A Projection-Based Rao-Blackwellized Particle Filter to Estimate Parameters in Conditionally Conjugate State-Space Models

摘要

Particle filters constitute today a well-established class of techniques for state filtering in non-linear state-space models. However, online estimation of static parameters under the same framework represents a difficult problem. The solution can be found to some extent within a category of state-space models allowing us to perform parameter estimation in an analytically tractable manner, while still considering non-linearities in data evolution equations. Nevertheless, the well-known particle path degeneracy problem complicates the computation of the statistics that are required to estimate the parameters. The present paper proposes a simple and efficient method which is experimentally shown to suffer less from this issue.