Cell-Free Massive MIMO SWIPT with Beyond Diagonal Reconfigurable Intelligent Surfaces

2026-04-11

AI Summary

This work addresses the resource efficiency bottleneck in cell-free massive MIMO (CF-mMIMO) systems supporting simultaneous wireless information and power transfer (SWIPT). To this end, it introduces, for the first time, a block-diagonal reconfigurable intelligent surface (BD-RIS) into the CF-mMIMO-SWIPT architecture. A novel access point (AP) grouping strategy is proposed: subsets of APs are dedicated exclusively to BD-RIS-assisted energy receivers (ERs), while others serve information receivers (IRs), thereby eliminating time-frequency resource contention. A protective partial zero-forcing precoder is designed to suppress non-coherent inter-group interference, and AP selection, power allocation, and BD-RIS scattering matrix are jointly optimized using statistical channel state information. The solution integrates closed-form performance modeling, heuristic scattering design, successive convex approximation, and deep reinforcement learning. Experimental results demonstrate that, compared with conventional diagonal RIS, the BD-RIS enhances average total harvested energy by up to 7×, achieving a superior trade-off between spectral efficiency and energy harvesting performance.

Technology Category

Intelligent Robots: Multi-Robot SystemsMultiagent Systems: ApplicationsSearch and Optimization: Mixed Discrete/Continuous Search

Application Category

Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environmentsUser Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendationSearch and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

We investigate the integration of beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) into cell-free massive multiple-input multiple-output (CF-mMIMO) systems to enhance simultaneous wireless information and power transfer (SWIPT). To simultaneously support two groups of users-energy receivers (ERs) and information receivers (IRs)-without sacrificing time-frequency resources, a subset of access points (APs) is dedicated to serving ERs with the aid of a BD-RIS, while the remaining APs focus on supporting IRs. A protective partial zero-forcing precoding technique is implemented at the APs to manage the non-coherent interference between the ERs and IRs. Subsequently, closed-form expressions for the spectral efficiency of the IRs and the average sum of harvested energy (HE) at the ERs are leveraged to formulate a comprehensive optimization problem. This problem jointly optimizes the AP selection, AP power control, and scattering matrix design at the BD-RIS, all based on long-term statistical channel state information. This challenging problem is then effectively transformed into more tractable forms. To solve these sub-problems, efficient algorithms are proposed, including a heuristic search for the scattering matrix design, as well as successive convex approximation and deep reinforcement learning methods for the joint AP mode selection and power control design. Numerical results show that a BD-RIS with a group-or fully-connected architecture achieves significant EH gains over the conventional diagonal RIS, especially delivering up to a 7-fold increase in the average sum of HE when a heuristic-based scattering matrix design is employed. Index Terms-Beyond diagonal reconfigurable intelligent surface (BD-RIS), cell-free massive multiple-input multiple-output (CF-mMIMO), deep reinforcement learning (DRL).

Problem

• Enhance SWIPT in CF-mMIMO using BD-RIS for energy and information transfer
• Optimize AP selection, power control, and BD-RIS design for efficiency
• Manage interference between ERs and IRs with protective precoding techniques

Innovation

• BD-RIS enhances SWIPT in CF-mMIMO systems
• Protective partial zero-forcing manages interference
• DRL optimizes AP selection and power control