Abstract

Sequential Monte Carlo particle filters (PFs) are useful for estimating nonlinear non-Gaussian dynamic system parameters. As these algorithms are recursive, their real-time implementation can be computationally complex. In this paper, we analyze the bottlenecks in existing parallel PF algorithms, and propose a new approach that integrates parallel PFs with independent Metropolis-Hastings (PPF-IMH) resampling algorithms to improve root mean-squared esti-mation error (RMSE) performance. We implement the new PPF-IMH algorithm on a Xilinx Virtex-5 field programmable gate array (FPGA) platform. For a one-dimensional problem with 1,000 particles, the PPF-IMH architecture with four processing elements uses less than 5% of a Virtex-5 FPGA's resource and takes 5.85 μs for one iteration. We also incorporate waveform-agile tracking techniques into the PPF-IMH algorithm. We demonstrate a significant performance improvement when the waveform is adaptively designed at each time step with 6.84 μs FPGA processing time per iteration.

Original languageEnglish (US)
Pages (from-to)211-227
Number of pages17
JournalJournal of Signal Processing Systems
Volume65
Issue number2
DOIs
StatePublished - 2011

Fingerprint

Particle Filtering
Parallel Implementation
Waveform
Sensing
Particle Filter
Field Programmable Gate Array
Metropolis-Hastings
Field programmable gate arrays (FPGA)
Metropolis-Hastings Algorithm
Sequential Monte Carlo
Iteration
Array processing
Resampling
Dynamic Systems
Dynamical systems
Integrate
Roots
Real-time
Resources
Processing

Keywords

  • Field programmable gate array
  • Parallel architecture
  • Particle filter
  • Target tracking
  • Waveform-agile sensing

ASJC Scopus subject areas

  • Hardware and Architecture
  • Information Systems
  • Signal Processing
  • Theoretical Computer Science
  • Control and Systems Engineering
  • Modeling and Simulation

Cite this

Algorithm and parallel implementation of particle filtering and its Use in waveform-agile sensing. / Miao, Lifeng; Zhang, Jun Jason; Chakrabarti, Chaitali; Papandreou-Suppappola, Antonia.

In: Journal of Signal Processing Systems, Vol. 65, No. 2, 2011, p. 211-227.

Research output: Contribution to journalArticle

@article{b312d05886d8441984a1bdb07424a13a,
title = "Algorithm and parallel implementation of particle filtering and its Use in waveform-agile sensing",
abstract = "Sequential Monte Carlo particle filters (PFs) are useful for estimating nonlinear non-Gaussian dynamic system parameters. As these algorithms are recursive, their real-time implementation can be computationally complex. In this paper, we analyze the bottlenecks in existing parallel PF algorithms, and propose a new approach that integrates parallel PFs with independent Metropolis-Hastings (PPF-IMH) resampling algorithms to improve root mean-squared esti-mation error (RMSE) performance. We implement the new PPF-IMH algorithm on a Xilinx Virtex-5 field programmable gate array (FPGA) platform. For a one-dimensional problem with 1,000 particles, the PPF-IMH architecture with four processing elements uses less than 5{\%} of a Virtex-5 FPGA's resource and takes 5.85 μs for one iteration. We also incorporate waveform-agile tracking techniques into the PPF-IMH algorithm. We demonstrate a significant performance improvement when the waveform is adaptively designed at each time step with 6.84 μs FPGA processing time per iteration.",
keywords = "Field programmable gate array, Parallel architecture, Particle filter, Target tracking, Waveform-agile sensing",
author = "Lifeng Miao and Zhang, {Jun Jason} and Chaitali Chakrabarti and Antonia Papandreou-Suppappola",
year = "2011",
doi = "10.1007/s11265-011-0601-2",
language = "English (US)",
volume = "65",
pages = "211--227",
journal = "Journal of Signal Processing Systems",
issn = "1939-8018",
publisher = "Springer New York",
number = "2",

}

TY - JOUR

T1 - Algorithm and parallel implementation of particle filtering and its Use in waveform-agile sensing

AU - Miao, Lifeng

AU - Zhang, Jun Jason

AU - Chakrabarti, Chaitali

AU - Papandreou-Suppappola, Antonia

PY - 2011

Y1 - 2011

N2 - Sequential Monte Carlo particle filters (PFs) are useful for estimating nonlinear non-Gaussian dynamic system parameters. As these algorithms are recursive, their real-time implementation can be computationally complex. In this paper, we analyze the bottlenecks in existing parallel PF algorithms, and propose a new approach that integrates parallel PFs with independent Metropolis-Hastings (PPF-IMH) resampling algorithms to improve root mean-squared esti-mation error (RMSE) performance. We implement the new PPF-IMH algorithm on a Xilinx Virtex-5 field programmable gate array (FPGA) platform. For a one-dimensional problem with 1,000 particles, the PPF-IMH architecture with four processing elements uses less than 5% of a Virtex-5 FPGA's resource and takes 5.85 μs for one iteration. We also incorporate waveform-agile tracking techniques into the PPF-IMH algorithm. We demonstrate a significant performance improvement when the waveform is adaptively designed at each time step with 6.84 μs FPGA processing time per iteration.

AB - Sequential Monte Carlo particle filters (PFs) are useful for estimating nonlinear non-Gaussian dynamic system parameters. As these algorithms are recursive, their real-time implementation can be computationally complex. In this paper, we analyze the bottlenecks in existing parallel PF algorithms, and propose a new approach that integrates parallel PFs with independent Metropolis-Hastings (PPF-IMH) resampling algorithms to improve root mean-squared esti-mation error (RMSE) performance. We implement the new PPF-IMH algorithm on a Xilinx Virtex-5 field programmable gate array (FPGA) platform. For a one-dimensional problem with 1,000 particles, the PPF-IMH architecture with four processing elements uses less than 5% of a Virtex-5 FPGA's resource and takes 5.85 μs for one iteration. We also incorporate waveform-agile tracking techniques into the PPF-IMH algorithm. We demonstrate a significant performance improvement when the waveform is adaptively designed at each time step with 6.84 μs FPGA processing time per iteration.

KW - Field programmable gate array

KW - Parallel architecture

KW - Particle filter

KW - Target tracking

KW - Waveform-agile sensing

UR - http://www.scopus.com/inward/record.url?scp=84905277674&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84905277674&partnerID=8YFLogxK

U2 - 10.1007/s11265-011-0601-2

DO - 10.1007/s11265-011-0601-2

M3 - Article

AN - SCOPUS:84905277674

VL - 65

SP - 211

EP - 227

JO - Journal of Signal Processing Systems

JF - Journal of Signal Processing Systems

SN - 1939-8018

IS - 2

ER -