Performance evaluation for particle filters

Remi Chou Yvo Boers Martin Podt Matthieu Geist 1
1 IMS - Equipe Information, Multimodalité et Signal
UMI2958 - Georgia Tech - CNRS [Metz], SUPELEC-Campus Metz
Abstract : Performance evaluation in particle filtering problems is commonly performed via point estimator comparison. However, in non-Gaussian cases, this can be not always meaningful and entire particle clouds need to be compared. The Kullback-Leibler divergence (KLD) can be used for such a particle cloud comparison. In contrast to KLD estimates commonly used in particle filtering applications, we present an estimator of the KLD being applicable to any cloud of particles. This estimator is applied to a performance evaluation scheme generally relevant to any particle filter, of which abilities are equal to no other known scheme in the literature. Through simulations and concrete examples, we will show that it is suitable to practically compare particle clouds, which have a limited number of particles, have a different size, are close to each other and have an high dimensionality.
Document type :
Conference papers
Complete list of metadatas

https://hal-supelec.archives-ouvertes.fr/hal-00652168
Contributor : Sébastien van Luchene <>
Submitted on : Thursday, December 15, 2011 - 9:01:01 AM
Last modification on : Wednesday, July 31, 2019 - 4:18:03 PM

Identifiers

  • HAL Id : hal-00652168, version 1

Collections

Citation

Remi Chou, Yvo Boers, Martin Podt, Matthieu Geist. Performance evaluation for particle filters. FUSION 2011, Jul 2011, Chicago, United States. pp.1-7. ⟨hal-00652168⟩

Share

Metrics

Record views

96