Improving Energy Efficiency in PIN-Diode-Based RIS-Assisted Wireless Systems

This letter addresses the challenge of maximizing energy efficiency in reconfigurable intelligent surface (RIS)-assisted wireless systems, specifically focusing on RISs constructed with PIN diodes. Prior works often relied on simplified power models, yielding inaccurate results. Recent advancements addressed this issue by proposing more accurate models that capture the power dissipation discrepancies between the ON and OFF states of PIN-diode elements. These models facilitated the formulation of energy efficiency optimization problems for PIN-diode-based RIS-assisted systems. However, existing optimization techniques, which typically combine zero-forcing precoding with semidefinite relaxation or maximum gradient search methods, incur significant computational complexity or provide limited performance gains. Additionally, these approaches are restricted to RISs with 1-bit resolution. To address these limitations, we propose a novel probability learning-based algorithm that leverages the cross-entropy optimization framework. This algorithm is applicable to PIN-diode-based RISs with arbitrary resolution and demonstrably achieves superior energy efficiency with reduced computational complexity across diverse scenarios.