Partitioned linear estimation algorithms: Discrete case
DEMETRIOS G. LAINIOTIS · IEEE Transactions on Automatic Control · 1975
Using the "partitioning" approach to estimation, fundamentally new, robust, computationally effective and fast filtering and smoothing algorithms have been obtained. The new algorithms are given in explicit expressions of a partitioned nature in terms of decoupled forward filters. The "patitioned" algorithms are especially advantageous both from a computational as well as an analysis standpoint. They are the natural framework for studying observability, controllability, unbiasedness, and especially in deriving robust, fast, and effective numerical algorithms for Riccati equations.