Comparison of Different Approaches to the Nearly Gaussian Filtering
Mehdi NOURI-SHIRAZI, Nobuo Sannomiya, Yoshikazu Nishikawa · Transactions of the Society of Instrument and Control Engineers · 1987
In this paper we consider the state estimation problem of a discretetime linear stochastic system for a fairly general case where both the system disturbance and the observation noise are nearly Gaussian. First, the techniques developed hitherto are introduced for special cases where either the system disturbance or the observation noise is nearly Gaussian while the other one is Gaussian. Secondly, two possible filtering approaches are proposed for the general case by extending the filters constructed for the special cases. The performances of various approximate filters are compared with that of the Gaussian sum filter through Monte-Carlo simulations. A discussion is made on reducing on-line computational difficulties in the nearly Gaussian filtering.