channels, you use the same principles they use to create real aircrafts.
I love these technical videos of yours. I cannot tell you how refreshing it is to have a real long form video to watch rather than some clickbait BS.
I love how you add an autopilot to anything that moves. Very skilled.
After seeing what the plane did with the stick/lever mechanism, I had this thought: apply the stick mechanism to a ‘normally’ shaped ground effect vehicle/plane but instead of thinking of control surfaces for an airplane consider it to be more like pressure control valves.
If each side of the vehicle had a few sticks that linked to aileron like ‘vent’ flaps, as that side/area of the wing chord came up to a wave/high spot the stick/lever would open the vent in that area reducing the air pressure in that same area keeping the wing level.
It shouldn’t take a very large vent flap to reduce the pressure. The same principle, of multiple small vents along the trailing edge of each side of the wing to reduce pressure, should be able to be applied to an electrical/radar/servo arm mechanism also.
Great video as always.
That would reduce the efficiency of the wing but what we want by flying low is efficiency.
I think maybe with the rolls on the water surface it might help to add dwell time to the surface indicated position as you would for temperature or water pressure. Those examples can take time to stabilizes as does the water level with a rolling surface. Up In Smoke has nothing on you and yes the license is on the back of the car.
I think the camera shots in this video proved that aviation is so beautiful whether it be RC or full scale
My knowledge of the ground effect is that it's the wind that's normally kicked off of the wing reflecting back off of the ground then interacting with the wing underside and the flow of air, forming a standing wave. A high sweep flying wing should be able to catch some of that air bouncing back on the tips, and the stability can be modified with dihedral and wing twist nicely.
Oh my dear, i love your experiments, thoughts and your efforts. You should be a teacher, but not in school (is boring) but in a free class for all those children (and adults) who are willing to learn for their life and have fun! Thank you so much for sharing!! 🥰😘
The fog machine in the car. I had a Cheech and Chong flashback moment!
You are a goldden god. Being able to fly from a speeding boat in rough water, and all the ground sensor elevator control, golden.
One thing you could try is making the rods of more flexible material. Even more so you could have them of greatest flexibility at the point and gradually decrease flexibility down the length if the rod. This could be done in part by varying the width of a given material
In full size aircrafts that are aerodynamically stable one always talk about ”pitch changes speed and throttle changes altitude”, atleast throttle changes vertical speed. Have you tought about connecting the hight measurement to throttle more than to pitch? Might work or might completely fail, might be something to try out as well😊 and as always, great video! Much fun to watch and learn from
you cant tie the altitude error directly to the elevator. pitch controls vertical speed, which in turn controls altitude. so the altitude error must be used to generate a vertical speed command, which in turn generates an elevator command. i spent a fair bit of time myself dinking around with altitude control loops for uavs, getting a balance of stability and responsiveness is for sure challenging
Just to throw a curve ball, assuming fixed throttle, pitch also controls airspeed. As airspeed drops, so does climb rate. Among other things, this makes the physical response time vary based on the control output.
The mechanical control system on a foiling Moth (similar to the sailboat you showed), relies on what we call a "fast point", where you have maximum control input (rate) at the desired height, with diminishing sensitivity (rate of control surface change for a given unit of wand rotation) the further you get from the desired height. This is key to preventing the porpoising and oscillation you're seeing.
ie if it's a constant rate throughout the arc of your wand, it will oscillate.
The issue is the input from your "sensors" and the output from your plane. Because the input from your "sensors" gets more intense the closer to the ground, and they seem to get way more intense before the plane has time to correct. So being able to adjust the sensitivity, or a delay or something along those lines would be the path forward.
I agree with this completely. Basic damping of the control surfaces should aid greatly. Maybe even something as simple as a friction damper attached to the control horns/surfaces. Maybe a rate slowing of the reaction speed in your duino? Or a steepening response rate based on altitude? Slow response with small altitude deviations, faster responses the further it deviates from its correct altitude.
I don't normally pay much attention to sponsors, but that battery looks tempting. I'm sure others can be charged from a car's 12V socket, but it's the first one I remember seeing a person specifically mention having that feature.
I'm so impressed with the quality of your videos, thank you for making such interesting stuff without getting stuck in gimmick territory.
You may wish to notice that PID works well only when there are no other hidden integrators in the system. Integrating twice produces a nice oscillator. Hidden integrators may come in form of inertia of any kind. Things to do about it is either turning off the I component of a regulator, or shifting its frequency window a full order of magnitude away from the existing hidden integrator.
Daniels channel is the most interesting RUclips channel there is. I even watch his ads. LOL! Seriously, this channel always is interesting. Even when he was a kid.
In the sonar code try adding a smoothing code RNGFND1_WSP_MEDF: Moving Median Filter = 0-255 unless you’re using a custom code in a ardupilot. I think some of the oscillation is the waves
Sets the window size for the real-time median filter. When MEDF is greater than 0 the median filter is active
The shot of the plane next to the boat with the seagull in the background is cool.
11:39 Your canard issues might stem from the sprung mechanism's reaction force, causing a mix of aerodynamic and mechanical forces in your feedback loop. Test this by balancing the plane on your fingertips and lowering it onto a surface with the mechanism activated; if it pitches up, there's a mechanical force in your feedback loop.
Yeah, she's bouncing off the rubber bands.
FMS are making some gorgeous aircraft lately. I've got the Moa 1500mm. Love it to bits!
I LOVE your videos!! Keep them coming!
A Close Coupled Canard may be better for ground effect. IGE pressure under the wing increases which causes an upwash at the leading edge which will increase the lift of the close coupled canard, causing a pitch up away from the "ground".
Also a swept wing pitches down less IGE than a straight wing because of less center of pressure shift.
In fact 25⁰ to 45⁰ swept wings with a tall OGE T tail will also work!
Wow what a great channel, so pleased I've found it. Can't wait to see what the next couple of years brings!
You make just such high quality content and explain everything perfectly
Honestly I think using the skimmers on the flaps was the mistake, I think had you used that directly on the elevator with a switch to disable them for “normal flight” would’ve solved your problem. This is such an interesting “problem.” Awesome video, love your channel
You really need to take some classes in control engineering. Cutting down the canards didn't only reduce the p-gain, it also lowered the saturation limit.
A better solution would have been to lower the gear ratio.
I would assume that the wands touching the ground actually reduce the lift required by taking up some of the weight. Aircraft is extremely light so imagine they would have an impact..
The wand/canard looked promising at first, don't know if anyone has already mentioned but it could be worth changng the feedback loop by adjusting the throw ratio rather than the flap size. You could do that by changing your gear ratio, but probably easier to switch to a coupled control arm set with different hole positions.
To apply the flaps to the controll loop is the right step. You create lift without loosing energy for pitch motion of the entire aircraft. The flaps need to be much bigger though, at least by factor 5! The canards are different, because of their leverage to the centre of gravity and it's coupled feadback loop with two different sets of time constants for the same prameters. The surest way to figure out a stable parameter loop is to keep the number of degres for freadom as los as posible. For instance, for ground effect mode - freeze the pich attitude and work only with flaps and rudder. The pich should be controlled by a gyro, but still keept fixed, when flying in the ground effect mode.
My wet grinder, besides using a similar non-fluid oil, has pads of felt to retain the lube along the shaft. Two of them form long tubes/rings around it. It withstands being submerged. It has a third trick too, gears on the ends with another parallel shaft to pick it up. Whole thing is in a bolt down cover (commercial manufacturer). It is legit waterproof.
Something in there might help.
I really enjoy your experiments! On the topic of water surface skimming vehicles, I'd like to challenge your thinking here a little: perhaps the optimal foils for this application are hydrofoils rather than airfoils. Smaller foils, and the control loop being more naturally stable n simple to control... I suppose that makes it a watercraft not an aircraft, but those are just words. Keep being excellent!
Cool progress in the ground effect experiments. Those fog shots of tip vortices is really cool.
Couch man here:
The problem I see with the stick changingthe flaps is the stick:
-I would try a longer and more nimble stick.
My guess is, that would definitely work.
Currently the stick is way too long.
Just scale the length and size to the water craft, you showed.
(Compared to that you are putting a tree into the water to adjust steering)
Great video, if you change the size of the gears at the front of the plane so that they are different size then you can reduce the amount that this will cause the plane oscilate as the distance moved by the lever could be reduced by having a larger gear control the pitch against a smaller one attached to the leader as the smaller gear will have to travel further to move the larger gear. A 2:1 gear ratio would halve the movement
the air to air thrown camera shots were great man
Your deep cord tests are more effective because it takes more time to bleed the pressure off from under the aircraft. Think of it like a tarp gathering wind and billowing out. To stop the billowing you can put cuts in the tarp and the tarp will billow less with each cut.
For a lesser cord there is a shorter path to release pressure despite more generated ground effect BUT the accumulated total air pressure under the wing surface would be less because of the amount that can escape relatively easily.
Double-check the barometer and right around where the barometer hole is. This behavior the fox glider had where it just moves up and down a couple feet randomly sounds like a problem I had where a little tiny piece of tape was being blown into and out of the little bay where the FC was, and it was changing something about how the pressure would change in there, and making the quadcopter bounce up and down a foot at random times.
As you said at the end scale seems to be the biggest success factor. Go big.
Those camera throw shots were suprisingly cool, with a 360 cam this would likely be an actually cool and good format.
Nice! Also, your initial plane mod (before the prop offset) made it look incredibly photogenic, no wonder the seagulls were envious
I think that for a proper feedback loop you will need wands with progressive flex that match the weight of the aircraft. Think of it like a fishing pole. At the tip, the wand would deflect more allowing for less input. As the aircraft gets closer to the ground the wand provides more input because the wand gets progressively stiffer.
As always, very fascinating concepts and solutions to your hypotheses! Love your videos Daniel...they always inspire me and provoke much thought!
Changing the wands so that they are curved, may help with the oscillation. As the plane gets closer to the ground, the curve would cause the flaps to be turned less for the same unit of distance closer.
That analogue altitude control worked so well!
Dude! Another awesome video! A++ all the way around. Content, editing, narrative. You’re nailing bro!
Thanks for inspiring my inner engineer to play. RC, micro controllers, 3D design & printing. PID. Checking all my favorite boxes. Love it! 👍
I can tell you as a Flexwing pilot our wing aoa is high on the take off roll, as soon as it lifts the wing aoa has to be reduced fast to increase air speed, or risk a stall. Your aircraft is oscillating because it is stalling.. the wand is producing too high an aoa regardless of control surface area. .Its a tricky one to solve dumb mechanically but you need to find a way to return to level flight and get your airspeed back at the top of the bounce. I would suggest use your sonar to activate a servo to bring the aircraft to level flights at anything ABOVE say 18 inch?. Wand is up control.. Sonar is level control. I'm rambling now 😅
The larger the model the more you can "feel" the ground effect, i have a 5 metre wingspan rc canadair waterbomber and you have to wait until it slowed enough for it to descend through ground efect and touch down nicely, if you force it it will get pio very easily, it has a deep cord flat bottom wing
This is fascinating, I believe that the efficiency gains of ground effect can influence the stability quite a bit which could be the cause of the oscillation, just my two pence.
A commercially available gyro would fix all your problems, definitely worth a go, nice work,
Just a theory, for tuning the control loop you may make headway by considering both power and elevator commands.
They say, "speed is life, altitude is life insurance" and while the saying goes for surviving (engine) failures, the same is true when landing. Landing aircraft must stabilize their approach, so that the aircraft's energy at the touchdown point is in a specific range. If they're going too fast, they can pull up and exchange the speed for altitude, and this exchange is done fairly quickly; a few knots for a few meeters is a balance between airspeed and altimeter readings. Their sum will calculate the aircraft's net kinetic potential energy, which you can add to by increasing power. You might have luck in your control loops by using altitude as input for the aileron loop, and altitude plus airspeed for the throttle. This would allow the plane to react to the situation where it starts too high, pitches down while adding power, and ends up low with too much speed.
I would personally just use simple gains for both, but logic would say you'd need a gain for mass (speed and altitude multiplier), drag (speed multiplier) and should tune the rest with PID.
I fly a powered glider similar to his plane. I was surprised when he said he didn’t feel(?) the ground effect of the plane before he modified it. Well you can see the ground effect. The plane just won’t stop it keeps going and going flying along a few feet off the ground with out power. That can be a problem for new pilots trying to land. Here’s it comes great glide slop, almost down but it just floats right on by you slow and happy.
Such a picturesque scene with the plane on the water in the fog!
The Kasperwing U/L utilized a single surface airfoil between it's floats, it was very effective in shortening water takeoffs. Ground Effect was also noticeable with the wing between the floats, it might be worth a try on your R/C float plane.
The moment u started to throw the camera I scrolled down and tried to sub, yet I've been a sub for ages and u deserve more dude. Creatively your content is unique.
Honestly lad keep up the good work, ya made us proud son.
This is where machine learning could really be interesting - with two ultrasound sensors (one facing down, one slightly forward) and a way to detect hitting water (capacitatively, or via an accelerometer), you could feasibly train a model to follow the water as close as possible, while predicting its future path (and increasing altitude for waves). It would also naturally learn to filter out noise from the sensors and water surface fluctuations without losing sensitivity to a sliding window average (like a regular control loop would).
Since the ground effect pushes against the airplane you could add Barometer airpressure sensors on tail and each wing tso servos react better.
2:00 Inventor invents inventive way to film, and it's brilliant.
You packed some incredible shots into this video! 💚
Love your videos!
I believe that the surface wand touch point with the water, has to be far ahead of the CG in order to have stability. This is evident in all of the foil boats that use this control method. When it's behind or close to the CG, the pitching motion may be causing a reversal in wand displacement angle.
Put another way, think of the wand controlling aircraft pitch, not altitude. Suttle difference maybe, but it matters for stability.
I would say add a reduction gear between the wands and the control surface. Print some different ratios and continue testing. It will depend on the wind condition and airspeed which configuration will work best.