I finally managed to get access to a decent camera on a day with good weather! The following video is mostly demonstrating the quadrotors stability under the new four-state Kalman filter I have implemented, that estimates both the current pitch/roll and two gyroscope biases. Most of the flight is hands-off, with the occasional input to correct for wind and inevitable position drift.
I feel this angle control is pretty much perfect, so I will now be moving on to integrating extra sensors to allow for direct velocity and position control. I’m excited to see just how complex I can make the navigation system before the Pi cracks, but it seems for now despite using unoptimized double precision floating point math for everything it still has plenty of headroom to go!
The (rather messy) code is available at https://github.com/big5824/Picopter. Now that the first stage is finished I can start going through cleaning up the spaghetti of test code, but that will have to wait for now.
You need to be a member of diydrones to add comments!
Comments
The diydrones nazis took my pi autopilot blog post down. There's probably a competing product coming from 3drobotics. It's 1 of those gadgets that every embedded computing project obviously needed to be ported to.
Very nice - very smooth....I look forward to seeing your next progressions....
Congrats!
Nice work. Wish I could claim to understand Kalman filtering, but every time I try, I realize why I am not a mathmetician.