Scholarlink

Unmanned Aerial Vehicles (UAVs) are envisioned to integrate terrestrial and aerial infrastructure in three-dimensional Sixth-Generation (6G) networks. When combined with MEC resources (MEC-UAV), they can expand the coverage and computational capacity of terrestrial infrastructures, enabling services such as computational offloading for users. However, as UAVs are battery-powered devices, energy-efficient management is essential. In this respect, UAV trajectory optimization plays a key role, as it impacts not only the system’s operational lifetime without recharging, but also the quality of service provided by MEC-UAV systems. This work proposes a RL-based solution for MEC-UAV trajectory optimization, considering the return of remote processing results to users and the impact of the MEC-UAV trajectory on energy consumption, the proportion of admitted offloaded tasks, and the timing of result delivery to users. Our solution aims to maximize the number of tasks admitted for processing on the MEC-UAV while also increasing the proportion of successfully completed tasks and reducing energy consumption during flight operations. Results demonstrate that our approach achieves a better balance across the metrics, including low energy consumption, a high percentage of admitted and completed tasks, and a more consistent MEC-UAV trajectory, compared with existing approaches from the literature.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement Learning Based Trajectory Optimization for MEC-UAV

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Filipe Samuel da Silva,
  • Renata Kellen Gomes dos Reis,
  • Andson Marreiros Balieiro

摘要

Unmanned Aerial Vehicles (UAVs) are envisioned to integrate terrestrial and aerial infrastructure in three-dimensional Sixth-Generation (6G) networks. When combined with MEC resources (MEC-UAV), they can expand the coverage and computational capacity of terrestrial infrastructures, enabling services such as computational offloading for users. However, as UAVs are battery-powered devices, energy-efficient management is essential. In this respect, UAV trajectory optimization plays a key role, as it impacts not only the system’s operational lifetime without recharging, but also the quality of service provided by MEC-UAV systems. This work proposes a RL-based solution for MEC-UAV trajectory optimization, considering the return of remote processing results to users and the impact of the MEC-UAV trajectory on energy consumption, the proportion of admitted offloaded tasks, and the timing of result delivery to users. Our solution aims to maximize the number of tasks admitted for processing on the MEC-UAV while also increasing the proportion of successfully completed tasks and reducing energy consumption during flight operations. Results demonstrate that our approach achieves a better balance across the metrics, including low energy consumption, a high percentage of admitted and completed tasks, and a more consistent MEC-UAV trajectory, compared with existing approaches from the literature.