业务能力可预见的分布式认知无线网络在线决策机制研究

本课题以认知无线网络分布式在线学习与决策为研究背景，针对目前理论研究以及系统实用化进程中所面临的三个挑战："时空覆盖失度"、"动态调度失衡"和"机会调度失序"，有效解决目前信道特征学习困难、动态适应性差、网络调度与协同开销巨大的问题。课题研究过程中，首先探索在线认知学习方法，利用在线认知的反馈信息实现"利用中学习"和"学习中利用"，根据残缺的、局部的、非确定性的，甚至是存在错误的认知信息，结合随机决策优化模型和非确定信息处理的方法，解决信道学习和利用的矛盾；接着通过准确有效的控制节点的行为，包括最优接入门限和网络状态分析，实现多用户在多维信息条件下的连续动态接入；最后，提出"可预见业务能力"的概念，优化网络的负载能力，追求在可靠性与稳定性之间达到平衡，减小节点之间的协议开销。在此基础上，能够增强系统的动态敏感性与抗扰动能力，有效提高信息的传播效率，从而有效解决机会调度困难，效率低下的问题。

In this proposal, we study the online learning and decision making problem in distributed wireless cognitive radio network. In tackling the three challenges : "spatia temporal covery failure", "dynamic scheduling efficiency" and "opportunistic scheduling disordering", the proposal aims at solving these problems and build an effcient and adaptive online learning and decision making pradigm. First, we propose the "using channels while learing" and "learning channels while using" methods, which could effectively utilize the imperfect channel knowledge. Secondly, using the opitmal access thereshold and network status analysis method, we can moderately control the cogntive node, where the continuous and dynamic accessing can be achieved in multi-dimesional information scenarios. Finaly, we propose the "predictable service" concept in achieve an optimized network workload and scheduling paradigm, which aims at achieving a perfect balance between reliablity and stability, minimizing the protocol overhead between nodes. Leveraging these methods, the system sensibility and stablity can be achieved at the same time, and the coordinative information overhead can be minimized, which will be helpful to solve the inefficient scheduling problem in opportunistic spectrum accessing network.