On the Performance of ML-RSMA-Aided Cell-Free Massive MIMO System

Leveraging its inherent flexibility in interference management, single-layer rate-splitting multiple access (SL-RSMA) has demonstrated superior rate performance compared to conventional space division multiple access (SDMA). In this work, we extend the SL-RSMA framework to a multi-layer RSMA (ML-RSMA) architecture, specifically designed for downlink cell-free massive multiple-input multiple-output (MIMO) systems. By considering the practical constraints of imperfect channel state information and spatially correlated Rician fading channels, we analytically derive a tight closed-form expression for the achievable downlink rate. To address the computational complexity associated with ML-RSMA, we propose a user grouping algorithm that exploits channel similarity, thereby achieving a favorable trade-off between performance gains and computational efficiency. Extensive numerical results demonstrate that the proposed ML-RSMA scheme significantly outperforms existing multiple access techniques, including generalized RSMA (G-RSMA), SL-RSMA, and SDMA, in terms of achievable rate.