Approximation and Limit Results for Nonlinear Filters Over an Infinite Time Interval (Q4699155)

The authors consider the signal process \(X \)which takes values in a compact subset of \(\mathbb{R}^{r}\) and satisfies the Ito equation \[ dX=a(X)dt+\sigma (X)dW, \] and the observation process \(Y,\) defined by \[ Y(t)=\int_{0}^{t}g(X(s))ds+B(t), \] where \(W\) and \(B\) are independent standard vector-valued Wiener processes. The cost of interest is the pathwise error per unit time over a long time interval. The authors construct the optimal filter for a process which suitably approximates \(X\) and for which the optimal filter is feasible, using the actual observations. They show that the limit of the pathwise average cost for the approximating filter is what one would get if one used the true optimal filter.