Green Access Network Technology Based On Dynamic Spectrum Management

Over the last decade, applications like voice over IP (VoIP), video-conferencing and highdefinition TV (HDTV) have become very popular, and social services such as e-health ande-government are starting to gain a lot of importance. The demand for accessing Internet withhigh speed and supporting multi-media applications has greatly accelerated the developmentof broadband access networks. However, with the energy becoming less and less and theworsening environment, when one evaluates the communications ecosystem from a carbonfootprint and energy cost perspective, the results are startling. The deteriorating economiclandscape combined with the emerging emphasis on stewardship to the environment has madethe “greening” of broadband access technologies an imperative. With more than half of allbroadband access subscribers, the DSL technology is still the dominating broadband accesstechnology, so the “greening” of the DSL systems is crucial to “green communication”.One of the essential factors that impacting the energy efficiency of current DSL systemsis the electromagnetic coupling among different twisted pairs within the same cable bundle,which is also referred to as crosstalk. Multi-user coordination techniques such as spectrumand signal coordination are recognized as the key approaches for crosstalk mitigation. In thelast few years, dynamic spectrum management (DSM) based on spectral coordination hasbeen an active area of research, and numerous optimal and suboptimal algorithms have beendeveloped. However, they mainly focused on crosstalk mitigation so as to maximize physicallayer data rates. Differently, the core issue of Green DSL vision is to reduce overall energyconsumption within framework of optimizing system capacity and maintaining user Qualityof Service. Therefore, the main topic of this thesis is the development of energy efficiencyresource management. Moreover，considering that the power consumption of overheadchannel is an crucial issue, this paper intended to develop optimization algorithms that can beimplemented more distributed.On the whole, the paper can be roughly divided into two parts. The first three chaptersare about theory introduction. Chap.4and Chap.5are mainly the innovation of the paper. Chap.6summarizes the acquired research achievements and presents how to discuss theproblem further.In the part of theory introduction, Chap.1is the overview on the background,significance, and the existing solution. Chap.2states the characteristics of line fading,crosstalk and all kinds of noises in the multi-user DSL systems, discusses discrete multi-tone(DMT) modulation and analyzes the system’s input-output model. Chap.2analyses thecharacteristics, classification, problem formulation of Dynamic Spectrum Management, andillustrates the systemic resolution of these problems.Based on the above analysis, this paper develops two innovations according to theenergy-efficient dynamic spectrum management problems:Firstly, through extending the traditional data rate maximization approach towards theenergy efficient DSM problem, this part studies the trade-off between the power saving andthe loss of the data performance. Specifically, a new DSM problem that can quantify thetrade-off is established. Then by associating duality theory for non-convex programming,sub-gradient algorithm with the Modified Iterative Water-filling algorithm, we construct adistributed feasibility algorithm. Simulation results show that we can reduce the consumedpower by higher percentage with just smaller sacrifice of data rate.Secondly, based on the fact that the actual system could only implement discretebit-loading, the optimal discrete bit-loading problem for energy saving objective is studied.More specifically, proceeded from the general power-minimization DSM problem, a newoptimal bit allocation problem is formulated through introducing the pricing factors on eachtone. To reduce the energy consumption of overhead channel, we propose a bit-loadingalgorithm with the rate priced factors on each tone. Combined dual decomposition and EE-LC(Energy Efficient Levin-Campello) bit-loading algorithm, a practicable algorithm isdeveloped. Results in practical VDSL environment illustrate the comparison between themethod and others. On the basis of simulations, we can see the efficiency of the proposedalgorithm. More significantly, through the different settings of weighting factor, this designcan achieve sub-optimal bit-loading vectors for the system of all kinds of topology, whichwould address the limitations of existing algorithms.