The Research On Self-optimization In Wireless Ubiquitous Network Based On Context Aware

This thesis investigates the self-optimization method and theory in the wireless ubiquitous network. Based on the analysis of the network architectures, requirements, and characteristics, this thesis proposes different perspectives on self-optimization mechanisms and algorithms in wireless ubiquitous network to improve the network performance. To study the self-optimization in the wireless ubiquitous network, data analysis, pattern recognition, information theory and cybernetics are emploied. The research of this thesis has important theoretical and practical significance in the application, development and research of wireless ubiquitous network.Firstly, this thesis studied method for context awareness by the network parameters. There are large amounts of data in wireless Ubiquitous Network, and effective usage of the information in these data will greatly enhance the performance of the network. This thesis studies the algorithm of multi-resolution analysis in wavelet analysis to process the received signal for contextual information. Then, to enhance the accuracy of the access selection process, the equalizer is used to collect information on surrounding environment. To deal with time-varying characteristics of the environment, a dynamically weight updating algorithm is proposed for access selection process. Simulation results show that with more efficient usage of the data in the network, the performance, reliability, and adaptability of the wireless ubiquitous network can be improved without extra equipment.In order to provide better services, wireless ubiquitous network needs to address the information like "good","bad" or "normal","abnormal". But traditional data analysis can not effectively deal with such conceptual information. In this thesis, many new methods are proposed to deal with conceptual information in Wireless Ubiquitous Network based on the idea of pattern recognition. Aimed at handoff failure recognition, quality of service (QoS) evaluation and quality of experience (QoE) evaluation, this thesis employs hypothesis testing, support vector machines and Bayesian networks respectively. The theoretical analysis and simulation both show that pattern recognition technology can effectively improve the performance of the wireless ubiquitous network.Then, from a higher level, this thesis studies the network optimization through an information theoretic perspective. Starting from the basic concepts of information theory, the way to reduce the information loss in the transmission of information based on Fanbo inequality is propoesd. Then, this thesis investigates the effect of uncertainty of information on the communication process; and a randomized algorithm is then proposed to improve system performance. This thesis also employs the concept of Kolmogorov complexity to investigate the relationship between information entropy and the computational complexity. Finally a method to improve the computational efficiency by independent component analysis is proposed.More macro, wireless ubiquitous network is a complex dynamic system, in order to ensure the reliability of such a system and realize the self-optimiztion; it is necessary to keep on carrying out the control of the system. This paper employs cybernetics to analyze the control process in the network. This thesis uses the variational methods and functional theory to deal with two kinds of typical power control problems in wireless ubiquitous network. Then, based on the principle of minimum optimal; then studies the TCP transmission from the cybernetic view to achieve the optimal control of the process.