面向无线网络智能管控的节点自主决策

Nodes in large-scale heterogeneous dynamic wireless network are envisioned to be endowed with more autonomic functions for high-efficiency network management. We shall investige the node autonomic decision approaches subject to various untertain information and network cost with node-selfishness provisions, thus attaining the network-wised performance. Firstly, the criteria of the quality of information (QoI) is conceived in terms of the completeness, the timeliness as well as the accuracy of the information. The context information along with the exchanged information will be filtered with respect to the QoI ceriteria. Next, we shall formalate the network costs in terms of mobilitymanagement, topology control, information exchange, selfishness management, as well as the complexity of the algorithm, etc. A tradeoff between network performance and cost is expected to be derived for practical implementations. Furthermore, we shall quatify the relationship between node-selfishness and node's resouces or its to be delived traffic load. It is cruical to make clear what kind of intrinsic and extrinsic information, which influence the degree of the node-selfishness. The prevision of the node's belief for other nodes' selfishness will be conceived for system wide selfishness management. Finally, we shall formulate the impact of fuzzy information, cost and node-selfishness on the solution of the autonomic decision. The dynamic resouce allocation schemes will be proposed for high network spectral efficiency and the content recommendation machanism will be conceived for fast-service-access and throughput-orient applications by virtue of the inteligent approaches, e.g., as predictive matching, self-learning, fuzzy infering, Markov Baysian network etc. It is also interesting to investigate the dynamic resource sharing with respect to the user's quality of experience. The proposed node autonomic decision approach will achieve the globle uitility by virtue of the accumulation of the individual contributions, thus benefiting system-wised performance and simplifying the network management.

针对集中管控大规模动态异构无线网络导致的高复杂度和低效率问题，智能管控赋予无线节点更多自主权并寄望节点发挥更大的个体能动性。本课题探究无线节点如何在各种信息和开销约束的条件下，配合节点自私性管制进行自主决策，以便于全网管控来实现最优网络收益。具体地，依据支撑决策的信息的完整性、时效性和精确性等指标确定决策信息的质量准则，以此对多维多尺度多用途决策信息进行分类处理和筛选。研究网络开销的表征，量化其对自主决策的影响，控制网络开销实现网络性能与开销的折衷。量化节点自私性与其自有资源及其承载业务的关系，推测节点对其它节点的信任形成对网络自私性的认知。研究决策信息、开销控制及自私管制等不确定性因素分别对自主决策的影响关系；基于匹配预测和自学习研究内容分发推荐；基于局部信息交互确定端到端传输路径；面向用户体验研究动态资源分配。无线节点自主决策以个体效用与全局效用对齐为准绳，使全网智能管控成为可能。

应用服务的持续膨胀的多样性需求使得无线网络越来越动态异构和规模宏大，导致网络管控复杂、网络效用低下。针对集中管控大规模动态异构无线网络导致的高复杂度和低效率问题，智能管控赋予无线节点更多自主权并寄望节点发挥更大的个体能动性，是提升网络效用的必然手段。主要研究内容为，依据支撑决策的信息的完整性、时效性和精确性等指标确定决策信息的质量准则，联合优化感知和传输资源，实现感知业务的端到端QoS保障。研究网络开销的表征，量化节点移动、数据帧长度、ARQ机制对路由机制的影响，实现网络性能与开销的折衷。研究网络参数、资源、及外部因素的统一描述，量化节点自私性与其自有资源及其承载业务的关系，推测节点对其它节点的信任形成对网络自私性的认知。基于自学习研究内容分发推荐；基于局部信息交互确定端到端传输路径；面向用户体验研究多时间尺度动态资源分配。重要成果有，实现对网络状态不确定条件下的网络情景的统一描述，支撑网络情景分析和挖掘；在异构复杂大规模网络中，考虑无线链路的不可靠性和网络动态等复杂因素，可通过优化资源配置提升网络效用。可支撑在复杂环境下网络协同的开销控制，支持时敏协同感知业务的综合应用。突破了动态规划算法的高复杂度问题，可保证在瞬时时刻算法收敛单次的算法实现，这允许在计算和通信资源受限的情况下，具有较好的可实现性。课题成果明确了开销与网络参数的关系、确定了网络情景信息的统一表征，实现了网络资源的动态优化，给出了优化性能界，并结合复杂度分析具有可实现性。是复杂网络和动态管控系统中的关键技术，在网络协同感知、车联网、无人机/机器人组网、工业互联网等领域具有重要的应用价值。