I'm looking for some general techniques about supervisor for Kalman filter (Extended filter, unscented filter, cubature kalman filter and so on). Is there any general approach to make sure that Kalman state estimator is working in a proper way?
As I know, there is chi-square test for it that works as follows:
- Compute Mahalanobis distance from innovation and covariation matrix
- Check is this distance more than Chi-square criteria for defined alpha
- If it's so, than do some stuff (reject this measurement, make an adaptive step for robustness improvement and etc, or just say that filter is failed)
- If Mahalanobis distance is less than Chi-square criteria, then go for further steps of Kalman filtering.
Is there any another approach not based on Mahalanobis distance? Any general pipeline to check consistency of the estimation?
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix