On the design of a stable adaptive filter for state estimation in high dimensional systems (Q5930058)

The paper studies reduced order filters for high dimensional systems when some parameters of state and observation noise statistics are unknown. For plants satisfying detectability assumptions a stable adaptive filter is proposed whose gain is chosen on the basis of a suitable stabilizing structure and with parameters estimated directly from the observations by imposing the minimization of the prediction error. An application connected with the analysis of oceanographic data is finally illustrated.