This paper discusses how to optimize the phase shifts of intelligent reflecting surface (IRS) to combat channel fading without any channel state information (CSI), namely blind beamforming. Differing from most previous works based on a two-stage paradigm of first estimating channels and then optimizing phase shifts, our approach is completely data-driven, only requiring a dataset of the received signal power at the user terminal. Thus, our method does not incur extra overhead costs for channel estimation, and does not entail collaboration from service provider, either. The main idea is to choose phase shifts at random and use the corresponding conditional sample mean of the received signal power to extract the main features of the wireless environment. This blind beamforming approach guarantees an $N^2$ boost of signal-to-noise ratio (SNR), where $N$ is the number of reflective elements (REs) of IRS, regardless of whether the direct channel is line-of-sight (LoS) or not. Moreover, blind beamforming is extended to a double-IRS system with provable performance. Finally, prototype tests show that the proposed blind beamforming method can be readily incorporated into the existing communication systems in the real world; simulation tests further show that it works for a variety of fading channel models.

翻译：本文探讨如何在不依赖任何信道状态信息（CSI）的情况下，优化智能反射面（IRS）的相位偏移以对抗信道衰落，即盲波束赋形。与大多数基于“先估计信道、后优化相位偏移”的两阶段范式不同，我们的方法完全由数据驱动，仅需用户终端处的接收信号功率数据集。因此，该方法无需额外开销进行信道估计，也无需服务提供商的协作。核心思想是随机选择相位偏移，并利用对应的接收信号功率条件样本均值来提取无线环境的主要特征。这种盲波束赋形方法可保证信噪比（SNR）实现$N^2$倍的提升，其中$N$为IRS的反射单元（RE）数量，且无论直射信道是否为视距（LoS）信道均适用。此外，该盲波束赋形方法被扩展至双IRS系统，并具有可证明的性能。最后，原型测试表明，所提出的盲波束赋形方法可便捷地集成至现实世界的现有通信系统中；仿真测试进一步表明，该方法适用于多种衰落信道模型。