面向非对称多媒体边缘计算的高效任务迁移机制研究

Driven by the urgent needs of high transmission bandwidth and low processing delay for a large number of multimedia applications, the "edge computing" that can handle these demands rises in recent years. As most existing researches on edge computing cannot guarantee a good performance in the asymmetrically informed network, this project tries to give some answers to this problem by interpreting the influence of the asymmetric network with “task demand”, “edge performance”, and “user trajectory” on the computation offloading for edge computing. Three mechanisms are proposed, including the semantic features aware asymmetric gaming based model, the partially informed edge performance learning based model, and the indeterministic trajectory aware participant recruitment model. These models can help form efficient computation offloading mechanisms for the asymmetric multimedia edge computing system. Subsequently, the project will also deduce the expression of typical performance indicators in the edge computing network environment and build the field multimedia edge computing measurement platform to verify the proposed mechanisms in a typical edge computing scenario. The research results of this project will not only become a necessary supplement to the existing research of computation offloading mechanism for edge computing, but also provide a series of new theoretical foundations for the deployment of next generation communication and computing network.

在海量多媒体应用对于传输带宽和处理时延需求益发苛刻的驱动下，将任务迁移至靠近用户计算服务节点的“边缘计算”悄然兴起，在大大提升处理效率的同时减轻了云端的负荷。考虑到现有边缘计算相关研究在“非对称”网络环境下所体现出的不足，本项目将从解读“任务需求”、“边缘性能”以及“终端轨迹”等的非对称性对边缘计算任务迁移机制所产生各类影响的本质出发，拟分别构建基于语义化特征的异质任务迁移非对称博弈模型、基于非对称边缘服务性能重建的任务迁移预测优化理论框架以及针对不确定轨迹的移动多媒体观测调度机制以指导非对称多媒体边缘计算系统的模块化设计。随后，本项目也将推导边缘计算网络环境中迁移机制典型性能指标的表达式，并利用多媒体任务迁移应用的验证测试平台，在典型场景中对研究成果进行验证。本项目的研究成果不仅将成为现有边缘计算任务迁移机制研究工作的必要补充，也将为物联网与新型计算网络的部署提供一系列新的理论依据。

考虑到现有边缘计算相关研究在“非对称”网络环境下所体现出的不足，本项目从解读“任务需求”、“边缘性能”以及“终端轨迹”等的非对称性对边缘计算任务迁移机制所产生各类影响的本质出发，分别构建基于语义化特征的异质任务迁移非对称博弈模型、基于非对称边缘服务性能重建的任务迁移预测优化理论框架以及针对不确定轨迹的移动多媒体观测调度机制以指导非对称多媒体边缘计算系统的模块化设计。随后，本项目也利用多媒体任务迁移应用的验证测试平台，在典型场景中对研究成果进行了相应的验证。依据本项目的研究成果，申请了2项具有自主知识产权的发明专利、发表了11篇国际重要期刊和国际重要学术会议论文。本项目的研究成果不仅将成为现有边缘计算任务迁移机制研究工作的必要补充，也将为物联网与新型计算网络的部署提供一系列新的理论依据。