应我校电信学院于启月教授邀请，浙江大学信息与电子工程学院蔡云龙教授将在2024年8月16日上午9:00，在科创大厦基础科研楼-J503会议室带来讲座“助力6G时代MIMO通信：先进的学习和优化技术”，探讨交流其团队在下一代 MIMO通信系统中机器学习和优化技术方向的最新研究进展，欢迎感兴趣的师生前来交流学习。
报告人简介：
蔡云龙于2010年在英国约克大学获得电子工程博士学位。2010年至2011年期间，他在法国CNAM大学电子与通信实验室担任博士后研究员。自2011年2月起，他在浙江大学信息与电子工程学院任职，现为教授。他曾在佐治亚理工学院、麦吉尔大学和加利福尼亚大学欧文分校进行学术访问。其研究兴趣包括通信中的信号处理与优化、多天线系统的收发器设计、无人机通信以及通信中的机器学习。他在JSAC、SPL、TWC、TCOM、JSTSP和TVT等著名IEEE期刊上发表了130多篇论文，其中若干篇被列为基本科学指标（ESI）前1%的高被引论文。此外，他还在国际会议上发表了超过100篇论文。蔡教授目前担任TCOM期刊的副编辑，以及SPL的高级领域编辑。他曾担任IEEE JSAC特刊“Next Generation Advanced Transceiver Technologies”的首席客座编辑。他还曾担任2022年10月在杭州举办的第十八届IEEE ISWCS的大会主席。
报告内容：
Title: Enhancing MIMO Communications in the 6G Era: Advanced Learning and Optimization Techniques
Abstract: In the 6G era, innovative MIMO technologies are crucial for enhancing system robustness, reliability, and transmission rates while addressing spectrum shortages, complexity, and network delays. This talk will present our recent advancements in machine learning and optimization techniques for next-generation MIMO systems.
First, we introduce a deep-unfolding framework that uses an iterative algorithm-induced deep-unfolding neural network in matrix form to tackle communication system challenges. This approach maximizes the sum-rate in massive MIMO systems via beamforming design by transforming iterative algorithms into a layer-wise structure with trainable parameters. We extend this deep-unfolding algorithm to jointly design channel acquisition and hybrid analog-digital (AD) beamforming through end-to-end learning in mmWave MIMO systems.
Second, we propose a double-reconfigurable intelligent surface (RIS)-assisted radar-communication coexistence system. Deploying two RISs enhances communication signals and reduces interference. Our goal is to optimize RIS and radar beamforming to maximize communication performance while preserving radar detection. We simplify this problem using auxiliary variables and develop a penalty dual decomposition (PDD)-based optimization algorithm.
Lastly, we explore a semantic-centric MIMO system with our Feature Allocation for Semantic Transmission (FAST) framework. FAST addresses the mismatch between feature importance and wireless channels by integrating an importance evaluator, employing channel prediction for future CSI estimation, and allocating transmission time slots to features. This framework enhances semantic communication performance in image transmission, applicable to both precoding-free and precoding-based MIMO systems in the space-time domain.
