Outliers in bound-based state estimation and identification

In the observation-based computation of bounds on the parameters or state of a model with specified bounds on output error or process and observation respectively, outliers may be defined as observations giving rise to clashes of the parameter or state bounds. Possible ways to detect and locate outliers in this context are considered. Static and dynamic problems are distinguished. Upper bounds, independent of the record length, on the numbers of isolated outliers or k-member outlier sets are given. Two possible approaches for locating outliers in the dynamic case are suggested. An efficient algorithm is presented for locating parameter or state bounds responsible for clashes