The Bounded Influence Filters for Dynamic-Measurement Systems

How to overcome the sensitivity of the Kalman filters to outliers in samples? Much attention is paid to, for this question, in resistance analysis of state estimator. Based on an innovation process, a quite large filter class, which is called W-filter class, is reasonably constructed in this paper. In order to control breakdown influ-ence of outliers to state filtering, we mianly deal with a bounded-influence subset, and find out the optimal φ-functions, under some constraints about the bound of influences.