A branching particle system approximation for nonlinear stochastic filtering (Q370932)

Numerical methods for filtering problems have been studied extensively, although much of the work has been done under the assumption that the observation noise is independent of the signal. A lot of different approximation methods (of the signal process, for the filtering equation etc.) are developed. In this paper, the optimal filter of a stochastic signal is approximated by a sequence of measure-valued processes defined by branching particle systems in a random environment. The location and weight of each particle are governed by stochastic differential equations driven by the observation process, which is common for all particles, as well as by an individual Brownian motion, which applies to this specific particle only. The branching mechanism of each particle depends on the observation process and the part of this particle itself during its short lifetime \(n^{-2\alpha},\) where \(n\) is the number of initial particles and \(\alpha\) is a fixed parameter to be optimized. The uniform convergence (as \(n\to\infty\)) during any fixed finite time interval is proved. The main contribution of this paper is that the approximation does not involve any stochastic integral, which makes the numerical implementation readily available.