Machine Learning and Artificial Intelligence for Future Generation Communication Networks

随着第五代移动的系统（5G）的实施和物联网（IoT）的出现，预计连接到互联网的设备数量将在未来几十年内大幅增长。与此同时，机器学习（ML）和人工智能（AI）正在成为在移动的设备上运行的许多应用程序的组成部分。这就引出了两个有趣的问题。首先，随着现代通信网络的复杂性，如何利用ML和AI来改进这些系统的设计？第二，通信网络如何帮助改进分布式ML和AI？ 该研究计划在三个关键领域使用ML和AI开发物联网的尖端技术：a）ML和AI如何帮助改善物理层的定位和认证; B）ML和AI如何用于检测复杂船舶网络中的化学源;以及c）深度神经网络可以如何在IoT设备，边缘节点，以及云服务器，存在引入错误和延迟的通信链路。这项研究提出的基于ML和AI的技术有望显著提高无线网络的速度、性能和动态适应性，从而增强多个行业内物联网设备和基础设施之间的通信。这将对智能城市、智能健康监测以及其他需要分布式传感、定位、认证和深度学习的领域（例如制造业、机器人和自动驾驶汽车）产生深远的影响。

With the implementation of the fifth generation of mobile systems (5G) and the emergence of internet-of-things (IoT), the number of devices connected to the Internet is expected to grow tremendously over the next few decades. At the same time, machine learning (ML) and artificial intelligence (AI) are becoming an integral part of many applications that run on mobile devices. This raises two interesting questions. First, with the complexity of the modern-day communication networks, how can ML and AI be leveraged to improve the design of these systems? And second, how can communication networks help improve distributed ML and AI? This research program develops cutting edge technologies for IoT using ML and AI in three key areas: a) how ML and AI can help improve localization and authentication at the physical layer; b) how ML and AI can be used to detect chemical sources in complex networks of vessels; and c) how deep neural networks can be distributed and trained over IoT devices, edge nodes, and cloud servers in the present of communication links that introduce errors and delay. The ML- and AI-based technologies arising from this proposed research are expected to significantly improve the speed, performance and dynamic adaptability of wireless networks, thereby enabling enhanced communications among IoT devices and infrastructure within multiple sectors. This will have a profound impact in smart cities, smart health monitoring, and other areas where distributed sensing, localization, authentication, and deep learning are required (e.g. manufacturing, robotics and autonomous vehicles).