移动边缘计算环境下高效可信的协同计算关键技术研究

Mobile edge computing has recently received considerable attention, due to its reduction in computational latency, energy consumption, and the improvement in users’ service experience. Nevertheless, the diversity of computing users and the decentralization of mobile resources bring new challenges in mobile resource management, task allocation and personal privacy preservation, etc. In this context, this project will study cooperative task computation in mobile edge computing. By utilizing algorithm design, optimization theory and game theory, we aim to design efficient and reliable task allocation algorithm, incentive mechanism, and private preservation scheme. The main researches can be summarized as follows: 1) for different cooperative task models, we investigate the execution/cooperation cost, and design cost-aware task allocation mechanism for mobile edge computing; 2) we further study the participation level of service providers as well as the cooperation cost, based on which, we design a participation-level-aware incentive mechanism; 3) to protect the personal privacy of mobile user, this project proposes a novel distributed resource trading mechanism, where resource requesters and owners locally make their decisions using their limited information. Besides that, for centralized resource trading model, we also propose a privacy preservation mechanism based on homomorphic encryption. The research of this project will significantly improve the service experience of mobile users, and realize efficent system management.

移动边缘计算在实现移动计算低延迟、高能效，提高移动用户服务体验上具有显著优势。然而，移动边缘计算中服务节点的多样性、动态性以及资源分布的分散性给移动计算网络资源管理、任务分配、隐私保护等提出了挑战。本项目将使用算法理论、最优化理论以及博弈理论等科学方法，针对移动边缘计算环境中的协同计算关键问题，如任务分配、提高服务节点参与意愿以及保护用户消费隐私等问题，展开研究。具体研究内容包括：1）探索不同协同计算任务模型下服务节点的执行代价、协同开销，设计代价、开销感知的协同计算任务分配策略；2）设计服务意愿、协同开销感知的服务节点激励机制，从而提高边缘计算节点参与协同计算的积极性、诚实性；3）在保护移动用户消费隐私上，分别设计分布式资源交易策略以及基于同态加密技术的移动用户消费隐私保护机制。本项目的研究将有效提高移动边缘计算环境下用户的服务体验，实现系统的高效管理。

本项目以移动边缘计算系统为应用背景，综合考虑边缘计算节点的动态性、计算能力的异构性、资源分布的分散性，以及节点可能存在的网络行为，针对不同协同计算任务模型，设计高效可信的协同计算任务分配方案、服务节点激励策略，以及移动用户的隐私保护机制。首先，项目探索协同计算任务需求，设计移动计算任务分配方案与资源配置策略。其次，针对边缘计算节点参与意愿问题，研究协同计算中资源定价策略以及高效可信的拍卖机制，提高边缘计算节点提供服务的积极性、可靠性。进一步地，为保护移动用户数据隐私，课题进一步优化基于联邦学习的分布式模型训练算法。本项目的研究为移动边缘计算系统的大规模推广应用提供有效的理论基础和技术支撑。在本项目的支持下，项目负责人在国际期刊/会议如IEEE Transactions on Mobile Computing、IEEE Transactions on Network Science and Engineering、IEEE Transactions on Vehicular Technology、IEEE Transactions on Emerging Topics in Computational Intelligence等上发表高水平学术论文15篇，申请发明专利2项目，授权发明专利1项。依托本项目，负责人指导毕业硕士研究生十余名。