GPS tracking flight

[Mike Callas]

Has GPS progressed to the point a tracking unit can be placed inside a stunter to record the flight for post flight review/training purposes?
I believe current single frequency units are in the 10 to 15 ft accuracy range. Dual frequency Rx down to a couple of inches.

I had a GPS camera on a formula ford that drew the course the car took lap after lap. It is a great tool to see how car and driver are performing.

[Dave Hull]

How big/heavy is what you have?

Divot McSlow

[David Hoover]

Eagletree has a GPS module that connects to their data logger.
http://www.eagletreesystems.com/index.php?route=product/product&product_id=54

[Mike Callas]

How big/heavy is what you have?

Divot McSlow

Its a Garmin camera like a Go Pro, so its too big.

[Dave Hull]

Not clear what you get beyond a data stream from the Eagletree GPS/Alt sensors. For this to be useful or entertaining, we need a graphical output, with tools. And, at 10 Hz, the corners might be undersampled a bit. Maybe not enough to degrade the output, since the airplane track would be curve-fit anyway.

Dave

[TDM]

Anything is worth a try. But my first instinct is that with 15ft accuracy your level laps can be from -10 to plus 20ft and there is not much I can do with the received data. I also am looking in to Video to record flights for training purpose. What I would use that logger for is to log power plant related data.
Can anyone loan me one?

[Tom Vieira]

GPS altitude tracking tends to lag severely, and with the rate of change we go through, it would be less than ideal.

some sort of inertial/accelerometer based system would be far better, but that would pose it's own issues since we are working around a hemisphere rather than an "open space" condition.

we use inertial, GPS, and barometric tracking in model rockets to track everything from flight events, x/y/z position, velocity, etc....  they work pretty good.  but, all three sets of data must be combined and filtered to get the "best" reading.  we are also working in a (hopefully) linear fashion on the way up, and a very gentle descent with some drift due to wind.

I don't think barometric would be ideal in our case, as there isn't much pressure difference between ground level and 60', and there are other challenges associated with that as a result of the plane passing through the air, propellers throwing air all over the place, flaps and elevators causing all sorts of pressure changes in addition to the speed changes, it all turns in to a pretty crazy circle.  the same happens with rockets as well, but it only gets really bad in the transonic realm, and it's easier to filter the data (going back to the linear thing and all of the real crazy thrust stuff is coming out of the back).

[Lauri Malila]

I don't know about necessary resolution, but most devices are not accurate enough. Especially altitude can vary several meters.
It's enough for logging free flight. In the picture is logged a bunch of F1A flights at last free flight world champs (Hungary). Icare tracker. L

[Dave Hull]

Now that is interesting!

[Tim Wescott]

GPS merged with inertial would probably be enough.  Google "sensor fusion".  If you could record the inertial data fast then as long as you know the delay to a gnat's eyelash then even delayed, 10Hz GPS would be enough to significantly correct it.

[Keith Renecle]

Hi Mike,

The sensor in the model concept is the way to go. You just need someone with enough knowledge, experience and time to do this. In the meantime you can use Alberto Solera's video tracker and that is more than good enough for practice. There is a thread on this forum on it with the download links and all. Here is the link: https://stunthanger.com/smf/open-forum/videof2b-at-landres-2018/msg528491/#msg528491

Keith R

[Dave Hull]

After doing a couple of basic calculations, a 10 Hz data rate is not going to come close. This assertion is based on a model flying 5.2 second laps on 65 ft lines and doing the theoretical 5 foot radius corner. At 10 Hz you could expect one sample during the turn. At 100 Hz you would probably be in business. You need enough samples to Kalman filter, and need to be on the right side of Nyquist. Still, this would take some good software, which was why I was asking above about existing code and outputs.

With the above approach, you don't need as much absolute accuracy as one might otherwise think. It is done all the time in guidance systems. Fusion brings other obvious advantages as Tim noted. At the very least, it allows the two data sets to be time registered with only one accelerometer, and it is easy to do much more than that. One of the inputs that might reduce the computations for data fitting would be to input the actual flight radius; centerline of plane to center of pilot rotation.

This is way too much like work---why don't we just get our buddies to point out all the errors. After all, in my experience, they are going to tell you about the lumpy loop or wobbly wingover whether you want to know about it or not. I need to learn to fly better though--I don't have a full hour between flights to get their full report. They usually just get to the errors in the outside loops before it is my turn again.

Divot McSlow


Dave

[Tim Wescott]

---why don't we just get our buddies to point out all the errors....

Because when I do get a chance to do practice flights, it's from 8 AM to 10:30 AM.  I'm usually leaving the field about the time that everyone else is showing up, so I get no coaching.  And the field where I fly has enough background clutter that a video would be useless, and enough people around that I don't want to leave a camera unattended.  I would love to have a means to record a flight so that I could either play it back myself to look at it, or send it to a remote coach.

I have a data recorder with built-in IMU that I'm working on, but it's on a back burner in favor of other projects (I do get SOME work on it, though, Howard).  I think there's a chance that just knowing the plane is orbiting around a more-or-less fixed point would be enough to reconstruct maneuvers decently, but I either need to fake up a set of data (lots of work) and give it a whirl, or record some flights (see above) and give it a whirl.

[Mike Callas]

After doing a couple of basic calculations, a 10 Hz data rate is not going to come close. This assertion is based on a model flying 5.2 second laps on 65 ft lines and doing the theoretical 5 foot radius corner. At 10 Hz you could expect one sample during the turn. At 100 Hz you would probably be in business. You need enough samples to Kalman filter, and need to be on the right side of Nyquist. Still, this would take some good software, which was why I was asking above about existing code and outputs.

With the above approach, you don't need as much absolute accuracy as one might otherwise think. It is done all the time in guidance systems. Fusion brings other obvious advantages as Tim noted. At the very least, it allows the two data sets to be time registered with only one accelerometer, and it is easy to do much more than that. One of the inputs that might reduce the computations for data fitting would be to input the actual flight radius; centerline of plane to center of pilot rotation.

This is way too much like work---why don't we just get our buddies to point out all the errors. After all, in my experience, they are going to tell you about the lumpy loop or wobbly wingover whether you want to know about it or not. I need to learn to fly better though--I don't have a full hour between flights to get their full report. They usually just get to the errors in the outside loops before it is my turn again.

Divot McSlow


Dave

Hmmm, my old kart had an AIMsport Mychron gauge. Among other functions, it used accels to monitor lateral G forces. The data was used to draw a track map. It ran at 200hz or so.

My goal is to have the data plotted in front of me like the Castle ESC data dump. With my laptop I can access ESC data after every flight in about 2 minutes.

Maybe this is an Igor question. I bought the Jeti spin 66 and accel from him a few years ago and have yet to use it.

[Tim Wescott]

You need gyros, accelerometers, and something to keep their results from drifting off into infinite rotations and infinite speed.  That "something" is usually a navigation fix, like a GPS reading.  Just an IMU (three axis gyro, three axis accelerometer) is not enough unless you're using something way bigger than will fit into a CL plane.  I'm not sure how small they've gotten unaided inertial nav devices down to, if such a thing is even a thing any more, but it's probably more than 50 pounds.  Even a once a second GPS reading is enough to discipline a cheap consumer-grade IMU to within a few feet of where it's supposed to be, though.

[Dave Hull]

It is still a thing.

[Tim Wescott]

It is still a thing.

So, how heavy a thing are they these days? The fanciest one I ever worked with was a little gyro about an inch in diameter and two long. It was good enough for airborne gimbaled imagers, but not for navigation.

[Keith Renecle]

Hi Tim,

You're dead right about the size of the "thing". I checked info on the Red Bull Air Racing tracker system and it's certainly a big box of tricks. I've been looking at this concept for a looooooong time and unfortunately this sort of design is way above my level of knowledge and experience. One of our local C/L guys is an electronic design engineer who has designed a few trackers for cars and trucks and is also looking at a stunt tracker, but is also stuck for the necessary time. He did send me the data on a high res GNSS chip that incorporates a whole bunch of tricks. Maybe this could help?  https://www.u-blox.com/en/product/zed-f9p-module.

On your comments of trying to eliminate some of the actual 3-D co-ords with known factors, I thought that maybe all we need is the angle of elevation of the model, like the line angle, and then the angle in the circle which will be 0 to 360°. The rules are in fact written to show shapes of maneuvers from the pilots point of view which is close to a 2-D view, and in that case the line length is irrelevant. Is that possible, or is drift still a factor?

Keith R

[Tim Wescott]

On your comments of trying to eliminate some of the actual 3-D co-ords with known factors, I thought that maybe all we need is the angle of elevation of the model, like the line angle, and then the angle in the circle which will be 0 to 360°. The rules are in fact written to show shapes of maneuvers from the pilots point of view which is close to a 2-D view, and in that case the line length is irrelevant. Is that possible, or is drift still a factor?

I believe there's a chance that setting up a Kalman filter based on the knowledge that the aircraft is roughly bound to fly on a spherical surface is enough to get elevation, and to track each individual maneuver.  But there wouldn't be a fixed "north/south" reference with just an IMU.  In other words, the "north" reference would be arbitrary and drifting.  An IMU plus just about any unambiguous direction indicator (compass, sun tracker, etc.) would work.  A sun tracker is just weird (and could easily get confused); a compass would work if you could keep the compass from going crazy, but not only is there a matter of board space, but you want this to work right next to this electrical system with high and varying currents.

[Brett Buck]

I believe there's a chance that setting up a Kalman filter based on the knowledge that the aircraft is roughly bound to fly on a spherical surface is enough to get elevation, and to track each individual maneuver.  But there wouldn't be a fixed "north/south" reference with just an IMU.  In other words, the "north" reference would be arbitrary and drifting.  An IMU plus just about any unambiguous direction indicator (compass, sun tracker, etc.) would work.  A sun tracker is just weird (and could easily get confused); a compass would work if you could keep the compass from going crazy, but not only is there a matter of board space, but you want this to work right next to this electrical system with high and varying currents.

          Don't jump to the conclusion that a Kalman Filter is the best solution. So far, I haven't seen a single example of a real system that was better off with varying gains as opposed to fixed gains. In fact, I have spent an awful lot of time trying to figure out how to defeat or overcome the variation of the covariance with time and samples, and usually winding up with fixed gains or severely denuded Kalman systems with, say, a 6x4 or 9x4 update matrix overridden with mostly zeros.

    In any case, both any current MEMs IMU package small enough for a stunt plane doesn't have sufficient performance, and you sure aren't going to use the sun, magnetic field, or horizon successfully. MEMS rate random walk is probably OK, short term drift stability is probably OK, but the scale factor performance is typically terrible, on the order of thousands or tens of thousands of PPM, and without also providing a temperature controlled-oven, you are going to have problems with the drift and scale factor.

   The photogrammetry techniques seems the most promising to me, if you think this is a critical issue. You might have to require the installation of targets to ensure reliable tracking, but you can get systems that would probably do the job within 1/16th" - the problem being that they are FAR out of the price range of model airplane people, and would required astronomical amounts of post-flight data reduction. Then there is the matter of what to do with it...

      Brett

[Tom Vieira]

what about motion capture ala Hollywood?

a few reference points on the plane, a camera, and the associated software?  it's designed to track stuff...  any info you want would be derivable from the footage and knowing the distances?  a couple of reference markers on the craft, and you should be set.  you could probably "cheat" somehow and use a canopy, the wheel(s), outboard wingtip, spinner, and rudder.

[Brett Buck]

what about motion capture ala Hollywood?

a few reference points on the plane, a camera, and the associated software?  it's designed to track stuff...  any info you want would be derivable from the footage and knowing the distances?  a couple of reference markers on the craft, and you should be set.  you could probably "cheat" somehow and use a canopy, the wheel(s), outboard wingtip, spinner, and rudder.

    Motion capture and photogrammetry methods are more-or-less the same thing, although photogrammetry is generally a static system. But, it boils down to optically measuring the position of something in space at a high sample rate.

    Brett

[Ken Culbertson]

And when we have spent millions to perfect an automated judging system the same people will place in the same order as before but they will do it with scientific precision.  We could then put electronics in our planes that would fly the pattern automatically and just ship our planes to Muncie and watch it all on TV.

Let's say we do find an automated judging system.  What are you going to do with all of us old farts that are on the downside of our reflexive skills and enjoy keeping in the mainstream of the sport through judging? 

I think the idea of tracking through video or other means is a great idea, but for training, not for judging.  If Keith R's work eventually leads to maneuver descriptions that can be drawn and flown then the ability to film your pattern and play it back over those images will be fabulous.

Ken

[Brett Buck]

And when we have spent millions to perfect an automated judging system the same people will place in the same order as before but they will do it with scientific precision.  We could then put electronics in our planes that would fly the pattern automatically and just ship our planes to Muncie and watch it all on TV.

    Don't take a discussion of the practicality as an endorsement of the concept. Of course it's not going to change anything in the results.

   The only reason to even look at it is to finally shut up the whiners, but they are mostly gone at this point anyway. And it wouldn't work - when the whiners don't win, completely objectively, they (or worse, their sychophants) will still gas on about how the scoring algorithm was designed by a bunch of cheaters to favor their buddies. That's more-or-less what the seeding had turned into, every year, and the judge selection, every year.

   The initial alleged "problem" was that Shareen and Warren were doing it to select favorable judges for Ted, etc. Paul and Howard's completely objective seeding and judge selection algorithms do more-or-less the same thing, with no human input, and it's *still* somehow unfair. And it had about the sorts of results you might have expected  - no change at all. So even if you came up with a completely objective system to score the flight, they still wouldn't be satisfied.

      It's an interesting thought experiment to see how it might be done, and also, to see what you might do with completely accurate position information to get a score. But of course it's not a good idea, and more importantly, we already have a good method that has more-or-less worked fine for 60 years.

    Brett

[Tim Wescott]

... the scoring algorithm was designed by a bunch of cheaters to favor their buddies...

And that's why I want to design the algorithm!  Because it's easier to win if you've had a hand in designing the game!

[Brett Buck]

And that's why I want to design the algorithm!  Because it's easier to win if you've had a hand in designing the game!

  The PAMPA Sekrit Kabal will of course, be closely monitoring the source code, including the Pacific Ocean proximity weighting factor, to ensure "correct" results.

    Brett

[Mike Callas]

Thanks for all the info. I was looking for a means to record/review/grade a flight and work on the shapes, height etc. without having to bring along a judge. Perhaps the video tracking I saw in another thread is the answer.
Mike

[Brett Buck]

Thanks for all the info. I was looking for a means to record/review/grade a flight and work on the shapes, height etc. without having to bring along a judge. Perhaps the video tracking I saw in another thread is the answer.
Mike

   I think it is much more likely to work and any inertial or inertial/GPS system, and a few dropouts (like we saw in the WC videos) wouldn't be a big problem for coaching purposes.

    Brett

[Ken Culbertson]

Thanks for all the info. I was looking for a means to record/review/grade a flight and work on the shapes, height etc. without having to bring along a judge. Perhaps the video tracking I saw in another thread is the answer.
Mike

Something I have been doing when I practice alone is to put up a marker like the ones they use in FAI dead down wind from the center of the circle about 10' back with a wind streamer on the top. then I set up my camera (I use an old cell phone) on the opposite side back far enough that I just get the edges of the circle in view and about 3' off the ground aimed at the marker.  You need to point the camera up or you miss the overhead stuff.  The effect is like you are the judge sitting in your chair.  If you are consistent in the placement you will get great comparative video's that you can play back later to see if you are making any progress.  One thing that I have found is that corners are tighter and intersections better if viewed at full speed!  Slow motion is a humbling experience.

One side effect of having the camera is that it gives you the same awareness in practice that someone is watching and helps make maneuver placement more consistent.

Good Luck - Ken

[Tim Wescott]

... including the Pacific Ocean proximity weighting factor ...

Oh, well, if you want that then we'll need the GPS no matter what -- you know as well as I do that you can't deduce longitude information from inertial measurements alone.