Channel Training for RIS-Aided Indoor Terahertz MIMO Systems

Reconfigurable intelligent surface (RIS) is a powerful tool in assisting large-scale tera-hertz (THz) communication, especially when the line-of-sight (LoS) propagation link is blocked. However, when the base station (BS) antenna array and RIS panel are very large, the THz channel shifts from the well-known far-field model to the less-investigated near-field model. Conventional far-field based channel training mismatches with the near-field channel characteristics, which leads to significant performance loss. In this letter, we investigate the near-filed channel training for an indoor RIS aided THz communication system consisting of a THz BS, an RIS, and an user equipment (UE) that are geographically close to each other. A BS channel training method is designed based on particle swarm optimization (PSO), where the unified near-field channel model is accommodated. Simulation results verify the effectiveness of the proposed channel training. Moreover, with RIS illumination, the UE received signal-to-noise ratio (SNR) can be further improved under imperfect channel and localization.

I. Introduction

In recent years, wireless data traffic has been increasing tremendously with the development of mobile networks [1]. To meet the requests for ultra high data rate, higher frequency and larger spectrum should be utilized [2]. Tera-hertz (THz) band needs to be considered and deployed in future communication systems to support ample spectrum, massive connectivity and ultra-dense network [3]. THz has been recognized to be applicable for indoor scenarios [4]. A fundamental difference with the outdoor communication is the characterization of channels [5]. For long-range case, the channel can be represented by planar-wave approximation, i.e., the combinations of steering vectors which are related with spatial angles [6]. However, for indoor scenario, the approximation brings large deviation [7]. In this case, the beam alignment in near-field range is different.