Transformer-Empowered Parallel Channel Prediction for Fast-Paced and Dynamic RIS-Aided Wireless Communication Systems
Guoxu Xia, Hongjun Liu, Ken Long
Abstract
In 6th generation (6G) communication systems, reconfigurable intelligent surface (RIS) is envisioned as a revolutionary technology for future wireless communication networks. However, in high-speed mobile RIS-aided system, the acquisition of channel state information (CSI) poses formidable challenges attributed to adverse transmission environments and sequential error propagation. In this letter, to surmount these challenging problems, we propose a two-timescale channel prediction framework and introduce a transformer-based parallel channel prediction scheme. The sparse-connected multi-causal convolution attention transformer (SC-MCCATrasformer) can simultaneously decompose the prior cascaded channel and parallelly predict the next several channels. Furthermore, relying on the improved attention module, more local context features can be extracted to predict more realistic channels. Simulation results validate the superior performance of the proposed scheme over other channel prediction methods. Moreover, our proposed methods demonstrated enhanced accuracy in CSI acquisition.