6DMA-FIM-Enabled User-Centric Cell-Free Massive MIMO With Two-Layer RSMA

Abstract

This paper proposes a novel user-centric (UC) cell-free massive MIMO (CF-mMIMO) architecture empowered by six-dimensional movable antennas (6DMA) and flexible intelligent metasurfaces (FIM), integrated with a two-layer rate-splitting multiple access (RSMA) framework. Unlike conventional fixed-position deployments, the proposed architecture jointly exploits large-scale statistical channel characteristics and instantaneous CSI by dynamically optimizing the three-dimensional positions, orientations, and surface morphologies of FIM-enabled 6DMA (6DMA-FIM) access points (APs), while hierarchically mitigating inter- and intra-cluster interference through two-layer RSMA. This unified spatial-electromagnetic-signal-domain design significantly enhances propagation conditions, array gain, and interference suppression in dense UC networks. The resulting joint optimization over antenna motion, metasurface configuration, beamforming, power allocation, and RSMA precoding yields a highly coupled and non-convex high-dimensional problem. To efficiently address this challenge, we develop a Meta-distributional Langevin Soft Actor-Critic with Diffusion-Augmented sampling (Meta-LSAC-DA) framework. The proposed learning architecture leverages meta-learning to extract transferable geometric and electromagnetic priors across heterogeneous channel realizations, incorporates distributional reinforcement learning for uncertainty-aware value estimation, and employs Langevin-based and diffusion-enhanced sampling to ensure stable exploration in constrained continuous action spaces. Numerical results demonstrate that the proposed 6DMA-FIM-assisted two-layer RSMA architecture substantially outperforms conventional fixed-antenna, rigid 6DMA, single-layer RSMA, and SDMA-based CF-mMIMO systems in terms of spectral efficiency (SE), scalability, and robustness. Furthermore, Meta-LSAC-DA achieves faster convergence and superior generalization across diverse user distributions, highlighting its effectiveness for intelligent and adaptive 6G wireless networks.