Approximate Nonlinear Filtering for a Two-Dimensional Diffusion with One-Dimensional Observations in a Low Noise Channel (Q4442963)

The authors study the problem of estimating the signal \((x^{(1)}_t, x^{(2)}_t)\) satisfying the system \[ \begin{aligned} dx^{(1)}_t &= f_1(x^{(1)}_t, x^{(2)}_t)\,dt,\\ dx^{(2)}_t &= f_2(x^{(1)}_t, x^{(2)}_t)\,dt+ \sigma(x^{(1)}_t, x^{(2)}_t)\,dw_t,\end{aligned} \] when the observation process is modelled by the equation \[ dy_t= h(x^{(1)}_t)\,dt+ \varepsilon d\overline w_t, \] where \(w_t\) and \(\overline w_t\) are standard independent Wiener processes and \(\varepsilon\geq 0\) is a small parameter. Their aim is to work out a rigorous study of an approximation of the optimal filter proposed by \textit{I. Yaesh}, \textit{B. Z. Bobrovsky} and \textit{Z. Schuss} [``Asymptotic analysis of the optimal filtering problem for two-dimensional diffusions measured in a low noise channel'', SIAM J. Appl. Math. 50, No. 4, 1134--1155 (1990; Zbl 0701.93098)].