Dimensional reduction in nonlinear filtering: a homogenization approach (Q389065)

At first, the authors mention that due to dimensionality issues and computational complexities that arise in representing the signal density using a high number of particles, the problem of particle filtering in high dimensions is still not completely resolved. In order to solve this problem, the present paper gives the theoretical basis for a novel particle filtering method that combines the homogenization with filtering techniques. The presented results are set within the context of slow-fast dynamical systems, where the rate of change of different variables differ by the order of magnitude. The paper addresses the effects of the multiscale signal and observation processes via the study of the Zakai equation. A formal asymptotic expansion of the multiscale Zakai equation is provided. In general, the paper provides rigorous mathematical results that support the numerical algorithms based on the idea that stochastically averaged models provide results helpful in developing inexpensive lower dimensional filtering.