Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing

Cellular networks have gained importance in today's mobile communication scenario. The deployment of first and second-generation cellular systems has involved extensive planning, both to achieve continuous RF coverage and to ensure the provision of capacity in the right locations. The process h...

Full description

Saved in:
Bibliographic Details
Main Author: Angela, Amphawan
Format: Thesis
Published: 2003
Subjects:
Tags: Add Tag
No Tags, Be the first to tag this record!
id my-mmu-ep.469
record_format uketd_dc
spelling my-mmu-ep.4692010-06-21T02:51:59Z Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing 2003-01 Angela, Amphawan LB2361 Curriculum Cellular networks have gained importance in today's mobile communication scenario. The deployment of first and second-generation cellular systems has involved extensive planning, both to achieve continuous RF coverage and to ensure the provision of capacity in the right locations. The process has been a costly one, accounting for 50-70% of the total network cost. With the explosive growth in the number of cellular subscribers, operators are constantly striving to meet demand. Call blockage during the peak traffic hour is one of greatest concerns of operators. The traffic increase may be accommodated by either adding more carriers to existing cell-sites, or by splitting the cell into several amaller cells. Both methods require the operator to derive a new frequency plan. Additionally , if cell splitting is performed, a new coverage plan will be required to ensure intercell interference is kept to a minimum. The cost of installing new hardware (transceivers and base stations) and devising new frequency and coverage plans is considerable. In big cities the peak traffic tends to occur to the base stations in the city center compared to those at the periphery of the city. Also, the ratio of traffic volume of the busiest to the quietest hour is large. It is therefore vital for an operator to ensure that their resources ( both hardware and spectrum) are utilized to their full potential. An even traffic distribution would ensure much higher utilization of their infrastructure. Obviously, it is not easy to influence user behavior, however it is possible to establish a more flexible network. The objective of this research is to study the potential gain of the dynamic cellular network system in the optimization of network resources and to further improve its performance through a cell priority selection algorithm and a load prediction model. A comprehensive study of the motivation for dynamic cell sizing and its mechanism is made. The capacity gain in the forward link and quality are analyzed. The total forward link capacity increase due to shedding and reduction of power on overhead channels is found to be approximately 16%. However, the gain is to the expense of an increase in the forward gain for the same SNR. Also, coverage holes cause deterioration of quality of service (QoS). Diverse cell interactions of cells in the dynamic cell sizing environment set forth the introduction of a cell priority selection algorithm, named the Virtual Community Parallel Genetic Algorithm (VV-PGA). The algorithm designates priorities to cells which in turn reduces the occurrence of coverage holes. A load prediction model is incorporated in the VC-PGA model to aid in the prediction process of future traffic volumes. The Evolving Fuzzy Neural Networks (EFuNN) computational model is employed for the prediction process. Simulation results prove that EFuNN is to be able to learn traffic sequences in an adaptive way through one pass learning and automatically adapt their parameter values as they operate. The performance of the EFuNN mobile traffic prediction model has been shown to be satisfactory for incorporation into the VC-PGA model. 2003-01 Thesis http://shdl.mmu.edu.my/469/ http://myto.perpun.net.my/metoalogin/logina.php masters Multimedia University Research Library
institution Multimedia University
collection MMU Institutional Repository
topic LB2361 Curriculum
spellingShingle LB2361 Curriculum
Angela, Amphawan
Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
description Cellular networks have gained importance in today's mobile communication scenario. The deployment of first and second-generation cellular systems has involved extensive planning, both to achieve continuous RF coverage and to ensure the provision of capacity in the right locations. The process has been a costly one, accounting for 50-70% of the total network cost. With the explosive growth in the number of cellular subscribers, operators are constantly striving to meet demand. Call blockage during the peak traffic hour is one of greatest concerns of operators. The traffic increase may be accommodated by either adding more carriers to existing cell-sites, or by splitting the cell into several amaller cells. Both methods require the operator to derive a new frequency plan. Additionally , if cell splitting is performed, a new coverage plan will be required to ensure intercell interference is kept to a minimum. The cost of installing new hardware (transceivers and base stations) and devising new frequency and coverage plans is considerable. In big cities the peak traffic tends to occur to the base stations in the city center compared to those at the periphery of the city. Also, the ratio of traffic volume of the busiest to the quietest hour is large. It is therefore vital for an operator to ensure that their resources ( both hardware and spectrum) are utilized to their full potential. An even traffic distribution would ensure much higher utilization of their infrastructure. Obviously, it is not easy to influence user behavior, however it is possible to establish a more flexible network. The objective of this research is to study the potential gain of the dynamic cellular network system in the optimization of network resources and to further improve its performance through a cell priority selection algorithm and a load prediction model. A comprehensive study of the motivation for dynamic cell sizing and its mechanism is made. The capacity gain in the forward link and quality are analyzed. The total forward link capacity increase due to shedding and reduction of power on overhead channels is found to be approximately 16%. However, the gain is to the expense of an increase in the forward gain for the same SNR. Also, coverage holes cause deterioration of quality of service (QoS). Diverse cell interactions of cells in the dynamic cell sizing environment set forth the introduction of a cell priority selection algorithm, named the Virtual Community Parallel Genetic Algorithm (VV-PGA). The algorithm designates priorities to cells which in turn reduces the occurrence of coverage holes. A load prediction model is incorporated in the VC-PGA model to aid in the prediction process of future traffic volumes. The Evolving Fuzzy Neural Networks (EFuNN) computational model is employed for the prediction process. Simulation results prove that EFuNN is to be able to learn traffic sequences in an adaptive way through one pass learning and automatically adapt their parameter values as they operate. The performance of the EFuNN mobile traffic prediction model has been shown to be satisfactory for incorporation into the VC-PGA model.
format Thesis
qualification_level Master's degree
author Angela, Amphawan
author_facet Angela, Amphawan
author_sort Angela, Amphawan
title Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
title_short Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
title_full Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
title_fullStr Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
title_full_unstemmed Evolving Fuzzy Neural Network Based Load Prediction Model In Dynamic Cell Sizing
title_sort evolving fuzzy neural network based load prediction model in dynamic cell sizing
granting_institution Multimedia University
granting_department Research Library
publishDate 2003
_version_ 1747829141417230336