Two-Timescale Design for STAR-RIS Aided NOMA Systems
Chenyu Wu,Changsheng You,Yuanwei Liu,Shuai Han,Marco Di Renzo
DOI: https://doi.org/10.1109/tcomm.2023.3321737
IF: 6.166
2024-01-01
IEEE Transactions on Communications
Abstract:Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs) have emerged as a promising technology to reconfigure the radio propagation environment in the full space. Prior works on STAR-RISs have mostly considered the energy splitting operation protocol, which has high hardware complexity in practice. Moreover, the full and instantaneous channel state information (CSI) is always assumed available for designing the STAR-RIS nearly passive beamforming, which, however, is practically difficult to obtain due to the large number of STAR-RIS elements. To address these issues, we study the mode switching design in STAR-RIS aided non-orthogonal multiple access (NOMA) communication systems. Moreover, two efficient two-timescale (TTS) transmission protocols are proposed for different channel setups to maximize the respective average achievable sum-rate. Specifically, 1) for the case of line-of-sight (LoS) dominant channels, we propose the beamforming-then-estimate (BTE) protocol, where the long-term STAR-RIS transmission and reflection coefficients are optimized based on the statistical CSI only, while the short-term power allocation at the base station (BS) is designed based on the estimated effective fading channels of all the users; 2) for the case of rich scattering environments, we propose an alternative partition-then-estimate (PTE) protocol, where the BS first determines the long-term STAR-RIS surface-partition strategy based on the path-loss information only, with each subsurface being assigned to one user; and then the BS estimates the instantaneous subsurface channels associated with the users and designs its power allocation and STAR-RIS phase-shifts accordingly. For the two proposed transmission protocols, we further propose efficient algorithms to solve the respective long-term and short-term optimization problems. Moreover, we show that both proposed transmission protocols substantially reduce the channel estimation overhead as compared to the existing schemes based on full instantaneous CSI. Last, simulation results validate the superiority of our proposed transmission protocols as compared to various benchmarks. It is shown that the BTE protocol outperforms the PTE protocol when the number of STAR-RIS elements is large and/or the LoS channel components are dominant, and vice versa.