IMU noise

141 views
Skip to first unread message

Wu Jiang

unread,
Jul 19, 2017, 11:19:03 AM7/19/17
to LINOROBOT
Looks like my GY85 has a lot of noise even after calibration. Sometimes the angular value is higher than 0.03 or below -0.03. The problem is that if I increase the noise threshold, it will ignore the regular low values from IMU, like when it rotates slowly. Any suggestions? Thank you.

Juan Miguel Jimeno

unread,
Jul 19, 2017, 11:35:17 AM7/19/17
to LINOROBOT
How high does the noise gets? I've encountered the same noise that's why I've included this "filter" in the odometry code. https://github.com/linorobot/linorobot/blob/master/src/lino_base_node.cpp#L34 . You might want to try tweaking 0.3 to the max noise value of your IMU. I think anything below 0.1 can still be considered as negligible. Might be a dirty hack but so far, this works for me. Maybe the rest can comment on this.

Wu Jiang

unread,
Jul 19, 2017, 12:23:21 PM7/19/17
to LINOROBOT
Yea, I saw the filter you added. There is a problem if the filter threshold value is high because it will ignore regular low value which is not noise too. The noise can go up to 0.045. Hmm, might be my PID is not tuned well. I had my PID tuned to let speed reduce slowly. So when it reaches the goal, the last few seconds, it turns very slow and the gyroscope reading is below 0.03.

Juan Miguel Jimeno

unread,
Jul 21, 2017, 9:12:45 PM7/21/17
to LINOROBOT
Have you tried running the robot at its minimal speed and measure the angular velocity? With that you can see what's the max you can go for your filter.

Wu Jiang

unread,
Jul 27, 2017, 3:34:07 PM7/27/17
to LINOROBOT
understood. thank you. Mine lost angular accuracy after a few turns.

Wu Jiang

unread,
Jul 27, 2017, 10:05:09 PM7/27/17
to LINOROBOT
Now I'm thinking it might be because lidar has miss matching which caused pose not being accurate. I still see "Scan matching failed. Use odometry" when generating a slam map.
Reply all
Reply to author
Forward
0 new messages