Distributed Resource Allocation in Competitive Wireless Networks

Abstract: The recent advancements in wireless technology and the developmentof efficient spectrum sharing strategies have had a great impacton today's communication systems. It is not uncommon that severalnetworks are deployed in the same area, each serving multipleusers who expect ubiquitous access to various voice andinformation services. While traditional cellular network operatorsare centrally managed and license their part of the radiospectrum, some portions of the available bandwidth are being opento anyone. In these bands, several communication units tend tocoexist and create mutual interference to each other as there isno centralized control unit managing the resources. Suchinteraction can be modelled as an interference channel andefficient decentralized spectrum sharing strategies need to bedevised.In this thesis, communication environments where two or morecommunication pairs coexist in the same spectrum are studied andtheir interaction is modelled as strategic non-cooperative games.Each pair operates as an independent unit and greedily allocatesits transmit power across the available spectrum such that theindividual link rate is maximized. We seek to achieve fixed-pointscorresponding to Nash equilibria (NE), where each user isunilaterally optimal. To attain such equilibria, a sequential orsimultaneous Iterative Water-Filling Algorithm (IWFA) is employedand its convergence properties are studied. Depending on thesystem parameters and the update strategy, different NE can beattained and some of these may result in poor overall spectrumutilization. The achievability of different NE is discussed forvarious system settings and the effect of initialization on theconvergence of IWFA is investigated. To improve the overall systemperformance in terms of sum rate, a class of modified utilityfunctions is introduced. In the long run, each link benefits fromsuch a utility since a more efficient spectrum usage results inbetter individual performance.The methodology and concepts introduced in this thesis are usefulwhen designing and analyzing future decentralized spectrumallocation schemes. In particular, the effects of initializationshould be considered as well as the convergence behavior resultingfrom usage of different update strategies.