Smoothed Particle Filtering for Dynamic Bayesian Networks

Particle filtering (PF) for dynamic Bayesian networks (DBNs) with discrete-state spaces includes a resampling step which concentrates samples according to their relative weight in regions of interest of the state-space. We propose a more systematic approach than resampling based on regularisation (smoothing) of the empirical distribution associated with the samples, using the kernel method. We show in our experiments that the smoothed particle filtering (SPF) leads to more accurate estimates than the PF.

• Type: Conference
• Year: 2006
• Where: ECAI
• Authors: Theodore Charitos
• Keywords: Artificial Intelligence, Dynamic Bayesian Networks, ECAI 2006, Particle Filtering, Resampling