The paper presents an IRS-assisted bi-layer multi-UAV system designed to enhance the reliability of wireless networks by addressing both data freshness and security concerns. The system comprises Computational-UAVs (C-UAVs) providing Mobile Edge Computing (MEC) services and IRS-aided UAVs (I-UAVs) operating at higher altitudes to create virtual Line-of-Sight (LoS) communication.
The key highlights include:
Incorporation of exponential AoI metrics and secrecy rate optimization to tackle eavesdropping and jamming threats, achieving a balance between data freshness and security.
Introduction of a transformer-enhanced Deep Reinforcement Learning (DRL) approach to optimize the task offloading process, including trajectory planning for C-UAVs and determining optimal beamforming vectors for I-UAVs.
Comparative analysis with existing algorithms showcasing the superiority of the proposed scheme in terms of data freshness and security performance.
The system model considers a distributed mobile user network with UEs, C-UAVs, I-UAVs, a base station, and the presence of a jammer and an eavesdropper. The optimization problem aims to minimize the threshold AoI violation and AoI penalty while maximizing the achievable secrecy rate, subject to energy and collision avoidance constraints.
The proposed solution utilizes a decentralized multi-agent DRL framework with a Gated Transformer (GTr) architecture for efficient temporal modeling and joint optimization across multiple agents. The framework consists of independent actors (UAVs and IRS) and a central learner, enabling decentralized learning and collaborative execution.
The simulation results demonstrate the effectiveness of the GTr-DRL approach in improving data freshness metrics and achieving higher average secrecy rates compared to benchmark schemes, highlighting the crucial trade-off between these two performance objectives.
Naar een andere taal
vanuit de broninhoud
arxiv.org
Belangrijkste Inzichten Gedestilleerd Uit
by Poorvi Joshi... om arxiv.org 04-09-2024
https://arxiv.org/pdf/2404.04692.pdfDiepere vragen