Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks

By Arnaud Doucet, Nando de Freitas, Kevin P. Murphy, Stuart J. Russell. Conference on Uncertainty in Artificial Intelligence, 2000

1,530 citations recorded by the source corpus.

Particle filters (PFs) are powerful sampling-based inference/learning algorithms for dynamic Bayesian networks (DBNs). They allow us to treat, in a principled way, any type of probability distribution, nonlinearity and non-stationarity. They have appeared in several fields under such names as "condensation", "sequential Monte Carlo" and "survival of the fittest". In this paper, we show how we can exploit the structure of the DBN to increase the efficiency of particle filtering, using a technique known as Rao-Blackwellisation. Essentially, this samples some of the variables, and marginalizes out the rest exactly, using the Kalman filter, HMM filter, junction tree algorithm, or any other finite dimensional optimal filter. We show that Rao-Blackwellised particle filters (RBPFs) lead to more accurate estimates than standard PFs. We demonstrate RBPFs on two problems, namely non-stationary online regression with radial basis function networks and robot localization and map building. We also discuss other potential application areas and provide references to some finite dimensional optimal filters.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/b9cbf264042dcbb04ac421697620268ccd382727

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Plan Recognition as Planning (2009, methodology): “This difficulty is bridged relying on the well–founded and efficient Rao–Blackwellised particle filter proposed by (Doucet et al., 2000) to perform approximate probabilistic inference over DBNs.”
Towards robotic assistants in nursing homes: Challenges and results (2003, methodology): “First, the robot path Ü Ø is estimated using a particle filter, as in the Monte Carlo localization (MCL) algorithm for mobile robot localization (7).”
Bayesian Nonparametric Inference of Switching Dynamic Linear Models (2010, methodology): “One could leverage the conditionally linear dynamics and use Rao-Blackwellized particle filtering (RBPF) [48].”
Unifying Map and Landmark Based Representations for Visual Navigation (2017, background): “Methods like particle ﬁltering maintain a probability distribution over the robot’s location and update this distribution based on a generative model of the sensors [8].”
Improved Techniques for Grid Mapping With Rao-Blackwellized Particle Filters (2007, background): “In the literature, the mobile robot mapping problem is often referred to as the simultaneous localization and mapping (SLAM) problem [4, 6 , 9, 15, 16, 26, 29, 32, 39].”
GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models (2008, methodology): “This is mitigated somewhat by using Rao-Blackwellised PFs ( Doucet et al. 2000 ).”
A framework for learning semantic maps from grounded natural language descriptions (2014, methodology): “…perform resampling based upon the effective number of particles, as proposed by Liu (1996), Neff = 1n∑ i=0 w2i (23) When the effective number of particles Neff falls below the threshold N/2, where N is the number of particles, we resample using the algorithm described by Doucet et al. (2000).”
Particle methods for change detection, system identification, and control (2004, background): “The same idea can be used for dynamic Bayesian networks with a large state space [13].”
Learning models for following natural language directions in unknown environments (2015, methodology): “When the particle weights fall below a threshold, we resample particles to avoid particle depletion [18].”