Pervasive Wireless Intelligence Beyond the Generations

据国际电信联盟估计，全球只有63%的人口可以上网，尽管4G已经覆盖了近88%的人口。基于此，本课题的目标是在下一代空-空-地综合网络（SAGIN）的背景下，通过寻找帕累托最优解来优化无线网络的效率，这构成了一个有前途的架构，将无线宽带红利扩展到迄今未连接的30亿人。本项目重点研究了以下问题：1）面对雷达与通信不可避免的共存，如何在SAGIN中实现雷达与通信的一体化？2)如何在不新建基站的情况下扩大覆盖范围？3)如何使SAGIN达到Shannon-Hartley信道容量定理所预测的近容量性能？在此背景下，该项目在以下方面取得了新的突破：1）为集成传感和通信（ISAC）提出了新的波形，这将是第一个可以改善这两种功能的解决方案。2)针对SAGIN中ISAC的覆盖范围，在双选择性衰落条件下对可重构智能表面（RIS）的新材料进行了优化。该项目建议免除RIS信道估计，这有利于在构思RIS辅助高移动性SAGIN的相干/非相干自适应性方面取得重大突破。3)新的深度学习（DL）工具被设想用于解决SAGIN的多目标优化问题。具体来说，我们的定制DL架构将针对任何给定的纠错码进行智能调整，以便SAGIN可以实现接近容量的性能，为超越几代人的普遍无线智能铺平道路。

The International Telecommunication Union estimates that only 63% of the world population have Internet access, despite the fact that almost 88% have already been covered by 4G. This clearly indicates the urgency in a paradigm shift from pursuing data rate or coverage to the global optimization of the entire wireless network.Motivated by this, the objective of this project is to optimize the wireless efficiencies by finding the Pareto-optimal solutions in the context of next-generation space-air-ground integrated networks (SAGIN), which constitutes a promising architecture that extends wireless broadband dividend to the hitherto unconnected 3 Billion. This project places emphasis on the following questions: 1) In the face of their inevitable coexistence, how can radar and communication both benefit from their integration in SAGIN? 2) How to extend their coverage without building new base stations? 3) How to achieve the near-capacity performance for SAGIN that is predicted by the Shannon-Hartley's channel capacity theorem?Against this background, this project breaks new grounds in the following aspects: 1) New waveforms are proposed for integrated sensing and communication (ISAC), which will be the first solutions that can improve both functionalities. 2) New materials of reconfigurable intelligent surfaces (RIS) are optimized in the face of doubly selective fading in order to extend the coverage of ISAC in SAGIN. This project proposes to dispense with RIS channel estimation, which facilitates a major breakthrough in conceiving coherent/non-coherent adaptivity for RIS assisted high-mobility SAGIN. 3) New deep learning (DL) tools are conceived for solving the multiobjective optimization problems of SAGIN. Specifically, our bespoke DL architecture will be intelligently tuned for any given error correction code, so that the near-capacity performance can be achieved for SAGIN, paving the way for pervasive wireless intelligence beyond the generations.