A Novel Q-stem Connected Architecture for Beyond-Diagonal Reconfigurable Intelligent Surfaces

TL;DR

This paper introduces a Q-stem connected BD-RIS architecture that balances system performance and circuit complexity, along with algorithms for efficient RIS matrix design, demonstrating comparable or improved gains over existing methods.

Contribution

It proposes a novel Q-stem connected BD-RIS architecture and two algorithms for RIS matrix design, optimizing the trade-off between performance and complexity.

Findings

Achieves sum channel gain comparable to fully connected RIS.

LS-based quasi-Newton algorithm outperforms baseline methods.

Reduces circuit complexity while maintaining high performance.

Abstract

Beyond-diagonal reconfigurable intelligent surface (BD-RIS) has garnered significant research interest recently due to its ability to generalize existing reconfigurable intelligent surface (RIS) architectures and provide enhanced performance through flexible inter-connection among RIS elements. However, current BD-RIS designs often face challenges related to high circuit complexity and computational complexity, and there is limited study on the trade-off between system performance and circuit complexity. To address these issues, in this work, we propose a novel BD-RIS architecture named Q-stem connected RIS that integrates the characteristics of existing single connected, tree connected, and fully connected BD-RIS, facilitating an effective trade-off between system performance and circuit complexity. Additionally, we propose two algorithms to design the RIS scattering matrix for a Q-stem connected RIS aided multi-user broadcast channels, namely, a low-complexity least squares (LS) algorithm and a suboptimal LS-based quasi-Newton algorithm. Simulations show that the proposed architecture is capable of attaining the sum channel gain achieved by fully connected RIS while reducing the circuit complexity. Moreover, the proposed LS-based quasi-Newton algorithm significantly outperforms the baselines, while the LS algorithm provides comparable performance with a substantial reduction in computational complexity.