You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It should be noted that for both KalmanNet and Split-
KalmanNet, measurement yt is not available at every moment
in sensor fusion tasks. To make KalmanNet work in this case,
we use yt = ŷt|t−1 to compute F3 and F4
Is there any other approach you would use/have used ?
Hi @XiaoyongNI
Thanks again for the great work!
I have a question regarding missing observation data scenarios : In case, at a particular time-stamp I don't get observation value (but I do have the corresponding Ground truth), how do I input this to the model. Something like this, where 'T' is True and 'F' is False : T T T T F F T T
One way I could think of is what it mentioned in this paper (https://www.researchgate.net/publication/382417846_Practical_Implementation_of_KalmanNet_for_Accurate_Data_Fusion_in_Integrated_Navigation) :
Is there any other approach you would use/have used ?
Thanks!