Font Size: a A A

Critical Techniques Of Nonlinear Optimization In Communication Networks

Posted on:2010-11-27Degree:MasterType:Thesis
Country:ChinaCandidate:Z SunFull Text:PDF
GTID:2178360278965495Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
During the past years, nonlinear optimization framework has already been applied to a broad region of communication systems, ranging from high speed Internet, to wireless Ad hoc networks, from equalization and coding to broadband access, from information theory to network topology modeling. In this graduation thesis, the author tries to provide a description of the state-of-art techniques in the topic of nonlinear optimization in communication networks.Firstly, the thesis considers the application of nonlinear optimization methods to the radio resource allocation problem in OFDMA networks. Specifically, we deal with a subcarrier allocation problem under equal power allocation assumption. By defining utility functions, the subcarrier allocation problem is modeled as a nonlinear objective function optimization problem. And by controlling and tuning the weights of real-time and non-real-time services in subcarrier allocation, the central controller was able to adaptively coordinate the fairness between real time service and best-effort service. This control theory-inspired method could be seen as a highlight and innovation of our work.Then, the thesis considers the general utility definition in utility maximization problems of nonlinear optimization. General utility definition is such a hot topic in nonlinear optimization that broadly exists in researches like Network Utility Maximization, and Stochastic Network Games. It tries to understand if there exists a method to define utilities to guarantee the convexity of the objective problems while fully characterizing services. In this part, based on a paper of the author, we propose a suite of utility definition criteria in OFDMA networks, which could be regarded as a reference to other researches in this field.A very difficult problem in nonlinear optimization is its hard solving, especially when the objective function is nonconvex. Multiple alternative decomposition algorithms in Network Utility Maximization (NUM) problems have been proposed. NUM believes that network itself is an optimizer; nodes in a network can be involved in the process of optimization of the network. Research in recent years addresses vividly how to decompose a NUM problem, so that nodes could join the process. In this part, we summarize the author's study during one graduate year, and provide numerical simulations to strengthen our discussion.The theory of nonlinear optimization is a general approach to the optimization of communication networks, and could be used in not only centralized networks, but also distributed networks. For example, the distributed data fusion in Wireless Sensor Networks (WSN) can be well modeled as a classic nonlinear optimization problem. Data fusion is an old topic in signal processing. As the application of WSN becomes general, how to design efficient distributed data fusion algorithms attracts a number of researchers. And good methods, including the ones using convex optimization and relevant algorithms, such as incremental subgradient algorithm and projective subgradient algorithms, have been proposed. In this part, we introduce one distributed spiral optimization algorithm, and prove its advantages.At last, a summary and a expectation of future works are given.
Keywords/Search Tags:Nonlinear optimization, wireless communications, OFDMA, Wireless sensor networks
PDF Full Text Request
Related items