Parallel Partitioning Estimation
Dominick Andrisani, Ching-Fu Gau · 1984
The estimation algorithm described in this paper solves the linear estimation problem as a two stage (or multistage) estimator. The first stage is a Kalman filter initialized with one set of initial conditions and process noise intensity. The residuals or innovations of this estimator become the measurements for the second stage Kalman filter estimator which has different initial conditions and process noise intensity. The interconnections between this estimator structure and the more familiar one stage optimal Kalman filter are discussed. Applications to decentralized estimation, bias estimation, and parameter identification are described.