search

osurdml / issues

0 stars 0 forks source link

Quadcopter Position Drift #14

Closed skeel3r closed 10 years ago

skeel3r commented 10 years ago

The quadcopter position drifts fairly significantly, ~1/6 meter a second. This may be a symptom of sensors needing to be calibrated.

skeel3r commented 10 years ago

This went away after calibration. The gyro calibration was off, causing the persistent drift we encountered.

kylc commented 10 years ago

Actually it looked like it was fixed but it wasn't. The drift depends on the yaw of the quadcopter. X drifts when the quad is facing E/W, Y drifts when facing N/S.

kylc commented 10 years ago

I thought it may have had something to do with a bad accelerometer calibration because the roll and pitch were always offset by a constant value. I tuned the accelerometer offsets by hand to bring the readings to zero, but the drift is still present.

yoos commented 10 years ago

Can you make the drift worse by incorrectly calibrating some of the sensors? On Aug 21, 2014 4:50 PM, "Kyle Cesare" notifications@github.com wrote:

I thought it may have had something to do with a bad accelerometer calibration because the roll and pitch were always offset by a constant value. I tuned the accelerometer offsets by hand to bring the readings to zero, but the drift is still present.

— Reply to this email directly or view it on GitHub https://github.com/osurdml/issues/issues/14#issuecomment-53003837.

skeel3r commented 10 years ago

My guess is that the magnetometer correlates to yaw, can you disable the magnetometer, or atleast check that the values are reasonable?

kylc commented 10 years ago

The FMU believed it was landed, which would cause it to ignore flow data. Fixed by not resetting estimates when landed.