基于增量调整和智能性自主合作的移动无线网动态管理关键技术研究

The prevalence of mobile devices and wireless communications makes the mobile wireless networks be the foundations of the future computing environments. The efficiency of the mobile wireless networks will directly determine the effect of the applications on such new computing environments. The fundmental characteristic of mobile wireless networks is the uncertain dynamic movement, and this characteristic will result in the very bad effect for a recent optimal configuration. The autonomous management is the basic technique to solve the problem, and this is the background of our research proposal. The state of arts show that there are many key problems are need to be solved,such as the structure of the cooperation configuration, the incentive mechanisms for cooperation, the decrease of managing cost, the treatment to uncertain information, the implementation of the protocols and their backward compatibility, etc. Therefore, this proposal will research the dynamic management in mobile wireless networks based on the incremental adjustment and the intelligent autonomous cooperation, and will study the following concrete contents: (1)the increamental reconfiguration structure is provided based on the network embeded theory and the queuing networks; (2)the trade-off of the management cost and adjustment effect is guaranteed by the localization of the adjustment in temporal and spatial dimension; (3)the uncertainty of information in adjustment is considered via the Markov decision process and optimal stopping theory; (4)the selfness and the heterogeneity of selfness are analyzed on the structure of the dynamic coalitional game theory; (5)distributed reconfiguration protocols and the autonomous management framework based on service will make the dynamic management proposed in this project for mobile wireless networks more flexible, more easy to implement and more esay to work with legacy of old protocols.

移动设备和无线通信的普及使得移动无线网成为基础计算环境，其能否高效工作直接决定上层应用的效果。移动无线网的基本特性就是不确定的动态变化，会让一个优化配置马上变得效果很差，而自主管理就成为解决该问题的根本手段，这是本项目的研究背景。目前的研究成果在合作配置结构、合作激励、降低管理代价、应对信息不确定、协议实现及向后兼容等方面仍有很多根本问题有待解决，因此本项目就以增量调整和智能性自主合作作为基本方法对移动无线网络的动态管理进行深入研究：(1)以网络嵌入和排队网络为基础提出一个在线的增量式重配置结构；(2)在时间、空间上让调整局部化来实现调整效果和代价的折中；(3)用马尔科夫决策和最优停止理论解决调整时的信息不确定性；(4)以动态联盟博弈为结构解决节点自私问题以及自私的异构性；(5)基于强化学习的分布式重配置协议和基于服务的自主管理框架使得本项目的动态管理灵活、易于实现，能和现有协议共存。

手机，智能平板，智能车辆都是以无线网为基础计算环境，其自身所处无线环境各自不同，导致像有线网络那样的统一且简单配置和调度无法达到很好的优化效果。通过实现无线网络自主管理使得无线网络及无线节点能够充分适应所处环境和环境变化（节点移动，无线通讯干扰变化，链路连接变化等）。本项目以无线网络的自适应性为目标完成以下研究内容：（1）以吞吐最大化为目标，通过Lyaponuv模型分析，提出最大权链路排队调度算法；（2）在时间和空间范畴考虑，实现了面向无线网络高移动性和不稳定性的自适应路由及链路修复算法；（3）利用不确定模型理论对于LBSNs信息分析动态网络位置信息变化；（4）基于用户网络接入行为分析实现面向解决用户节点自私问题的异构网络激励策略；（5）开发了能够采集无线通讯指标并进行自我学习分析和配置调度的软件定义无线网络实验平台，其可以实现本项目所提出多种通讯配置并能和现有协议同存，且具有动态管理，易于实现和灵活扩充等特点。