system is divergent after the hardware(vio-sensors) stay static for a while?

[Wangxuefeng92]

hi, I got some problem with open-vins.
I carried my vio-sensor, and walked around for a long-term, then I stayed static for a while. then I found that the trajectory is divergent, also the residual is divergent...
hoping for your reply!

[goldbattle]

This can happen if you lose track of your SLAM features and there isn't a zero velocity update. We will hopefully open source the zero velocity update soon (not sure on the timeline). But you can avoid this by making sure you track a lot of SLAM features and they are able to be continuously tracked when you are still.

The underlying computer vision issue is that you cannot triangulate any new features if you are still since there is no parallax between features.

[Wangxuefeng92]

thanks for your reply！I am also wandering that how to detect the “STATIC” condition for a vio system. Because the vision or imu measurement is so noisy, and the vio system can also slight jitter sometimes.

[rpng-guest]

PR #79 should address this, feel free to try it out (does require some tuning of the detection thresholds).