NeTS: Small: Designing Agile and Scalable Self-Healing Functionalities for Ultra Dense Future Cellular Networks

蜂窝网络受到几种类型的蜂窝中断的影响。完全中断通常是由设备故障引起的；部分中断或性能下降通常是由参数配置错误引起的。中断率与小区密度和基站复杂性成正比。从1G到4G，这两个因素一直在上升。即使对于今天的网络，目前半自动化的蜂窝中断管理方法也被证明是不充分和效率低下的，而且对于以超密集蜂窝部署和日益增加的复杂性为标志的未来蜂窝网络来说，肯定是不可行的。如果不采取干预措施，蜂窝中断管理可能成为未来蜂窝网络（如5G）的主要挑战。为了解决这个问题，该项目将开发一种先进的蜂窝中断管理（ACOM）框架，用于在未来的超密集、异构蜂窝网络中实现完全自动化的蜂窝中断检测和补偿。ACOM将通过开发和集成三种新颖的解决方案来构建：1)自主宏单元中断检测和根本原因分析（MOD）；2)自主小细胞停机检测和根本原因分析（SOD）；3)自主异构单元中断补偿（HOC）。在ACOM中，由MOD和SOD检测和诊断的中断将被HOC中的优化过程所利用，从而将未来的超密集、异构蜂窝部署转变为完全自我修复的系统。ACOM的关键特点将包括敏捷性、稳定性和灵活性，以适应不同的用户密度、小区大小和无线电信道条件，以及小型小区的超密集部署。该方案结合了剖面分析、异常检测、预测、稀疏矩阵完井、多维缩放和动态处理。它采用了机器学习、优化和博弈论。项目成果将利用真实网络数据进行验证，同时利用全面的户外5G测试平台。如果成功，该项目肯定会对依赖蜂窝网络可靠性的数字社会发展的各个方面产生强烈影响。该项目的另一个关键影响是，它提供了开展拟议研究所需的急需的多学科技能的强有力的劳动力培训，同时确保妇女和其他代表性不足的群体的参与，以及K-12的外展。该项目还将利用与蜂窝生态系统中的国内和国际利益相关者的合作，最大限度地提高其对标准化的影响。

Cellular networks are subject to cell outages of several types. Complete outages are often caused by equipment malfunctions; partial outages or degraded performances are often caused by parameter mis-configurations. Outage rates are proportional to cell density and base station complexity. Both of these factors have been consistently on rise from 1G to 4G. Current semi-automated approaches to cell outage management have proven inadequate and highly inefficient even for today's network, and are surely unfeasible for future cellular networks marked by ultra-dense cell deployment and mounting complexity. If no intervening measures are taken, cell outage management may become a primary challenge for future cellular networks, such as 5G. To remedy this, this project will develop an Advance Cell Outage Management (ACOM) framework for fully automating cell outage detection and compensation in future ultra-dense, heterogeneous cellular networks. ACOM will be built by developing and integrating three novel solutions: 1) Autonomous Macro Cell Outage Detection and root cause analysis (MOD); 2) Autonomous Small Cell Outage Detection and root cause analysis (SOD); and 3) Autonomous Heterogeneous Cell Outage Compensation (HOC). In ACOM, the outage detected and diagnosed by MOD and SOD will be exploited by the optimization process in HOC to transform future ultra-dense, heterogeneous cellular deployments into fully self-healing systems. Key distinct features of ACOM will include agility, stability and flexibility to accommodate varying user densities, cell sizes, and radio channel conditions, and ultra-dense deployments of small cells. The proposed plan lies at the nexus of profiling, anomaly detection, prediction, sparse matrix completions, multidimensional scaling, and dynamics handling. It employs machine learning, optimization, and game theory. Project outcomes will be validated using data from real network data while leveraging a full scale outdoor 5G testbed. If successful, this project is certain to make strong impact on all aspects of evolving digital society that count on reliability of cellular networks. Another key impact of this project is that it offers strong workforce training in a highly sought-after multi-disciplinary skill set needed to conduct proposed research, while ensuring participation of women and other underrepresented groups, and K-12 outreach. The project will also leverage collaboration with national and international stake holders in the cellular ecosystem to maximize its impact on standardization.