A Sequential Learning Framework for Resource Management in Wireless Networks

由于无线频谱的稀缺性和对无线数据服务的不断增长的需求，资源管理是从城域网到个人区域网的具有非常不同特征的无线系统中的一个永久性问题。当系统状态和参数的知识未知或不完整时，顺序学习或特别是多臂强盗框架提供了一个优雅的解决方案来平衡开发（实践迄今为止最知名的决策）和探索（找出其他决策的效用）的权衡。我的研究计划的主要目标是推进对网络物理系统以及无线和移动的数据网络的有效鲁棒解决方案的基本理解和开发。该项目的具体目标是为无线网络中的各种资源管理问题开发一个顺序学习框架。该项目的贡献将是三方面的。首先，我们将推进顺序学习理论，并设计新的算法来解决无线网络带来的实际挑战。其次，我们将应用所提出的顺序学习方法来开发有效的解决方案，在网络监控，认知无线电频谱接入和路由延迟容忍网络的资源管理问题。第三，通过测试平台开发和系统实施，该项目将为HQP提供参与培训的机会，不仅在算法设计和分析方面，而且还在最先进的无线网络硬件和软件方面提供实践经验。

Due to the scarcity of wireless spectrum and growing demands for wireless data services, resource management is a perpetual problem in wireless systems with very diverse characteristics from metropolitan scale networks to personal area networks. When the knowledge of system states and parameters is unknown or incomplete, sequential learning or specifically the multi-armed bandit framework offers an elegant solution to balance the trade-offs in exploitation (practicing the best-known decisions thus far) and exploration (finding out the utility of other decisions).****The primary goal of my research program is to advance the fundamental understanding and development of efficient robust solutions for cyber physical systems, and wireless and mobile data networks. The specific objective of the project is to develop a sequential learning framework for a variety of resource management problems in wireless networks. The contribution of the project will be three-fold. First, we will advance the theory of sequential learning and devise new algorithms to address practical challenges arising from wireless networks. Second, we will apply the proposed sequential learning methodologies to develop effective solutions to resource management problems in network monitoring, cognitive radio spectrum access and routing in delay tolerant networks. Third, through testbed development and system implementation, this project will offer HQP involved training opportunities not only in algorithm design and analysis but also hand-on experiences with state-of-the-art wireless network hardware and software.****