Improve the outlier detector for TDoA #375
Milestone
Comments
krichardsson
added a commit
that referenced
this issue
Oct 16, 2018
The call to the outlier filter has been moved to the kalman filter. The outlier filter is based on the estimated distance from the current estimated position to the candidate sample. A number of counters are used to keep track of how many samples that are within certain error levels. The counters in turn are used to determine which error level to accept and if a sample should be passed on to the kalman filter or not.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The current outlier detector is super simple and identifies outliers based on anchor positions only which lets almost everything through. In a small system this is OK since the anchors are close to each other. In a larger system this approach will accept too large errors and the estimated position will be affected.
An improved outlier filter could use the estimated position to determine if a new tdoa sample is close enough to be used. Care has to be taken to not discard too many samples which will affect the estimated position, and thus the outlier filter. There should also be functionality to "open" the filter up completely when the kalman filter has not converged, to make it possible to recover the state.
The text was updated successfully, but these errors were encountered: