Rate-Overhead Tradeoff for IOS-Aided Beam Training: How Large Codebook Is Enough for the IOS?

Abstract

Intelligent omni-surface (IOS) is a novel metasurface to provide full-dimensional services by jointly engineering its reflective and refractive properties. In this letter, we study an IOS-aided system where the beamforming scheme is developed via beam training with codebooks at the base station (BS), IOS, and users to avoid the extremely high overhead of acquiring perfect channel state information (CSI). A larger codebook size brings a higher peak data rate but requires more spectrum and time resources for beam training, leading to low system throughput. To investigate the tradeoff between the peak data rate and training overhead, we mathematically model the relationship between the codebook size of the IOS and the throughput. Given different user distributions, we discuss the codebook size under the constraint of the asymmetric power of the reflective and refractive signals. Moreover, the influence of the power ratio between reflective and refractive signals on the codebook size is also analyzed. Simulation results verify our theoretical analysis.