ghowardeng / attitudeestimator Goto Github PK

Will be more efficient. Only convert to degrees for final output.

More sophisticated approach than what is currently used.

Don't depend on Accel vectors/angles to transition attitude between limits. E.g. roll change from 180 to negative 180 should be done post filtering and not dependent on z Accel magnitude.

Should try only using correction at low rates and attitudes. Accelerometer becomes unreliable at certain angles and under dynamic acceleration. Gyro should be good enough to propagate estimate until returning to more level attitudes.

Could also have dynamic gains on observer based on reliability of Accel given attitude.

This will be unreliable during large dynamic accelerations.

Would be a nice way to simplify filtering and the management of buffers etc.

pitch rate = cos(abs(rollAngle)*M_PI/180)gyroRate[X] + sin(rollAngleM_PI/180)*gyroRate[Z]

When using accelerometer attitude measurement over the full -180 to +180 range, the pitch angle is unstable while the device is rolled over. Don't see the behavior in the roll axis. Check angle calculation.