Adapting Particle Filter Algorithms to the GPU Architecture

Master thesis (2011)

Authors

Contributors

A. Van Amesfoort (mentor)
A. Simonetto (mentor)
T. Kevizcky (mentor)
H.J. Sips (mentor)
Programme
Parallel and Distributed Systems (EEMCS) (TU Delft)
Copyright: © 2011 Chitchian, M.M.
Particle filter CUDA GPU Real-time estimation GPGPU Bayesian estimation Bayes filter Non-linear estimation Distributed particle filter Parallel particle filter
More Info
expand_more

Published Date

20-05-2011

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Programme

Parallel and Distributed Systems

Abstract

The particle filter is a Bayesian estimation technique based on Monte Carlo simulations. The non-parametric nature of particle filters makes them ideal for non-linear non-Gaussian systems. This greater filtering accuracy, however, comes at the price of increased computational complexity which limits their practical use for real-time applications. This thesis presents an attempt to enable real-time particle filtering for complex estimation problems using modern GPU hardware. We propose a GPU-based generic particle filtering framework which can be applied to various estimation problems. We implement a real-time estimation application using this particle filtering framework and measure the estimation error with different filter parameters. Furthermore, we present an in-depth performance analysis of our GPU implementation followed by a number of optimisations in order to increase implementation efficiency.