光纤通信与大容量高频无线通信深度融合是未来第六代移动通信（6G）的核心技术底座，对于构建“沉浸式通信、泛在连接、通信人工智能（AI）一体化”等6G典型场景具有重要意义。文章梳理了优化光纤无线融合传输系统架构和提升频谱效率的主流技术及其实现方案，对研发团队在这些方面取得的部分进展进行了总结。首先，面向新一代沉浸式通信的大容量需求，借助商用数字相干光模块（DCO），提出了一种“光纤－无线－光纤”一体融合传输系统新型架构，率先完成了光子太赫兹100/200/400 GbE实时无线传输通信实验，最高实现了2×240.558 Gbit/s的线路速率；其次，面向覆盖范围广和灵活部署的应用场景，将数字副载波复用（DSCM）技术引入光纤无线融合接入系统，文章设计并搭建出同时支持最多32路固定宽带接入和32路W波段毫米波无线接入的点对多点（P2MP）100 Gbit/s相干无源光网络（PON），能够灵活调整速率且便于后续迭代升级；最后，面向通信AI一体化需求，提出了一种基于似然感知的矢量量化（VQ）变分自编码器（VAE），基于AI技术对光纤无线融合通信系统进行端到端优化，在无需太赫兹功率放大器的情况下，成功演示了净速率为366.4 Gbit/s的双偏振（DP）2×2多输入多输出（MIMO）太赫兹信号6.5 m无线传输和20 km标准单模光纤（SSMF）传输。上述技术在未来6G典型场景中具有巨大的应用潜力，此外，文章还从大容量、长距离、集成化和智能化等方向对超100 Gbit/s光纤无线融合传输技术进行了展望。

The deep integration of optical fiber communications and high-capacity, high-frequency wireless communications is a core technology for the future 6th Generation Mobile Communication (6G). It is crucial for establishing typical 6G scenarios such as immersive communications, ubiquitous connectivity, and integrated communications Artificial Intelligence (AI). This paper reviews mainstream technologies and solutions for optimizing the architecture and improving the spectral efficiency of fiber-wireless integrated transmission systems. It also summarizes the progress made by our team in these areas. Firstly, to meet the high-capacity needs of next-generation of immersive communications, a novel architecture for seamless "fiber-wireless-fiber" integrated transmission system is proposed using commercial Digital Coherent Optics (DCO). It marks a pioneering achievement in real-time wireless transmission at 100/200/400 GbE with photonics, achieving a maximum line rate of 2×240.558 Gbit/s. Secondly, for application scenarios requiring broad coverage and flexible deployment, the Digital Sub Carrier Multiplexing (DSCM) technology is incorporated into the optical fiber-wireless convergence access system. It results in a coherent Passive Optical Network (PON) supporting up to 32 channels of fixed broadband and 32 channels of W-band millimeter-wave wireless access. This coherent PON could reach a rate of 100 Gbit/s and is adaptable for flexible rate adjustments and future upgrades. Finally, addressing the demand for communication AI integration, a likelihood-aware Vector-Quantization (VQ) Variational Auto Encoder (VAE) is proposed based on AI technology for end-to-end optimization of fiber-wireless integrated communication systems. Without the need for terahertz power amplifiers, we successfully demonstrate the wireless transmission of Dual-Polarization (DP) terahertz signal at a net rate of 366.4 Gbit/s over 6.5 m 2×2 Multiple Input Multiple Output (MIMO) wireless link and 20 km Standard Single Mode Fiber (SSMF) link. These technologies exhibit tremendous potential for future 6G applications. Moreover, we briefly explored the possibilities for beyond 100 Gbit/s fiber-wireless integrated transmission technology, focusing on high capacity, long distance, integration and intelligentization.