【作者】 任海豹；

【导师】 周武旸；

【作者基本信息】 中国科学技术大学 ， 通信与信息系统， 2014， 博士

【摘要】 随着移动互联网的发展,无线业务出现爆炸性增长。这种增长不仅体现在数据量的增加,也体现在业务种类和QOS需求的多样性变化。为了满足业务需求、提升网络容量,需要从传输技术和网络架构两个方面进行下一代无线网络设计。先进的物理层技术,如OFDM、MIMO,在增大频谱效率的同时,带来了更大的能量开销。未来网络站点密集化、异构化不仅增加站点数目、增加耗能,还提升网络干扰强度,降低了整网的能量效率。为了降低能量开销,减少碳排放,需要在保证业务需求的前提下对基站进行节能设计。本文先从业务QOS的表征参数出发,将时延作为最重要的QOS影响因素,并按照业务的时延要求,将业务分为严格时延约束和时延容忍两大类。针对这两类业务,分别从链路级角度设计时延保障的节能调度器,并考虑实际多小区系统,设计分层调度器,在保证时延要求的前提下最小化基站耗能,并协调相邻小区间休眠子帧,以降低网络间干扰。所有的这些研究都是基于实际的功耗模型进行的。由于基站功耗模型非连续,调度问题在多数场景下没有最优解。本文研究了严格时延约束业务的高能效调度策略。首先针对AWGN信道,研究如何在给定时间内传输不同时延要求的数据包使得系统所需的能量最小的问题,利用拟凸优化提出了低复杂度最优传输速率和休眠控制策略,在此基础上,提出因果系统的高能效传输调度器设计；进一步需要考虑衰落信道下存在信道估计耗能时的数据包级能效最优传输控制和调度策略,通过联合考虑信道的统计特性、导频耗能和数据包长度,提出包含剩余时间项、信道相关项和能量相关项的次优调度算法进行能效资源调度。本文针对时延容忍业务进行能效传输设计,提出两种休眠策略并利用排队论中带休假的M/G/1模型,计算出不同休眠控制策略下的能耗和业务平均时延的闭式表达,利用凸优化理论和拉格朗日乘子法对问题进行迭代求解,并提出给定拉格朗日乘子下的迭代凸搜索的最优功率分配和休眠控制算法；在多用户场景下,分析ARQ对时延影响,依据业务的统计特性建立优化问题,利用组合优化理论进行能效资源分配,在保证业务平均时延要求下最大化系统能效。本文最后研究实际多小区中满足用户QoS需求的子帧级协调休眠调度策略,提出两层调度器模型,利用线性控制理论设计上层调度器的等效滤波器,为每帧分配数据以保证QoS,底层进行高能效资源分配；先针对单用户,将活跃子帧连续化,利用拟凸优化证明问题存在最优解,通过迭代搜索求得最优功率和休眠子帧数目；然后考虑多用户系统,利用时间共享条件和拟凸优化可以获得休眠子帧数目,利用边缘自适应算法分配子信道、功率；最后对于多小区,依据单小区的资源分配结果确定各小区在每帧内所需传输数据量,依据上一时隙信道状态和干扰强度分布结果,动态协调休眠子帧,使得强干扰邻小区休眠子帧正交,进而提升网络能量效率。更多还原

【Abstract】 With the development of mobile Internet, the traffic demands of wireless appli-cations increase explosively. At the same time, the kinds of service and QoS (Qual-ity of Service) requirements become diverse. To satisfy traffic demand and improve the capacity of network, the designment of next generation network should consider the transmission technology and network architecture. Advanced physical technolo-gies such as OFDM and MIMO result in more energy cost while improving the spectral efficiency(SE). Densification deployment requires more number of sites and makes the network interference limited and decrease the energy efficiency(EE). To reduce the en-ergy consumption and carbon emission, BSs need energy saving design.This thesis starts from the characterization parameters of QoS and finds that delay is the most important effect factor of QoS. We group the wireless applications into delay constraint traffic and delay tolerant traffic and design energy efficient schedular for these two kinds of traffic with delay providing from link level. In addition, we propose a two-level schedular with QoS guaranteeing and energy saving for the practical cellular networks. When the traffic load is low, BSs can coordinate the micro sleep subframes to minimize interference and improve EE. All these works consider practical power model. There are no optimal solutions for most of the cases because the power function is discontinuous.Firstly, we study the energy efficient scheduling policies for delay constraint ser-vice. We focus on how to minimize the total energy consumption when transmit the non-causal independent arriving and delay constraint packets. Using the quasiconvex optimization theory, we propose an optimal rate adaptive and sleep control algorithm with low complexity. Then, based on the results, we propose a heuristic online schedul-ing algorithm for the causal Poisson arriving process. In addition, we consider the scheduling policy for fading channel in which the transmitter should send pilot signal to estimate the channel state and this process is not only time consuming but also energy consuming. The original optimization problem is NP-Hard and we propose a subopti-mal scheduling algorithm consisting three parts:a time-dependent weighted sum of a delay associated term, an opportunistic term and a practical energy consumption related term.Then, we also study the energy efficient transmission scheme for delay tolerant traffic. For the single channel, we propose two sleep strategies and deduce the closed form of energy consumption with the M/G/1queueing theory. Then we propose an optimal iterative algorithm using biconvex optimization theory and Lagrangian method. For the multi channel multi user scenario, we deduce the average and variance of user’s delay for the ARQ process. With the constraint of average delay, we propose a heuristic resource allocation algorithm based on the traffic statistical information and channel state information.In the end, we study the multi BSs cooperative subframe-level sleeping strategy with QoS guaranteeing for the practical cellular networks. We propose a two-level schedular model and use the linear control theory to design the upper level schedu-lar which segments the packets into multiple small ones and transmit each one over different frames. The lower level schedular is used for the energy efficient resource allocation. For the scenario with single BS and single user, we formulate the prob-lem as a mix integer programming problem to minimize the energy consumption while satisfying average data rate requirement. We relax the integer of sleep subframes to a continuous variable and use convex theory to demonstrate the existence of a unique globally optimal resource allocation solution and develop an iterative algorithm to ob-tain it. For the scenario with single BS and multiple users, we use the property of time sharing condition for multiple channel system to obtain the optimal sleep subframes for the BS. Then we allocate the power and subchannel for each user using the traditional Marginal Adaption problem. For the multiple BSs system, each BS allocates the re-source independently based on the channel state and interference level of the last frame. Then the BSs coordinate sleep subframes with each other as orthogonal as possible to reduce interference and improve the energy efficiency of network.更多还原