When is preview beneficial?
Maxim Kristalny, Leonid Mirkin · 2015
We analyze the H2 performance of the fixed-lag smoothing problem when the measurement noise intensity is a function of the smoothing lag (preview window). We derive computable necessary and sufficient conditions on the rate of the measurement noise intensity growth as a function of the smoothing lag, under which minuscule preview improves the estimation performance. A sufficient condition in terms of the spectrum of the associated Kalman filter are also derived.