Gluonpilot Version .5 Beta

[Mitch]

Flew a few flights today with the Katana and the new ver .5 beta. I only used circle relative in a loop but all went well. There was about a 5 knot wind from the east and some significant thermal/turbulence. Only real comment is I had to really dial down the altitude error -> pitch P gain to about .05 to mitigate extreme pitch oscillations. I adjusted it after the first flight below. I might still go a bit lower. This was noticible on the ground when I moved the airplane attitude as very jerky elevator movement. During the second flight I let it fly hands off for about five orbits.

Circle shape is much improved over ver .3 and the circle radius measures very close to the input value.

A few minor details with gluonconfig. After a CIRCLE_REL command is set. When it is recalled for edit the lat and longs are reversed. Also negative longitudes cannot be entered for absolute coordinates. Yes, I live way over here!

It would be very nice to be able to save and load flight plans to/from a file and to invoke the "To KML" function to be able to preview a flight plan after it is built.

Thanks Tom - great progress. Now I'm going to fly some more complex flight plans and see how that goes.

first flight - altitude error -> pitch P gain was way too high

first flight

flt1.jpg (79.5 KiB) Viewed 10367 times

second flight

second flight

flt2.jpg (84.52 KiB) Viewed 10367 times

[Tom]

Hi Mitch, thanks for your test flight!

Altitude hold seems to be pretty good at an altitude of only 60 meters and a circle radius of about 75 meter. It is a bit jittery, so I guess your P-values are maybe still a bit too high.
Remark that there is also a 4th PID loop: altitude error to pitch. You might change that one if everything seems fine in stabilized mode.

You are right about those 2 bugs in the editor! I'll fix it ASAP!

Tom

[Mitch]

Thanks for the quick fixes with ver .5b2. The navigation tab export to kml feature is outstanding!

[magellan]

Hi Mitch,

Congrats on this achievement, looks good.
Are you using your magnetometer ?

Regards,

Olivier

[Mitch]

I'm not using the magnetometer for this application because Tom has done such a great job integrating the board and the firmware without it. I have experimented with it to try and achieve full three axis stability for a static application such as a vtol, quad, or heli without the requirement for gps movement input. The best routine so far is vector matching of the accelerometer and magnetometer triads integrated with a quaternion kalman filter See Prof Elkaim's papers. He has a gluonpilot board and is looking at porting the SLUGS project onto it. Working well but very calculation intensive.

[Tom]

Mitch,

Did you integrate a magnetometer module on the gluonpilot?

[Mitch]

Not literally on the Gluonboard. I did use it on one of my old boards that uses the same dspic. So I do have an I2C library for hardware I2C (dspic hardware I2C can be kind of buggy) and code that I wrote for the HMC5843 3 axis magnetometer, including the calibration stuff. HMC5843 seems to be coming into favor with hobbyists and is inexpensive and readily available in several breakouts.

I posted some of the code here last spring. It could be easily integrated into gluonpilot code using gluonboard I2C. If you want to do something with it I can post the whole package at the end of the week when I get home.

[Tom]

Yes that would be great! I plan to release v0.5 soon.
With a funjet flying smoothly, the next goal will be to navigate the quadrocopter. This will require a magnetometer (I only have the HMC5842, so I hope it will be enough).

[Mitch]

I think the advantage of three axis is that one can get the entire magnetic field vector rather than just the projection in two axis. Makes things simpler at attitudes other than level. The reason I was using three axis is I was using the mag vector and the accelerometer vector into a filter. Sparkfun has a breakout (their's has had a problem with one of the capacitors) and I've seen a different breakout on a DIY website store.

[Tom]

I bought mine before the 3-axis version came out, but never got to use it. Hopefully the I2C interface will be about the same. I also hope that knowing the heading during zero roll and small pitch is enough