Multi-sensing Data Fusion: Target tracking via particle filtering
 |
 |
 |
@mastersthesis{controparticlefiltering2018,
abstract = {In this thesis Multisensing Data Fusion is firstly introduced, with a focus on perception and the concepts that are the base of this work, like the mathematical tools that make it possible. Particle filters are one class of these tools that allow a computer to perform fusion of numerical information that is perceived from real environment by sensors. For this reason they are described and state of the art mathematical formulas and algorithms for particle filtering are also presented. At the core of this project, a simple piece of software has been developed in order to test these tools in practice. More specifically, a Target Tracking Simulator software is presented where a virtual trackable object can freely move in a 2-dimensional simulated environment and distributed sensor agents, dispersed in the same environment, should be able to perceive the object through a state-dependent measurement affected by additive Gaussian noise. Each sensor employs particle filtering along with communication with other neighboring sensors in order to update the perceived state of the object and track it as it moves in the environment.},
amseprint = {16835},
amspractice = {1249361},
author = {Contro, Alessandro},
available = {2018-04-01},
cosupervisor = {Ciatto, Giovanni},
cycle = {LM},
description = {Several architectured & techniques have been proposed into the litterature to face the problem integrating data coming from heterogeneous sources. Among the manies, the particle filter (PF) \[1\] tecnique is of particular interest since it provides a probabilistic way to estimate or track the state of a dynamic system during the system functioning, i.e. as soon as observations become available. Variants of the PF moving towards distributed approaches have been proposed too \[2\], liying under the umbrella of Distributed Particle Filters (DPFs). The general idea is to make several distriuted computational entities -- possibly observing the system from different perspectives -- cooperate to achieve a common view of the current system state and its evolution. The goal of this thesis is to understand if DPF and MASs may take advantage of each other by creating a bridge between the two worlds. 1.1 References: \[1\] Doucet, A., & Johansen, A. M. (2011). A Tutorial on Particle filtering and smoothing: Fiteen years later. The Oxford Handbook of Nonlinear Filtering, (December 2008), 656--705. https://doi.org/10.1.1.157.772 \[2\] Hlinka, O., Hlawatsch, F., & Djuric, P. M. (2013). Distributed Particle Filtering in Agent Networks: A Survey, Classification, and Comparison. IEEE Signal Process. Mag., 30(1), 61--81. https://doi.org/10.1109/MSP.2012.2219652},
language = {en},
month = oct,
start = {2018-04-20},
supervisor = {Omicini, Andrea},
title = {Multi-sensing Data Fusion: Target tracking via particle filtering},
type = {Master's thesis},
year = 2018
}
abstract = {In this thesis Multisensing Data Fusion is firstly introduced, with a focus on perception and the concepts that are the base of this work, like the mathematical tools that make it possible. Particle filters are one class of these tools that allow a computer to perform fusion of numerical information that is perceived from real environment by sensors. For this reason they are described and state of the art mathematical formulas and algorithms for particle filtering are also presented. At the core of this project, a simple piece of software has been developed in order to test these tools in practice. More specifically, a Target Tracking Simulator software is presented where a virtual trackable object can freely move in a 2-dimensional simulated environment and distributed sensor agents, dispersed in the same environment, should be able to perceive the object through a state-dependent measurement affected by additive Gaussian noise. Each sensor employs particle filtering along with communication with other neighboring sensors in order to update the perceived state of the object and track it as it moves in the environment.},
amseprint = {16835},
amspractice = {1249361},
author = {Contro, Alessandro},
available = {2018-04-01},
cosupervisor = {Ciatto, Giovanni},
cycle = {LM},
description = {Several architectured & techniques have been proposed into the litterature to face the problem integrating data coming from heterogeneous sources. Among the manies, the particle filter (PF) \[1\] tecnique is of particular interest since it provides a probabilistic way to estimate or track the state of a dynamic system during the system functioning, i.e. as soon as observations become available. Variants of the PF moving towards distributed approaches have been proposed too \[2\], liying under the umbrella of Distributed Particle Filters (DPFs). The general idea is to make several distriuted computational entities -- possibly observing the system from different perspectives -- cooperate to achieve a common view of the current system state and its evolution. The goal of this thesis is to understand if DPF and MASs may take advantage of each other by creating a bridge between the two worlds. 1.1 References: \[1\] Doucet, A., & Johansen, A. M. (2011). A Tutorial on Particle filtering and smoothing: Fiteen years later. The Oxford Handbook of Nonlinear Filtering, (December 2008), 656--705. https://doi.org/10.1.1.157.772 \[2\] Hlinka, O., Hlawatsch, F., & Djuric, P. M. (2013). Distributed Particle Filtering in Agent Networks: A Survey, Classification, and Comparison. IEEE Signal Process. Mag., 30(1), 61--81. https://doi.org/10.1109/MSP.2012.2219652},
language = {en},
month = oct,
start = {2018-04-20},
supervisor = {Omicini, Andrea},
title = {Multi-sensing Data Fusion: Target tracking via particle filtering},
type = {Master's thesis},
year = 2018
}
