| Abstrakt: |
With the assistance of unmanned aerial vehicles (UAVs), wireless communication networks can provide connectivity to edge devices (EDs) even in challenging signal conditions. However, the forthcoming 6th generation mobile networks (6G) will demand increased energy, while EDs typically have limited energy resources. Furthermore, EDs often struggle to accurately track dynamic UAV positions due to time-lag mechanisms, hindering their ability to adapt emission energy dynamically. Additionally, fixed emission power settings on EDs contribute to their limited endurance. To address these challenges, we propose a deep learning-based energy-aware optimization technique (DEO) in this study. DEO dynamically adjusts the emission power of EDs to ensure that the received energy at the mobile relay UAV closely matches the receiver's sensitivity, while minimizing energy consumption. The edge server plays a crucial role by providing computational infrastructure for this task. Our approach employs the enhanced gradient-based graph recurrent neural network (Gradient GRNN) deep learning technique to predict the dynamic locations of relay UAVs. Based on these predictions, the emission energy of EDs is adaptively modified, enabling reliable connections with mobile relay UAVs while conserving energy. Through extensive simulations, we evaluate the effectiveness of various predictive networks under different time-delay conditions (0.4 s, 0.6 s, and 0.8 s). The results demonstrate that, even with communication delays, the algorithm achieves low weighted mean absolute percentage errors (WMAPE) of 0.54%, 0.80%, and 1.15%, respectively. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink