Nonlinear Filtering and Control of a Switching Diffusion with Small Observation Noise (Q4210182)

The paper considers the problem of nonlinear filtering and control of a switching diffusion \(x(t)\) whose evolution is described by \[ \begin{aligned} dx(t)&= b(t,\alpha(t),x(t),u(t))dt+\sigma(t,\alpha(t))dW(t),\\ dy(t) &= h(t,\alpha(t),x(t))dt+\varepsilon dV(t),\quad 0<t<T. \end{aligned} \] Here \((\alpha(t))\) is an unobservable Markov chain, \(W(t)\) and \(V(t)\) are Wiener processes. The author proposes two methods for designing an approximate (with small \(\varepsilon\)) filter of the process \(x(t)\) based on two estimation methods of \((\alpha(t))\). These filters are asymptotically optimal as the observation noise parameter \(\varepsilon\) vanishes. The filters are used to construct an asymptotically optimal control.