New design of estimators using covariance information based on an innovations approach (Q1180389)

This paper presents a filter and fixed-point smoother using covariance information. The design is based on the innovations approach. The filter is optimal in the linear least-squares estimation sense. The fixed-point smoother is also optimal for white Gaussian observation noise but suboptimal for white Gaussian plus coloured observation noise. It is shown that the proposed optimal fixed-point smoother for white Gaussian observation noise and the suboptimal fixed-point smoother for white Gaussian plus coloured noise are feasible.