Check out the ArduPilot Discuss forums. I put a lot of info there. Mower is tip top and running great.

Thanks much! Funny you mention waning interest. I got that impression for a bit, too, and my own interest faded slightly to be honest. But the dev team seems to have some renewed vigor over the past few months, and the updates are fast and furious. It's been a pleasure to work with them. I feel like interest is on the rise again. Also a bit ironic that your impression is that I'm helping Kenny...those servos everyone asks about were HIS idea (and actually how I found him in the first place). He helps me at least as much as I hope I reciprocate. I've briefly considered using some image recognition for obstacle avoidance but have not delved into it too much. I have a Google Coral device for another little indoor robot that I've been meaning to mess with, but the mower has dominated much of my time to tinker lately.

Why don't you join the discussion at discuss.ardupilot.org ? What you're asking is a monumental task. I have little interest in documenting every single piece and part of my mower so you can replicate it.

I don't think a GPS based attitude indicator is practical. The antennas need to be pointed skyward, so any attitude beyond about 30-45 degrees in pitch/roll will produce anomalies. Also, the update rate is only 5Hz, which is far slower than any flight controller needs to compute attitude information. You could potentially use RTK corrected GPSs to remove bias/drift in the INS (which is exactly how GPS-for-yaw works), but it's not a practical attitude indicator/controller solution. For yaw (heading) information, I'm using to uBlox Zed-F9P modules measured 0.366m left and right of centerline (on the roll bar uprights). At the turn rates my mower is capable of producing, the 5Hz update rate + accelerometer input keeps the nav controller's heading information very accurate. In a roughly 2D environment, I have the luxury of knowing the antennas are always oriented toward the sky.

@@YurisHomebrewDIY Not talking aerobatics here!. For gliders where the small ones have a 15m span, 7 - 8 meter nose to tail. 20 seconds to do a complete turn. On the bigger gliders, you will get big flex in the wings, but that will, for the most part, be symmetric [g loading] You might go to 50 degrees of bank but mostly it's 45 or level. You certainly have AHRS systems in them now. So I was wondering if throwing in extra information to the Kalman filters they use works. Certainly, as you say, deals with drift. So two things spring to mind. 1. DGPS from a broadcast signal. The problem I see here is that it's going to be 20-50K distances or more. Does that give another measure of altitude and in the particular accurate rate of climb that is free from some of the noise of pressure based sensing? Typical filtering would be average over 1-2 seconds anyway for those. So a 5 Hz sample rate isn't an issue. 2. Given the three GPS sensors, can they operate as a local DGPS set-up, where you just want the difference, accurately between P(left), P(right) and P(tail)? I might email ublox.

@@Nickle314 Your creative idea seems like a solution in search of a problem. You can buy a toy drone with reliable 6-axis gyro stabilization for under $30, and you're talking about using well over $600 of precision GPS equipment with disadvantageously placed antennas to accomplish similar. I applaud the out-of-the-box approach and would certainly be curious if you find a good use case and method to employ it, but I'm afraid I just don't quite see the practical application here.

@@YurisHomebrewDIY I have a mower set up very similar, how do you get 90 degree shift to have heading facing forward ?

@@cesmaeili Simple mathematics?