基于机会移动网络的移动数据卸载关键技术研究

Recently, due to its promising potentials, mobile data offloading through Opportunistic Mobile Networks has attracted much research attention. Most of existing works investigated how to select initial seeds to maximize the offloading amount, without considering the impact of data freshness on the offloading performance. Furthermore, existing works mainly considered mobile data offloading problem in the collaborative environment, without considering the impact of node selfishness on the offloading performance. In this project, we attempt to use the approach of theoretical analysis and experimental verification to study the key problems mentioned above. Specifically, we will first model the mobile data offloading problem to ensure data freshness, and derive the optimal data injection strategy. Based on the optimal strategy, we will propose an effective initial seeds selection algorithm. Then, on the base of the previous research, we further study the multiple mobile data offloading problem with different node interests and limited node buffer size. To solve this problem, we propose effective initial seeds selection algorithm and node caching strategy based on the defined data utility. Finally, we study the mobile data offloading problem with considering node selfishness, and design a dynamic social trust model to evaluate the social relationship among nodes. Based on the proposed dynamic social trust model and detected temporal communities, we design an effective mobile data offloading strategy to increase the offloading performance in the selfish environment. This project would provide important theoretical and technical supports for the research and application of mobile data offloading through Opportunistic Mobile Networks.

近年来，基于机会移动网络的数据卸载受到大量研究者重视，具有广泛的应用前景。大部分现有工作研究如何选取种子节点以使数据卸载量最大，没有考虑数据新鲜度对于数据卸载性能的影响。另外，现有研究主要考虑协作环境下的数据卸载问题，没有考虑节点自私行为对于数据卸载性能的影响。因此，本项目拟采用理论分析和实验验证相结合的途径，对上述关键问题展开研究。项目首先对保证数据新鲜度的数据卸载问题进行建模，推导出最优数据注入策略；基于最优策略，设计有效的种子节点选择算法。然后，进一步研究节点兴趣不同和缓存有限的多数据卸载问题，根据数据效用值设计有效的种子节点选择算法，以及每个数据和节点的缓存策略。最后，研究考虑节点自私行为的数据卸载问题，设计动态社交信任模型评估节点间的社交关系；基于提出的模型和检测的暂态社团设计有效的数据卸载策略。本项目的研究将为基于机会移动网络的数据卸载技术的研究与应用提供重要的理论和技术支撑。

项目组按照预定的研究计划对所提的各项研究内容展开了研究。首先，调研了已有的移动数据卸载研究成果，特别是对种子节点选择、边缘缓存、激励机制、深度强化学习方法等进行了分析、总结。然后，在此基础上，分析初始种子选择的最优策略，提出了基于新鲜度感知的种子节点选择方法。同时，根据内容流行度的动态特性和内容替换策略，设计了一种分层边缘缓存体系结构，并提出了基于分布式多智能体强化学习的边缘缓存管理方法。最后，基于预测的节点社交关系和反向拍卖机制，提出了一种基于深度强化学习的内容缓存算法以最大化服务提供商节省的成本，同时提出了一种基于标准VCG 框架的支付规则保证了个体合理性和真实性。利用Python做了大量的仿真实验，对所提出的算法进行了性能评估，通过比较实验验证所提出的算法的优越性。按照这一技术路线，项目组经过4年的努力，已在包括IEEE Journal on Selected Areas in Communication, IEEE Transactions on Parallel and Distributed System, IEEE Transactions on Cloud Computing, IEEE Transactions on Wireless Communications, IEEE MASS等在内的各类国际权威期刊和重要会议上发表论文27篇，其中SCI论文18篇，EI检索国际会议论文8篇,入选ESI 高被引论文 1篇，超额完成预定的各项研究内容。