Analytic Methodology For OFDMA Resource Allocation

OFDMA is a core technique for next generation mobile communications, wherethe resource allocation problem has attracted much attention from both industry andacademia. Previous works often formulate the OFDMA resource allocation into op-timization problems. However, the system performance can still not be evaluated inan analytic way by this kind of approach. The intrinsic connections among systemparameters are still unknown. From both theoretical frameworks and practical algo-rithms, this dissertation demonstrate the disciplines which dominates the performanceof OFDMA resource allocation by using an analytic methodology so as to lay the foun-dation for the OFDMA system design.This dissertation focuses on the following three aspects of system model, per-formance evaluation, and analysis method. It will address the following five issues:OFDMA system modeling, characterizing the performance of resource allocation, theperformance limitation of OFDMA resource allocation, the performance evaluation ofjoint coding, and the algorithm design for subcarrier allocation. The following mathe-matical tools are applied in this dissertation: random graph theory, generalized match-ings, large deviations principle, and Meijer's G-function. The main achievements in-clude the following: the OFDMA system is formulated into a random bipartite graphmodel; the concept of outage exponent is defined to illustrate a compete performancepicture of OFDMA resource allocation; the methods of uniformly generating maxi-mum matchings, maximum constraint H-matchings and maximum f-matchings areproposed for continuous subcarrier allocation systems, non-continuous subcarrier al-location un-coded systems, and non-continuous subcarrier allocation coded systems,respectively; the outage probability and the optimal diversity-multiplexing tradeo? arecalculated for these matching methods; the outage exponent for the joint coding in thefrequency domain is derived; the optimal subcarrier allocation algorithms are designedwith low complexity. The theoretical framework proposed in this dissertation incorporates the conceptsof outage probability, error exponent, transmission rate, outage capacity, ergodic ca-pacity and diversity-multiplexing tradeo?. Therefore, it fully re?ects the intrinsic con-nections of the coherence bandwidth, signal to noise ratio, the number of users, andthe performance of resource allocation. The used generalized matching method doesnot only fully utilize the available channel diversity gain and multiplexing gain, butalso make full use of multi-user diversity. If the coherence bandwidths are less thanthe number of users, the OFDMA system can achieve the optimal outage performanceby using no joint coding schemes. The induced subcarrier allocation algorithms enjoya complexity of O(N2/3) (N is the number of subcarriers), which are much lower thaniterative or heuristic algorithms. The proposed algorithms only require 1 bit channelstate information for each subcarrier, which yields a low requirement on channel es-timation and small signaling overhead. The method of outage exponent also solvesthe outage probability calculation problem for block-fading channels which was intro-duced in early 1990's.