Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks

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.

[1]  D. Mayne,et al.  Monte Carlo techniques to estimate the conditional expectation in multi-stage non-linear filtering† , 1969 .

[2]  Alan S. Willsky,et al.  A survey of design methods for failure detection in dynamic systems , 1976, Autom..

[3]  Hiromitsu Kumamoto,et al.  Random sampling approach to state estimation in switching environments , 1977, Autom..

[4]  Jitendra K. Tugnait,et al.  Detection and estimation for abruptly changing systems , 1981, 1981 20th IEEE Conference on Decision and Control including the Symposium on Adaptive Processes.

[5]  B. Anderson,et al.  Optimal Filtering , 1979, IEEE Transactions on Systems, Man, and Cybernetics.

[6]  Alfred V. Aho,et al.  Data Structures and Algorithms , 1983 .

[7]  Brian C. Williams,et al.  Diagnosing Multiple Faults , 1987, Artif. Intell..

[8]  D. Rubin.  Using the SIR algorithm to simulate posterior distributions , 1988 .

[9]  Brian C. Williams,et al.  Diagnosis with Behavioral Modes , 1989, IJCAI.

[10]  Keiji Kanazawa,et al.  A model for reasoning about persistence and causation , 1989 .

[11]  Peter Struss,et al.  "Physical Negation" Integrating Fault Models into the General Diagnostic Engine , 1989, IJCAI.

[12]  Eduardo D. Sontag,et al.  Mathematical Control Theory: Deterministic Finite Dimensional Systems , 1990 .

[13]  A. Hasman,et al.  Probabilistic reasoning in intelligent systems: Networks of plausible inference , 1991 .

[14]  Peter S. Maybeck,et al.  Reconfigurable flight control via multiple model adaptive control methods , 1991 .

[15]  Michael A. West.  Mixture Models, Monte Carlo, Bayesian Updating and Dynamic Models , 1992 .

[16]  Peter Mueller.  Posterior Integration in Dynamic Models , 1992 .

[17]  Leland Stewart,et al.  Use of Bayesian belief networks to fuse continuous and discrete information for target recognition, tracking, and situation assessment , 1992, Defense, Security, and Sensing.

[18]  Alan E. Gelfand,et al.  Bayesian statistics without tears: A sampling-resampling perspective , 1992 .

[19]  Luca Console,et al.  Readings in Model-Based Diagnosis , 1992 .

[20]  N. Gordon,et al.  Novel approach to nonlinear/non-Gaussian Bayesian state estimation , 1993 .

[21]  Jun S. Liu,et al.  Sequential Imputations and Bayesian Missing Data Problems , 1994 .

[22]  Hisashi Tanizaki,et al.  Prediction, filtering and smoothing in non-linear and non-normal cases using Monte Carlo integration , 1994 .

[23]  Stuart J. Russell,et al.  Stochastic simulation algorithms for dynamic probabilistic networks , 1995, UAI.

[24]  P. Pandurang Nayak.  Automated Modelling of Physical Systems , 1995 .

[25]  P. Green.  Reversible jump Markov chain Monte Carlo computation and Bayesian model determination , 1995 .

[26]  G. Kitagawa.  Monte Carlo Filter and Smoother for Non-Gaussian Nonlinear State Space Models , 1996 .

[27]  Craig Boutilier,et al.  Context-Specific Independence in Bayesian Networks , 1996, UAI.

[28]  Michael Isard,et al.  Contour Tracking by Stochastic Propagation of Conditional Density , 1996, ECCV.

[29]  P. Pandurang Nayak,et al.  A Model-Based Approach to Reactive Self-Configuring Systems , 1996, AAAI/IAAI, Vol. 2.

[30]  G. Casella,et al.  Rao-Blackwellisation of sampling schemes , 1996 .

[31]  Jun S. Liu,et al.  Metropolized independent sampling with comparisons to rejection sampling and importance sampling , 1996, Stat. Comput..

[32]  Y. Bar-Shalom,et al.  Multiple-model estimation with variable structure , 1996, IEEE Trans. Autom. Control..

[33]  H. Uhlig.  Bayesian vector autoregressions with stochastic volatility , 1997 .

[34]  Jun S. Liu,et al.  Sequential Monte Carlo methods for dynamic systems , 1997 .

[35]  R. Pons.  Causal or-dering for multiple mode systems , 1997 .

[36]  Michael A. West,et al.  Bayesian forecasting and dynamic models (2nd ed.) , 1997 .

[37]  Simon J. Godsill,et al.  On sequential simulation-based methods for Bayesian filtering , 1998 .

[38]  Brian C. Williams,et al.  Decompositional, Model-based Learning and its Analogy to Diagnosis , 1998, AAAI/IAAI.

[39]  Pierre Del Moral,et al.  Discrete Filtering Using Branching and Interacting Particle Systems , 1998 .

[40]  Jane T. Malin,et al.  INTERACTIVE SIMULATION-BASED TESTING OF PRODUCT GAS TRANSFER INTEGRATED MONITORING AND CONTROL SOFTWARE FOR THE LUNAR MARS LIFE SUPPORT PHASE III TEST , 1998 .

[41]  Xavier Boyen,et al.  Tractable Inference for Complex Stochastic Processes , 1998, UAI.

[42]  Kevin P. Murphy,et al.  Bayesian Map Learning in Dynamic Environments , 1999, NIPS.

[43]  Jie Chen,et al.  Robust Model-Based Fault Diagnosis for Dynamic Systems , 1998, The International Series on Asian Studies in Computer and Information Science.

[44]  Jun S. Liu,et al.  Sequential importance sampling for nonparametric Bayes models: The next generation , 1999 .

[45]  M. Pitt,et al.  Filtering via Simulation: Auxiliary Particle Filters , 1999 .

[46]  Christophe Andrieu,et al.  Sequential Bayesian Estimation And Model Selection Applied To Neural Networks , 1999 .

[47]  P. Vidoni.  Exponential family state space models based on a conjugate latent process , 1999 .

[48]  Reuven Bar-Yehuda,et al.  Random Algorithms for the Loop Cutset Problem , 1999, UAI.

[49]  A. Doucet,et al.  Sequential MCMC for Bayesian model selection , 1999, Proceedings of the IEEE Signal Processing Workshop on Higher-Order Statistics. SPW-HOS '99.

[50]  David J. Spiegelhalter,et al.  Probabilistic Networks and Expert Systems , 1999, Information Science and Statistics.

[51]  Pieter J. Mosterman,et al.  Building Hybrid Observers for Complex Dynamic Systems Using Model Abstractions , 1999, HSCC.

[52]  Michael I. Jordan,et al.  Probabilistic Networks and Expert Systems , 1999 .

[53]  Gomes de Freitas,et al.  Bayesian methods for neural networks , 2000 .

[54]  Simon J. Godsill,et al.  On sequential Monte Carlo sampling methods for Bayesian filtering , 2000, Stat. Comput..

[55]  A. Gehin,et al.  Structural Analysis of System Reconfigurability , 2000 .

[56]  S. Narasimhan,et al.  Efficient diagnosis of hybrid systems using models of the supervisory controller , 2000 .

[57]  Neil J. Gordon,et al.  Editors: Sequential Monte Carlo Methods in Practice , 2001 .

[58]  Feng Zhao,et al.  Distributed Monitoring of Hybrid Systems: A model-directed approach , 2001, IJCAI.

[59]  Nando de Freitas,et al.  Sequential Monte Carlo Methods in Practice , 2001, Statistics for Engineering and Information Science.

[60]  Arnaud Doucet,et al.  Particle filters for state estimation of jump Markov linear systems , 2001, IEEE Trans. Signal Process..

[61]  Visakan Kadirkamanathan,et al.  Particle filtering based likelihood ratio approach to fault diagnosis in nonlinear stochastic systems , 2001, IEEE Trans. Syst. Man Cybern. Part C.

[62]  Brian C. Williams,et al.  Mode Estimation of Probabilistic Hybrid Systems , 2002, HSCC.