The previous problem was roll,pitch and yaw angle from quaternion.
Now it is fixed. 1-2-3 Euler representation is used.
Also accelerometer sign change has been applied.
Here are few observations:
- When the system initialized, roll angle is initially reversed.
As filter converged, it becomes normal.
- I put a negative sign on roll, yaw. It should naturally has right
sign, but I do not know why for now. Let me investigate again.
- Gain : I do not know what gain is good for quadrotor flight.
Let me take a look Ardupilot gain in the later.
Anyway, you can fly with this attitude estimator.
Hyon Lim (Retina)