GA-Optimized BD-RIS-Assisted Network Slicing for Uplink eMBB and URLLC Coexistence

This paper investigates the impact of Beyond-Diagonal Reconfigurable Intelligent Surfaces (BD-RIS), considering single-, group-, and fully-connected architectures, on the achievable rate region of an uplink system with coexisting eMBB and URLLC services under OMA, NOMA, and RSMA-based RAN slicing. A Genetic Algorithm (GA)-based optimization method is proposed to design the BD-RIS beamforming. Numerical results demonstrate that BD-RIS significantly improves system performance, even under random phase configurations. In particular, the fully-connected architecture achieves the highest eMBB sum rates, while RSMA provides superior interference management capabilities in multiuser scenarios.