面向智能服务的移动边缘智能网络及高效感知-接入-处理机制研究

随着5G无线网络、物联网、人工智能等技术不断成熟并加速融合，许多新兴的智能服务模式不断涌现，将在智慧城市、智能制造、智慧农业等众多领域得到广泛应用。为此，本项目研究面向智能服务的新型移动边缘智能网络体系架构，采取“边缘计算赋能、端边云协同、人工智能驱动”的创新思路，有机协同端边云多层次分布式资源，推进计算、通信、智能的深度融合，以解决感知-接入-处理多层面的关键挑战。首先，项目提出了智能服务应用结构与时空动态多维资源的精准感知方法，为性能优化提供高精度信息。其次，项目提出了边缘智能驱动的高可靠传输与大规模接入技术，为高质量数据传输提供保障。项目进一步设计了端边云多层次协同的高效智能处理机制，为实时分布式任务处理与多层次联邦学习提供支撑。最后，项目通过智慧牧场与智能工厂示范应用进行算法测试与功能验证。

As 5G wireless networks, Internet of things, artificial intelligence and other technologies continue to mature and begin to converge rapidly，many new smart services are emerging, which will be widely used in many application domains including smart cities, intelligent manufacturing and intelligent agriculture. For well supporting the smart services, in this proposal we develop a novel mobile edge intelligent network architecture, and leverage the novel design principles of “Enabled by Edge Computing, Synergy of Device-Edge-Cloud, and Powered by Artificial Intelligence”, in order to propel the deep convergence of computing, communication and intelligence for solving the key challenges from the perception-access-processing perspectives. First, we propose accurate perception methods for acquiring the information of the smart service application structures and the multidimensional resources with spatial-temporal dynamics, which are critical for performance optimization. Second, we propose edge intelligence driven approaches for highly reliable transmission and massive access, which guarantee high-quality data transmissions. We further design efficient intelligent task processing mechanisms, aiming at provisioning the capabilities of real-time distributed computing and hierarchical federated learning. Finally, we evaluate the performance of the proposed algorithms and mechanisms above through pilot applications of smart ranch and intelligent factory.