信息不确定下边缘算网协同动态资源管理研究

边缘计算作为下一无线网络关键支撑技术，可提供低时延、本地化的计算与内容服务，成为近年研究热点。良好的算网协同资源分配可有效提升边缘计算网络的资源利用率，降低网络开销。现有资源分配方法主要关注无线与计算资源协同且都依赖于精确的网络状态参数和完备的开销模型，较少考虑算网复杂耦合、网络参数及模型动态不确定等问题，实际应用有局限。因此，为提高算网协同资源管理的有效性和鲁棒性，降低网络运行开销，本课题将针对信息不确定下算网协同资源分配问题展开研究。主要研究包括：1.探究不确定参数对算网协同的影响并基于鲁棒性优化设计资源分配方案；2. 基于在线凸优化研究目标模型不确定下在线算网协同机理与高效资源分配算法；3. 研究转移概率参数不确定下算网协同边缘服务推送与迁移机理。本课题为边缘计算网络资源协同共享、开销优化及服务质量保证提供理论基础与技术支持，对推动算网协同实际应用具有重要的理论意义。

Edge computing, as a key enabling technology for the next generation wireless network, can provide low-latency and localized computing as well as content services and become a hot topic in recent years. A suitable scheme of collaborative resource allocation in an edge computing network can effectively improve the resource utilization of and reduce network costs. Existing resource allocation methods mainly focus on the cooperation between wireless and computing resources and rely on certain network state parameters as well as well-defined cost models. However, the complex coupling of computing networks, uncertain network parameters, and dynamic models have been largely ignored, which results in limited practical implementations. In order to improve the effectiveness and robustness of collaborative resource allocation in edge computing networks and reduce network operating costs, this research studies the issues and solutions of collaborative resource allocation in edge computing networks with uncertain information as follows. 1. The influence of uncertain parameters on the collaboration of edge computing networks is explored. Resource allocation schemes based on robust optimization are developed accordingly. 2. Based on the online convex optimization, considering uncertain and unknown network costs, an online collaboration mechanism with efficient resource allocation algorithms are studied. 3. The mechanism of pushing and migration of edge services in edge computing networks without certain transition probability parameters is explored. This research provides theoretical and technical support for collaborative research sharing, cost optimization and service quality assurance in edge computing networks. It can promote the practical application of collaboration in edge computing networks significantly.