Performance analysis of parallel/distributed particle filters

Zhang, Xudong; Mohamed, Ali Wagdy; Nguyen, Linda D.; Gu, Feng

doi:10.1145/3213187.3213192

Cited by 5 publications

(6 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Parallel or distributed particle filters distribute particles among different processing units. How to distribute (route) particles and how much is gained for various routing policies is investigated in [ 64 ]. The work of [ 65 ] investigates using multiple local particle filters, that are correlated and low dimensional, in high dimensional problems using the phenomenon referred to as ’decay of correlations‘.…”

Section: Solutions To the Particle Filter Challengesmentioning

confidence: 99%

“…Computational complexity of algorithms handling OOSMs can often be reduced at the expense of additional assumptions. The work of [ 71 ] offers a more efficient alternative to the optimal solution of [ 64 ] by assuming a Gaussian posterior. In [ 72 ], mixed linear/nonlinear state-space models are assumed such that Rao–Blackwellization can be used.…”

Section: Solutions To the Particle Filter Challengesmentioning

confidence: 99%

See 1 more Smart Citation

Particle Filters: A Hands-On Tutorial

Elfring

Torta

Molengraft

2021

Sensors

View full text Add to dashboard Cite

The particle filter was popularized in the early 1990s and has been used for solving estimation problems ever since. The standard algorithm can be understood and implemented with limited effort due to the widespread availability of tutorial material and code examples. Extensive research has advanced the standard particle filter algorithm to improve its performance and applicability in various ways in the years after. As a result, selecting and implementing an advanced version of the particle filter that goes beyond the standard algorithm and fits a specific estimation problem requires either a thorough understanding or reviewing large amounts of the literature. The latter can be heavily time consuming especially for those with limited hands-on experience. Lack of implementation details in theory-oriented papers complicates this task even further. The goal of this tutorial is facilitating the reader to familiarize themselves with the key concepts of advanced particle filter algorithms and to select and implement the right particle filter for the estimation problem at hand. It acts as a single entry point that provides a theoretical overview of the filter, its assumptions and solutions for various challenges encountered when applying particle filters. Besides that, it includes a running example that demonstrates and implements many of the challenges and solutions.

show abstract

Section: Solutions To the Particle Filter Challengesmentioning

confidence: 99%

Section: Solutions To the Particle Filter Challengesmentioning

confidence: 99%

Particle Filters: A Hands-On Tutorial

Elfring

Torta

Molengraft

2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Particles are sent and received between CU and PUs according to the transferring scheme. Different centralized resampling routing policies and corresponding complexity analysis could be found in (Bai et al 2016;Zhang et al 2018). Unlike the centralized resampling, the decentralized resampling does not have a CU for particle transfers, which results in reducing communication costs, but might be harmful on system convergence and estimation accuracy.…”

Section: Parallel/distributed Particle Filtersmentioning

confidence: 99%

“…However, there exist different strategies about how the extra particles are sent to those PUs with shortage of particles to achieve the load balance in resampling stage. Although effective and efficient particle routing strategies are able to achieve speedups to some extent, they are still suffering from high communication costs (Zhang et al 2018). Some decentralized resampling algorithms are proposed to further improve the performance of parallel/distributed particle filters (Bolic, Djuric, and Hong 2005).…”

Section: Introductionmentioning

confidence: 99%

Adaptive Particle Sampling and Resampling in Parallel/distributed Particle Filters

2019

High Performance Computing (HPC 2019)

View full text Add to dashboard Cite

Particle filters have been widely used in estimating the states of dynamic systems by using Bayesian interference and stochastic sampling techniques. Parallel computing techniques were introduced to improve the performance of sequential particle filters with multiple processing units (PUs). However, the unavoidable communications between Pus, lower the performance. The hybrid and adaptive resampling algorithms were proposed to improve the performance of parallel/distributed particle filters by reducing the communication costs without loss of estimation accuracy. In this paper, we propose an adaptive sampling and resampling technique in particle filters. In the proposed algorithm, the number of particle is dynamically adjustable based on the model convergence. As a result, less particles will be used if the current convergence is good and more particles will be used if the convergence is getting bad. The experimental results show the improved performance by using less particles and reducing the communication cost compared with other algorithms.

show abstract

“…In the centralized resampling, the transfers of weights and particles increase the communication cost and lower the speedup factor. Although some efficient particle routing algorithms [20] were introduced to improve the performance of the centralized resampling, it is still suffering from the high communication cost [22].…”

Section: Introductionmentioning

confidence: 99%

Performance Analysis of Resampling Algorithms of Parallel/Distributed Particle Filters

et al. 2021

Self Cite

View full text Add to dashboard Cite

Particle filters have been widely used in various fields due to their advantages in dealing with non-linear and/or non-Gaussian systems. A large number of particles are needed to guarantee the convergence of particle filters for the state estimation, especially for large-scale complex systems. Therefore, parallel/distributed particle filters were adopted to improve the performance, in which different paradigms of resampling were proposed, including the centralized resampling, the decentralized resampling, and the hybrid resampling. To ease their adoptions, we analyze time consumptions and speedup factors of parallel/distributed particle filters with various resampling algorithms, state sizes, system complexities, numbers of processing units, and model dimensions in this study. The experimental results indicate that the decentralized resampling achieves the highest speedup factors due to the local transfer of particles, the centralized resampling always has the lowest speedup factors because of the global transfer of particles, and the hybrid resampling attains the speedup factors between. Moreover, we define the complexity-state ratio, as the ratio between the system complexity and the system state size to study how it impacts the speedup factor. The experiments show that the higher complexity-state ratio results in the increase of the speedup factors. This is one of the earliest attempts to analyze and compare the performance of parallel/distributed particle filters with different resampling algorithms. The analysis can provide potential solutions for further performance improvements and guide the appropriate selection of the resampling algorithm for parallel/distributed particle filters.

show abstract

Performance analysis of parallel/distributed particle filters

Cited by 5 publications

References 19 publications

Particle Filters: A Hands-On Tutorial

Particle Filters: A Hands-On Tutorial

Adaptive Particle Sampling and Resampling in Parallel/distributed Particle Filters

Performance Analysis of Resampling Algorithms of Parallel/Distributed Particle Filters

Contact Info

Product

Resources

About