FARS: Elevating Rate-Splitting Multiple Access in Non-Territorial Networks With Intelligent Fluid Antenna System

Non-Terrestrial Networks (NTNs) are essential for 6G communication, providing connectivity in remote areas and addressing coverage gaps. However, NTNs face challenges such as limited frequency spectrum and significant channel uncertainty due to dynamic operating environments. This paper proposes Fluid Antenna-enabled Rate-Splitting system (FARS), an integration of Fluid Antenna Systems (FAS) with Rate-Splitting Multiple Access (RSMA), to address these challenges. FAS reconfigures antenna shapes and positions, optimizing beamforming while enhancing diversity and multiplexing gains, making it well-suited for NTNs. RSMA is employed to better manage interference and improve reliability, particularly under imperfect Channel State Information (CSI), which is common in NTN scenarios. Under the proposed FAS-enabled RSMA framework, we formulate a stochastic optimization problem to jointly optimize antenna port locations, precoder design, and rate allocation with the aim of maximizing user fairness. A deep learning-based solver utilizing Mamba is proposed to efficiently and scalably solve the formulated problem. Additionally, an unsupervised transfer learning framework is proposed to ensure the adaptivity and flexibility of the Mamba-based solver. Simulation results demonstrate that the proposed FAS-enabled RSMA system significantly outperforms traditional approaches, achieving superior spectral efficiency and enhanced reliability, validating its effectiveness for NTN communication systems.