Optimal Transportation Methods in Nonlinear Filtering: The feedback particle filter

TL;DR

This paper situates the feedback particle filter within optimal transportation theory, showing how it improves nonlinear filtering by using optimal couplings and feedback control, leading to better performance than traditional particle filters.

Contribution

It introduces a coupling-based framework for particle filters, extending optimal transportation concepts to derive and analyze the feedback particle filter algorithm.

Findings

Feedback particle filter employs optimal couplings for improved filtering.

The FPF algorithm benefits from feedback control, enhancing performance.

Comparison shows FPF outperforms conventional particle filters.

Abstract

Feedback particle filter (FPF) is a Monte-Carlo (MC) algorithm to approximate the solution of a stochastic filtering problem. In contrast to conventional particle filters, the Bayesian update step in FPF is implemented via a mean-field type feedback control law. The objective for this paper is to situate the development of FPF and related controlled interacting particle system algorithms within the framework of optimal transportation theory. Starting from the simplest setting of the Bayes' update formula, a coupling viewpoint is introduced to construct particle filters. It is shown that the conventional importance sampling resampling particle filter implements an independent coupling. Design of optimal couplings is introduced first for the simple Gaussian settings and subsequently extended to derive the FPF algorithm. The final half of the paper provides a review of some of the salient aspects of the FPF algorithm including the feedback structure, algorithms for gain function design, and comparison with conventional particle filters. The comparison serves to illustrate the benefit of feedback in particle filtering.