Channel Estimation for RIS-Assisted Massive MIMO with Diffusion Model

Reconfigurable intelligent surface (RIS) has received wide-spread attention as a critical enabling technology for next generation wireless communications. Accurate channel state information (CSI) is fundamental for RIS to reach its full potential. Thanks to the powerful latent representation from data, generative artificial intelligence (AI) has the potential as a strong driver for the highly intelligent and digital twin of 6G. In this paper, we propose a generative diffusion model (DM)-based cascaded channel estimation (CE) algorithm through unsupervised learning for RIS-aided massive multiple-input multiple-output (MIMO) systems. The DM can effectively capture the implicit prior of the cascaded channel, and the received signal is used as conditional information to precisely guide the channel recovery. Simulation results validate that the proposed algorithm achieves superior estimation accuracy with reduced pilot overhead.