Channel estimation strategies for STAR-RIS-Aided NOMA: robustness to imperfections and complexity trade-offs for 6G wireless systems

Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RIS) have attracted increasing attention as an effective means to improve coverage and spectrum utilization in future wireless systems. This paper examines channel estimation strategies for STAR-RIS-assisted non-orthogonal multiple access (NOMA) networks, highlighting both estimation accuracy and computational burden. Five representative schemes are considered: time-switching (TS), ON/OFF training, energy-splitting (ES) in ideal and practical forms, and a simplified two-phase method. The normalized mean square error (NMSE) expressions are derived and compared, together with their pilot overhead and arithmetic complexity. To reflect realistic deployment, the analysis is extended to scenarios with imperfect channel state information (CSI) due to pilot contamination and noise mismatch, as well as residual interference arising from imperfect successive interference cancellation (SIC). Numerical results show that ES-based approaches achieve substantially lower NMSE than TS and ON/OFF, maintaining up to an order-of-magnitude improvement under strong CSI errors. The two-phase ES method achieves a favorable balance, offering near-ideal performance with reduced complexity. System-level evaluations further reveal that NOMA-STAR-RIS provides higher ergodic sum rate and spectral efficiency than its OMA counterpart, and the relative gain becomes more pronounced when imperfections are present. These results demonstrate the robustness and efficiency of STAR-RIS channel estimation under NOMA and confirm its potential for beyond-5G and 6G networks.