A Federated Deep Learning Framework for Cell-Free RSMA Networks

TL;DR

This work tackles resource allocation in downlink cell-free RSMA networks by jointly optimizing AP selection and RSMA precoding under a max-min rate objective. It introduces a three-block, federated deep reinforcement learning framework (FDRL) that combines PCA-based AP assignment with DRL-based precoder design, executed cooperatively by distributed AP agents and a central CPU. The AP-selection step uses PCA to linearize the objective and an ILP solved by standard solvers, while the precoder design operates in two DRL phases, with federated aggregation to preserve data privacy and reduce CPU burden. Simulation results show that FDRL-RSMA-MIMO closely matches centralized DRL performance while offering improved security and lower processing demand, indicating strong practical potential for scalable, privacy-preserving resource management in future wireless networks.

Abstract

Next-generation wireless networks are poised to benefit significantly from the integration of three key technologies (KTs): Rate-Splitting Multiple Access (RSMA), cell-free architectures, and federated learning. Each of these technologies offers distinct advantages in terms of security, robustness, and distributed structure. In this paper, we propose a novel cell-free network architecture that incorporates RSMA and employs machine learning techniques within a federated framework. This combination leverages the strengths of each KT, creating a synergistic effect that maximizes the benefits of security, robustness, and distributed structure. We formally formulate the access point (AP) selection and precoder design for max-min rate optimization in a cell-free MIMO RSMA network. Our proposed solution scheme involves a three-block procedure. The first block trains deep reinforcement learning (DRL) neural networks to obtain RSMA precoders, assuming full connectivity between APs and user equipments (UEs). The second block uses these precoders and principal component analysis (PCA) to assign APs to UEs by removing a subset of AP-UE connections. The final block fine-tunes the RSMA precoders by incorporating the associated APs into a second DRL network. To leverage the distributed nature of the cell-free network, this process is implemented in a Federated Deep Reinforcement Learning (FDRL) structure operating through the cooperation of APs and a central processing unit (CPU). Simulation results demonstrate that the proposed FDRL approach performs comparably to a benchmark centralized DRL scheme. Our FDRL approach, provides a balanced trade-off, maintaining high performance with enhanced security and reduced processing demands.

Figures (8)

• Figure 1: The relation between the KTs and KPs incorporated in our proposed structure.
• Figure 2: The system model of a cell-free MIMO network including APs, UEs, CPU and fronthaul connections. The APs transmit both common and private signal parts using the RSMA, illustrated by wide yellow beams for common signals and narrow beams for private signals.
• Figure 3: FDRL process in the cell-free RSMA network. Each agent $n$ shares its NN's weights $\boldsymbol{\theta}_n$ with the CPU. After the aggregation of all agents' parameters, the updated weights of the global NN are transmitted back to the agents.
• Figure 4: A three-stage precoder design and AP selection procedure involved at the training phases of the proposed FDRL algorithm. While the FDRL precoder pre-training and fine-tunning are executed at the agent level, the AP selection algorithm is conducted at the CPU level.
• Figure 5: Minimum UE rate per $P^{\mathrm{max}}$ in various methods for $(N,K)=(4,8)$, and $(N,K)=(8,12)$.

Theorems & Definitions (1)

• Remark 1