A Stroock Varadhan support theorem in non-linear filtering theory (Q908579)

In this interesting paper the support of the law of the unnormalized filter is shown to satisfy a similar result to that of Stroock-Varadhan for diffusion processes. Suppose \((\rho_ t\phi)\), \(0\leq t\leq 1\), is the unnormalized filter of the \(C^ 2\)-function \(\phi (x_ t)\), where \(x_ t\) is the signal process observed via a, possibly correlated, observation process \(y_ t\). Then \(\rho\) \(\cdot \phi\) is considered as a stochastic partial differential equation with values in \({\mathcal C}(R^ n,R)\). The support of the law of \(\rho\) \(\cdot \phi\) is shown to be the closure in the Fréchet space \({\mathcal C}([0,1]\), \({\mathcal C}(R^ n,R))\) of the set of trajectories obtained from the Zakai equation by replacing the Stratonovich differential \(\circ dy\) by an \(H^ 1\) control.