Sparsity-Aware RF Cartography for Cognitive Networks

ECCS-1002180明尼苏达大学Georgios Giannakis认知网络稀疏性感知射频制图摘要智能优点：无线认知无线电(CR)技术具有巨大的潜力，可以有效地解决带宽利用不足与频谱稀缺的两难困境，这使得固定接入通信网络效率低下。因此，需要在关键认知基础设施中进行基础研究，以感知、学习和适应认知无线电网络运行的环境。这项提议旨在通过射频制图的新概念来发展这一基础设施，以便全面了解情况。与路由表的成功并行，我们的愿景是联合利用干扰和信道增益图来识别机会可用频段以供重复使用和切换操作；本地化和跟踪用户活动；以及控制资源分配和路由决策。该方法借鉴了稀疏性感知回归、压缩采样、基扩展、样条插值和克里格卡尔曼滤波等方面的当代进展。该项目利用这些工具来调查：(A)用于地图估计和跟踪的分布式、在线和自适应算法；(B)基于训练的和盲制图选项；(C)时空频谱的重复使用，以及在存在多径和阴影效应的情况下的定位；以及(D)制图驱动的网络效用最大化，以实现CR网络的最佳跨层设计。在更广泛的意义上，稀疏性感知回归、克里格法、样条法和频谱制图方面的进展将渗透到广泛的领域，如社交网络的机器学习和数据挖掘、动态磁共振成像、使用无线传感器网络的监视以及导航和安全系统。对网络和本地化应用的认知将进一步为本科生提供有意义的经验，并整合所获得的专业知识，以增强研究生课程的内容。通过短期课程、研讨会和会议中的教程以及学生-教职员工合作，可以接触到政府和工业部门。

ECCS-1002180Georgios GiannakisUniversity of MinnesotaSparsity-Aware RF Cartography for Cognitive NetworksABSTRACTIntellectual Merit: Wireless cognitive radio (CR) technology holds great promise to address fruitfully the perceived dilemma of bandwidth under-utilization versus spectrum scarcity, which has rendered fixed-access communication networks inefficient. Accordingly, the need arises for fundamental research in critical cognition infrastructure to sense, learn, and adapt to the environment CR networks operate. This proposal aims to develop this infrastructure for comprehensive situation awareness through the novel notion of RF cartography. Paralleling the success of routing tables, the vision is to jointly utilize interference and channel gains maps to identify opportunistically available bands for re-use, and handoff operation; localize and track user activities; as well control resource allocation and routing decisions. The approach draws from contemporary advances in sparsity-aware regression, compressive sampling, basis expansions, spline interpolation, and kriged Kalman filtering. The project leverages these tools to investigate: (a) distributed, online, and adaptive algorithms for map estimation and tracking; (b) training-based and blind cartography options; (c) spatio-temporal spectrum re-use, and localization in the presence of multipath and shadowing effects; and, (d) cartography-driven network utility maximization for optimal cross-layer design of CR networks.Broader Impacts: This research is of interest to software radio designs with IEEE 802.11 compliant standards. In a broader sense, advances in sparsity-aware regression, kriging, splines, and spectrum cartography will permeate benefits to a gamut of areas as diverse as machine learning and data mining for social networks, dynamic magnetic resonance imaging, surveillance using wireless sensor networks, as well as navigation and safety systems. Cognition in networking and localization applications will further provide meaningful experiences to undergraduates and integration of the expertise gained to enhance the content of graduate classes. Outreach to the government and industrial sectors will be possible through short courses, tutorials in workshops and conferences, and student-faculty-staff collaboration.