This chapter covers the particle filter, which handles severe nonlinearity as well as non‐Gaussianity. Particle filter deploys the sequential Monte Carlo method as a numerical approximation scheme to approximate the corresponding distributions by a set of particles, which are random samples. These samples are drawn from a proposal or importance density, which has the same support as the distribution of interest. A set of normalized weights is associated with the set of particles. The prior, the likelihood, or a Gaussian approximation of the posterior, which is provided by the extended Kalman filter or the unscented Kalman filter, can be selected as the proposal distribution. Resampling is used to address the degeneracy problem in the sequential importance sampling. To cope with the sample impoverishment problem due to resampling, regularization and resample‐move algorithm can be deployed. Simultaneous localization and mapping is reviewed as an application of particle filtering algorithms.
Particle Filter
Nonlinear Filters ; 71-83
2022-04-12
13 pages
Article/Chapter (Book)
Electronic Resource
English
Gaussian Diffracted Particle Filter
British Library Online Contents | 2010
|Quadrature Kalman Particle Filter
British Library Online Contents | 2009
|Distributed Particle Filter Based on Particle Exchanges
IEEE | 2023
|European Patent Office | 2021
|