Great job. Congratulations Christian!
1 Like
Another great test day! All the systems are finally starting to synch and move in concert. Something I noticed though… It was hard to extend the lines of the kite and fly it at the same time… I kept wondering what the issue was and why I couldn’t control it, when I realized the hand crank is on the wrong side so I was flying with reverse controls!
Luckily the design is modular enough for me to swap the controls to the other side. I’ll be out testing the larger 3 line kite tomorrow. Cheers everybody
3 Likes
Well done Christian. I can’t wait to see the rest of the tests with this promising device.
1 Like
Brilliant to see steady progress @ChristianH
Well done!
Is there force sensing on the lines ? Or is that all done via winch torque?
1 Like
Steady… we’ll see… I dont know how the rest of you guys balance everything. I originally was going to use strain gauges, but I think the winch/motor torque is a much better solution. I just need a way to tell the difference between the two lines. I did another stellar test today, results soon.
I’d be happy to see one of these days continuous loops at a stabilized altitude. I think it’s coming soon. And congratulations again for this innovation.
2 Likes
Another fantastic test day. Pierre, we are getting closer to the “yo yo” maneuver you spoke of.
Stabalized loops will be very, very hard. Something like this is easier for a powered craft, but for a tethered craft we have to respond to the changing wind. For instance a stabilized loop would be very difficult on a gusty day because the oncoming wind is unpredictable.
I think it is possible of course. There’s a lot that can be learned from stunt pilots. I’m really curious about this because it’ll play a big factor in automation.
4 Likes
Fantastic test video well done
Quite often a line tension sensor is just a load cell connected to the axle of a sheave which the line runs over at a known pair of in and out angles … free body analysis, triangulating the load cell reading with the line vectors (assume equal tension each side) lets you calculate the line tension force.
I’d also love to know if it’s possible to listen to the vibration of a kite line and translate that into wind speed. But that’s gonna be a whole other world of complex.
It looks that you achieved this between about 3:53 and 4:09. You have pulled and released the fixed green line several times in succession. Is this correct?
By soon Daisy rotor(s) will be controllable. And a little later it will be yo-yo control. Or vice versa. Mechanical means and automation will follow, as well as other tests.
It may be fairly easy to get IMU data from your kite In order to co-ordinate the rotation of the barrel of the OKE winches with the rotation of the kite.
We have been attempting something very similar with the rotation of the kite turbine and the generator power take off wheel, and that’s spinning a lot faster . TBH not done enough work on this yet but we have some general ideas and small tests run. Far too much talking not enough time making
Try to run a video through a Yolov_ app. It may be able to find your kite with almost zero effort. After that, maybe a python script is sufficient to communicate revolutions from a camera to the motors on the OKE device
2 Likes
I think visual accompanied by pressure data will be the course I go, at least primarily.
Tesla also uses camera data more than any other sensor
Does any AWES company rely on visual?
Tesla’s don’t have to look so far, nor directly into the sun, or cope with birds shite on their lens, fog ,
Do they really only rely on camera, night too?
This is actually amusing because I don’t really like the idea of visual either, but I have a vested interest in keeping all critical systems on the ground, there’s a bevy of reasons I can get into later. A camera can work at night with reflective foil and infrared. I do worry about fog, and other visual obstructions. But hey, humans use sight to fly a kite as well. and visual algorithms are very sophisticated, I think it would be arguably easier to train a model to recognize a unique kite with fewer obstructions in the sky.
This is all to say I think visual is just the easiest starting point, because as soon as we talk data, we need to talk about transmitting that data, and then we need to talk about power, and then you eventually end up with a conductive cable or batteries and that’s more weight aloft.
3 Likes
Video is the best fit for your system imo. Maybe add a IR light and reflective surface for better performance, but judging the video, that fine tuning is probably a few years away from top priority
2 Likes
Hey guys, I put together a YouTube series that investigates potential designs for record-breaking sailboats. The mission of OKE is to develop the necessary tools to turn these designs into reality. This gives us the flexibility to start small, iterate on our designs, and gradually progress towards creating a consumer product. Hopefully it’s a practical approach to fine-tuning our technology and eventually introducing innovative consumer tech to the marine industry.
I feature Luc Armant of Ozone Paragliding in this edition. In the future I speak with Don Montague of Project.Kiteboat, and I speak about my experience with Grin Technologies, the BC based Electric Bike Company where I sourced my regen motor controllers
3 Likes
3d force sensor might be an option as well if you want the direction of your pull . I think Beyond The Sea is using them.
I think these sensors are quite poor in meeting with realities… first off, the tether is far from straight. Next high low force varies, and pull may not be in the direction the 3d sensor prefers.
[edit,added the following]
Next, you may need to place the winch on top of the sensor. Then the mass of the winch, vibrations, wiring etc will affect the force reading.
I do think though for short tethers it can be useful, as long as the reading is not used as the sole input for feedback control of the yaw axis. In that case, the tether slack, bending, sensor inaccuracy, winch mass and vibrations together will be too much to compensate in feedback control and still follow a path consistently in the face of varying winds.
This is my gut feeling. Specific sensor data and control requirements would be necessary to make an accurate analysis.
All this is why I prefer video as a starting point, because the non-linearities of the sensor and tethers are gone and you can do feedback control in fair visibility conditions. Once there, consider other sensing options. At that point the system is probably pretty sophisticated overall, and you may already have power and telemetry on the kite. In which case adding a sensor on the kite would be the preferred option for feedback control.
1 Like
I disagree. Now you’re using (motorized…expensive) cameras to get a delayed and imperfect position and direction of your kite that you then have to process further to get more accurate data out of, which you would have to learn how to do, and hope the result is good enough to use. Much better to skip that and just put some sensors on your kite that transmit accurate information to your ground station, or other kites, with millisecond delay. After you’ve figured out how to control your kite, you can then see if you can skip this, but you probably shouldn’t as there is no real cost to this and it adds redundancy.
You would still need cameras for foreign object detection though I think.
It’s like in VR inside-out vs outside-in tracking: https://pimax.com/pose-tracking-methods-outside-in-vs-inside-out-tracking-in-vr/ or like how in autonomous vehicles it would be good if every object broadcast its position. See also: Swarm Robotics.
I agree this is the simpler and more straightforward way. It does however introduce the need for something onboard the kite, which is not necessarily all good. Eg handling can get more cumbersome. You cant drop the kite in the sea. Battery life issues. Kite and bridle tangles in zero wind. Its just a matter of choosing the better option really, case by case