Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion
The main objective of this research is to reduce the chain length of Markov Chain Monte Carlo (MCMC) to track maneuvering vehicles that undergoes overlapping situation. Overlapping situation will cause to the lost of observable vehicle information whereas maneuvering situation will give varying vehi...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English |
Published: |
2013
|
Subjects: | |
Online Access: | https://eprints.ums.edu.my/id/eprint/11546/1/mt0000000625.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
my-ums-ep.11546 |
---|---|
record_format |
uketd_dc |
spelling |
my-ums-ep.115462017-11-07T06:56:23Z Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion 2013 Kow, Wei Yeang QC Physics The main objective of this research is to reduce the chain length of Markov Chain Monte Carlo (MCMC) to track maneuvering vehicles that undergoes overlapping situation. Overlapping situation will cause to the lost of observable vehicle information whereas maneuvering situation will give varying vehicle outlook which increases the tracking difficulties. MCMC is capable of tracking objects under various conditions by estimating the position of the target object with the sampling of probability distributions. The computations of MCMC are highly depending on the sampling process where the tracking error will be escalated if the sampling is not computed accurately. As a result, conventional MCMC with fixed chain length is facing difficulties to determine the appropriate length to accurately track the target vehicle undergoing various situations. Thus, convergence diagnostic algorithm is embedded into MCMC to quantitatively and qualitatively determine the steady state of MCMC samples. In addition, introduction of genetic operators to the adapted MCMC has further reduced the chain length by improving the convergence speed of the MCMC. Evaluation and assessment of the MCMC tracking algorithm have been carried out under multiple overlapping and maneuvering situations. Implementation results have shown that the qualitative and quantitative convergence diagnostic algorithm had successfully reduced the MCMC computational time by 58.24% and 67.14% respectively compare to fixed chain length MCMC. Subsequently, the implementation of genetic operator has further reduced the computational time of the qualitative and quantitative adaptive MCMC at 36.16% and 13.57% respectively. Thus, the reduction of variances between the MCMC samples by the genetic operator has successfully improved the convergence speed of MCMC which reduced the computational time for tracking the target vehicle accurately under overlapping and maneuvering situations. 2013 Thesis https://eprints.ums.edu.my/id/eprint/11546/ https://eprints.ums.edu.my/id/eprint/11546/1/mt0000000625.pdf text en public masters Universiti Malaysia Sabah School of Engineering and Information Technology |
institution |
Universiti Malaysia Sabah |
collection |
UMS Institutional Repository |
language |
English |
topic |
QC Physics |
spellingShingle |
QC Physics Kow, Wei Yeang Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
description |
The main objective of this research is to reduce the chain length of Markov Chain Monte Carlo (MCMC) to track maneuvering vehicles that undergoes overlapping situation. Overlapping situation will cause to the lost of observable vehicle information whereas maneuvering situation will give varying vehicle outlook which increases the tracking difficulties. MCMC is capable of tracking objects under
various conditions by estimating the position of the target object with the sampling of probability distributions. The computations of MCMC are highly depending on the
sampling process where the tracking error will be escalated if the sampling is not computed accurately. As a result, conventional MCMC with fixed chain length is facing difficulties to determine the appropriate length to accurately track the target vehicle undergoing various situations. Thus, convergence diagnostic algorithm is
embedded into MCMC to quantitatively and qualitatively determine the steady state of MCMC samples. In addition, introduction of genetic operators to the adapted MCMC has further reduced the chain length by improving the convergence speed of the MCMC. Evaluation and assessment of the MCMC tracking algorithm have been carried out under multiple overlapping and maneuvering situations. Implementation results have shown that the qualitative and quantitative convergence diagnostic algorithm had successfully reduced the MCMC computational time by 58.24% and 67.14% respectively compare to fixed chain length MCMC. Subsequently, the implementation of genetic operator has further reduced the computational time of the qualitative and quantitative adaptive MCMC at 36.16% and 13.57% respectively. Thus, the reduction of variances between the MCMC samples by the genetic operator has successfully improved the convergence speed of MCMC which reduced
the computational time for tracking the target vehicle accurately under overlapping and maneuvering situations. |
format |
Thesis |
qualification_level |
Master's degree |
author |
Kow, Wei Yeang |
author_facet |
Kow, Wei Yeang |
author_sort |
Kow, Wei Yeang |
title |
Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
title_short |
Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
title_full |
Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
title_fullStr |
Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
title_full_unstemmed |
Adaptive Markov Chain Monte Carlo techniques to estimate vehicle motion |
title_sort |
adaptive markov chain monte carlo techniques to estimate vehicle motion |
granting_institution |
Universiti Malaysia Sabah |
granting_department |
School of Engineering and Information Technology |
publishDate |
2013 |
url |
https://eprints.ums.edu.my/id/eprint/11546/1/mt0000000625.pdf |
_version_ |
1747836420386455552 |