Federated multi-agent deep reinforcement learning with digital twin-assisted cross-layer optimization for secure energy-aware massive MIMO-NOMA mobile edge computing systems
TL;DR
Experimental results demonstrate that the proposed FMADRL with digital-twin assistance significantly outperforms existing methods in energy savings, reduced delay, improved reliability, and robustness against security threats.
Abstract
The rapid growth of Mobile Edge Computing (MEC) and massive device connectivity in next-generation wireless networks has intensified the need for intelligent, secure, and energy-efficient resource management strategies. This paper proposes a novel Federated Multi-Agent Deep Reinforcement Learning (FMADRL) framework integrated with Digital Twin (DT) - assisted Cross-Layer Optimization for secure and energy-aware Massive Multiple-Input Multiple-Output–Non-Orthogonal Multiple Access (MIMO-NOMA) MEC systems. The introduction highlights the limitations of existing centralized approaches, which suffer from high latency, scalability constraints, and privacy risks due to extensive data sharing. Several critical issues are identified including inefficient power allocation, dynamic interference, heterogeneous user demands, and vulnerabilities in data transmission. To address these challenges, the proposed methodology employs Federated Learning (FL), enabling multiple distributed agents to collaboratively learn optimal policies without sharing raw data, thereby preserving privacy. A DT model is developed to replicate the real-time network environment, enabling predictive analysis and adaptive optimization. The multi-agent deep reinforcement learning framework performs cross-layer optimization by jointly managing communication, computation, and security parameters, including task offloading, user clustering, and power control. The primary objective of this research is to enhance energy efficiency, reduce latency, improve spectral efficiency, and ensure secure communication in complex MEC-enabled networks. Experimental results demonstrate that the proposed FMADRL with digital-twin assistance significantly outperforms existing methods in energy savings, reduced delay, improved reliability, and robustness against security threats. This framework offers a scalable and efficient solution for future intelligent wireless communication systems.
