Over the past decades cellular technology has developed through user demand and influence into the Third Generation (3G) of mobile communication systems that are in use today. These networks have adopted a microcellular network approach in densely populated areas with cells only a few 100 meters wide resulting in users roaming between cells at a much higher rate due to the smaller cell sizes compared to legacy cellular networks. This places a fundamental challenge to the handover process, the most frequently requested, complex and time-critical function of a cellular network. As it ensures the continuity of a connection it has a direct impact on the quality perceived by users and is a key factor when measuring system performance and efficiency as well as Quality of Service provided to users. With a user base that is expected to continue rising an efficient and scalable handover scheme is therefore essential to ensure user satisfaction in Third Generation mobile communication systems now and in the future. One way to achieve this is to consider future user movement with the aim to predict the next visited cell to forward data destined for a mobile device and allocate resources prior to a hand over being initiated which is being considered in this thesis. Abstract Following a mathematical description of the problem to be solved to allow the implementation of scenarios for experimental analysis, this dissertation is focusing on the development of an adaptive learning based prediction scheme using Neural Networks (NN) to predict the future movement steps of mobile devices using historical movement information to support the handover process. The aim of the developed scheme is to improve system efficiency by supporting the delivery of a constant Quality of Service to mobile devices by facilitating resource reservation in the future visited cells prior to their arrival while at the same time keeping additional overhead and resource requirements low. The focus of this research undertaken within this thesis is different to previous work as it takes into consideration the changes to mobility behaviour of mobile devices over time. As part of this a detailed performance analysis of the developed system is conducted to study its behaviour in the presence of changing mobility patterns. This study resulted in the development of an adaptive Genetic Algorithm (GA) based retraining scheme tailored to the NN based prediction system to counteract the effects of changing mobility pattern and to add robustness. The NN based prediction system with GA based retraining is then further developed providing a description of a hierarchical implementation within a cellular communication system of the Third Generation. The concepts of the developed schemes have been investigated through experimental as well as simulation work within a framework developed in the simulation environment OPNET to study the behaviour and establish their efficiency. Obtained results confirm the success of the developed concepts with respect to their predictive properties and ability to retrain the NN based prediction system.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:551190 |
| Date | January 2010 |
| Creators | Bauer, Carolin Isabel |
| Publisher | Staffordshire University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0023 seconds