移动用户群体聚集行为模型及其高能效资源配置方法  

The model of large-scale mobile user behavior and energy efficiency resource allocation method

作　　者：杨坤[1] 张兴[1] 杨居沃 王硕[1] 梅承力[2] 王文博[1] 

机构地区：[1]北京邮电大学泛网无线通信教育部重点实验室,北京100876 [2]中国电信股份有限公司技术创新中心,北京100032

出　　处：《中国科学：信息科学》2017年第5期620-636,共17页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61372114;61631005);国家重点基础研究计划(973)(批准号:2012CB316005);北京市科技新星计划(批准号:Z151100000315077)资助项目

摘　　要：由于用户社会属性的存在,复杂蜂窝移动网络的业务特征和用户行为在时域、空域和内容等多维度上的分布都呈现出以群体为特征的聚集行为规律.以往静态、孤岛式的网络资源配置方法造成了网络资源的巨大浪费,因此利用用户群体行为特征规律将存在巨大的能效和资源利用提升空间.基于对实际运营的蜂窝移动通信系统中数据的采集和测量,首先从空间、时间等多个维度对用户群体聚集行为进行了深入分析研究,得到了基站流量在空域、时域和空–时联合的分布规律.研究表明,业务在空间符合Log-normal分布,其参数与典型区域类型有关;用户数及其产生的业务量随着时间变化具有明显的规律性,正弦叠加模型能够很好地反映出现网实际业务量的变化情况.其次,通过对空域和时域的联合分析,得到了能精准预测基站业务变化的空–时联合分布模型.与实际数据对比发现,该模型准确度可以达到93%以上.为了更明确地表征用户群体聚集行为,利用经济学中的基尼系数对用户群体聚集行为进行了数学定义和定量描述.最后,基于所提出的业务空–时模型和用户群体行为聚集模型,提出了几种高能效的无线网络资源配置方法、传输控制方法和基站分级休眠策略,探索利用用户群体行为规律提升无线网络能效的新途径.For the existence of user social patterns, the traffic characteristics and user behavior in complex cellular mobile network show clustering behavior regularity characterized by groups in multi-dimensional distri- butions, such as time, space~ and content. Previous static, island-like network resource allocation methods lead to a massive waste of network resources. Thus, the use of user behavior characteristics can have a huge im- pact on energy efficiency and resource utilization. Based on data collection and measurements from real cellular mobile communication systems, we analyze the behavior of user group aggregation in spatial, time, and other dimensions, to obtain traffic distributions across base stations in the space, time, and spatio-temporal domains. The study shows that traffic distribution in the space domain conforms to a Log-normal distribution and that its parameters are related to typical region types. The number of users and the amount of traffic they produce show regularity over time, and the Sine superposition model can reflect changes in the traffic volumes in a network. Next, through joint analysis of the space and time domains, we obtain a spatio-temporal joint distribution model that can accurately predict changes in base station traffic. When compared with real data, it is shown that the accuracy of the model reaches over 93%. In order to characterize user group clustering more clearly, we use the Gini coefficient to present a mathematical definition and quantitative description of user group behavior. Finally, based on the proposed spatio-temporal model and user group behavior model, we propose several energy-efficient wireless network resource allocation methods, transmission control methods, and a base station hierarchical sleep strategy for exploring new approaches using user group behavior regularity to improve wireless network energy efficiency.

关 键 词：用户群体聚集行为 资源分配 能量效率 移动蜂窝网络 

分 类 号：TN929.5[电子电信—通信与信息系统]