ON THE OPTIMALITY OF TWO-STAGE KALMAN FILTERING FOR SYSTEMS WITH UNKNOWN INPUTS

This paper is concerned with the optimal solution of two-stage Kalman filtering for linear discrete-time stochastic time-varying systems with unknown inputs affecting both the system state and the outputs. By means of a newly-presented modified unbiased minimum-variance filter (MUMVF), which appears to be the optimal solution to the addressed problem, the optimality of two-stage Kalman filtering for systems with unknown inputs is defined and explored. Two extended versions of the previously proposed robust two-stage Kalman filter (RTSKF), augmented-unknown-input RTSKF (ARTSKF) and decoupled-unknown-input RTSKF (DRTSKF), are presented to solve the general unknown input filtering problem. It is shown that under less restricted conditions, the proposed ARTSKF and DRTSKF are equivalent to the corresponding MUMVFs. An example is given to illustrate the usefulness of the proposed results.