search

ethz-asl / kalibr

The Kalibr visual-inertial calibration toolbox
Other
4.33k stars 1.4k forks source link

Some question about "Bad IMU Rate". #654

Closed liaozihang closed 8 months ago

liaozihang commented 10 months ago

In the wiki, there is a mention about bad IMU timestamps , I have a question about the impact of bad timestamps.

Can anyone explain me，a big question is what is the impact of a bad imu rate on vio, which seems to have no relevant explanation found on the wiki or issue.

I would be thankful for any answer.

Thanks

goldbattle commented 9 months ago

I think this all depends on the noise and the integration methods used. You want to have a reasonable rate ~100-500Hz IMU with low noise.

The problem with "bad timestamps" is that they are inconsistent, i.e. it is unknow what the true measurement timestamp of each IMU reading is. In this case, I believe basically all VIO / inertial method will fail.

liaozihang commented 9 months ago

I think this all depends on the noise and the integration methods used. You want to have a reasonable rate ~100-500Hz IMU with low noise.

The problem with "bad timestamps" is that they are inconsistent, i.e. it is unknow what the true measurement timestamp of each IMU reading is. In this case, I believe basically all VIO / inertial method will fail.

Thanks for the reply !

Based on your reply, my understanding is that "a good imu driver output should have stable time intervals.If the actual output imu time intervals appear to be unstable, there may be a problem with the driver or the sensor itself (although based on the clock mark a time, it may not coincide with the actual time it happens), which is not good for any VIO method"

Is that a correct understanding? Thanks

goldbattle commented 8 months ago

Correct!