随着智能手机的普及和短距离无线通信技术的快速发展，人类社会进入群智感知时代。普通用户利用其随身携带的移动智能设备作为基本感知单元，通过有意识或无意识的群体协作实现对感知任务的分发和海量感知数据收集，从而完成大规模、复杂的社会感知任务。由于群智感知网络无需专门部署，且具有便捷、灵活、传输成本低等优点，一经出现便备受关注；但携带用户的频繁移动性、设备资源的受限性以及网络拓扑的高度时变性等特征都为群智感知网络中海量数据的传输带来极大的挑战，亟需高效的路由算法对用户间的数据投递进行约束、引导、优化和支撑。然而，现有机会路由算法一贯采用的贪婪传输模式则存在网络资源不均衡、节点负载不均衡、网络性能不均衡等问题，从而使其无法直接应用于群智感知网络的任务数据收集。因此，本文围绕群智感知网络的高效数据传输问题，从其特定的网络环境入手，分别从资源均衡、负载均衡和性能均衡等三个方面对感知数据的路由问题展开研究，并由此提出相应的模型、算法和机制。具体来讲，本文主要工作和研究成果体现在以下几个方面： 1.针对传统谱聚类算法中需要预先指定聚类个数而存在的聚类精确度不足等问题，结合模块度的概念，依据最优的聚类划分应具有最大模块度的特征，提出自适应谱聚类算法，实现对网络中移动节点快速、准确的聚类划分，从而完成对节点社区属性的精确提取，为高效机会路由算法的设计奠定基础。 2.针对当前机会数据传输中终端设备资源的受限性和多备份投递造成的数据传输冗余性之间的矛盾，深入分析了移动节点的设备资源对机会数据投递性能的影响，并在此基础上，将节点的剩余能量、剩余缓存等资源属性以及节点的中心度、社区等社会属性通过效用函数进行融合，以最大化均衡网络的资源利用率为目标，提出了基于自适应谱聚类的资源均衡机会路由协议。该协议能够在给定投递质量约束下有效提升节点资源消耗的均衡度，提高节点存活率，延长网络寿命，进而保证数据投递的性能。 3.针对参与用户要求的投递负载公平性和数据传输质量要求的中继节点选择性之间的矛盾，分析了可能导致节点传输负载不均衡的各个因素；通过量化节点当前的传输负载、节点活跃度、节点剩余序列长度和节点间的相似度等因素，将数据包扩散过程建模为一个多属性决策制定问题；进而，以上述模型为基础，结合最优副本停止理论，提出基于多属性决策的负载均衡机会路由协议。该协议旨在保证给定投递质量约束的同时，均衡节点间的负载分布，提高节点在数据传输过程中任务承担的公平性，从而避免个别节点由于负载过高而成为整个网络的瓶颈，进而影响网络的投递性能。 4.针对现实应用对数据的传输可靠性、传输时延和投递代价等综合质量需求与现有算法的投递性能指标极度不均衡之间的矛盾，分析了网络中节点移动的规律，将两节点的相遇次数模型化为一个参数为两节点间接触频率的泊松过程；据此，预测数据包在任意节点对之间的投递成功概率和投递时延，考虑泛在的多节点接触，将数据包扩散过程建模为一个多人合作博弈问题，以性能均衡度最大化为博弈目标，提出基于多人合作博弈的性能均衡机会路由协议。该协议旨在针对特定应用需求在保证投递性能基本约束的前提下在数据可靠性、投递实时性和传输有效性之间达到一个折中，从而提高各性能指标间的均衡。 5.为有效评估前述工作，在Visual Studio2010上利用VC++构建群智感知网络框架，并分别在各种真实数据集和TVCM合成数据集上，针对三种不同应用场景对本文所提协议和现有相关协议进行大量仿真和性能评估，从多角度验证本文所提协议的合理性和有效性，为群智感知网络的广泛应用提供技术和理论的支撑。 关键词：群智感知网络；机会路由；自适应谱聚类；多属性决策；多人合作博弈 论文类型：应用基础研究 选题来源：国家自然科学基金(61772175)

 With the popularity of smart mobile phones and the rapid development of short distance wireless communication technologies, the human beings have entered the era of crowd sensing. Mobile smart devices carried by common users can be acted as basic sensing units, achieve sensing task distribution and sensing data collection, and complete large-scale and complicated social sensing tasks by utilizing the collaboration among nodes consciously or unconsciously. For the advantages of no need for special deployment, convenience, flexibility and low transmission cost, crowd sensing network has attracted much attention once it appears. But, the frequent mobility of equipment, the limitation of equipment resources and the highly time-varying network topology have brought great challenges to the transmission of massive data in such network. Efficient routing algorithms are urgently needed to constrain, guide, optimize and support data delivery among users. Nevertheless, the greedy transmission mode usually adopted in existing opportunistic routing is prone to lead to a series of problems, such as network resource utilization imbalance, node load imbalance, network performance imbalance and so on, which make them unable to be applied directly to the collection of sensing data in crowd sensing network. Thus, starting from the specific networking environments of crowd sensing network, we combine with the latest research on opportunistic routing in the world, and conduct our research from three aspects including resource balancing, load balancing and performance balancing. As a result, some model, strategy and scheme have been proposed accordingly. Specifically, our main work and contributions are as follows: 1.In order to solve the problem of insufficient clustering accuracy caused by the cluster number which is specified in advance in traditional spectral clustering algorithm, an adaptive spectral clustering algorithm is proposed based on the feature that the optimal clustering division should have the maximum modularity, combining with the concept of modularity. With this algorithm, mobile nodes in the network can be clustered quickly and accurately and node community attributes can be extracted precisely, which lays the foundation for the design of efficient opportunistic routing algorithm. 2.Aiming at the contradiction between the limited resource of mobile terminal and the multi-backup delivery of sensing data in the current opportunistic routing, the influence of resource restriction on the performance of opportunistic data delivery is deeply analyzed. On this basis, resource attributes such as residual energy, residual cache, and social attributes such as node centrality, community are fused by utility functions. And a resource balanced opportunistic routing based on adaptive spectral clustering is proposed, which aims at maximizing the balance of network resource utilization. This protocol can effectively improve the balance of node resource consumption, improve node survival rate, prolong network life, and ensure the performance of data delivery under given delivery quality constraints. 3.Aiming at the contradiction between the fairness of delivery load required by users involved and selectivity of relay nodes required by data transmission quality, the factors that may lead to the imbalance of transmission load between the nodes are analyzed. The factors such as current transmission load of nodes, node activity, residual sequence length, and similarity between nodes, have been quantified and the data packets diffusion process is modeled as a multi-attribute decision-making problem. Then, based on the above model, combined with the optimal copy stopping theory, a load-balancing opportunistic routing based on multi-attribute decisionmaking is proposed. This purposed protocol is to balance the load distribution between nodes while guaranteeing the given delivery quality, and to improve the fairness of the tasks undertaken by the nodes in the process of data transmission, so as to avoid the individual nodes becoming the bottleneck of the whole network because of the excessive load and affect the delivery performance of the network. 4.Aiming at the contradiction between comprehensive quality requirements for data transmission reliability, transmission delay and delivery cost in practical applications and the imbalance between delivery performance indicators of existing algorithms, the law of node movement in the network is analyzed, and the encounter times between any two nodes are modeled as a Poisson process with a parameter that is contact frequency between two nodes. According this, the probability and delivery delay of packets between any nodes pair are predicted and the process of data exchange is modeled as a multi-player cooperative game problem on the basis of the ubiquitous multi-node contact at the same time. Taking the maximization of performance balancing as the goal of the game, a performance-balanced opportunistic routing protocol based on multi-player cooperative game is proposed. The aim of this protocol is to achieve a compromise between data transmission reliability, transmission delay and delivery cost, and improve the balance between performance indicators while guaranteeing given delivery quality. 5. In order to evaluate the work mentioned above effectively, a framework of crowd sensing network is constructed using VC++ in Visual Studio 2010. For three different application scenarios, extensive simulations and performance evaluation are carried out for our proposed protocols and existing related protocols based on various real data sets and a TVCM synthetic trace, and the rationality and validity of the proposed protocols are verified from many aspects. It provides the technical and theoretical support for the wide application of crowd sensing networks. KEY WORDS: Crowd sensing networks; Opportunistic routing; Adaptive spectral clustering; MADM; Multi-player cooperative game Dissertation type: Basic research for application Subject source: National natural science foundation of China(61772175