IMU rotation drift

[PaulBouchier]

Hi @hippo5329 - for the first time, I've turned on the IMU, to test the wifi port to upstream. I'm going through the SLAM & Nav2 sections of your wiki and seeing something I don't know how to deal with - can you help?

After startup, /imu/data shows the orientation is constant, but there is a non-zero (but small) twist around the z axis. This would be the expected drift in the IMU. /odom/unfiltered shows constant unchanging pose/orientation and twist/orientation is zero - this is expected because the robot is stationary. However, /odom shows the orientation changing slowly (z and w change). I believe this is because the EKF filter is fusing IMU and wheel odometry and deciding that the drift in the IMU means the robot is rotating. Consequently, SLAM shows the lidar scan slowly rotating around the robot in the map frame.

This doesn't seem like correct operation. Do you know how to fix it? Is it documented in the wiki and I'm not seeing it?

BTW I made some updates to the wiki headings to more clearly state that running Nav2 SLAM & navigation is a smoke test, as stated in the text. I also clarified some of the instructions in the Linorobot SLAM & navigation sections.

[PaulBouchier]

Well, it was happening this morning, but now it's not, so closing till I can get a better handle on it. Rate of rotation in Twist message this morning was about 0.02 rad/s

[hippo5329]

Which IMU do you use? IMU will be calibrated during initialization. The robot must not move during the calibration.

[PaulBouchier]

MPU6050. I wonder if I was moving the robot when I powered it on. It seems to be stable now, but I'll watch for a recurrence.