Energy-efficient DDPG-based Federated Learning-assisted Fluid Antenna Systems with MC-NOMA

Citations

WEB OF SCIENCE

0

초록

This paper proposes a novel approach to minimize energy consumption in multi-carrier non-orthogonal multiple access (MC-NOMA) uplink communication federated learning-assisted fluid antenna systems (FL-FAS) by jointly optimizing power distribution, local iteration allocation, and antenna positioning using deep deterministic policy gradient (DDPG). Currently, federated Learning (FL) has garnered considerable attention and research interest in large-scale data communication networks, primarily due to its potential to enhance data security and preserve user privacy. Unlike traditional fixed-position antenna systems, the fluid antenna system (FAS) enables dynamic antenna reconfiguration to exploit spatial diversity and improve energy efficiency. We formulate the energy minimization problem subject to FL performance and latency constraints and develop a DDPG with a hybrid action space-based solution (FAS-EEDDPG) that adapts to channel variations while maintaining computational efficiency. Simulation results demonstrate that our approach reduces user energy consumption compared to conventional optimization methods while maintaining comparable system performance. The proposed DDPG framework effectively handles the non-convexity of the joint optimization problem and provides a practical solution for energy-efficient FAS deployment in sixth-generation (6G) networks.

키워드

Deep reinforcement learning; energy efficiency; fluid antenna systems
제목
Energy-efficient DDPG-based Federated Learning-assisted Fluid Antenna Systems with MC-NOMA
저자
Ho, Manh Cuong; Cho, Sungrae
DOI
10.1109/GLOBECOM59602.2025.11432285
발행일
2025-12
유형
Proceedings Paper
저널명
IEEE Global Communications Conference, Exhibition, & Industry Forum (GLOBECOM)
페이지
6514 ~ 6519